LC/MS applications in drug development

LC/MS APPLICATIONS IN DRUG DEVELOPMENT Mike S. Lee and Edward H. Kerns Milestone Development Services, Pennington, New Jersey 08534-0813 Received 28 January 1999; revised 22 July 1999; accepted 24 July 1999 I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . A. Emerging Analytical Needs . . . . . . . . . . . . . B. Integration of LC/MS into Drug Development C. Partnerships and Acceptance . . . . . . . . . . . . D. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 189 189 191 192 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 193 193 194 194 195 195 III. Accelerated Development . . . . . . . . . . . . . . . . . . . . . . A. Accelerated Development Strategies . . . . . . . . . . . . B. Quantitative and Qualitative Process Elements . . . . . C. Quantitative Process Pipeline . . . . . . . . . . . . . . . . . D. Qualitative Process Pipeline . . . . . . . . . . . . . . . . . . E. Motivating Factors . . . . . . . . . . . . . . . . . . . . . . . . F. Analysis Opportunities for Accelerated Development 1. Full-Time Equivalent . . . . . . . . . . . . . . . . . . . . 2. Sample Throughput Model . . . . . . . . . . . . . . . . 3. Elimination Model . . . . . . . . . . . . . . . . . . . . . . 4. Rate-Determining Event Model . . . . . . . . . . . . . G. Accelerated Development Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196 196 197 198 199 199 200 200 200 201 202 203 IV. Development of LC/MS . . . . . . . . . . . . A. The Elements of LC/MS Application 1. HPLC . . . . . . . . . . . . . . . . . . . . 2. Mass Spectrometry . . . . . . . . . . . 3. LC/MS Interface . . . . . . . . . . . . B. Growth of LC/MS . . . . . . . . . . . . . . II. Drug Development Overview . . . . . . . . . A. Analysis Perspectives . . . . . . . . . . . . B. The Four Stages of Drug Development 1. Drug Discovery . . . . . . . . . . . . . . 2. Preclinical Development . . . . . . . . 3. Clinical Development . . . . . . . . . . 4. Manufacturing . . . . . . . . . . . . . . . V. Strategies . . . . . . . . . . . . . . . . . . . A. Standard Methods . . . . . . . . . . . B. Template Structure Identi®cation C. Databases . . . . . . . . . . . . . . . . D. Screening . . . . . . . . . . . . . . . . E. Integration . . . . . . . . . . . . . . . . F. Miniaturization . . . . . . . . . . . . . G. Parallel Processing . . . . . . . . . . H. Visualization . . . . . . . . . . . . . . I. Automation . . . . . . . . . . . . . . . J. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 203 203 203 204 204 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206 206 208 210 210 212 212 213 213 214 216 ÐÐÐÐ Correspondence to: Mike S. Lee. Dedicated to the memory of Jerry R. Allison. Mass Spectrometry Reviews, 1999, 18, 187± 279 # 1999 by John Wiley & Sons, Inc. CCC 0277-7037/99/030187-93 & LEE AND KERNS VI. LC/MS Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Drug Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Peptide Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Glycoprotein Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Natural Products Dereplication . . . . . . . . . . . . . . . . . . . . . 4. Bioaf®nity Screening . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. Open-Access LC/MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. In Vivo Drug Screening . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. Metabolic Stability Screening . . . . . . . . . . . . . . . . . . . . . . B. Preclinical Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Metabolite Identi®cation . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Impurity Identi®cation . . . . . . . . . . . . . . . . . . . . . . . . . . . 3. Degradant Identi®cation . . . . . . . . . . . . . . . . . . . . . . . . . . C. Clinical Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Quantitative BioanalysisÐSelected Ion Monitoring . . . . . . . 2. Quantitative BioanalysisÐSelected Reaction Monitoring . . . 3. Quantitative BioanalysisÐAutomated Solid-Phase Extraction 4. Quantitative BioanalysisÐAutomated On-Line Extraction . . . 5. Metabolite Identi®cation . . . . . . . . . . . . . . . . . . . . . . . . . . 6. Degradant Identi®cation . . . . . . . . . . . . . . . . . . . . . . . . . . D. Manufacturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. Impurity Identi®cation Using Data-Dependent Analysis . . . . 2. Peptide Mapping in Quality Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216 216 217 220 224 226 227 230 234 235 236 240 242 246 248 249 251 254 255 256 257 258 259 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261 263 264 264 265 265 266 VIII. Perspectives on the Future Growth of LC/MS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266 IX. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 X. Glossary of Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270 VII. Future Applications and Prospects A. Workstations . . . . . . . . . . . . B. Multidimensional Analysis . . C. Miniaturization . . . . . . . . . . . D. Information Management . . . E. Strategic Outsourcing . . . . . . F. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The combination of high-performance liquid chromatography and mass spectrometry (LC/MS) has had a signi®cant impact on drug development over the past decade. Continual improvements in LC/MS interface technologies combined with powerful features for structure analysis, qualitative and quantitative, have resulted in a widened scope of application. These improvements coincided with breakthroughs in combinatorial chemistry, molecular biology, and an overall industry trend of accelerated development. New technologies have created a situation where the rate of sample generation far exceeds the rate of sample analysis. As a result, new paradigms for the analysis of drugs and related substances have been developed. The growth in LC/MS applications has been extensive, with retention time and molecular weight emerging as essential 188 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . analytical features from drug target to product. LC/MS-based methodologies that involve automation, predictive or surrogate models, and open access systems have become a permanent ®xture in the drug development landscape. An iterative cycle of ``what is it?'' and ``how much is there?'' continues to fuel the tremendous growth of LC/MS in the pharmaceutical industry. During this time, LC/MS has become widely accepted as an integral part of the drug development process. This review describes the utility of LC/MS techniques for accelerated drug development and provides a perspective on the signi®cant changes in strategies for pharmaceutical analysis. Future applications of LC/MS technologies for accelerated drug development and emerging industry trends are also discussed. # 1999 John Wiley & Sons, Inc., Mass Spec Rev 18: 187±279, 1999 LC/MS APPLICATIONS I. INTRODUCTION Current trends in drug development emphasize highvolume approaches to accelerate lead candidate generation and evaluation. Drug discovery-based technologies that involve biomolecular screening and combinatorial chemistry paved the way, resulting in shortened timelines and the generation of more information for more drug candidates. The impact on the overall drug development cycle has been signi®cant, creating unprecedented opportunities for growth and focus, particularly in the analytical sciences. A. Emerging Analytical Needs Perhaps a major cause of these opportunities is the fact that the rate of sample generation far exceeded the rate of sample analysis. To put this factor in perspective, consider the following example that deals with combinatorial chemistry. Prior to the advent of combinatorial chemistry technologies, a single bench chemist was capable of synthesizing ca. 50 ®nal compounds per year, depending on the synthesis. Today, chemists are capable of generating well over 2,000 compounds per year, using a variety of automated synthesis technologies. If traditional approaches to analytical support were maintained, then analysts would outnumber chemists by nearly 40 to 1! The reality of the situation has become evident: without analytical tools that could keep pace with new benchmarks for sample generation, the advantages would not be fully realized. Thus, the relationship between sample generation and analysis is a major issue in the pharmaceutical industry. Clearly, traditional approaches for analysis are not capable of meeting specialized needs created by dramatic improvements in sample generation. New technologies ®gure prominently in the success of drug development and directly impact pharmaceutical analysis activities. The recent integration of sample generation technologies such as combinatorial chemistry workstations, for example, created distinctly new requisites for analysis. Rapid, high throughput, sensitive, and selective methods are now a requisite for pharmaceutical analysis. Also, the ability to analyze trace-mixtures, using an instrumental con®guration compatible with screening approaches, emerged as an important feature. As requirements for analysis rapidly adapted to breakthroughs in sample generation, a new scienti®c and business culture aimed at decreasing costs and accelerating development became entrenched in the pharmaceutical industry. These factors combined to produce more frequent, and perhaps, new demands on analysis. In particular, these demands underscored the importance of analytical instrumentation and the creation of novel analysis strategies. For example, to keep pace with & emerging needs, the timely evaluation of new tools and applications appropriate for pharmaceutical analysis is essential. Once evaluated, the effective integration of these analysis tools represents an equally signi®cant hurdle. Here, the development of novel strategies for analysis has been an effective approach for introducing new technologies and for creating opportunities for streamlined drug development. These recent trends have been complemented by the need to determine or predict molecular and physicochemical properties of an unprecedented number of structurally diverse molecules faster than previously required and at earlier stages in the drug development cycle. Prospective methods for investigating pharmaceutical properties were born, along with data-mining techniques to search large databases. Furthermore, new experimental approaches typically generated samples that contain small quantities of analyte in complex mixtures. This combination placed a tremendous burden on the existing methods for pharmaceutical analysis. Many recent industry initiatives feature the integration of sample generating and analysis activities, resulting in new paradigms for the discovery, evaluation, and development of pharmaceuticals. The basic idea of these initiatives is to ``do more with less''. Invariably, more resources tend to be awarded to activities involved with sample generation, whereas less is received for analysis. As a result, a wide variety of analysis-based applications have been implemented. These applications emphasize ef®ciency and throughput. Three common themes arose from these activities: * * * An earlier availability of information leads to faster decision making. Integration of instrumentation with information networks is a popular approach for combining high throughput analytical information generation with drug-candidate screening. Software is a powerful resource for the coordination of analysis events and the management and visualization of data. A considerable growth in analysis methods resulted, with the primary focus being on the acceleration of drug development. New tools and strategies for analysis combined with technologies such as biomolecular screening, combinatorial chemistry, and genomics have positioned the pharmaceutical industry to harvest discovery and manufacture development opportunities. B. Integration of LC/MS into Drug Development Liquid chromatography/mass spectrometry (LC/MS)based techniques provide unique capabilities for pharma189 & LEE AND KERNS ceutical analysis. LC/MS methods are applicable to a wide range of compounds of pharmaceutical interest, and they feature powerful analytical ®gures of merit (sensitivity, selectivity, speed of analysis, and cost effectiveness). These analytical features have continually improved, resulting in easier to use and more reliable instruments. These improvements were timely and coincided with the aforementioned developments in the pharmaceutical industry. An important perspective of these events, improvements in LC/MS technology and industry change, is how LC/MS techniques became so widely accepted within every stage of drug development. It can be argued that the proliferation of LC/MS occurred not by choice but by need. For example, if a nuclear magnetic resonance (NMR)-based approach existed for the quick, sensitive, and ef®cient analysis of combinatorially derived mixtures in the early 1990's, then LC/MS would certainly have had a limited role in this area of drug development. However, at the time LC/MS provided the best performance without any rival or complement. The signi®cance of this fact is two-fold. First, LC/MS has, indeed, become the method of choice for many pharmaceutical analyses. Because the utilization of analysis technology in the pharmaceutical industry is highly dependent on perception, the breakthroughs and barriers that LC/MS has overcome provided opportunity for acceptance and a widened scope of application. Currently, LC/MS is widely perceived in the pharmaceutical industry to be a viable choice, as opposed to a necessary alternative, for analysis. Second, these events led to an increased understanding of LC/MS in such a way that practitioners and collaborators have become more diverse. The result of this diversity is a mutually shared sense of purpose within the industry, inspiring creativity and generating new perspectives on analysis. Along with timing and perception issues, there are four technical elements that have been critical for the recent acceptance of LC/MS-based techniques in the pharmaceutical industry. The ®rst is separation sciences. Simply put, the chromatographic method de®nes the pharmaceutical analysis. Chromatography provides analytical criteria to compare, re®ne, develop, and control the critical aspects of developing and manufacturing highquality drug products. Thus, it is common in industry to see LC/MS methods distinguished by the chromatographic technology and features rather than by mass spectrometry performance and capabilities. Indeed, the effective combination of a wide variety of HPLC technologies and formats with mass spectrometry played a vital role in the acceptance of LC/MS. This achievement is signi®cant because HPLC-based methods are a universally recognized analysis ``currency'', and perhaps, 190 the ®rst to be utilized throughout every stage of drug development. The second element that allows for industry acceptance of LC/MS techniques is mass spectrometry. The analytical ®gures of merit dealing with sensitivity and selectivity provide a powerful platform for analysis. However, it was not until these analytical attributes could be harnessed into a reliable, reproducible, rugged, and high throughput instrument that mass spectrometry techniques could be taken seriously as an integral tool for drug development. Though perhaps indirect, the pioneering work performed with LC/MS interfaces that use moving-belt (Smith & Johnson, 1981; Hayes et al., 1983; Games et al., 1984), direct liquid introduction (DLI) (Yinon & Hwang, 1985; Lee & Henion, 1985; Lant et al., 1985), thermospray ionization (TSI) (Blakely & Vestal, 1983; Irabarne et al., 1983), and electrospray ionization (ESI) (Whitehouse et al., 1985; Bruins et al., 1987; Fenn et al., 1989) approaches certainly played a signi®cant role in the acceptance of mass spectrometry as a routine tool for pharmaceutical analysis. Furthermore, added dimensions of mass analysis provide enhanced limits of detection for the analysis of complex mixtures and unique capabilities for structure identi®cation. The third element is information. The rate of analysis and subsequent distribution of results has grown tremendously due to the increased utilization of LC/MS and other information-rich technologies. From strictly an analysis perspective, LC/MS has demonstrated a unique capability for maintaining high quality performance and a rapid turnaround of samples. Yet, it is the accurate and ef®cient processing of information that has been essential for LC/ MS use and acceptance. As a result, LC/MS has developed unique partnerships with tools responsible for sample tracking, interpretation, and data storage. Consequently, LC/MS has become an information-rich, informationdependent technology in the pharmaceutical industry. It is highly dependent on software to integrate key analysis elements that deal with sample preparation, real-time analysis decisions, and the distribution of results. The pharmaceutical industry has bene®ted from this trend and, as a result, the derived information has been easily translated into a form that many professionals can understand, interpret, and base their decisions on. Finally, the fourth element is a widened scope of application. The fact that LC/MS is now routinely used during every stage of drug development is a powerful benchmark for acceptance. The increased performance of applications that incorporate LC/MS have, in turn, stimulated new performance levels for sample preparation, high speed separations, automated analysis, information databases, and software tools, to name a few. It is the drive for new applications, motivated by unmet industry needs, which has stimulated tremendous growth in LC/MS APPLICATIONS pharmaceutical analysis marked by invention and creativity. C. Partnerships and Acceptance What has happened in the pharmaceutical industry during this relatively short time span is truly remarkable. With the advent of advanced technologies responsible for increasing the rate of sample generation, there is strong motivation to respond with LC/MS-based analysis techniques. The understanding of principles, fundamentals, operation, and maintenance enabled researchers to improve analytical performance. Here, the power of ``seeing is believing'' led to lower barriers of acceptance as well as to a new breed of practitioners. Chemists, biologists, and other industry professionals are becoming more familiar and comfortable with LC/MS and its corresponding data as an everyday tool for analysis. The vast technical advances with LC/MS, along with a renewed emphasis on sharing, collaboration, and mutual understanding amongst disciplines, have helped researchers increase ef®ciency and overall productivity. At the same time, highly trained, highly skilled analysts are continually challenged with learning new principles in chemistry, molecular biology, and pharmaceutical development. Of course, all of the above-mentioned successes would not have been possible without basic research and the ultimate design and manufacture of analytical instrumentation. Basic research and the manufacture of high performance instruments have each played a signi®cant role in the drug development process. Continued relationship and partnership with universities and instrument manufacturers help to increase awareness, better understanding, and bridge the gaps among research, discovery, and the development of high quality pharmaceutical products. & The seven ages of an analytical method ®rst described by Laitinen (Laitinen, 1973) can be used to depict the important partnerships among academia, instrument manufacturers, and the pharmaceutical industry. These partnerships are responsible for the widened scope of application and acceptance of LC/MS that are experienced in the pharmaceutical industry today. Here, the ages of an analytical method are translated into stages of LC/MS events that lead to its routine use in the pharmaceutical industry [Table 1]. The various stages represent a continuum for LC/MS advancement, beginning with basic research performed in universities, followed by the design and manufacture of instruments, and concluding with industry benchmarks for acceptance. The ®rst and second stages involve the conception of the fundamental principles and experimental validation of the analytical potential, respectively. The conception stage of LC/MS methods was emphasized during the 1970's± 80's via basic research conducted in universities. For example, the fundamentals of interfacing an HPLC with a mass spectrometer were studied (Arpino et al., 1974; Carroll et al., 1975; Arpino, 1982) and mechanisms of ionization were characterized (Thomson & Iribarne, 1979; Blakely et al., 1980; Whitehouse et al., 1985). The validation stage of the analytical method represents the convergence of interest among research, instrumentation, and potential application. The results and interest generated from the basic research that deal with LC/MS led to signi®cant investments in technology from instrument manufacturers. Applications dealing with pharmacokinetic (Covey et al., 1986) and biomolecular (Wong et al., 1988) analysis showed signi®cant promise, insight, and direction. The market potential of an LC/MS instrument, providing expanded capabilities over gas chromatography/mass spectrometry (GC/MS) and HPLC methods for pharmaceutical analysis, was realized. The availability of commercial instruments provided the TABLE 1. The seven stages of the LC/MS analytical method that result from partnership within academia, instrument manufacturers, and industry. Stage Event Conception Validation Fundamental principles outlined Analytical potential experimentally validated Availability Instruments developed/manufactured Foundation A platform of performance established Application A widened scope of application Acceptance Used as a routine, standard method Senescence Replaced by newer methods Activity Basic research Basic research; applied research; technology investments; product development; targeted pharmaceutical applications Commercial instruments solid; method development; applied research Method development/re®nement; analysis benchmarks; quantitative bioanalysis methods established; new product development Unique methods developed to address sample generating technologies and traditional analyses for the identi®cation of biomolecules, metabolites and natural products Development of fully automated methods for high throughput analysis; open-access instruments; standard methods; outsourcing Decline in applications, utility and popularity? 191 & LEE AND KERNS pharmaceutical industry with LC/MS capabilities plus training, service, and technical support. Applied research directed towards meeting current industry needs ensued, with active participation and collaboration from university; manufacturing, and pharmaceutical-led research groups (Covey et al., 1991; Weintraub et al., 1991; Aebersold et al., 1992; Weidolf & Covey, 1992). The ability to reliably develop and re®ne LC/MS-based methods helped to establish a solid fundamental foundation of this technique. Here, the utility of LC/MS methods for quantitative bioanalysis was benchmarked as the industry standard in the early 1990's for performance and ef®ciency (Fouda et al., 1991; WangIverson et al., 1992). New products were designed and developed exclusively for LC/MS performance. A widened scope of application occurred with the development of unique LC/MS-based methods for the analysis of novel pharmaceuticals. Analysis methods were easily developed and re®ned in the pursuit of opportunities created by the use of traditional, time-consuming procedures. Applications that deal with biomolecule analysis, drug metabolism and pharmacokinetics, natural products research, and combinatorial chemistry represent some important areas of LC/MS diversi®cation and will be discussed in the following sections of the review. Perhaps the most signi®cant benchmarks for industry acceptance of LC/MS appeared when fully automated methods were developed for high throughput analysis and when collaborators (i.e., sample generators) themselves became analysts via the purchase of instruments or routine use of open access instruments (Taylor et al., 1995; Pullen et al., 1995). These methods and approaches were developed primarily in response to samplegenerating technologies. And this step represents the present stage of LC/MS methods in the pharmaceutical industry. Although the scope of application continues to grow, the routine use of LC/MS is now embraced by pharmaceutical researchers. Standard methods that incorporate highly specialized features are routinely developed for a variety of novel applications. Furthermore, many LC/MS applications that deal with quantitative bioanalysis (i.e., pharmacokinetic studies) are frequently outsourced to contract analytical laboratories. Thus, the routine use of LC/MS is a benchmarked commodity for drug development. The ®nal stage, senescence, does not appear to be a prospect in the near future, but a decline in popularity and application will likely occur sometime. Perhaps the onset of this stage will be triggered by the divergence of academic, instrument manufacture, and industry interests. However, the current industry trends highlight the tremendous challenge of drug development and an expanding need for tools that provide for fast, sensitive, 192 and selective analysis of drugs and drug related compounds. D. Overview This review focuses on LC/MS applications in drug development. The role of LC/MS in the pharmaceutical industry during the past decade is examined, and key elements for recent success are illustrated to include signi®cant advances in instrumentation, methodology, and application. The applications are highlighted with reference to the analysis opportunity and analysis strategy implemented. Examples that depict unique advantages of LC/MS during speci®c stages of drug development are selected to capture the signi®cant events and/or initiatives that occurred in the pharmaceutical industry during this time. In many instances, an analysis is provided to illustrate the result or development situation if LC/MS was not utilized. In these cases, the impact (number of samples) and value (cost) on drug development is highlighted independent of the technical features of LC/ MS analysis. We hope that these unique industry perspectives will offer a ``currency'' and assist in understanding the events that resulted in the proliferation of LC/ MS throughout the drug development cycle. The review concludes with perspectives on future trends and some thoughts on the future direction of LC/ MS applications in the pharmaceutical industry. New standards of analytical performance are discussed with regard to throughput and capacity. A prospective look at how higher standards of analytical performance in the pharmaceutical industry will affect relationships with academia and instrument manufacturers is featured. These sections extend the initial thesis of accelerated development to include new analysis bottlenecks and perspectives on analysis issues and industry needs. II. DRUG DEVELOPMENT OVERVIEW Drug development may be de®ned as a series of specialized events performed to satisfy criteria, internal (i.e., competitive industry benchmarks) and external (i.e., regulatory compliance), to yield a novel drug. Much attention has so far been given to the various activities in drug development. These accounts primarily have a sample-generating perspective. For example, the timely review of innovations in automated synthesis stimulated new paradigms for drug discovery (Gallop et al., 1994; Gordon et al., 1994; Desai et al., 1994). The combined vision and depth of knowledge has had a profound effect on the pharmaceutical industry, helping to promote a greater understanding of technology and the development of new strategies for discovering novel lead candidates. LC/MS APPLICATIONS A. Analysis Perspectives Traditionally, the role of analytical technologies is to respond to a pharmaceutical event, rather than to lead one. A complementary perspective from an analytical point of view can provide substantial insight into relevant drug development issues. This insight may not be intuitively obvious from a sample-generating (i.e., chemistry, biology) approach. And, when sample analysis activities are taken into consideration as an equal partner with sample-generating activities, global, and perhaps integrated, strategies for drug development may be derived. The view here is that analysis insights provide unique perspectives and opportunities to contribute to the design, development, and manufacture of high-quality drug products. This statement does not intend to imply that this process does not occur in the pharmaceutical industry, only that there is opportunity for more such interaction and collaboration. With that said, sample analysis can be viewed as a dependent partner with sample generation. Without analysis, sample generation yields no information for satisfying drug development criteria, and vice versa. Therefore, no matter how quickly or ef®ciently samples are generated, the bene®ts are not realized unless they are analyzed in an equally ef®cient manner. Identical, or perhaps, matched criteria for performance (i.e., speed, throughput, compatibility) is, therefore, required for sample-generating and sample-analysis responsibilities. B. The Four Stages of Drug Development In recent years, the drug development process has become more complex and highly competitive while the sample analysis contributions have become increasingly important. This perspective recognizes the impact of sample analysis activities and the corresponding information that must be accumulated throughout the various stages of development. At present, drug development consists of four distinct stages: (1) drug discovery; (2) preclinical development; (3) clinical development; and (4) manufacturing [Table 2]. Each development stage is geared toward the swift accomplishment of goals and objectives. Each stage & culminates in a speci®c corresponding milestone: lead candidate; investigational new drug (IND)/clinical trial application (CTA); new drug application (NDA)/marketing authorization application (MAA); and sales. The IND and NDA are the required regulatory documents ®led in the United States; the CTA and MAA are required in Europe. For the successful completion of each milestone, a diverse array of analyses is required. The focus is generally unique to the speci®c stage of development and is a determining factor for criteria for analysis. For example, drug discovery approaches typically require rapid, high-throughput screening methods with the purpose of selecting a lead candidate from a large number of diverse compounds. Analyses that emphasize quick turn-around of results are desirable. As the discovery of lead candidate moves forward through the drug development cycle, the analysis requirements become more focused. In preclinical development, the main goal is directed towards the swift ®ling of the IND/CTA. Thus, analyses are aimed at providing more speci®c and detailed information that evaluate drug properties. This stage of drug development is also the ®rst point at which regulatory issues are addressed; therefore, the use of validated analytical methods and the compliance with Food and Drug Administration (FDA) guidelines are critical. For example, pharmaceutical scientists interact with regulatory agencies to establish impurity limits so that development and approval phases can proceed in a predictable fashion. Thus, the generation and analysis of drug products are conducted in accordance with FDA good manufacturing practice (GMP) and good laboratory practice (GLP) regulations, respectively. During the clinical development stage, the lead candidate (now an IND or CTA) is fully characterized in humans. Subsequent analyses continue to be performed under strict protocol and regulatory compliance to register the drug for NDA/ MAA. Once the NDA/MAA is approved, analyses are focused on speci®cations to provide regulatory compliance and to assure quality during the manufacturing stage. Below is a brief summary of the four stages of drug development. Signi®cant events are highlighted with respect to their relationship to analysis requirements. TABLE 2. The four stages of drug development and their corresponding milestone and analysis emphasis. Development Stage Milestone Analysis emphasis Drug discovery Lead candidate Screening Preclinical development Clinical development Manufacturing IND/CTA ®ling NDA/MAA ®ling Sales Evaluation Registration Compliance LC/MS analysis activities Protein identi®cation; natural products identi®cation; metabolic stability pro®les; molecular weight determination for combinatorial/medicinal chemistry support Impurity, degradant, and metabolite identi®cation Quantitative bioanalysis; structure identi®cation Impurity and degradant identi®cation 193 & LEE AND KERNS 1. Drug Discovery The goal of the drug discovery stage is to generate a novel lead candidate with suitable pharmaceutical properties (i.e., ef®cacy, bioavailability, toxicity) for preclinical evaluation. The drug discovery process is often initiated with a decision to begin research on a new biological target. Studies are performed to characterize and de®ne the target to establish the biological rationale. Highthroughput screening assays are developed in conjunction with a formal medicinal chemistry program. Potential lead compounds contained in natural product sources or from the extensive database of a synthetic compound library are screened for activity. Lead compounds identi®ed from screening efforts are optimized in close collaboration with exploratory metabolism programs and drug safety evaluations. In 1997, it was estimated that the synthesis and screening of ca. 100,000 compounds is typically required for the discovery a single quality lead compound (Baxter, 1997). The process of identifying a lead compound can take up to 2±4 years. Optimization of the resulting lead may take an additional 1±2 years. The drug discovery stage culminates with a decision to advance a lead candidate for preclinical development studies and more extensive evaluation. Thus, the drug discovery stage involves three primary analysis activities: target identi®cation; lead identi®cation; and lead optimization. A recent survey forecasted the impact of new technologies on drug discovery and preclinical development activities (Banerjee & Rosofsky, 1997). Figure 1 illustrates the maximum and minimum development times in 1996 and projections for the year 2000. The results suggest that lead identi®cation activities would decrease from an average of 15 months in 1996 to just over 6 months in 2000. Generally, a shorter and more predictable time-scale is projected for drug discovery-related activities. 2. Preclinical Development The preclinical stage of drug development focuses on activities that are necessary for ®ling an IND/CTA. The completed IND/CTA contains information that details the drug's composition and the synthetic processes utilized for its production. The IND/CTA also contains animal toxicity data, protocols for early phase clinical trials, and an outline of speci®c details and plans for evaluation. Process research, formulation, metabolism, and toxicity are the major areas of responsibility in this development stage. Analysis activities that feature LC/MS primarily focus on the identi®cation of impurities, degradants, and metabolites. Generally, preclinical development activities are completed in 10±15 months; however, shorter timelines are predicted (Banerjee & Rosofsky, 1997). Preliminary data from early animal toxicology and pharmacokinetic studies are obtained to determine the optimal doses and dosage form for the initial phase I clinical trials (see next Section on clinical development). These early studies also provide insight into the extent of safety monitoring necessary during phase I. These data support the IND/ CTA submissions and clinical development for all indications. All issues that are expected to attract the attention of regulatory agencies are identi®ed at this time, and are addressed in the clinical plan. During preclinical development, the structure, physical, and chemical characteristics, and stereochemical identity of the IND/CTA candidate are fully characterized. This information, for example, is required for the FIGURE 1. The maximum and minimum development times for a drug in 1996 and projections for the year 2000 (Banerjee & Rosofsky, 1997). 194 LC/MS APPLICATIONS Chemical Manufacture and Control (CMC) section of the IND. Appropriate bioanalytical methods are developed for the evaluation of pharmacokinetics, typically a series of studies focusing on absorption, distribution, metabolism, and excretion (ADME) in toxicology species, as well as systemic exposure and metabolism in toxicological and clinical studies. Characterization of the new drug substance is then initiated, which includes preliminary information on stability, preparation, and control for manufacturing purposes. Preliminary information on the composition, manufacture and packaging, and control of the investigational drug product are obtained. Registration dossiers require a full description of the manufacture and control of the new drug substance. Stability of the new drug substance and drug products for at least 6 months is required. Appropriate data con®rming the stereochemical homogeneity of the drug substance during stability studies, validation of analytical methods, and manufacture of the drug products are also required. 3. Clinical Development The clinical development stage is comprised of three distinct components or phases (I±III), and culminates in the ®ling of the NDA/MAA. Each phase involves process scale-up, pharmacokinetics, drug delivery, and drug safety activities. During phase I clinical development, the compound's safety and pharmacokinetic pro®le is de®ned. The determination of Cmax, AUC, elimination half-life, volume of distribution, clearance and excretion, and potential for drug accumulation is made in addition to studies that provide estimates of ef®cacious doses. Dose levels typically range from 10 to 2000 mg, with half the patients on placebo. Patients are carefully observed, monitored, and questioned about side effects. Plasma samples are obtained at appropriate time-points following administration of the drug, from which plasma-time concentration curves are determined. Urine is collected just prior to drug administration and at subsequent time points to provide an estimate of the rate of urinary excretion of drug and/or metabolites. Urine is also collected during the study to provide insight into metabolic stability. Typical pharmacodynamic evaluations include blood glucose monitoring and blood pressure. Safety evaluations include physical examinations and clinical laboratory tests (i.e., liver function tests) performed before dosing and before discharge. After acceptable safety and pharmacokinetic data are observed in phase I trials, phase II studies are initiated with the goal of establishing ef®cacy, determining the effective dose range, and obtaining safety and tolerability data. In phase II, the dose and dosing interval to be used in the patient population are de®ned as well as the estimated & no-effect dose. Phase II studies may require 1±1.5 years to complete. The goal of phase III is to complete the human safety and ef®cacy programs and to secure approval. Programs are designed to demonstrate clinical ef®cacy superior to a placebo. Placebo-controlled, double-blind, randomized trials are typically performed on 150±200 patients that last up to six months. Additional studies with comparative agents may be performed to satisfy registration requirements and to help to determine marketing and pricing strategies. 4. Manufacturing As with earlier stages of drug development, the transition to the manufacturing stage begins while the previous clinical development activities are moving toward its milestone (NDA/MAA). Plans begin well in advance to ensure manufacturing capability for the production of large quantities of synthesized drug substance and drug product. Once formulated, the drug is packaged and readied for distribution to pharmacies. Manufacturing processes and facilities undergo a preapproval regulatory review and periodic inspections once production is in progress. Analytical procedures and information databases are formalized into standard operating procedures (SOPs) and product speci®cations. This information and technology are formally transferred for routine monitoring and release by quality control (QC) scientists in manufacturing groups. Several other events occur simultaneously with these activities. Some events focus on extending therapeutic applications and formulations. Clinical studies are conducted to extend the diseases (indications) for which the drug is proven ef®cacious and safe. For example, TAXOL1 was initially approved for the treatment of ovarian cancer, and was later extended for the treatment of breast cancer after follow-on clinical studies demonstrated ef®cacy for the new indication. In addition, new product formulations are investigated to extend the routes of administration for patient convenience, increased bioavailability, and new disease therapies. For example, a drug that was initially developed as an injectable product may be formulated as a tablet for oral administration. Some manufacturing events are triggered by business considerations. Changes in processes, such as the synthetic production of a drug previously isolated from natural sources, can ensure the expanded supply and a more economical production. During the manufacturing stage, comparisons are made to other drug products in the same category, including stability, bioavailability, and purity. With the direct advertising of pharmaceuticals and more wide-spread information on drugs, patients are taking a more active role in therapy decision-making. 195 & LEE AND KERNS Thus, comparative information is of interest during the transition from exclusive patent-protected drugs to the open generic market. Also, companies monitor for the infringement of process patents by other organizations. Other events that occur during the manufacturing stage indicate an immediate need for analytical troubleshooting. Long-term stability studies (LTSS) may reveal new degradants in retained lots, such as particulates in an injectable. Adverse patient events are reported and investigated, and consumer complaints about off-taste or odor are immediately addressed. Manufacturing interruptions occur due to contamination by packaging materials or unexpected impurities that exceed product speci®cations. Also, with the growing use of outsourced services for the product manufacturing of intermediates, drug substance and drug product, out-of-speci®cation results must be immediately addressed. III. ACCELERATED DRUG DEVELOPMENT Accelerated drug development strategies focus on producing drug candidates and accomplishing goals in less time than with traditional development approaches. The key elements of accelerated development strategies involve the early identi®cation of the most promising drug candidates. Opportunity exists to expose weaknesses early in the drug development cycle and to make decisions on how to address (i.e., re®ne) or drop from further development. The premise of this approach focuses on maximizing the return on investment via the cost-effective application of resources. The return on investment is captured by bringing a pro®table drug to market faster and by utilizing resources more ef®ciently. A. Accelerated Development Strategies From large-scale sample-generating efforts, two accelerated development strategies involving analysis have emerged [Table 3]. The ®rst involves quantitative process approaches aimed at achieving high throughput analysis. The focus is on sample volume with the primary objective of accommodating increases in sample generation. This approach is typically accomplished with the addition of more resources and/or improved methods for analysis, and is highly effective when a ``go'' decision is made for lead candidate development. The activities associated with faster analysis are generally independent of sample-generating approaches. The incorporation of an automated task into an existing method for analysis is the example of a quantitative process approach. The second strategy involves the use of qualitative process approaches which are supposed to eliminate candidates that have unsuitable characteristics weak from the drug development pipeline. Here, analyses that focus on pharmaceutical properties are preformed during the earlier stages of drug development. This approach usually requires the development of a new application that is highly integrated with the sample-generating responsibilities that lead to faster decisions to ``stop'' development activities. Predictive in vivo and in vitro models of metabolic stability are examples of a qualitative process approach. Accelerated development exploits the relationship between quantitative and qualitative process approaches. TABLE 3. The characteristics of quantitative and qualitative process approaches for accelerated drug development, featuring analysis aspects. Accelerated development strategy Objective Focus Quantitative process approach High throughput analysis Sample volume Qualitative process approach Pharmaceutical properties 196 Elimination of week candidates Analysis features Provide increased support when a ``go'' decision is made for development Increased resources and/or improved methods of analysis Independent of sample generation Intra-laboratory integration (i.e. automation) Generate information of ``stop'' development Development of a new application and implemented during early stages of development Integrated with sample generation Inter-laboratory integration (i.e., predictive models) LC/MS APPLICATIONS Often, it is the balance between the two approaches that creates new opportunities for development success as well as signi®cant challenges for analysis. Typically, one approach is developed in response to the other, followed by re®nement and integration. B. Quantitative and Qualitative Process Elements To help illustrate the dynamics of quantitative and qualitative process approaches to accelerated drug development, a hypothetical pipeline is constructed in Fig. 2. This pipeline represents a ``snap-shot'' of drug development activities during a 12-month period. Here, the focus is on the drug discovery, preclinical, and clinical development stages. Quantitative process elements are de®ned as the actual number of compounds (i.e., lead compounds, lead candidates, IND/CTA's, NDA/MAA's) in each development stage. Qualitative process elements are de®ned as the development activity (i.e., metabolism, pharmacokinetics, toxicity) utilized to evaluate and select compounds for advancement to the next stage. Quantitative process approaches are typically benchmarked by productivity, derived from the number of compounds (or samples) in each development stage, whereas qualitative approaches are benchmarked by ef®ciency, corresponding to the rate at which drug candidates (or samples) ¯ow through the various stages in the pipeline. The relationship of these accelerated development elements provide a useful tool to highlight the features of quantitative and qualitative process approaches, and are important factors in identifying strategic analysis opportunities for increased productivity and ef®ciency. The application of these approaches to accelerated drug development has become more essential due to aggressive sample-generating technologies such as combinatorial chemistry. The ability to reach higher levels of performance (i.e., high throughput) without sacri®cing the quality of data (i.e., accuracy) is desirable. These & approaches typically involve the re®nement of an existing activity or the creation of an entirely new one. Re®nement approaches lead to a decreased cycle time via the faster and more ef®cient analysis of samples. Here, automation is an obvious and desirable goal to speed up the analysis, to optimize the measurement, and to coordinate diverse tasks. A tremendous emphasis is placed on certain aspects of analysis such as sample preparation and data processing, and data management. Once considered to be peripheral to the actual analysis, these activities have become important elements of high throughput analysis. The creation of new analysis approaches is a strategic complement to re®nement. Here, the object is not necessarily focused on replacing an existing method, but rather supporting it by providing an opportunity to screen and/or predict the likelihood of success. This approach is effective to generate useful information while simultaneously providing a measure of relative order or ranking. Though qualitative process approaches to accelerated development actually add a step to the drug development cycle, they provide a highly ef®cient method to make decisions on a compound or a series of compounds to move forward for further analysis. The re®nement or creation of new approaches may result in the elimination of existing activities. For example, con®rmation of the structure of newly synthesized lead compounds traditionally involved an extensive use of NMR. Once reliable LC/MS methodologies became available and their performance was benchmarked, they were soon accepted as an exclusive method for rapid structure con®rmation of lead compounds at an earlier stage of the lead identi®cation process. The likelihood of success, or failure, is an important strategic factor in drug development. Generally, drug development proceeds with a multitude of events and demands that superimpose onto organizational sequence, regulatory compliance, and diverse analysis needs. Thus, faster development timelines typically occur when out- FIGURE 2. Hypothetical drug development pipeline, illustrating activities during a 12-month period. Three stages of drug development are indicated at the bottom of the ®gure. The corresponding milestones for each stage are indicated at the top of the ®gure. 197 & LEE AND KERNS comes are predictable. Drug development is often slowed by the unpredictable. In either case, opportunities exist for new methodologies to address the unpredictable needs. Great skill, or perhaps fortunate circumstance, is required to anticipate needs and to devise effective plans. So what happens when the pipeline is ®lled? Two examples serve to illustrate the features of quantitative and qualitative process approaches, and the corresponding models for accelerated drug development are described below. Each model focuses on the dynamics of quantitative and qualitative approaches, and emphasizes opportunities that impact analysis and the overall accelerated drug development situation. C. Quantitative Process Pipeline Figure 3 (top) shows a quantitative drug development pipeline that contains 100,000 lead compounds. In this hypothetical model, ten lead candidates are generated in the drug discovery stage. In the subsequent preclinical development activities, 80% of the lead candidates are successfully transferred to clinical development. The resulting eight IND/CTA candidates are presented to the clinical development stage, which involves three phases of evaluation. During this stage, 50% successfully pass through to phase II, and 50% of the phase II are successfully transferred to phase III. A single NDA/MAA results at the end of the pipeline, corresponding to a 50% success rate from phase III. In this model, every 100,000 lead compounds generated in drug discovery results in a single NDA/MAA. The overall process can be analogously viewed as a multi-step synthesis. Each step involves the critical evaluation of candidate properties for conversion to the next development stage. Each development stage has a corresponding yield, which is indicative of productivity. Each process has a corresponding rate, which is indicative of ef®ciency. The yield and rate are dependent on the ``starting material'' (i.e., volume and quality of the lead compound) as well as on the analysis tools and strategy used to generate the information that is necessary to ``convert'' candidates to the next development stage. To illustrate the effects of a purely quantitative process approach to accelerated drug development, the pipeline is ®lled with more discovery leads while the qualitative process elements (i.e., rates of conversion) remain the same for each stage (bottom, Fig. 3). In the ®rst example, the number of lead compounds in drug discovery is doubled totaling 200,000. The results are arithmetic as each quantitative benchmark is doubled, resulting in the production of two NDA/MAA compounds. This model is extended further to illustrate the effects when 300,000 and 400,000 lead compounds are introduced into the drug development pipeline to generate three and four NDA/ MAA compounds, respectively. This approach to accelerated development emphasizes high throughput activities, and thus, sample volume is typically high. The approach targets speci®c end-points and benchmarks that aim at increasing productivity. The strategy is based on introducing more lead compounds into the pipeline to obtain more NDA/MAA ®lings. The logic is highly intuitive; if the number of compounds entering the pipeline is doubled, then the number of compounds leaving are doubled as well. Thus, a quantitative process pipeline is volume-enhancing, and is driven by ``thermodynamic'' properties. An assumption is made that ef®ciency will be maintained as the sample volume is increased. For this relationship to occur, improvements in analysis throughput are required. Without these improvements, quantitative process approaches FIGURE 3. A quantitative process pipeline model, illustrating the transfer of successful drug candidates from one stage to the next. Quantitative increases in drug candidate sample-generation volume are complemented by proportional increases in resources for sample analysis. 198 LC/MS APPLICATIONS to accelerated drug development require proportional increases in resources (i.e., personnel, instrumentation, space) for analysis. D. Qualitative Process Pipeline Qualitative process approaches to accelerated development target the activities involved with converting (or eliminating) the drug candidate through the various development stages within the pipeline. Using the same 12-month model as described for a quantitative process, a pipeline that contains 100,000 lead compounds is shown in Fig. 4. In this approach, identical benchmarks for performance may be obtained; however, the potential of improving the overall ef®ciency via the elimination of weak drug candidates exists by producing one NDA/MAA in less than 12 months. This approach is extended to target improvements with the rate of conversion between speci®c development stages. For example, Fig. 4 illustrates the effect on overall productivity if the number of lead candidates transferred to the IND/CTA stage are reduced to six, corresponding to a 60% conversion rate. With two less compounds to support during the clinical development stage, a signi®cant amount of resources can be saved. Figure 4 also illustrates the effect of qualitative process enhancements throughout the drug development pipeline without increasing the number of preclinical leads. A qualitative process approach to accelerated development emphasizes the elimination of drug candidates from the pipeline. This approach targets the advanced evaluation of speci®c pharmaceutical properties. The reward is not necessarily intuitive; introduce a qualitative & process change earlier in the drug development pipeline, resulting in the proactive identi®cation of promising lead compounds and the utilization of fewer resources during the costly clinical development phases. This approach is driven by ``kinetic'' properties with an acute focus on ef®ciency (i.e., time is money!). A key element of a qualitative process approach is the incorporation of new applications for drug candidate evaluation at early stages of drug development. This action provides a mechanism to regulate the ¯ow of drug candidates through the development pipeline. In this way, rational decisions are made, resulting in the selection of speci®c drug candidates for accelerated (or delayed) development. Without this approach, traditional methods of sample analysis would be left to deal with the bulk of drug candidates and their corresponding samples. E. Motivating Factors The motivation to implement quantitative and/or qualitative process approaches in drug development is understandable. Consider the fact that a $1 billion drug (annual) generates approximately $3 million sales per day. Therefore, the addition of an equivalent drug from a revenue standpoint is quite lucrative. This ®gure is derived from a pure quantitative process approach. When applied throughout an entire drug development pipeline, a qualitative process approach is equally profound. For example, within the framework of an organization operating with a $1 billion research and development budget and producing two NCE's per year, the cost of developing a single NCE per year is ca. $500 million per year. Therefore, development costs would equal $2 FIGURE 4. A qualitative process pipeline model, illustrating the elimination of unsuccessful drug candidates at earlier stages. Qualitative improvements in ef®ciency and the rate at which drug candidates ¯ow through the pipeline result from the incorporation of new applications for drug candidate evaluation at earlier stages of drug development. 199 & LEE AND KERNS million per day per NCE. This ®gure is consistent with recent cost estimates of bringing an NCE to market (Drews & Ryser, 1997). The notion of accelerating the development of a $1 billion drug is powerfully motivating, and can result in a signi®cant source of revenue. Furthermore, the bene®t from the early elimination of drug candidates for further development may result in considerable savings. This analysis suggests that future accelerated drug development activities require quantitative and qualitative process considerations. Experience suggests that there exists an iterative relationship, if not balance, between the two approaches. Whatever the case may be, future approaches to accelerated drug development is likely to continue to focus on the number of compounds/candidates in each development stage, and on the rate at which they ¯ow through the pipeline. These features are likely to combine elements of quantitative and qualitative process approaches. An extension of the previously described hypothetical pipeline is shown in Fig. 5 to illustrate the dynamics of a combined quantitative and qualitative process approach. Combined approaches provide a synergistic mechanism to focus on goals for productivity and benchmarks for ef®ciency. Finally, these models indicate that combined quantitative and qualitative process approaches to accelerated development require strategic changes (affecting sample-generating and sample-analysis activities) with only moderate increases in resources. F. Analysis Opportunities for Accelerated Development So what opportunities exist for accelerated development? Are these opportunities predictable? And what role can analysis techniques such as LC/MS play? Perhaps a straight-forward approach involves the determination of analysis costs. 1. Full-Time Equivalent The ®rst step in calculating the cost of analysis in the pharmaceutical industry is to determine the yearly cost of the analyst. This cost is referred to as a full-time equivalent (FTE). A typical FTE calculation is shown in Table 4. The cost per year can be estimated as about $250,000. This cost would include factors for salary, bene®ts, space (laboratory, of®ce, common rooms), management, support staff, capital equipment, and operating supplies, which increase or decrease the FTE cost. The time available to a researcher for sample analysis is estimated at ca. 200 days per year. This number takes into consideration the allotment of time for holidays, vacation, training, conferences, company business meetings, and maintenance and repair of laboratory equipment. This analysis provides a cost of about $1,250 per day for each productive researcher. 2. Sample Throughput Model Once the ®gures for an FTE are established, the cost corresponding to the number of samples that can be analyzed per day is calculated. In Fig. 6, a cost pro®le is illustrated for LC/MS analyses of up to 100 samples per day. This model indicates analysis throughput from a quantitative process approach, and provides a ®scal illustration of the impact the analysis may have on drug development. For example, LC/MS-based strategies, which have been demonstrated to increase the rate of FIGURE 5. A combined process pipeline model, incorporating quantitative and qualitative elements. Combined approaches for drug development result in increased ef®ciency and rate with only moderate increases in resources due to strategic process changes. 200 LC/MS APPLICATIONS & TABLE 4. A calculation of the costs associated with a full-time equivalent (FTE) analyst. Budget Annual Productivity days/yr Salary Bene®ts Space Support staff & management 65,000 28,000 52,000 15,000 Capital equipment Operating supplies 45,000 45,000 Working days Holidays Vacations Training Company meetings Conference Maintenance/repair Total 250,000 Cost 260 (13) (15) (5) (15) (5) (7) 200 days/yr $1,250/day $156/h FIGURE 6. Fiscal illustration of the impact of analysis ef®ciency on drug development costs, based on the FTE costs indicated in Table 4. An LC/MS application that increases the rate of sample analysis from 10 sample/day to 50 samples/day would reduce the cost of analysis from $125/sample to $25/sample. TABLE 5. A model that projects the yearly costs associated with the ®rst three stages of drug development. Drug discovery Number of drug candidates Estimated cost of developmenta Cost per drug candidate Preclinical development 100,000 $150 million $1,500 10 $50 million $5 million Clinical development 8 $300 million $40 million a Assuming that the cost to develop a single new drug equals $500 million/yr. sample analysis by two- to ten-fold in the pharmaceutical industry, can be expected to reduce the cost per analysis by a corresponding ratio. 3. Elimination Model A second model, which can be used to illustrate the cost advantages of LC/MS in drug development, highlights the ®scal bene®t of qualitative process approaches used to eliminate candidates from the drug development pipeline. In this example, the number of compounds in various stages of drug development is represented, using the previously described drug development pipeline models [Figs. 3±5]. From this model, the costs associated with each stage of drug development is estimated [Table 5]. Assuming that the cost to develop a single drug candidate equals $500 million per year, a ®gure for the cost per candidate is calculated for the three stages that lead up to NCE/MAA approval. For example, the cost of preclinical and clinical development is estimated at $5 million and $40 million, respectively. From this information, the ®scal impact of qualitative process approaches is evaluated. For example, the amount of savings that result from the early elimination of a 201 & LEE AND KERNS FIGURE 7. Cost saving associated with the early elimination of drug candidates that would have failed in clinical development. Early elimination of two drug candidates that would have failed in clinical development would save approximately $80 million within a year. single drug candidate that would have failed in clinical development is illustrated in Fig. 7. Clearly, there are strategic opportunities to select or eliminate candidates at earlier stages of drug development. This analysis reveals the savings that result from the elimination of two compounds from the clinical development stage, totaling $80 million. 4. Rate-Determining Event Model Recent trends in accelerated drug development illustrate the important role of LC/MS technology. As prospective LC/MS analysis approaches continue to be accepted, more activities that are identi®ed as ratedetermining can be investigated. In recent years, these FIGURE 8. Additional sales associated with an earlier introduction of a new drug due to saving days of work from rate-determining activities during the drug development process. Two days saved for each of ten ratedetermining activities would allow the introduction of the drug 20 days earlier, and would result in $60 million additional sale for a drug with $1 billion/year sales. 202 LC/MS APPLICATIONS rate-determining events correspond to proteomics for drug target identi®cation, combinatorial chemistry, and pharmacokinetics, to name a few. These areas continue to undergo quantitative and qualitative process re®nement for additional gains in productivity and ef®ciency. Future approaches to accelerated drug development may focus on activities that do not routinely utilize LC/MS. These activities may include screening, solubility, absorption, formulation screening, stability, process research, toxic mechanisms, and direct patient monitoring, for example. As these rate-determining activities are accelerated with the application of LC/MS techniques, the cost savings can be highly signi®cant. For example, if a hypothetical drug development pipeline contains a total of 20 rate-determining activities, then the application of LC/MS for accelerated development can be measured [Fig. 8]. Using the same model of a drug with $1 billion in annual sales ($3 million per business day), the additional income associated with earlier introduction of the drug is calculated. In this model, the application of LC/MS on ten rate-determining activities that result in two days of accelerated development per activity allows the launch of the drug to proceed 20 days earlier, for additional sales that total $60 million. & techniques (Snyder, 1995; Covey, 1995; Thomson, 1998) and the mechanistic aspects of ionization (Bruins, 1998) have been recently reviewed. Efforts to develop and re®ne an interface for introducing a ¯owing liquid HPLC system into a high vacuum mass spectrometry environment were fueled by a strong notion that the combination would be unique and would provide powerful advantages for analysis (Arpino et al., 1979; Arpino, 1982). The combined efforts and vision from a diverse group of pioneering researchers helped to create a unique combination for pharmaceutical analysis. A. The Elements of LC/MS Application From an applications standpoint, the partnership of HPLC and mass spectrometry bene®ted greatly from the tradition of HPLC within the pharmaceutical industry, and from the growing trend to obtain structural information during earlier stages of drug development. Ultimately, it has been the power of HPLC to resolve and the ability of mass spectrometry to identify that enabled LC/MS to integrate effectively with drug development and to solve problems. 1. HPLC G. Accelerated Development Perspectives The risk in drug development is tremendously high. Thousands of lead compounds are synthesized and tested. Perhaps only one in 100,000 lead compounds is introduced into clinical trials. Only one in ten lead candidates is registered and marketed. Such a high degree of failure is further exacerbated when one realizes that only one in three marketed drugs returns their R and D investment. Therefore, there exists a great deal of motivation to constantly evaluate drug development activities and to integrate new approaches that provide advantages for faster development and economical research. This motivation has led the pharmaceutical industry to search for new opportunities that deal with analysis and accelerated development. The successful application of new technologies to drug development is noteworthy. Although the end result cannot be reliably predicted, each application has its origin with some speci®c development crisis. Crisis creates a poignant stoppage in work ¯ow and a bottleneck in drug development. Bottlenecks demand a search for something new. IV. DEVELOPMENT OF LC/MS The signi®cant events and overcome obstacles that led to the development and application of LC/MS-based HPLC-based techniques have long been a traditional mainstay of the pharmaceutical industry. It is a powerful technology that allows complex mixtures to be transformed into separated components. It is highly sensitive, reproducible, accessible, and well-understood from an operator's standpoint. Perhaps the output from the HPLC is its unique characteristic that distinguishes it from all other analytical techniques. The output from an HPLC, the chromatogram, is de®ned and simple. Each peak is characteristic of a component; each chromatogram is diagnostic of an event or experiment associated with a drug development activity. When combined with the fact that nearly all compounds of pharmaceutical interest are amenable to HPLC methodologies and conditions, and that critical information on nearly all events in the drug development cycle can be derived from HPLC chromatograms, it becomes evident why HPLC is a universally accepted analysis tool in the pharmaceutical industry. 2. Mass Spectrometry Until the widespread commercial introduction of electrospray ionization (ESI)±LC/MS instruments in the early 1990's, mass spectrometry-based techniques had a functional yet limited role in drug development. Primarily geared toward a medicinal chemistry environment, mass spectrometry was a fairly routine tool for molecular weight determination and a specialty tool for complex 203 & LEE AND KERNS structure identi®cation problems. Applications and methods were typically de®ned by the ionization method of choice. Fast atom bombardment (FAB), desorption chemical ionization (DCI), chemical ionization (CI), and electron ionization (EI) were the predominant choices, with varying degrees of applicability. Once a molecule is ionized, the mass spectrometer provided separation of the resulting molecular ions and dissociation products, according to weight. These masses were assigned to corresponding substructures of the molecule. These approaches were primarily limited to the characterization of low molecular weight compounds (<500 Da) with varying degrees of polarity and thermal lability. Detailed analysis was de®ned by the depth of spectral interpretation for structure identi®cation purposes or by the resolving power of the instrument for exact mass, molecular formula, and purity assessment. 3. LC/MS Interface The LC/MS interface provides the connection between the HPLC and mass spectrometer. It is responsible for the reliable and ef®cient transfer of analytes from the solution phase to the gas phase. It is also responsible for a critical element of mass spectrometry analysis: ionization. For most pharmaceutical analyses, the ideal ionization technique would generate a single ion that corresponds to the molecular weight of the drug compound, with little or no fragment ions. The con®rmation of structure would be facile, and quantitation would proceed with a high degree of sensitivity. Elements of selectivity would be provided by the HPLC separation (i.e., drug components, biological matrix) and/or MS/MS (i.e., structure elucidation, enhanced quantitation). In the late 1980's, thermospray ionization (TSI) techniques offered what would be the precursor to a universal and reliable LC/MS interface for compounds of pharmaceutical interest. Conventional HPLC ¯ow-rates (1±2 mL/min) were accommodated by this interface, using volatile buffers that contain ammonium acetate. New applications were realized, and higher standards of analytical performance were established for pharmaceutical analysis (Voyksner et al., 1985; Beattie & Blake, 1989; Oxford & Lant, 1989; Malcolm et al., 1990). However, TSI±LC/MS applications did not completely capture the imagination of chemists and biologists in the pharmaceutical industry. TSI±LC/MS was overshadowed by its unpredictable performance and questionable ruggedness when compared to LC/UV. This perception was the case, in part, because the industry was ready for a universal LC/MS system, but with very few limits. Simple methods to handle small and large molecules, combined with a gentle technique for ionization, were needed. Researchers were not content with the unique capabilities 204 of TSI±LC/MS. They wanted a few limits and boundaries for applicability with familiar levels of analytical performance (i.e., similar to LC/UV). This requirement was not necessarily derived from an analytical perspective, but rather an industry perspective, which ultimately forms the basis of acceptance. Whether this requirement was fair or unfair is perhaps another topic for discussion; however, the developments associated with the TSI±LC/ MS interface were indeed quite signi®cant and made clear two signi®cant aspects of LC/MS technologies during the late 1980's: (1) the pharmaceutical industry was ready for LC/MS applications, and (2) clear performance benchmarks for acceptance were established for future instruments. From a medicinal chemist's perspective, NMR was still the analytical tool of choice, whereas mass spectrometry, IR, and elemental analyses completed the necessary ensemble of analytical structure con®rmation. Synthesis routines were capable of generating several milligrams of product, which is more than adequate for proton and carbon NMR experiments. For analyses that involved natural products, metabolites, or synthetic impurities, time-consuming and often painstaking isolation methods were necessary, followed by expensive scale-up procedures, to obtain the necessary amount of material for an NMR experiment. In situations that involved trace mixture analysis, radiolabelling approaches were often utilized in conjunction with various formats of chromatographic separation. The original work of Bruins (Bruins et al., 1987) and Fenn (Fenn et al., 1989) revealed the promise of an LC/ MS interface that satis®ed the industry requirements of applicability and performance. Examples ranged from the atmospheric pressure chemical ionization (APCI)±LC/ MS analysis of small drug molecules (Bruins, 1991; Baillie, 1992) to the ESI±LC/MS analysis of biomolecules (Hunt et al., 1992; Yates et al., 1993) and receptor± ligand interactions (Ganem et al., 1991; Ganguly et al., 1992; Smith & Light-Wahl, 1993). These applications demonstrated a widened scope of LC/MS utility for pharmaceutical analysis and set new standards for performance. Perhaps equally important, these early applications of APCI±LC/MS and ESI±LC/MS captured the imagination of pharmaceutical researchers, and created a powerful perception that MS-based techniques could have an expanded role as a versatile tool for drug development. B. Growth of LC/MS Once designed, tested, manufactured, and championed, an explosion in pharmaceutical applications involving LC/MS soon began. This growth is illustrated by the number of papers that feature LC/MS (Fig. 9) and LC/MS APPLICATIONS & FIGURE 9. The growth in presentations at the annual conference of the American Society for Mass Spectrometry featuring: all LC/MS applications (LC/MS); and LC/MS applications in drug development (LC/ MS Pharma). that are presented at the American Society for Mass Spectrometry (ASMS) conference during the past ten years. The total number of papers has steadily increased during this time, which was marked by signi®cant increases in LC/MS papers. In 1998, LC/MS presentations at the ASMS conference in Orlando, Florida comprised nearly 30% of the total papers. Pharmaceutical applications of LC/MS exhibited a similar growth, and represented nearly 15% of the total papers presented in 1998. The rapid growth of LC/MS experienced in the pharmaceutical industry was in many ways, fortuitous. Development of the LC/MS interface coincided with industry initiatives that made it imperative to understand, acquire, and integrate this technology into the drug development cycle. There did not exist an alternative method or approach to obtain LC/MS-derived information that had the same impact on drug development. The faster rate of sample generation combined with cost-saving initiatives that emphasized ef®ciency helped to make LC/ MS-based solutions an easily justi®able approach. The LC/MS instrument has become a highly productive analysis platform for drug development activities due to: (1) technological advances in the LC/MS interface, (2) improved HPLC and mass spectrometer performance, and (3) an increased need for rapid, high-throughput analysis of trace mixtures. During the past decade, we have witnessed the successful introduction of a variety of LC/ MS instruments into drug development that feature single quadrupole, tandem quadrupole, magnetic sector, time-of¯ight (TOF), and ion trap mass analyzers. From dedicated instruments for highly specialized analyses (i.e., high throughput quantitation) to instruments residing on the laboratory bench (i.e., open access, reaction-monitoring stations) are now a signi®cant part of the modern drug development landscape. Negative perceptions that mass spectrometry-based techniques were time-consuming, limited in application, and dif®cult to operate and maintain became a memory of the past. In a relatively short time, LC/MS has truly come a long way. Furthermore, its growth has helped set the stage for parallel analysis opportunities with matrix-assisted laser desorption/ionization (MALDI) (Stoeckli et al., 1999), capillary electrophoresis (CE) (Chakel et al., 1997; Figeys & Aebersold, 1998), LC/NMR (Ehlhardt et al., 1998; Vogler et al., 1998), and in general, simpler, less expensive instruments. Certainly, ``carry-over'' perceptions in the pharmaceutical industry still exist; but, if instrument sales are a good indicator, then there seems to be a growing population of believers in LC/MS-based tools for analysis. Today, LC/MS technologies continue to experience an expanding role in drug development. High premiums on fast analysis times and the rapid turn-around of results have generated immense challenges and opportunity. New standards for throughput and capacity continue to be established. The early and quick evaluation of drug candidates has now become the norm throughout the pharmaceutical industry. Everything seems faster. To reach the required level of analysis performance with LC/MS techniques, new strategies for analysis are required. The challenge of coordinating this activity cannot be underestimated because these strategies often determine the ultimate success of the analysis method and provide a mechanism for an entire organization to interact 205 & LEE AND KERNS with the information generated. Without an appropriate strategy, the results are useful to only those who generated the data. Thus, the creation and selection of effective strategies for analysis have become critical factors in drug development. V. STRATEGIES An exciting element of LC/MS growth has been the development of effective strategies for analysis. Unprecedented needs in drug development stimulated new ways of providing information. Shorter timelines and a greater number of drug candidates resulted in a tremendous focus on streamlined approaches that generate information for decision-making. This approach allows decision makers to readily obtain, or even request, the necessary information that leads to accelerated development. Thus, the recent emphasis in LC/MS strategies in drug development is on producing information appropriate for decision. Accelerated drug development applications incorporate key strategies that serve to de®ne the attributes of the resulting method. Selection and incorporation of these strategies into the LC/MS method is highly dependent on the speci®c need or task. As a result, the strategies that are ultimately incorporated into a method may seem obvious, and may not be often enumerated. However, a fundamental understanding of these strategies can assist with the development of highly effective methods for analysis. There are nine strategies that consistently appear in LC/MS-based methods for accelerated development. These strategies form a powerful set of tools to devise, construct, and re®ne pharmaceutical analysis methods. The proactive use of a single strategy, or the combination of several, produces unique information and inspires new perspectives on analysis. In this way, analysis methods create new mechanisms for information gathering, rather than passively waiting for a sample to be generated. And it is in this way that LC/MS analysis capabilities recently partnered with sample-generating disciplines for the creation of new accelerated development opportunities. Described below are the nine analysis strategies, along with brief examples of how they are typically used to accelerate drug development. Selected references are listed in Table 6 to illustrate participation with speci®c drug development stages. Many of the examples referenced here are described in the applications section. Each strategy represents a signi®cant process change that affects the number of samples analyzed and the rate of compounds that ¯ow through the development pipeline. A. Standard Methods Method development and validation are two of the most time-consuming aspects of analysis-related activities. Furthermore, if a system is used for more than one analysis, then the time required for changeover from one set of conditions to another can actually be more time- TABLE 6. The nine analysis strategies that consistently appear in LC/MS-based methods for accelerated drug development. Analysis strategy Drug development application Standard methods Open-access Pharmacokinetics, screening Metabolite identi®cation CYP450 metabolism Natural products Degradants DD DD PD CD DD CD Impurities CD Metabolites QC/proteins Library searching CD M DD Natural products dereplication DD Impurity identi®cation Substructure nomenclature Glycoprotein mapping Phosphorylated peptides Target screening Drug-protein binding PD PD DD DD DD DD Template structure identi®cation Databases Screening 206 Development stage Selected references Taylor et al., 1995; Pullen et al., 1995 Olah et al., 1997 Kerns et al., 1997 Ayrton et al., 1998a Lee et al., 1996 Qin et al., 1994; Volk et al., 1996; Volk et al., 1997 Nicolas and Scholz, 1998; Williamson et al., 1998 Dear et al., 1998 Chang et al., 1997 Henzel et al., 1993; Eng et al., 1994; Arnott et al., 1995; McCormack et al., 1997; Kleintop et al., 1998 Gilbert and Lewer, 1998; Janota and Carter, 1998 Kerns et al., 1994 Kerns et al., 1995 Carr et al., 1993 Ding et al., 1994 Hsieh et al., 1997 Tiller et al., 1995 (Continued) LC/MS APPLICATIONS & TABLE 6. (Continued) Analysis strategy Integration Miniaturization Parallel processing Visualization Automation Drug development application Development stage Pharmacokinetics DD Caco-2 models Metabolic stability DD DD In vitro metabolites Bioaf®nity screening Predictive models Toxicology Toxicology Sulfates/glucuronides Peptide mapping Multidimensional chromatography Drug discovery metabolism Metabolic stability DD DD DD PD CD CD DD DD DD PD Pharmacokinetics CD In vivo microdialysis sampling CD Biotransformation Direct analysis of plasma CD CD IEC Chiral analysis/on-line Desalting CD CD Organic extractables Peptide mapping Glycoprotein mapping Cellular uptake In vitro hydrolysis Biotransformation In vitro biotransformation Impurities Biotransformation Chiral separations 96-Well SPE Protein analysis Predictive models/degradants Off-line SPE M DD DD DD DD DD PD PD CD CD CD DD PD CD Natural products Combinatorial chemistry Peptide mapping/base Peak pro®les Protein analysis Metabolite identi®cation Impurities SPE extraction DD DD DD Toxic leachables M DD DD PD CD Selected references Olah et al., 1997; Bryant et al., 1997; Hop et al., 1998; Beaudry et al., 1998; Allen et al., 1998 Caldwell et al., 1998 Ackermann et al., 1998; Davis and Lee, 1998 Poon et al., 1996 Davis et al., in press Rourick et al., 1996; 1998 Stevens et al., 1997 Maurer et al., 1997 Bean and Henion, 1997 Heath and Giordani, 1993 Apffel et al., 1995; 1996 Li et al., 1995 Herron et al., 1995; Wood et al., 1996; Heath et al., 1997; Manini et al., 1998 Covey et al., 1986; Fouda et al., 1991; Wang-Iverson et al., 1992; Kaye et al., 1992 Scott and Heath, 1998; Wong et al., 1999 Chen et al., 1991 Ayrton et al., 1997; Needham et al., 1998 Cai and Henion, 1997 Kanazawa et al., 1998; Joyce et al., 1998 Wu et al., 1997 Liu et al., 1998 Liu et al., 1996 Kerns et al., 1998 Van Breeman et al., 1991 Tiller et al., 1998 Beattie and Blake, 1989 Liu et al., 1997 Ackermann et al., 1996b Zavitsanos and Alebic-Kolbah, 1998 Kaye et al., 1996 Houthaeve et al., 1995 Rourick et al., 1996 Allanson et al., 1996; Simpson et al., 1998 Ackermann et al., 1996a Takach et al., 1998; Tong et al., 1998 Arnott et al., 1995 Houthaeve et al., 1997 Lopez et al., 1998 Josephs, 1996 Allanson et al., 1996; Janiszewski et al., 1997; Simpson et al., 1998 Tiller et al., 1997a 207 & LEE AND KERNS consuming than the analysis! Increasingly, the desire for customized methods has yielded to the demands of increased sample generation. This demand resulted in the introduction of standard methods, sometimes referred to as ``generic methods'' (Ayrton et al., 1998b; Dear et al., 1998), to accommodate a wide range of compound classes. This strategy allows the same set of conditions to be used on diverse samples and reduces the time required for method development, validation, and experimental setup. Some reduction in chromatographic resolution may be experienced compared to a customized method; however, the separations are often suf®cient to provide the appropriate information for decision making. Generally, this approach does not replace existing methods; however, it provides for the necessary throughput plus new options for accelerated drug development. Also, methods are ``locked-in'' during early development stages for faster project start-up. This strategy eliminates iterative cycles of method development and re®nement. This approach has been recently demonstrated for lead optimization in the drug discovery stage (Olah et al., 1997) and in situations where a drug candidate enters a new stage of evaluation (Kerns et al., 1997; Bansal & Liang, 1998). One standard method approach that appears to have wide application in the pharmaceutical industry is the use of reversed-phase HPLC with a wide solvent gradient program. For example, a linear gradient from 95% aqueous/5% organic to 5% aqueous/95% organic at near neutral pH (i.e., 6±7) is typical. Often, a fast gradient (5 and 20 min) is utilized to provide rapid analysis. An important element of this strategy is the application of the ``80/20 Rule'', also known as Pareto's Law (Heller and Hindle, 1998), for method development and performance (Lee et al., 1995a; 1997). This approach basically targets a practical benchmark for performance where the method is applicable to about 80% of the samples analyzed. Analysis is initiated with minimal time spent on development and re®nement activities. The strategy is to spend 20% of analysis time on 80% of the samples, whereas the majority of time (i.e., collaboration, method development, interpretation) can be spent on more challenging analyses that represent 20% of the samples. The strategy can be re®ned and the scale set to the desired level of performance, such as 90/10 or 95/5. For the samples that are not successfully analyzed using this method, a backup 80/20 method is utilized (Lee et al., 1995a). Using this method-ensemble approach, maximum information is generated in the shortest amount of time and with minimal resources. A standard method allows information to be collected in a consistent manner. Thus, information generated on one day is reliably compared to information generated on another. This consistency is an important consideration, 208 particularly for those compounds that enter late stages of drug development, because useful information acquired 1±3 years earlier is used to accelerate development. Therefore, this strategy also results in a signi®cant reduction of re-analyses. An example that involves a series of process research experiments performed by a chemist over several weeks serves to illustrate this strategy. If, in the ®rst sample, a drug impurity is identi®ed using LC/MS and LC/MS/MS studies, then it may only be necessary in subsequent samples to con®rm the molecular weight of an impurity at the same retention time, using a single LC/MS or LC/UV analysis (Kerns et al., 1994). This approach has been successfully implemented for metabolite identi®cation during the preclinical development stage (Kerns et al., 1997) and for patent infringement issues during the manufacturing stage. B. Template Structure Identi®cation The identi®cation of unknown structures can often seem daunting, particularly when an authentic standard is not available. As a result, impurities, degradants, and metabolites have often been prospectively synthesized. To support accelerated development, where 40-times as many samples can be generated, a more streamlined approach for structure identi®cation must often be implemented. One strategy involves the use of the parent drug itself as a template for the interpretation of unknown structures (Perchalski et al., 1982; Covey et al., 1991). This approach often provides suf®cient information for early decision making in drug discovery. Con®rmation for registration purposes (i.e., NDA/MAA) is performed, using a synthesized standard, as a project enters the later stages of drug development. Template structure identi®cation is readily performed with LC/MS/MS using a consistent protocol. First, the parent drug is analyzed with LC/MS. Retention time and molecular weight information are obtained. Using LC/ MS/MS, a product-ion analysis of the parent drug is obtained, and speci®c product ions and neutral losses are assigned to substructures of the molecule. Thus, the unique fragment ions contained in the product-ion mass spectrum of the parent drug serves as the template for identi®cation. An example that illustrates this template structure identi®cation strategy is shown in Fig. 10 for paclitaxel (Volk et al., 1997). Subsequent analysis of samples via LC/MS and LC/MS/MS yields comparable information for the unknown compound(s). Further information is obtained, by the observation of sequential neutral losses, to determine the sequence of substructures or ``molecular connectivity'' within the analyte (Lee et al., 1996). This procedure is analogous to two-dimensional NMR techniques used to sequentially connect substructures. A familiar example of molecular LC/MS APPLICATIONS & FIGURE 10. Template structure identi®cation of a base-induced degradant of paclitaxel. (A) Product-ion spectrum of the ion at m/z 829 (H ‡ NH4) ‡ of a base-induced degradant of paxlitaxel. (B) Product-ion spectrum of the m/z 871 (H ‡ NH4) ‡ ion of paclitaxel used as a template. The product ions and neutral losses that correspond to speci®c substructures are indicated. Diagnostic product ions at m/z 286, 268, and 240 indicate the presence of the paclitaxel sidechain. The m/z 509 ion in the degradant is indicative of the deacetylated paclitaxel core ring structure. The presence of the m/z 527 ion instead of the m/z 569 ion of paclitaxel indicates the loss of an acetyl substructure from the core ring structure. This result is consistent with the MW difference of 42 Da of the degradant, which is indicative of an acetyl substructure (Volk et al., 1997). 209 & LEE AND KERNS connectivity is the determination of the amino acid sequence of a peptide. Speci®c neutral losses are diagnostic of speci®c amino acids, and the sequence of these losses identi®es the peptide (Roepstorff & Fohlman, 1984). The MS/MS identi®cation strategy is based on the premise that much of the parent drug structure will be retained in the metabolites, impurities, or degradants (Perchalski et al., 1982; Lee et al., 1986; Straub et al., 1987; Lee & Yost, 1988). In addition, the product ions associated with unique substructures are also expected to be retained. Direct comparison of molecular weight and product ions reveal substructural differences, and lead to an interpreted or proposed structure. This strategy is highly successful for impurity identi®cation (Kerns et al., 1994) during preclinical development. When information is stored within a comparative database, this approach is also highly effective for protein identi®cation (Arnott et al., 1995). In these applications, the characteristic fragmentation corresponding to amino acid residues provide the searchable template for identi®cation. This approach is particularly useful when identi®cation studies are required for vast numbers of compounds or for samples that contain many analytes of interest. C. Databases One strategy for taking advantage of (leveraging) processed information is to categorize it into databases. Although a modest amount of time and resources is required to implement this strategy, databases have two important bene®ts. First, it provides a reference-friendly format to search data. This feature is essential for the rapid identi®cation of known compounds. For example, the identi®cation of a metabolite may require only retention time and molecular weight information via LC/MS analysis when compared to the metabolite structure database compiled from previous LC/MS/MS studies (Kerns et al., 1997). The process of identifying a natural product in a plant extract, when the natural product has been previously identi®ed and recorded in the database, is often referred to as ``dereplication''. This approach has been recently described by Gilbert and Lewer (Gilbert & Lewer, 1998) and Janota and Carter (Janota & Carter, 1998). Databases are also used to search for known proteins (Eng et al., 1994). These databases are utilized extensively for the rapid identi®cation (dereplication) of proteins in studies involved with proteomics (Arnott et al., 1998; Kleintop et al., 1998). A second bene®t of databases is the ef®cient extraction of information. The database may be ``mined'' to detect trends that may not be otherwise noticed. This approach is used to reveal trends such as the metabolically 210 active sites of a molecule and/or substructures labile to degradative conditions (Kerns et al., 1997). This information is useful for candidate selection and development planning activities. Once established, a database is transferred to other laboratories that are participating in the speci®c drug development activity. The resulting databases are readily accessed via information intra-nets. Information is coordinated within the database, and a variety of scientists ``pool'' their information. When implemented early within the drug development cycle (i.e., drug discovery), valuable information for late stages of drug development are available (Kerns et al., 1994; 1997). Therefore, this approach provides a comprehensive method for information gathering whereby future projects are planned, coordinated, and ef®ciently supported. It should also be noted that database creation, modi®cation, and use are greatly bene®ted by the standard methods approach described previously. This approach produces reliable data that lend themselves to a highly consistent database format throughout a project lifetime. D. Screening The recent acceptance of standard methods for analysis, template structure identi®cation, and databases has allowed increased amounts of information to be generated in shorter periods of time. As researchers embraced approaches calling for the earlier collection of information on pharmaceutical properties, LC/MS emerged as an advantageous technique for screening. For example, to maintain adequate sample throughput and turn around, a variety of screening-based approaches is utilized. Results are either geared toward categories of performance (i.e., high, low) or absolute criteria such as a speci®c structure or targeted level of quantitation. Corresponding limits are pre-set during method development, and the LC/MS-based analysis merely sorts the information in a high volume fashion. Important elements of LC/MS screening feature quantitative and qualitative process approaches. Screening-based strategies are incorporated into quantitative bioanalytical assays to provide a highly selective approach for high throughput analysis. Highly automated methods featuring fast chromatographic separations in combination with either one or two dimensions of mass spectrometry analysis provide powerful methods for quantitative analysis. The two approaches are referred to as selected ion monitoring (SIM) and selected reaction monitoring (SRM). In experiments that use SIM, the mass spectrometer is set to detect a selected ion mass that corresponds to the drug molecule. Sensitivity is enhanced by using this selected ion drug screening approach as opposed to sampling a range of ions LC/MS APPLICATIONS via a conventional mass spectrum (Fouda et al., 1991; Wang-Iverson et al., 1992). To provide higher selectivity and to enhance the limits of detection (LOD), SRM is used to screen for a compound by monitoring the ion current between a speci®c precursor±product ion relationship (Covey et al., 1986). A speci®c product ion resulting from unimolecular decomposition or a collisionally induced dissociation (CID) of the protonated molecule ion, (M ‡ H) ‡ , is detected (Kaye et al., 1992). Recently, high throughput methods for metabolic stability assessment have been introduced early in the drug discovery stage in close proximity to compound synthesis activities (Cole et al., 1998; Ackermann et al., 1998). Analyses often incorporate a screening motif associated with the method of analysis, collection of information, and the subsequent interpretation of results. Similar approaches to rapidly generate and identify large numbers of related compounds (i.e., impurities, degradants, metabolites) employ ``predictive models'' (Rourick et al., 1996). Typically, these methods involve the incubation of the drug candidate in an in vitro model that is indicative of a condition encountered during drug development. For example, some methods incorporate metabolizing microsomes, acidic/basic/oxidizing chemical degradation conditions, or light/heat/humidity environmental conditions. Another LC/MS-based screening approach features the use of HPLC columns that contain an immobilized protein to determine relative binding. Tiller et al. demonstrated the use of this LC/MS method- & ology for the determination of drug-protein (serum albumins) binding (Tiller et al., 1995) [Table 7]. These screening approaches require minimum information for the evaluation of pharmaceutically relevant properties. Compounds that do not meet a predetermined criterion are eliminated from further consideration. Guidance is obtained for further structure optimization via synthetic modi®cation. Thus, pharmaceutical properties are assessed on the time-scale of compound synthesis, resulting in an optimization while programs are still active. Therefore, unfavorable candidates are eliminated early and resources are focused on more promising drug candidates. Two-dimensional mass spectrometry screening approaches have also been demonstrated for qualitative analysis. The various MS/MS scan modes of the mass spectrometer (Yost & Boyd, 1990) provide powerful methods of analysis and unique capabilities for information gathering. For example, the product-ion scan mode for substructure analysis provides information useful for structure identi®cation, as described earlier for template structure identi®cation strategies. The precursor-ion scan mode offers a highly speci®c approach to screen for targeted component due to the presence of a diagnostic substructure. Carr et al. demonstrated the use of precursorion experiments to identify N- and O-linked glycoproteins (Carr et al., 1993). The neutral loss scan mode is used as a class identi®er to also screen components that contain a common substructure. Barbuch et al. demonstrated this TABLE 7. Results from an LC/MS-based drug-protein binding screen using HPLC columns containing immobilized human serum albumin to determine the relative binding of drugs (Tiller et al., 1995). Drug Glucose Salbutamol Pyridoxin Tinidazole Paracetamol Phenacetin Cefuroxime±Axetil Triamterene Alprazolam Ondansetron Quinidine Lorazepam Quinine Salmeterol R-Temazepam S-Temazepam Diazepam R-Warfarin S-Warfarin Molecular weight 180 239 169 247 151 179 510 253 308 293 324 322 324 415 300 300 284 308 308 % Binding 0 7.5c 10b 12b 24b 33a 40a 57b 70a 74c 90a 90a 90a 95c 97a 97a 98a 99a 99a tr k0 1:15 1:24 1:26 1:24 1:33 1:53 1:59 3:19 4:43 8:14 7:19 7:10 7:19 7:37 4:42 6:19 7:56 15:42 20:54 ± 0.187 0.229 0.187 0.375 0.792 0.916 2.583 4.333 8.729 7.583 7.396 7.583 7.958 4.312 6.333 8.354 18.062 24.562 k 0 /(k 0 ‡ l) ± 0.157 0.186 0.157 0.273 0.442 0.478 0.721 0.812 0.897 0.883 0.881 0.883 0.888 0.812 0.863 0.893 0.947 0.961 a Clinical Pharmacokinetics. 1988. 15:254±282. Clarke's Isolation and Identi®cation of Drugs, The Pharmaceutical Press, 1986. c Data obtained in-house. b 211 & LEE AND KERNS approach for the class identi®cation of phytoestrogens (Barbuch et al., 1989), using TSI±LC/MS/MS. Brownsill et al. used a similar ESI±LC/MS/MS approach for the analysis of metabolites in rat liver slices (Brownsill et al., 1994). As new LC/MS-based technology improved the analytical ®gures of merit (i.e., sensitivity, selectivity, speed, cost effectiveness), these advantages approached performance benchmarks that are necessary for the routine generation of screening information. The application of LC/MS-based technologies as a screening tool is a powerful analysis strategy for quantitative and qualitative structure analysis. E. Integration Integration strategies often encompass separate events that involve instrumentation, methodology, and process. When viewed as integrated stages of analysis, similar to the examples described by Cooks and Busch (Cooks & Busch, 1982), increased selectivity is attained. For example, LC/MS/MS represents the integration of three powerful instruments: the HPLC and two mass spectrometers. Retention time information is obtained from the HPLC, and the mass spectrometer provides molecular weight information. A second stage of mass spectrometry yields structure information. The on-line nature of LC/MS/MS provides a high degree of selectivity and an ef®cient on-line analysis. In addition, integration affords unique opportunities for new experiments. Traditional methods of pharmaceutical analysis involve a series of multiple steps. For example, the identi®cation of natural products traditionally involves the scale-up of fermentation broths, solvent extraction, liquid/ liquid or column fractionation, chromatographic fraction collection, and spectroscopic analysis (usually NMR) of the individual components. Figure 11 illustrates the integration of these bench-scale steps into a dedicated LC/MS/MS system (Lee et al., 1997). Integration provides unique and powerful advantages for the on-line identi®cation of natural products (Kerns et al., 1994; Ackermann et al., 1996). Experiments that once required two weeks to perform with traditional approaches are now performed in half a day with LC/MS/MS (Gilbert & Lewer, 1998). Similar LC/MS integration approaches are described for bioaf®nity-based screening applications (Davis et al., in press) and automated sample preparation for in vivo pharmacokinetics screening (Beaudry et al., 1998). A recent review that describes multidimensional approaches for protein characterization highlights the important features of integration (Anderegg et al., 1997). Multidimensional applications that feature LC/MS instrumentation are also used to integrate sample preparation for online extraction via column switching (i.e., LC/LC) with quantitative bioanalysis (Needham et al., 1998), and detailed structure analysis of metabolites that use MSn approaches on an ion trap mass spectrometer (Lopez et al., 1998). Often, integration strategies involve the addition of a step into an existing analysis protocol, which at ®rst would seem less ef®cient. However, the additional step is strategically placed to provide signi®cant bene®ts in productivity. For example, the assessment of metabolic stability for large numbers of drug candidates during the drug discovery stage (Ackermann et al., 1998) incorporates an additional step, but provides unique and timely information that results in better decisions on which lead compounds are rejected, optimized, or selected for further development. This qualitative process change results in faster lead optimization. In the same manner, the creation of structure databases for lead candidates in the early preclinical development stage (Rourick et al., 1996) provides a faster approach for the identi®cation of impurities, degradants, and metabolites during preclinical development, clinical development, and manufacturing stages. F. Miniaturization As accelerated drug development approaches become more widespread, there is recognition that a single qualitative process change can have a pronounced effect FIGURE 11. The integration of high performance liquid chromatography and two mass spectrometers to form the LC/MS/MS instrument. The integrated bench-scale steps are shown at the top. 212 LC/MS APPLICATIONS in reducing the size or scale of analysis. Traditional protocols operate at the milliliter to liter volumes and micromole to millimole quantities. As analysis systems have become more sensitive and techniques for manipulating small volumes of liquid samples have become more reliable, experimental procedures have emerged that use smaller amounts of sample. The same is true from the opposite perspective, indicating that smaller sample sizes have resulted in miniaturized systems and formats for analysis. As a result, it is less time-consuming and less expensive to generate smaller sample sizes. Thus, analysis protocols that feature nanoliter to microliter volumes and picomole to nanomole quantities emerged. One example of a miniaturized LC/MS strategy is the use of 96-well sample plates (Kaye et al., 1996) for extraction. This sample extraction procedure combines batch sample processing within a miniaturized format. Increased sensitivity and decreased volume advances have fostered a new wave of ``scale-down'' models. Here, experiments that were formerly performed at the bench are, instead, performed at the microliter scale in the batch mode. For example, synthetic process research was traditionally performed manually with apparatus at the milliliter level. This approach involves the testing of a range of synthetic conditions for optimum yield and minimum impurity production. Now, process research conditions are tested in microliter levels to produce information on purity and structure (Rourick et al., 1996). This strategy requires less reagent and accelerates the evaluation of a wider range of conditions in a shorter time. Another example includes the direct analysis of samples from cell culture experiments (Kerns et al., 1997). Instrumental developments facilitate the miniaturization opportunity. Advances in chromatography led to the use of capillary HPLC techniques (Liu et al., 1993; Kassel et al. 1994; Arnott et al., 1996) for the powerful separations of increasingly smaller samples. For mass spectrometry, developments with the LC/MS interface that resulted in increased ion transmission and ion sampling helped to provide improved sensitivity. The development of micro- and nano-spray electrospray sources allows the production of intense ion current from nanoliter ¯ows (Emmett & Caprioli, 1994; Andren et al., 1995; Wilm & Mann, 1996; Davis & Lee, 1998). G. Parallel Processing Analysis methods often involve a sequential strategy in which individual compounds are analyzed either separately or in series. The data are evaluated, and the results are compared. This ``serial'' analysis approach consistently generates useful information for drug development. However, when a single analysis step is ratelimiting to the overall ef®ciency of the method, parallel & processing strategies are employed. Once identi®ed, the rate-limiting step is removed from the on-line approach and performed in a batch processing mode off-line. As the task is completed, the samples are re-introduced to the analysis procedure. Recent studies performed by Pleasance and coworkers (Allanson et al., 1996) and Wu and coworkers (Simpson et al., 1998) illustrate a highly effective parallel processing approach, using solid-phase extraction (SPE) techniques. This strategy for analysis is analogous to parallel-compound synthesis approaches for high volume lead optimization (Selway & Terrett, 1996). Alternatively, drug candidates may be combined for coanalysis in the same sample. Although this ``combinatorial analysis'' approach may have previously resulted in impossible analysis complexity, the capabilities of LC/MS and LC/MS/MS for providing detailed information from highly complex samples is an excellent ®t with ``combinatorial parallel processing''. An example of parallel processing using a ``combinatorial'' approach involves lipophilicity screening of a series of synthetic analogs (Lee & Kerns, 1998). Lipophilicity, as indicated by retention behavior (k 0 ) on an octadecylsilane (C18) bonded phase HPLC column, is often used as a predictor of pharmaceutical characteristics such as membrane transport and protein binding. In the past, k 0 studies have been conducted sequentially; using a 60 min isocratic LC/UV method; thus, ten compounds would require 10 h for analysis. A comparable parallel experiment would be to mix all ten compounds into a single ``combinatorial mixture'' and analyze them in a single 1 h analysis with LC/MS. The mass spectrometer is able to distinguish the compounds according to their molecular weight. Thus, LC/MS reduces the qualitative process enhancement from 10 h to 1 h. Similar applications (Volk et al., 1996) combine parallel processing strategies with binary screening (i.e., yes/no results) to rapidly analyze combined mixtures of drug candidates. In these studies, the combinatorial mixture is exposed to a chemical, physical, or physiological environment in vitro, and the resulting degradants/ metabolites are identi®ed and quantitated with LC/MS and LC/MS/MS techniques [Fig. 12]. H. Visualization The rapid growth in LC/MS productivity resulted in the production of massive amounts of data. Thus, with the increased productivity experienced with modern analysis systems, the bottleneck quickly shifted to data interpretation and management. Approaches that feature the visualization of data help to provide meaningful information for decision. One simple example of a visualization tool is the extracted ion current pro®le (EICP) or mass chromato213 & LEE AND KERNS applications. For example, methods for the analysis of combinatorial chemistry-derived samples provide visual representations of the 96-well plate [Fig. 13] (Takach et al., 1998). Following the LC/MS analysis, an automated analysis is performed, according to pre-established thresholds to search for the protonated molecule ion of the analyte. If the ion is found, then the visual representation of the corresponding well is marked with a distinguishing color scheme. In this way, the scientist quickly inspects the visual representation to make decisions. Diagrams provide another form of visualization that rapidly communicates pharmaceutical properties of a drug candidate. For example, the labile sites of a molecule can be diagrammed in simple format superimposed on the structure [Fig. 14] to rapidly communicate the labile sites or ``soft spots'' of the candidate structure (Lee & Kerns, 1998). This information can facilitate decisions for the further optimization of the candidate structure series and can result in molecules with superior pharmaceutical properties. FIGURE 12. The analysis of three compounds with a serial strategy and a parallel ``combinatorial'' analysis strategy (Volk et al., 1996). I. Automation gram. These plots provide meaningful information from complex LC/MS data, such as the EICP for the protonated molecule ion of the drug candidate during quantitative bioanalysis experiments. Also, methods for the rapid assessment of drug metabolites often include the plotting of EICP's that represent common metabolic transformations, such as [MH ‡ 16] ‡ , which corresponds to monohydroxyl metabolite structures. Other approaches for presenting information to facilitate the visualization of meaningful patterns for rapid decision involve combinatorial chemistry-related Many examples exist for the integration of robotic mechanisms with analysis instruments. Automation is quite successful for sample preparation stages, such as SPE applications for pharmacokinetic studies, as well as for the queuing of samples for instrument analysis. In many cases, signi®cant savings are realized in human labor expense and the reduction of routine operations. Furthermore, consistent robotic operations afford increased precision via reproducibility of operations from sample-to-sample compared to manual operations. Often, the robotic process takes longer than the manual process; however, signi®cant savings and acceleration of the process is realized by overnight or around- FIGURE 13. Visual representation of the results from LC/MS (TOF) analysis of compounds from a 96-well plate for combinatorial chemistry. Visualization of the information in terms of the presence of a compound and the mass accuracy of the analysis allows for rapid visual inspection of data (Takach et al., 1998). 214 LC/MS APPLICATIONS & FIGURE 14. Structure of buspirone with areas susceptible to metabolic or chemical degradation highlighted for rapid visualization of the ``soft spots'' of the structure (Lee & Kerns, 1998). the-clock operation. In these cases, the routine operations can be performed by the automated system, leaving the highly trained scientist to perform the more detailed aspects of the analysis that require experience and realtime decisions. Clearly, standard methods provide a necessary step toward the implementation of automated methods via their focus on routine process steps. The automated approach is implemented and validated by careful inspection and observation of performance. Typically, automated procedures are locked-in after they are performed manually. This approach allows for the application of validation procedures to ensure adherence to the desired performance speci®cations. Automation provides signi®cant support for those quantitative process approaches that are aimed at accelerating drug development. Automated analysis strategies are invaluable tools for keeping pace with automated sample generation methods. With automation, the opportunity exists to incorporate qualitative process changes for accelerated development. For example, automated approaches that feature standard methods, databases, and screening are locked-in early in drug development; that automation sets the stage for qualitative process changes to facilitate structure identi®cation and the visualization of data. It is important to note that signi®cant bene®ts are obtained from simple and widely available automation systems. For example, the HPLC autosampler is widely applied for tremendous advantages with LC/MS analysis. This system provides consistent operations, the queuing of large numbers of samples, unattended operation, and integration with analysis. Automated LC/MS techniques produce a wealth of data. The subsequent interpretation of data provides unique insight for decision. When LC/MS data are accumulated in a matrix format, separation is indexed by time and structural information is indexed by mass. Thus, the task of interpretation becomes systematic and relatively simple. The key for advances in automated interpretation appears to lie in the development of algorithms that embody the ``rules'' for discerning patterns and making reliable assignments, typically performed by highly trained analysts. Translation of these rules into software allows these bene®ts to be transferred to a wider group of LC/MS users via automated search routines as described for database strategies as well as for real-time decision during the analysis. Data-dependent software programs allow real-time decisions to be made during an analysis. These approaches feature pre-established thresholds for the detection of a peak during full-scan mass spectrometry and MS/MS scan modes. If a peak of interest is detected in real-time, then the mass spectrometer is switched from full-scan mode to another scan mode to obtain more information from the same analysis. For example, the system may be automatically switched to the product-ion mode during the analysis of a chromatographic peak to obtain substructural information. Thus, more detailed information is obtained in fewer analyses. This powerful feature was recently demonstrated for the identi®cation of leachables during the manufacturing stage of drug development (Tiller et al., 1997a). Analysis strategies that feature automation signi®cantly impact daily routines in the laboratory. For realtime analysis, this approach is referred to as ``in-process'' analyses. Open access analysis strategies make a highly effective approach for maintaining adequate support for high volume sample-generating activities such as medicinal and combinatorial chemistry. Another approach features a batch-mode of analysis. A 24-hour analysis is strategically viewed in three parts: morning, afternoon, and evening. The morning hours are used to review results and communicate with collaborators; meaning is derived and new experiments are devised. Afternoons are spent developing and setting up the actual experiment. Evenings 215 & LEE AND KERNS are dedicated to high volume analysis, using automated features for sample preparation, analysis, data reduction, and reporting. This iterative cycle maximizes the use of automation for highly ef®cient analysis schemes (evening) while creating optimum approaches for collaboration and method development (day). J. Summary The diverse applications of LC/MS in the pharmaceutical industry led to the development of novel strategies and new methods for analysis. The nine analysis strategies described are consistently used throughout the drug development cycle. Many applications focus on structure characterization activities for the quick con®rmation of synthesis products and for the identi®cation (quantitative and qualitative) of drugs and related compounds that are contained in complex mixtures such as metabolites in physiological ¯uids. Recent applications of LC/MS-based techniques in the pharmaceutical industry highlight the need for novel analysis strategies to keep pace with sample generation and provide unique information. It is important to emphasize the importance of chromatography for separating biological interferents from analytes of interest (Snyder & Kirkland, 1979). The valuable role of on-line puri®cation and separation in conjunction with mass spectrometry was recognized early by Johnson & Yost (1985) and Henion & Covey (1986), and are still emphasized in the LC/MS methods used today. Many of these important chromatography features are described in the following section. Interestingly, many of the current LC/MS approaches for pharmaceutical analysis are extensions of GC/MS (Foltz, 1978), mass spectrometry (Garland & Powell, 1981), and MS/MS (McLafferty, 1983)-based methods. With the introduction and utilization of LC/MS-based methods, these fundamental approaches for quantitative and qualitative structure analysis became more routinely applicable to a wider scope of pharmaceuticals. VI. LC/MS APPLICATIONS The LC/MS applications described in this section are organized into the respective drug development stages; drug discovery, preclinical development, clinical development, and manufacturing [Fig. 15]. A sequential illustration of pharmaceutical analysis activities provides a unique perspective on the contributions of LC/MS techniques in drug development. The speci®c selected examples help to illustrate the successful incorporation of effective LC/MS-based analysis strategies in the pharmaceutical industry, and to highlight their impact on accelerated drug development. Thus, the literature references highlight the application and strategic impact of LC/MS on drug development and are not intended to be comprehensive. Recent and earlier reviews deal with sample preparation (Henion et al., 1998), ionization (Bruins, 1998), mass spectrometry (Burlingame et al., 1998), LC/MS instrumentation (Niessen, 1998), MS/MS scan modes (Yost & Boyd, 1990), and the application of mass spectrometry techniques in the biological sciences (Burlingame & Carr, 1996). A. Drug Discovery In recent years, a signi®cant shift in the focus of drug discovery activities has been made from a detailed knowledge-gathering science to incorporate high volume events. Surely, approaches that emphasize detail still remain as a critical element of drug discovery; however, they have been overwhelmed by a vast array of screening events aimed at identifying target and lead compounds. As a result, drug discovery researchers are becoming increasingly more dependent on technologies that FIGURE 15. The stages of drug development and associated pharmaceutical analysis activities. 216 LC/MS APPLICATIONS enhance their ability to quickly generate, test, and validate a discovery. a. Analysis requirements. Due to the use of highvolume approaches to discover new lead compounds, analysis requirements focus on sensitive, high throughput, and robust methods. These analysis methods provide an excellent complement to rapid sample-generating methods. The functional requirement is to provide analysis information that keeps up with the rate of sample generation. In this way, the timely evaluation is made of chemical and biological parameters that in¯uence decisions for lead candidate selection. Thus, this stage of drug development is made extremely challenging by the performance required on a diverse range of compounds. b. Analysis perspectives. The drug development cycle is initiated by activities that deal with the discovery of new medicinal lead structures. Historically, these structures have originated from the isolation of natural products from microbiological fermentation, plant extracts, and animal sources. Other discoveries can be traced to the screening of the vast compound libraries of pharmaceutical and chemical companies. The application of mechanism-based, structure-based approaches to rational drug design (Hirschman, 1991; Patchett, 1993; Terrett et al., 1996) and screening approaches (Dower et al., 1991; Moos et al., 1993) led to three distinct activities that impact pharmaceutical analysis: (1) target identi®cation; (2) lead identi®cation; and (3) lead optimization. These activities have been the primary responsibility of drug discovery with the goal of identifying a novel lead candidate for preclinical development. c. LC/MS contributions. The use of LC/MS-based approaches have expanded rapidly in drug discovery & during the past several years. Applications that range from the quick molecular weight con®rmation of synthetic lead compounds (Taylor et al., 1995; Pullen et al., 1995) to novel and highly selective methods for structure identi®cation (Carr et al., 1993) have been demonstrated. Analysis strategies typically emphasize rapid turnaround of results with an applicability to a diverse range of compounds. This emphasis generated unique demands and criteria for performance associated with the LC/MS analysis, namely, sample preparation and data management. As a result, LC/MS applications in the drug discovery stage use highly automated methods, many of which feature the integration of sample generation and analysis activities. d. Overview. In this section, recent applications dealing with peptide mapping, glycoprotein mapping, natural products dereplication, bioaf®nity screening, open access combinatorial chemistry/medicinal chemistry support, pharmacokinetic screening, and metabolic stability are described [Table 8]. The features of each LC/MSbased analysis are highlighted with regard to the unique generation of information and the contributions to accelerated drug development. 1. Peptide Mapping Protein analysis represents a signi®cant challenge in the pharmaceutical industry. The identi®cation of proteins is essential for understanding biological process and the function of the protein during native and disease states. The resulting insights lead to the development of therapies for intervention, and ultimately, the cure of disease. The information and knowledge derived from this type of study are extremely valuable for activities involved with target identi®cation activities during drug discovery. TABLE 8. Applications of LC/MS in drug discovery. Drug discovery activity Analysis LC/MS application Target identi®cation Protein identi®cation Glycoprotein identi®cation Peptide mapping Glycoprotein mapping Lead identi®cation Natural products identi®cation Screening Combinatorial/medicinal chemistry support Natural products dereplication Bioaf®nity screening Open-access Pharmacokinetic screening In vivo drug screening Metabolic stability High throughput metabolic screening Automated metabolite identi®cation Lead optimization Selected references Arnott et al., 1995 Carr et al., 1993; Liu et al., 1996 Ackermann et al., 1995a Davis et al., in press Taylor et al., 1995; Pullen et al., 1995 Olah et al., 1997; Allen et al., 1998; Beaudry et al., 1998 Ackermann et al., 1998 Lopez et al., 1998 217 & LEE AND KERNS Two-dimensional gel electrophoresis (2-DGE) is the primary analysis tool used in the pharmaceutical industry to characterize the expression of proteins. Large numbers of proteins, mostly protein variants, are identi®ed with these methods, and highly expressed proteins are easily located. The resulting differences in protein expression due to treatment with various stimulating factors are the basis for comparative 2-DGE maps (Celis, 1991). Traditional approaches for protein identi®cation of 2DGE spots use automated Edman degradation, amino acid analysis, and immunostaining techniques. Mass spectrometry-based methods are emerging as an important partner with 2-DGE for correlating proteins to their function (Yates, 1998a). Identi®cation schemes that involve LC/MS and MALDI±TOF techniques have made signi®cant contributions to understanding a broad range of events that involve proteins such as covalent modi®cation, proteolytic cleavage, and binding. The use of LC/MS approaches for protein identi®cation in conjunction with 2-DGE offers distinct advantages such as the ability to handle low picomole (miniaturized) level samples, enhanced separation, detection, the amenability to Nterminally blocked proteins, and fast analysis. Usually, the focus of LC/MS methods for protein characterization is on four distinct goals: (1) con®rmation of putative sequence; (2) identi®cation of amino acid modi®cations; (3) identi®cation of known proteins; (4) sequence determination of unknown proteins. Approaches for identifying proteins separated by 2DGE use in situ digestion methods. Arnott et al. described an ESI±LC/MS-based approach where the proteins from normal neonatal rat cardiac myocytes are electroblotted to a membrane and identi®ed by staining (Arnott et al., 1995). Protein spots are cut from the blot, destained, reduced, alkylated, and digested with endoproteinase LysC (Henzel et al. 1993). Small diameter, packed capillary HPLC columns (<100 mm) are used to enhance performance by minimizing sample handling and maximizing sensitivity (Moseley et al., 1991; Kassel et al., 1994). The resulting capillary LC/MS and LC/MS/MS experiments performed on a triple quadrupole mass spectrometer are used to determine the identities of each protein in combination with a database searching routine. In this study, the differences in protein expression levels between normal and enlarged (hypertrophic) heart cells were pro®led as part of a study to determine factors that are involved with congestive heart failure. The capillary HPLC separation from a selected protein spot provides a base peak pro®le shown in Fig. 16A. The base-peak pro®le is similar to a total ion current (TIC) pro®le, but it contains only the most abundant mass spectral peak in each scan. The chromatogram is simpli®ed and the contributions from background ion abundances are eliminated, resulting in an enhanced 218 signal-to-ion ratio for an improved visualization of data. The molecular mass for each component is labeled along with corresponding amino acid residues. This format provides a comprehensive approach for peak selection and peptide identi®cation. Figure 16B shows the mass spectrum of the second major chromatographic peak (residue 60±70 peptide), which contains the triply charged (m/z 451.3) and doubly charged (m/z 676.7) molecular ions. Thus, a molecular weight of 1351 is assigned. The next step involves the selection of either one of the molecular ions (m/z 451.3, 676.7) to obtain amino acid sequence information via MS/MS. The resulting product-ion spectrum of m/z 451.3 is shown in Fig. 16C. This analysis features a reduced resolution on both mass analyzers (Q1 and Q3) to obtain maximum sensitivity. The corresponding amino acid residues are assigned by using a template structure identi®cation approach on the basis of mass differences between peaks. The partial sequence FQ[I,L]F is determined directly from the product-ion spectrum and entered into an in-house database searching program, FRAGFIT. Because leucine and isoleucine cannot be distinguished with low-energy collisions in a triple quadrupole mass spectrometer, FQIF and FQLF were both searched. The protein that corresponds to myosin light chain was matched by using this partial sequence information, peptide masses, and search parameters [Table 9]. The development and use of various protein sequence databases for automated search routines (Eng et al., 1994; Clauser & Baker, 1997) are an essential component of protein analysis that uses mass spectrometry techniques. These programs require only a few peptides for matching; therefore, the absence of a match for a particular peptide does not affect the search performance. The use of protein database searches provide an ef®cient way of con®rming a putative sequence from corresponding full-scan mass spectrometry and MS/MS data. The capillary LC/MS-based approach for peptide mapping performed by Arnott et al. (1995) features miniaturized sample-loading procedures, which are routinely amenable to small quantities of peptides. The reliable characterization of protein/peptide mixtures in conjunction with the widely used 2-DGE methods offers a powerful ®ngerprinting approach in the pharmaceutical industry. Low femtomole detection limits (typically <50 femtomole) with a mass accuracy of  0.5 Da provide unique advantages for protein identi®cation. Liberal parameters for mass range and unmatched masses are used for the initial protein search, whereas more conservative parameters are used to reduce the number of matches and to improve the con®dence in the search. The use of HPLC on-line or off-line is an essential feature for peptide mapping to integrate the removal of buffers and salts (puri®cation) and the separation of LC/MS APPLICATIONS & FIGURE 16. Identi®cation of proteins separated by 2-DGE. (A) ``Base peak pro®le'' from the capillary HPLC separation from a selected protein spot. (B) Mass spectrum of the second major chromatographic peak (residue 60±70). (C) LC/MS/MS product-ion spectrum of m/z 451.3 (Arnott et al., 1995). analytes (preconcentration) with mass spectrometry. With on-line LC/MS approaches, low ¯ow rates (<100 mL/min) have been demonstrated to provide maximum sensitivity with ESI techniques for the analysis of proteins. In the work performed by Arnott et al. (1998), a pre-column split of the mobile phase (200 mL/min) was used to obtain a ¯ow rate of 0.2±0.5 mL/min. This microcolumn HPLC method minimizes the column elution volume and maximizes the concentration of peptide that enters the ESI interface of the mass spectrometer (Kennedy & Jorgenson, 1989). Other examples of low-¯ow approaches include ``peak parking'' techniques, where the solvent ¯ow and gradient are stopped and the sample continues to ¯ow into the mass spectrometer (Davis et al., 1995). This procedure provides longer sampling times and an improved signal-to-noise ratio because the signal can be averaged for a greater length of time. On-line LC/MS approaches for peptide mapping are also used with ESI±TOF mass spectrometers. Banks and Gulcicek recently described the analysis of peptide mixtures with separations in less than 35 sec, using a nonporous silica support (Banks & Gulcicek, 1997). The 219 & LEE AND KERNS TABLE 9. FRAGFIT results for spot #6 from 2-DGE of normal neonatal rat cardiac myocytes, indicating a match with myosin light chain (Arnott et al., 1995). Data Summary 5 major peaks with [M ‡ H] ‡ 1783.8, 1352.1, 1751.8, 2017.6, 1992.6 4 smaller peaks with [M ‡ H] ‡ 2164.1, 1510.1, 3285.1, 3171.1 MS/MS data yields partial sequences for 1352.1 : FQ[I,L]F and 1783.8 : APE FRAGFIT parameters (chosen for all searches) Lys-C speci®city Oxidized Met allowed Partial digestion allowed One unmatched mass allowed Carboxymethyl Cys assumed A. Input: Masses of the 5 major peaks and mass tolerance of  0.5 Da., Output: 4 database hits: Myosin light chain 1, slow-twitch muscle b/ventricular isoform (22025.01 Da) Reticulocyte-binding protein 1, PvRBP-1 ˆ transmembrane-anchored (325851.41 Da) Reticulocyte binding protein 1 precursor--Plasmodium vivax (330217.57 Da) Titin--rabbit (fragment) (751113.32 Da) [4 specified mol wts matched. not found: 2017.60] B. Input: Masses of all peaks, mass tolerance of  0.5 Da, and partial sequences, Output: 1 database hit: Myosin light chain 1, slow-twitch muscle b/ventricular isoform (22025.01 Da) 1784.03 1510.68 0.23 0.58 19: AAPAPAAAPAAAPEPERPK 48: IEFTPEQIEEFK 1752.02 1352.53 0.22 0.43 46: IKIEFTPEQIEEFK 60: EAFQLFDRTPK 3MO 3284.80 0.30 75: ITYGQCGDVLRALGQNPTQAEVLRVLGKPK 2017.38 0.22 112: MMDFETFLPMLQHISK 1MO 1992.15 0.45 130: DTGTYEDFVEGLRVFDK 3171.51 0.41 147: EGNGTVMGAELRHVLATLGERLTEDEVEK 1MO 2164.37 0.27 176: LMAGQEDSNGCINYEAFVK C. Input: A single mass plus partial sequence: 1352.1 FQ[I,L]F, Output: 1 database hit: Myosin light chain 1, slow-twitch muscle b/ventricular isoform (22025.01 Da) resulting chromatographic peak widths are less than 1 sec wide. The unique nature of the TOF mass spectrometry allows for the complete analysis of these narrow peaks. Off-line approaches also feature sample clean-up, preconcentration, and the batch elution of peptides. This offline feature is commonly used with MALDI±TOF techniques for the simultaneous analysis of peptide mixtures. Because these methods are decoupled from the actual analysis, highly automated procedures can be used to prepare and manipulate samples (Houthaeve et al., 1995; Stevenson et al., 1998). 2. Glycoprotein Mapping Similar analysis strategies can be applied for the peptide mapping of glycoproteins (Medzihradszky et al., 1994; Reinhold et al., 1995; Hancock et al., 1996). Carr et al. combined in-source collisionally induced dissociation (CID) with LC/MS/MS to identify sites of N- and 220 O-linked glycosylation (Carr et al., 1993). This novel approach uses a series of LC/MS and LC/MS/MS experiments to generate peptide maps and to selectively screen for glycoproteins. The method is based on the characteristic fragmentation of glycoproteins to form a N-acetylhexosamine (HexNAc ‡ ) fragment ion at m/z 204. This fragment ion serves as a diagnostic marker for N- and O-linked glycopeptides. A precursor-ion experiment is performed on complex mixtures to detect precursor ions that fragment to produce the m/z 204 (HexNAc ‡ ). The resulting precursor ions correspond to the [M ‡ H] ‡ ions of glycopeptides. This LC/MS/MS-based strategy for mapping glycoproteins and proteins is shown in Fig. 17. A mixture of peptides and glycopeptides is generated when the glycoprotein (>75 pmol) is reduced, alkylated, and enzymatically digested. A series of three separate experiments, requiring ca. 25 pmol of digest per experi- LC/MS APPLICATIONS & FIGURE 17. LC/MS/MS strategy for mapping glycoproteins and proteins via reduction, alkylation, and enzymatic digestion, followed by three LC/MS and LC/MS/MS experiments (Carr et al., 1993). ment, is used to identify the N- and O-linked glycopeptides. The ®rst experiment uses LC/MS to provide a map of the peptide portion of the protein and to indicate the presence of glycopeptides. The second experiment involves the use of LC/MS/MS to screen the mixture for compounds that fragment to yield the diagnostic m/z 204 (HexNAc ‡ ). N- and O-linked glycopeptides are both identi®ed. A ®nal experiment is performed on a sample treated with peptide N:glycosidase F (PNGase F) to release the N-linked oligosaccharides and to identify the O-linked glycopeptides exclusively. The utility of this procedure was demonstrated on a digest of bovine fetuin, a 42 kDa glycoprotein. The TIC chromatogram and a reconstructed ion current (RIC) trace for m/z 204 from a trypsin/Asp-N digest of bovine fetuin are shown in Fig. 18A and B, respectively. In this experiment, the ori®ce voltage is ramped linearly from 120 to 65 V to enhance the production of the diagnostic m/z 204 (HexNAc ‡ ). The TIC chromatogram, corres- ponding to the precursor-ion scan of m/z 204 (HexNAc ‡ ), is shown in Fig. 18C. This TIC trace indicates the presence of at least 12 glycopeptides (N- or O-linked) that generate the m/z 204 product ion. The resulting spectra typically contain multiple charge states that correspond to triply and doubly protonated molecules. From the observed masses, the average molecular weight of the glycopeptide is calculated. Further treatment with PNGase F results in the selective removal of N-linked carbohydrates, and in the identi®cation of N- or O-linked glycopeptides. Liu et al. described a glycoprotein analysis strategy that features micro-LC/MS techniques for the highresolution chromatographic separations of complex peptide mixtures (Liu et al., 1993). By digesting biological samples with appropriate enzymes such as trypsin and PNGase F, comparative maps are generated and used for locating glycosylated peptide fragments. This example highlights the application of chromatography to achieve 221 & LEE AND KERNS FIGURE 18. Application of the LC/MS/MS strategy for mapping glycoproteins to bovine fetuin (42 kDa). A trypsin/Asp±N digest of bovine fetuin was analyzed using the strategy shown in Fig. 17. (A) Total ion chromatogram (TIC), scanning m/z 150±2000. (B) Reconstructed ion chromatogram (RIC) trace for m/z 204 (HexNAc ‡ ), scanning m/z 150±2000. (C) TIC screening precursors of m/z 204 (Carr et al., 1993). the necessary selectivity for identi®cation by mass spectrometry. This procedure for glycoprotein characterization involves the generation of a full-scan mass spectrum of the intact glycoprotein without pretreatment. From this spectrum, a pattern of glycosylation microheterogeneity is determined. Next, a full-scan mass spectrum of the sample is obtained after pretreatment with PNGase F. Most glycosylated peaks disappear, and the molecular mass of the deglycosylated protein is identi®ed. The possible glycoforms observed from this type of LC/MS experiment performed on bovine ribonuclease B are listed in Table 10. Diagnostic increments of Mr 162 and Mr 203 are observed, which correspond to hexose and N-acetylhexosamine residues, respectively. By comparing peptide maps prior to, and after enzymatic cleavage, peptide fragments attached to oligosaccharides may be 222 located and distinguished from non-glycosylated peptides. The comparative map of enzymatic digests obtained for bovine ribonuclease B is shown in Fig. 19. In this particular case, abundant signals are exhibited and assigned to the majority of expected peptide fragments. The bene®t of this comparative mapping scheme is the rapid visual survey of unique peaks indicative of glycosylation. This comparative map indicates a unique broad peak between 16 and 17.5 min. Based on the protocols for enzymatic treatment, this peak is related to the heterogeneity of glycoforms that are attached to a speci®c tryptic peptide. The tryptic fragments of bovine ribonuclease B are further reacted with PNGase F and analyzed with LC/MS and LC/MS/MS standard methods to establish the identity and attachment of the carbohydrate substructure. This micro-LC/MS approach provides a valuable method for the rapid microheterogeneity LC/MS APPLICATIONS & TABLE 10. Heterogeneity of ribonuclease B glycoforms observed by electrospray LC/MS analysis (Liu et al., 1993). Native protein Mr Polypeptide chain Polypeptide adduct (phosphate) Hex5-(HexNAc)2-polypeptide Hex5-(HexNAc)2-(polypeptide ‡ phosphate Hex6-(HexNAc)2-polypeptide Hex6-(HexNAc)2-(polypeptide ‡ phosphate Hex7-(HexNAc)2-(polypeptide ‡ phosphate Hex8-(HexNAc)2-(polypeptide ‡ phosphate adduct) adduct) adduct) adduct) 13,692 13,789 14,908 15,005 15,070 15,168 15,326 15,492 PNGase F-treated protein Polypeptide chain Polypeptide adduct (phosphate) HexNAc-polypeptide HexNAc-(polypeptide ‡ phosphate adduct) Mr 13,692 13,791 13,895 13,993 FIGURE 19. Comparative peptide map of enzymatic digests of bovine ribonuclease B. (A) TIC trace of digested RCM-glycoprotein with trypsin. (B) TIC trace of digested RCM-glycoprotein with trypsin and PNGase F (Liu et al., 1993). 223 & LEE AND KERNS screening of oligosaccharide substructures and the determination of carbohydrate content. Additional structural analysis of glycoproteins generally requires fragmentation by chemical or enzymatic cleavage and by separation or isolation. On-line LC/MS methods provide an integrated approach for this type of analysis. This approach is consistent with the early analyses of glycoproteins that were described by Covey et al. (1991) and Ling et al. (1991). Comparative maps of enzyme digests allow for a quick survey of the expected peptide fragments. Various LC/MS schemes may be used to indicate unique peaks related to the heterogeneity of glycoforms that are attached to a speci®c tryptic peptide (Reinhold et al., 1995) or to quantitate speci®c glycoprotein structures (Mazsaroff et al., 1997). 3. Natural Products Dereplication Natural products offer a source of unique chemical diversity for the pharmaceutical industry. Numerous successful drugs derived from natural products have been introduced for the treatment of cancer (Hung et al., 1996; Pettit et al., 1994), immunosuppression (Perico et al., 1996), cardiovascular therapy (Nadin & Nicolaou, 1994; Tomoda et al., 1994), and anti-infective therapy (Turner & Rodriguez, 1996). Traditional approaches for natural product screening in drug discovery involve the testing of crude extracts obtained from microbial fermentation broths, plants, or marine organisms. When activity above a certain level is detected, active components are isolated and puri®ed for identi®cation. This process is often time-consuming; where the physicochemical characteristics of the active components are determined, known compounds are identi®ed (dereplication), and the novel compounds are scaled-up for more detailed investigation. Shorter discovery timelines and accelerated development expectations have hindered the traditional approaches in natural products research. Furthermore, the recent emphasis on chemical diversity presents a great challenge in this area, particularly because traditional natural products screening programs focus on one source of chemical diversity such as microorganisms or plants. Still, the primary issue remains: how to assay this ideal source of new, biologically active compounds within the current timeframe necessary for modern drug discovery research. At the heart of this issue is the fact that traditional isolation and scale-up procedures are highly inef®cient and often become the bottleneck in natural products dereplication. A speci®c area of drug discovery research, which required an immediate need for highly sensitive and rapid analysis, dealt with antibiotic-containing mixtures from fermentation broths. The need for rapid dereplication of 224 bioactive compounds derived from natural sources is critical due to the overuse of antibiotics (Georgopapadakou & Walsh, 1996). This overuse has signi®cantly accelerated the rate of pathogen mutation, and has resulted in a growing resistance to current drugs (Davies, 1994). Furthermore, there is an increasing number of immunocompromised patients due to AIDS (Bates, 1995). Thus, novel antibiotic compounds are in high demand. Most antibiotics come from secondary metabolites of soil microorganisms that inhibit bacteria or fungi. Largescale screening of microorganism fermentations followed by isolation and structure elucidation is required. Because many natural products have been previously identi®ed, approaches that avoid time-consuming isolation and provide quick elucidation are essential. To address these needs, a new strategy was introduced by Ackermann et al. that uses an on-line ESI±LC/MS approach that integrates multi-component identi®cation, fraction collection, and sample preparation for bioactivity screening (Ackermann et al., 1996). Using this LC/MSbased approach, crude extracts are screened without extensive puri®cation and chemical analysis. Less material is required due to the sensitivity of the technique and chromatographic resolution is retained. This resolution would ordinarily be relatively poor with a fraction collection process. Furthermore, because the ESI acts as a concentration-dependent detector (Hopfgartner et al., 1993), the HPLC ef¯uent can be split between mass spectrometry analysis and fraction collection. Thus, molecular weight information is obtained, and approximately 90% of the fractionated material is recovered for biological testing. The key to rapid dereplication, using this approach, is dependable molecular weight determination. This information is used with existing natural product databases that contain information on: bioactive compounds, the physical descriptions of the microorganisms from which they come, their spectrum of activity, the method of extraction and isolation, and physical data (i.e., molecular weight, UV absorption maxima). Molecular weight is the most critical information for initial searches because of its link to structural speci®city. This information is used to make pivotal decisions on whether or not to proceed to more time-consuming isolation steps based on novelty of the compound. In the study reported by Ackermann et al. (1996a), LC/MS is used to increase sensitivity and accelerate analysis. These features serve to signi®cantly reduce labor. The instrumental con®guration of the LC/MS system developed by Ackermann et al. is shown in Fig. 20. This system features an HPLC, UV detector, fraction collector, ESI-triple quadrupole, and MALDI mass spectrometers. Filtered fermentation broths are extracted with butanol or ethyl acetate, and eluted on a gradient C18 reversed-phase HPLC separation. The eluent is split 1:10 between a LC/MS APPLICATIONS & FIGURE 20. Diagram of instrumental con®guration of the LC/MS system used for the characterization of crude fermentation extracts. The system consists of the following components: (1) HPLC; (2) loop injector; (3) guard column; (4) 5 mm C18 HPLC column (4.6 mm  25 cm); (5) zero dead volume tee; (6) UV detector; (7) fraction collector; (8) triple quadrupole mass spectrometer equipped with ESI interface; (9) ESI power supply and gas manifold; and (10) syring pump (Ackerman et al., 1996). quadrupole MS/MS instrument (scanning 250±2000 amu/ 3 sec in the full scan mode) and a single wavelength UV detector (254 or 230 nm). One-minute fractions are collected after the UV detector. Of these fractions, 20± 50 mL is used for MALDI±TOF analysis, and the remainder is concentrated for microbiological testing. The LC/UV chromatogram is compared to the bioactivity assay histogram to highlight the peaks that contain activity. The molecular weights of the active peaks are obtained for novelty assessment of the compounds. Using this approach for natural products dereplication, data are routinely obtained from 40 mg of crude extract. Performance examples include the identi®cation of 16 analogs of teicoplanin and 12 analogs of phenelfamycin from separate samples. The summary of results obtained for phenelfamycin is shown in Table 11. The correlation of fraction, retention time, and molecular weight provides the essential information for rapid dereplication and identi®cation. The time required to dereplicate natural product samples is about one week with this LC/MS-based method compared to several weeks by previous methods that involve traditional isolation steps. The use of this LC/MS-based methodology results in greater clarity and con®dent decisions for proceeding with the full structural study of an active component that is derived from a culture. Similar approaches that use on-line LC/MS and LC/ MS/MS techniques have been recently described for natural products dereplication (Gilbert & Lewer, 1998; Janota & Carter, 1998). In the method described by Gilbert and Lewer, approximately 100,000 natural product extracts are screened annually for in vivo and in vitro activity, using the system shown in Fig. 21. The approach for dereplication involves a comparison of retention time, full scan mass spectra (i.e., molecular weight information), and MS/MS spectra with those from known biologically active standards. Thus, previously identi®ed components are rapidly eliminated and do not require time-consuming structure elucidation studies. This saving of effort allows researchers to focus efforts on novel chemistries. Samples of novel compounds are infused into an ion trap mass spectrometer, and an MSn fragmentation map is generated. Using this LC/MS-based approach, the structure of homolelobanidine was identi®ed from an extract of Polygonum ¯acidum. The MSn feature provides unique structure identi®cation capabilities that allow for the facile assignment of neutral losses and characteristic substructures of the molecule. 225 & LEE AND KERNS TABLE 11. Mass spectral LC/MS and MALDI±TOF data summary for the novelty assessment of phenelfamycin sample GE21640 F VI 45 (Ackerman et al., 1996). Active fraction b Retention time (min) Molecular weight monoisotopic 8:40 9:15 n.a. 13:45 14:30 14:40 15:30 16:25 17:35 18:20 19:15 19:40 819.4 963.5 n.a. 829 973 937.5 1081.6 1225.6 937.5 1081.6 1225.6 927 8 9 12 13 14 14 15 16 17 18 19 19 [M ‡ K] ‡ ESI 858.2 1002.4 n.d. 868.2 1002.4 976.2 1120.3 1264.3 976.4 1120.2 1264.4 966.3 [M ‡ K] ‡ MALDI Identityc n.d. n.d. n.d. n.d. n.d. 976 1121 1263 976 1121 1263 n.d. UN-1 UN-2 unk unk unk A C E B D F unk Key fragment ions (m/z) 626,504 626,504 n.a. n.a. n.a. 744,622 744,622 744,622 744,608 744,608 744,608 n.a. a n.a. ˆ not applicable; n.d. ˆ not detected; unk ˆ unknown. See histogram in Fig. 6. c For structural assignments refer to Fig. 5. b FIGURE 21. Diagram of instrumental con®guration of the LC/MS system used for annually screening approximately 100,000 natural product extracts (Gilbert & Lewer, 1998). 4. Bioaf®nity Screening Combinatorial chemistry has changed the strategy of drug candidate synthesis. As a result, hundreds of thousands of compounds are now screened against a particular biological target. Once activity is determined for a mixture, the identi®cation of the active component(s) is necessary. One strategy is to iteratively re-synthesize sub-pools of the mixture. However, this approach requires a considerable amount of resources. To reduce the time and resources required for screening the large number of compounds produced by combinatorial chemistry, approaches featuring the parallel screening of mixtures of compounds (20±30) have been investigated. The recent studies performed by Anderegg and coworkers describe the use of bioaf®nity selection LC/MS methods for the identi®cation of active mixture component(s) (Davis et al., in press). This approach features an integrated bioaf®nitybased LC/MS screening method to separate and identify compounds from mixtures. This work is an extension of previously described studies performed by Nedved et al. (Nedved et al., 1996) and Dollinger and coworkers (Kaur et al., 1997), where 226 ligands are injected onto chromatography columns that contain target proteins to observe various degrees of ``selection''. Compounds that bind to the proteins are selectively bound to the column, and are eluted for identi®cation. Other studies have reported on the successful use of ultra®ltration membranes to selectively retain compounds that are bound to target proteins (Van Breemen et al., 1997; 1998). Unbound molecules pass through the membrane and the bound molecules are released and identi®ed. In the approach described by Anderegg and coworkers, LC/MS is incorporated as a bioaf®nity screening strategy for lead identi®cation in drug discovery. A mixture of compounds is incubated with the target protein and the components bound to the protein are selected by using a size exclusion chromatography (SEC) ``spin column''. In this experiment, the unbound compounds are retained on the column. The bound components are eluted and identi®ed with LC/MS. The spin column enrichment scheme is illustrated in Fig. 22. Increased speci®city is obtained by dissociating the bound compounds and performing a second equilibration incubation with the protein. This procedure preferentially selects for LC/MS APPLICATIONS & FIGURE 22. Procedure for bioaf®nity screening of combinatorial drug candidates libraries with two cycles of iterative size exclusion chromatography, using ``spin columns'' to separate the receptor and receptor±binder complexes from unbound ligands (Davis et al., in press). the compounds with higher af®nity, and results in an enhancement of the quantitative LC/MS response. Iterative stages of incubation, size-exclusion, and LC/MS allow the tighter binding components to be enriched relative to weaker binding components. In this study, the peroxisome proliferator-activated receptor (PPARg), which is a target for anti-diabetic drugs (construct molecular weight of 32,537 Da), is incubated with ten ligands that range in molecular weight from 283 to 587 units. A spin column of 6,000 Da cutoff is used for SEC purposes. The retained mixture of components is analyzed by fast perfusive chromatography (Regnier, 1991; Afeyan et al., 1991), using a standard full-scan LC/MS strategy. This analysis procedure allows for the identi®cation and quantitation of the protein and the ligands, compared to their responses prior to incubation. The ligand-protein complex that dissociated under the reversed-phase chromatographic conditions is selectively detected. The LC/MS response of the 10-compound mixture is shown in Fig. 23. After one pass through the spin column, several weak-binding ligands (B, F, H) disappear, and several others diminish in relative amount (D, E, G). After a second pass through the spin column, the three weak-binding ligands (D, E, G) are further diminished. Four ligands remain, corresponding to the tight-binders (A, C, I, J). This analysis scheme provides a quick measurement of binding af®nity, and serves as a screening tool during drug candidate selection. Spreadsheets are constructed and used to calculate the binding af®nity of the components. In this example, two incubation cycles followed by the SEC separation provide an enhancement of strong binders to weak binders. This LC/MS-based method provides a unique approach to obtain information in situations when lower concentrations of tighter binding ligands are present in the same mixture with higher concentrations of weaker binding ligands. Furthermore, this method is more ef®cient than synthetic deconvolution procedures and does not require the use of radioligands. 5. Open-Access LC/MS In the early 1990's, the notion that LC/MS could be developed into an automated, walk-up system for structure characterization of medicinal lead compounds was prevalent. At the time, mass spectrometers were becoming more stable and easier to use. Microprocessor control, robust ion optics, detectors, reliable LC/MS interfaces, and autosamplers created a high degree of sophistication as well. These improvements, combined with the realization that molecular mass was suf®cient for structure con®rmation of synthetic products, led to a qualitative shift in lead compound characterization. For many years, medicinal chemists were accustomed to open access or ``self service'' NMR and IR for routine and rapid con®rmation of molecular structure. An openaccess structure provides a convenient environment to follow synthetic schemes ``in process''and to facilitate the gathering of information, thus allowing chemists to quickly and ef®ciently proceed to the next step of 227 & LEE AND KERNS FIGURE 23. LC/MS extracted ion current pro®les for ten combinatorial drug candidate library components, using the bioaf®nity screening procedure shown in Fig. 22. (A) Before passing through a spin column. (B) After one cycle. (C) After two cycles. The enhancement of tight-binding ligands is evident (Davis et al., in press). synthesis. During this time, medicinal chemists experienced tremendous pressure to be productive. At the same time, analytical chemists became inundated with samples. Clearly, traditional NMR and IR approaches were not feasible due to sample quantity and throughput limitations. The mutual need was to devise alternate approaches. LC/MS methods, using an open-access format, were ®rst developed as a primary means of structure con®rmation (Taylor et al., 1995; Pullen et al., 1995). This approach validated the belief that monoisotopic molecular mass was suf®cient for the structure con®rmation of synthetic products, and that detailed spectral interpretation was not necessary during this drug discovery activity. In 228 general, these approaches feature integrated analysis strategies that emphasized simplicity with less detail. In the open-access LC/MS procedure described by Pullen et al., the samples are directly introduced from solution for ease of automation and sample preparation. Chemists prepare samples in solvent to a suggested concentration range, then log the samples into the system. The sample log-in is done at any time during the continuous automated queue. Autosampler vials are used to hold the samples, and autosamplers are used to directly deliver samples in solution to the mass spectrometer. The system uses a standard method to analyze the samples in queue, average spectra according to a pre-set scheme, and LC/MS APPLICATIONS print out a spectrum for the chemist. Fail-safe procedures for untrained users and instrument self-maintenance at start-up and shut-down were also developed. The LC/MS analyses were performed with either TSI or particle beam (PB) interfaces. These systems successfully analyzed the labile, polar, or higher mass compounds, whereas a complementary GC/MS system was used for volatile compounds. The LC/MS system proved to be widely applicable to a diverse range of compounds. The TSI and PB systems were both successful for 80± 90% of the compounds analyzed. Automated, open-access LC/MS analyses performed well because sample throughput was expected to reach 250,000 in 1995. This throughput corresponds to ca. 1,000 samples per day. Development of the method involved the installation of a system in an existing mass spectrometry laboratory and working with chemists for three months to determine speci®c needs and to develop a consistent, reliable procedure. The instrument was moved to an open-access laboratory and chemists were trained in its use. It should be emphasized that a key to making this approach a success is the fact that instrument down-time was kept to a minimum. Understandably, maintenance is done at offpeak times, and support mechanisms are put in place so that problems are immediately addressed. Training and education was highlighted as a key factor for the successful implementation of this LC/MS system to optimize performance and to reduce the possibility of instrument contamination. In 1995, Taylor and coworkers also described the use of an open-access LC/MS system for routine structure con®rmation, featuring APCI. This system featured dual personal computers (PCs) for automated instrument control and sample log-in. A ``system-PC'' is responsible for running the Windows NT for Workgroups operating system and interfaces with the network for instrument control. A separate ``log-in PC'', isolated from the LC/MS system, is used by the synthetic chemist to enter details about the samples. The analyst prepares the sample in an autosampler vial in one of several solvent options. The system speci®es where to place the sample vial in the autosampler, and following analysis via a standard method, spectra are automatically processed and printed without any chemist intervention. An average of 76 samples per day were analyzed over a 6-month period in 1995, for a total of approximately 8,000 samples. This rate compares with a 20±30 samples per day output prior to introducing the open-access LC/ MS system. The total cycle time is less than 4.5 min per sample. The APCI method, when 60 representative compounds were tested, generated more intense molecular ions, less fragmentation, and better signal/noise ratio than TSI. The ESI approach was found to be a softer ionization method, but produced multiply charged ions & and a higher chemical background than APCI did. Typically, [M ‡ 1] ‡ and [M ‡ 23] ‡ ions, corresponding to protonated and sodiated molecules, respectively, are observed with ESI. The APCI technique produces intense molecular ions for the variety of compounds with little fragmentation. No signi®cant memory effects are observed from sample to sample, due to the 4-min cycle, during which time the mobile phase ``self-cleans'' the system. As highlighted with the previous example, the ability to generate simple, predictable, easy-to-interpret mass spectra without signi®cant contamination from sample-to-sample is of paramount importance for success. Taylor et al. further demonstrated the value of openaccess LC/MS systems for generating a widened scope of pharmaceutical analysis applications, including: (1) characterization of synthetic intermediates and target compounds; (2) reaction monitoring; (3) reaction optimization; (4) analysis of preparative HPLC fractions; and (5) analysis of thin layer chromatography (TLC) plate spots. The availability of these methods led to the increased use of LC/MS for structural analysis. The short analysis time and reliable structure con®rmation resulted in the use of LC/MS as a ®rst choice for structure characterization for synthetic chemistry applications, as well as an expanded, and perhaps, integrated role of sample generator and analyst. Chemists now routinely use open-access LC/MS in the same way that they previously used TLC to monitor reaction mixtures for the desired product and to optimize reaction conditions. In practice, medicinal chemists require only molecular mass data, and are comfortable with a variety of ionization methods to obtain this information. However, con®dence in the actual method and procedure is a requisite. Today, molecular mass measurement has quickly become a preferred means of structure con®rmation over NMR and IR during the early stages of synthetic chemistry activities, where sample quantities are limited. Open-access LC/MS formats have spawned new dimensions in data management and access. Versatile software packages for data manipulation and processing have been a popular approach for integrating analysis and information (Tong et al., 1998). These software programs are ef®ciently implemented with stand-alone computers and servers that are networked with open-access mass spectrometer data systems (Fig. 24). In this con®guration, the data are generated, visualized, processed, and automatically reported to the chemist. The program compares a template of predicted molecular ions with the actual ions generated by ESI and APCI for the quick analysis of synthetic products, intermediates, reactants, reagents, and contaminants. A list of observed ions along with known artifact ions is generated and used to provide a measure of the quality-of-®t to the predicted product(s). 229 & LEE AND KERNS FIGURE 24. Software packages for data manipulation and processing, using standalone computers and servers that are networked with open-access mass spectrometer data systems (Tong et al., 1998). Open-access LC/MS systems provide an effective means for maintaining the high-throughput characterization of synthetic compounds. These systems offer an ef®cient integration of sample generation and analysis activities from laboratory- to bench-scale. Advances in analytical instrumentation and electronic communication have also played a major role in the emergence and acceptance of LC/MS as a front-line tool for structure characterization. From an industry perspective, these advances were signi®cant catalysts for the acceptance of LC/MS analysis responsibilities by mainstream sample generators (i.e., medicinal chemists). A recent review by Burdick and Stults indicates that a similar methodology may be adopted for the analysis of peptide synthesis products, using ESI±MS techniques (Burdick & Stults, 1997). The application of LC/MS for peptide analysis is similar to the previously described schemes for chemical synthesis±puri®cation and focuses on con®rmation of the desired peptide and identi®cation of synthetic by-products. A variety of ESI-based instruments, using quadrupole, ion trap, and TOF, was indicated to have a signi®cant impact due to the performance advantages of ESI±MS, continued lower price, and smaller size. 6. In Vivo Drug Screening Advances in technologies such as combinatorial chemistry, combined with initiatives calling for the faster evaluation of lead compounds, resulted in a tremendous interest in establishing technology for automated highthroughput bioanalysis. These analyses focused on methods for determining multicomponent mixtures of 230 drug candidates that involve pharmacokinetics and metabolic stability. These activities were traditionally (and still are) positioned during the preclinical and clinical stages of drug development. However, the need to provide an early assessment of these properties resulted in new strategies for in vivo drug screening during the drug discovery stage. Methods and approaches that can simultaneously determine mixtures of drug candidates provide a powerful approach for the selection of an optimal drug candidate within a speci®c therapeutic area or a targeted class of compounds. LC/MS/MS approaches, using relatively simple isolation and chromatography conditions, present an effective approach to evaluate new lead compounds for subsequent development. Consequently, the use of LC/ MS-based screens has emerged, and has been successfully applied to large numbers of substances derived from either plasma obtained from animals dosed with a mixture containing several test substances or from pooled plasma from an animal that was sampled at speci®c time points. In general, standard methods applicable to a vast majority of compounds of interest, to ensure throughput capabilities, are critical for LC/MS screens. Although not optimized for speci®city, standard conditions provide a systemic measure of control. This control results in data that have high quality, reliability, and comparability. With a strategic selection of compounds that have similar molecular weights, structural features, and chromatographic properties, the detection selectivity and precision are satisfactory for this particular type of analysis. The premise of this approach is to provide a mechanism for evaluating vast quantities of samples and for forwarding on a select group for further testing, using LC/MS APPLICATIONS traditional, more rigorous approaches. A key to screening is to establish an appropriate method that allows for highthroughput characteristics, while providing for an adequate level of discrimination. Development of an LC/MSbased screen involves a simpli®cation of the overall experiment (i.e., method, expected results, interpretation). For example, an LC/MS method is simpli®ed when it is widely applicable to the majority of compounds under investigation, and it provides yes/no or pass/fail results. This binary mode LC/MS screening is perhaps the simplest experiment for dealing with large amounts of data. Extensions to this approach can be designed to offer speci®c information from the screen, whether binary or more comprehensive. In any case, the threshold criteria for acceptance or failure de®nes the method. Typically, screen performance, emphasizing high throughput qualities, has less discriminating features. Incorporating more discriminating features may result in longer analysis times and/or more attendant data interpretation. Here, the strategy is to provide a screen that satisfactorily meets performance standards and maintains throughput requirements. Once an LC/MS-based screen is developed and locked-in for use (i.e., established as the standard method for subsequent analyses), the object becomes to arrive at the most promising lead compounds within a series quickly without the traditional, rigorous testing, and analysis of each candidate separately. The reward is a signi®cant savings in time and resources. The inherent risk is devising a discriminatory screen that actually discards legitimate drug candidates! The reality is that each screen will likely be unique and highly dependent on instrumentation, technology, method, scientist(s), and scienti®c/ business culture. The development of an LC/MS-based screening method is not intuitive. The object is to devise a method that can signi®cantly impact sample throughput without compromising the quality of data. The strategy is to forego details while maintaining quality. A realistic starting point for LC/MS screening is an approach that uses the ``80/20 Rule'' (see above) (Lee et al., 1997; Heller & Hindle 1998). This approach sets the criteria for method development, method re®nement, and screen performance. High-throughput bioanalysis screening approaches involve the characterization of full-scan mass spectra and MS/MS properties to determine the predominant molecular and product ions, respectively. This information is useful for the selection of appropriate ions for SRM experiments. Settings such as collision energies and CID pressure or gas thickness can be optimized as well. Typically, the most abundant product-ion is selected for SRM. Various acquisition software programs are used to perform the experiment, display the results, and process the data in an automated fashion. & Animal models, involving rats and dogs, have been developed by Olah et al. for pharmacokinetic screeningbased studies in drug discovery (Olah et al., 1997). In this application, LC/MS-based methods are used to simultaneously assay plasma concentrations of up to 12 substances. The plasma is obtained from either single animals dosed with mixtures of lead compounds, or from multiple animals dosed with a single lead compound, after which aliquots of plasma from common time-points are pooled. Essentially, simpli®ed, less stringent versions of preclinical/clinical development procedures are used for sample preparation, assay validation, and analysis. A plasma concentration±time pro®le obtained from a dog administered simultaneously with 12 compounds is shown in Fig. 25. Each compound is quickly evaluated relative to other compounds in the series and comparisons can be made. Desirable pharmacokinetic pro®les are assessed, and the total number of samples analyzed are reduced signi®cantly. In studies that involve the dosing mixtures of lead compounds, the number of animals required is signi®cantly reduced. This methodology provides a perspective on the traditional quantitative LC/MS/MS approaches for the analysis of drugs and their metabolites that are contained in biological ¯uids. Although traditional LC/MS approaches provide a straight-forward way of assessing preclinical pharmacokinetic properties, the volume of drug candidates generated via automated synthesis routines combined with the requirements for turn-around time made this approach costly and timeconsuming during drug discovery. The application of APCI±LC/MS techniques for the rapid determination of protein binding and pharmacokinetics during the discovery stage was recently described by Allen et al., using a single quadrupole instrument (Allen et al., 1998). A ``cocktail'' approach consisted of 4 experimental compounds and a control compound dosed orally at 1 mg/kg with plasma samples obtained at 0.5, 1, 2, 4, and 8 h post dose. To insure reproducibility, the control compound was tested with each cocktail. This approach generated timely systemic exposure (AUC and Cmax) data on 44 test compounds in three work days, using two laboratory scientists. The use of LC/MS/MS-based screening approaches for quantitative bioanalytical measurements allow a large, chemically diverse, range of potential drug candidates to be analyzed quickly and con®dently. The development of unique LC/MS-based systems for in vivo pharmacokinetic screening reduces the analysis to a manageable number of samples, and results in a cost-effective approach to evaluate new lead compounds. Approaches to this type of methodology will likely vary, according to the behavior of the molecules of interest, standard operating procedures (SOPs), performance capabilities of the mass spectrometer, and integration of automated sample preparation, 231 & LEE AND KERNS FIGURE 25. Plasma concentration-time pro®les of 12 compounds given orally as a mixture to a single dog (Olah et al., 1997). and data analysis procedures. Success will likely be dependent on the above parameters, as well as on the degree of tolerance to which the speci®c screen is set. The simultaneous pharmacokinetic assessment of multiple drug candidates in one animal has been termed ``n-in-one'' or ``cassette dosing''. As discussed for the previous example, this parallel approach results in an increased productivity for bioanalysis during drug discovery. Beaudry et al. (Beaudry, 1998) recently investigated the extension of this methodology to study larger 232 numbers of compounds in each mixture, and to integrate sample preparation with the LC/MS/MS system for increased ef®ciency. The number of analytes studied in parallel was extended to 63 plus an internal standard. A list of the 64 analytes and their method performance is shown in Table 12. The increased number of analytes is possible because of improvements to the collision region of the MS/MS system that lead to increased sensitivity and reduced ``memory effects''. In addition, robotic systems for LC/MS APPLICATIONS & TABLE 12. Results of an n-in-one experiment of the parallel LC/MS analysis of 64 analytes from the same plasma sample (Beaudry et al., 1998). Drug name Alprazolam Astemizole Beclomethasone Betamethasone Bromazepam Bromocriptine Bupropion Buspirone Butoconazol Carbamazepine Clemastine Clonazepam Clonidine Clotrimazole Cortisone Danazol Diltiazem Econazole Estazolam Famotidine Felodipine Fen¯uramine Fluconazole Flunitrazepam Fluoxetine Guanfacine Haloperidol Indapamide Indomethacine Indoprofen Itraconazole Ketoconazol Ketorolac Lidocain Lido¯azine Lisinopril Loperamide Lorazepam Lormetazepam Medazepam Miconazol Midazolam Naltrexone Nicardipine Nifedipine Nimodipine Nitrendipine Nitrazepam Nizatidine O¯oxcacin Omeprazole Oxazepam Oxybutinine Paroxetine Pen¯uridol Pentazocine Extracted LOD (pg) Analysis ng/mL Range values R-Values 5 100 25 25 25 2500 25 25 25 25 25 25 250 0.5±500 1.0±500 0.5±500 1.0±500 5.0±500 20±500 0.5±500 0.5±500 0.5±500 0.5±500 0.5±500 5.0±500 5.0±500 0.995 0.993 0.994 0.989 0.991 0.991 0.990 0.994 0.998 0.989 0.991 0.990 0.994 250 1000 25 25 25 1000 100 25 25 25 100 250 25 25 25 25 5.0±500 50±500 0.5±500 1.0±500 1.0±500 20±500 20±500 0.5±500 0.5±500 1.0±500 5.0±500 5.0±500 1.0±500 0.5±500 0.5±500 0.5±500 100 25 25 25 250 25 25 25 100 25 100 25 25 100 25 25 25 250 1000 25 25 25 100 100 25 5.0±500 0.5±500 0.5±500 0.5±500 5.0±500 0.5±500 1.0±500 0.5±500 5.0±500 1.0±500 1.0±500 0.5±500 0.5±500 1.0±500 1.0±500 1.0±500 0.5±500 5.0±500 20±500 1000 500 1000 500 100 25 1000 1000 1000 1000 1000 100 100 Signal too weak 100 10 1000 500 500 25 25 1000 1000 500 100 100 500 1000 1000 1000 Internal standard 100 1000 1000 1000 100 1000 500 1000 100 500 500 1000 1000 500 500 500 1000 100 25 Unstable signal 1000 100 100 100 1000 0.5±500 5.0±500 5.0±500 5.0±500 0.5±500 0.992 0.996 0.993 0.996 0.990 0.985 0.989 0.991 0.998 0.995 0.996 0.991 0.994 0.993 0.996 0.991 0.993 0.995 0.996 0.992 0.997 0.995 0.996 0.995 0.996 0.999 0.990 0.995 0.995 0.993 0.997 0.996 0.998 0.993 0.997 0.997 0.990 0.997 0.994 0.997 (Continued) 233 & LEE AND KERNS TABLE 12. (Continued) Drug name Extracted LOD (pg) Analysis ng/mL Range values R-Values Prazepam Propafenone Ranitidine Spironolactone Temazepam Terconazol Terfenadine Zopiclone 25 25 100 250 25 250 25 25 0.5±500 1.0±500 1.0±500 5.0±500 5.0±500 20±500 0.5±500 0.5±500 1000 500 500 100 500 25 1000 1000 0.997 0.993 0.997 0.999 0.992 0.995 0.992 0.996 Average 111 574 0.994 FIGURE 26. Diagram of a robotic ProspektTM system for on-line SPE automated extraction integrated with LC/MS/MS analysis of plasma samples (Beaudry et al., 1998). sample handling and on-line SPE extraction of plasma samples were integrated with the LC/MS/MS system [Fig. 26]. An isocratic reversed-phase HPLC method provided a cycle time of 4.5 min per sample. The on-line sample preparation and short analysis resulted in an increased sample throughput that required less time from the scientist. The method produced good performance, in terms of extraction ef®ciency, linearity, and LOD, and has the capability of analyzing 320±960 samples per day. The strategic emphasis is on providing high throughput LC/ MS methods for evaluating large numbers of drug candidates during the discovery stage to eliminate poor pharmacokinetic performers. 234 7. Metabolic Stability Screening Recently, the use of fast gradient elution LC/MS techniques was described for high throughput metabolic stability screening (Ackermann et al., 1998). The method uses a HPLC column-switching apparatus to desalt and analyze lead candidates incubated with human liver microsomes. The 12 overlaid SIM pro®les acquired by using a standard method for the test substrates spiked into blank human microsomal supernatant is illustrated in Fig. 27. These pro®les show a favorable comparison among the series of substrates injected over a 12 h period. Substrates were selected whose in vivo clearance is LC/MS APPLICATIONS & This feature is unique and avoids the multiple (2±4) injections that are necessary with other MS/MS con®gurations (e.g., tandem quadrupole). Along with the signi®cant savings in time, detailed structure information is generated, which enables a comprehensive analysis of substructure relationship to be constructed for each metabolite. These automated studies provide unique advantages during drug discovery, and provide an early perspective on the metabolically labile sites, or ``soft spots'' of a drug candidate. This knowledge is useful during lead optimization activities, and can lead to the initiation of proactive research efforts that deal with metabolism-guided structural modi®cation and toxicity. B. Preclinical Development The transition from a lead candidate to IND/CTA represents the interface between drug discovery and preclinical development. With the focus on accelerated development timelines, the relationship between these two functions has become less distinct and less formal than in the past. Today, preclinical development activities are routinely initiated during the mid-to-late stages of drug discovery. The emphasis is primarily on the ef®cient transition of a novel bench-scale lead compound to bulk drug supplies that are suitable for testing on animals and humans. FIGURE 27. SIM pro®les of 12 lead candidates incubated with human liver microsomes, using a high throughput LC/MS system equipped with column switching to desalt and analyze the incubates. The reproducibility of the system over 960 injections is demonstrated. Analysis cycle time is ca. 0.5 min. (Ackermann et al., 1998). controlled predominantly by phase I oxidative metabolism as opposed to phase II metabolism or renal clearance. In this way, the resulting data could be resolved into four categories of metabolic stability: high (  60%); moderate (  30±59%); low (  10±29%); and very low (<10%). The rapid structure identi®cation of metabolites is a powerful complement to previously described quantitative approaches. Davis and coworkers recently demonstrated the utility of an automated metabolite identi®cation approach, using LC/MS/MS with an ion trap mass spectrometer (Lopez et al., 1998). In this study, MSn analysis is automated to provide maximum structural information in combination with predictive strategies for biotransformation. Automated data-dependent scan functions are used to generate full scan, MS/MS, and MSn mass spectra of metabolites within a single chromatographic analysis. a. Analysis requirements. The previous section that highlights recent LC/MS-based analyses in support of drug discovery activities featured fast and highly ef®cient methods. High throughput, speci®city, and robust analysis characteristics were essential to meet the high sample volume (quantitative process) demands brought on by sample-generating technologies such as combinatorial chemistry. In preclinical development activities, analyses are more dependent on diversity of process rather than the diversity of samples. This dependency is because preclinical development analyses typically require the measurement of a diverse number of analytes rather than a single or targeted analyte. Thus, an essential feature of preclinical development analyses is the ability to reliably monitor and identify components of interest, and to relate the information to the corresponding chemical or physiological process. Traditional preclinical development analysis approaches involve the routine use of HPLC techniques. Preclinical development activities are typically slowed when HPLC retention times change, and require authentic standards or the re®nement of the chromatographic method to con®rm identity. The use of LC/MS during this stage of drug development is far less dependent on the use of standards, and provides reliable identi®cation in situations where the retention times vary. Perhaps the 235 & LEE AND KERNS most common type of structure identi®cation activity during preclinical development involves the analysis of metabolites, impurities, and degradants. The early knowledge of structure allows for the quick re®nement of the process directed toward improving a qualitative element of ef®ciency (i.e., yield, bioavailability, potency). Therefore, structure identi®cation plays a central role in prospective research, the acceleration of preclinical development activities, and ®ling of the IND/CTA. b. Analysis perspectives. Preclinical development is the ®rst stage of the drug development cycle, where regulatory issues are formally addressed and validated methods are required. The transition from a regulatoryfree environment in drug discovery presents some unique challenges for sample-analysis activities. For example, issues that deal with the conversion of HPLC methods that employ phosphate buffers to methods that use volatile buffers, which are more suited for LC/MS are common. Also, the evolutionary process of setting requirements and limits for impurities (i.e., speci®cations) is initiated based on analysis results for the IND/CTA. This stage of drug development involves pharmaceutical analysis activities directed toward the support of: (1) metabolism; (2) process research; (3) formulation; and (4) toxicology. These activities represent the ®rst formal interaction between science and policy (i.e., regulatory compliance) within the drug development cycle. c. LC/MS contributions. LC/MS-based approaches have ®gured prominently in this stage of drug development. Bench-scale mixture analysis methodology is integrated into a single micro-scale on-line technique that is driven by the coordinated use of LC/MS and LC/MS/ MS. Analysis strategies that emphasize an early lock-in of analytical methods provide a quick approach to begin support for a wide variety of analysis needs. This standard method feature is critical because method development and re®nement activities are very time-consuming and often delay the initiation of work. The chromatographic separation affords a pro®le of components with reproducible relative retention time, whereas molecular weight is determined and structural information is collected via LC/MS and LC/MS/MS methods, respectively. A template motif for structure identi®cation is used whereby the fragmentation pattern from the product-ion spectrum of the parent drug is used to deduce the structures of the unknown metabolites, impurities, or degradants. Interpretation is systematic because speci®c product ions and neutral losses are correlated with speci®c substructures of the parent drug molecule. Structure databases are generated and provide a quick reference to proposed structure, RRT, molecular weight, and diagnostic functional groups. The databases provide a comprehensive approach to organizing structure information and the basis for comparison. In this way, LC/ MS methods are used during the later stages of drug development to rapidly generate information in support of preclinical development and to provide valuable information in support of registration activities. d. Overview. In this section, the rapid structure identi®cation of metabolites, impurities, and degradants is described during the preclinical development stage. The use of structure libraries that contain information on metabolic and chemical stability is illustrated during metabolism, chemical process research, formulation, and toxicology events. Consideration and approaches for method development, method lock-in, and speci®cations are highlighted for structure database assembly and technology transfer. Table 13 provides a summary of the LC/MS-based applications featured in this section. 1. Metabolite Identi®cation Metabolite identi®cation is central to many of the activities in preclinical development. A more complete characterization of pharmacokinetic properties is performed in animals (typically, rats and dogs) during this stage. The knowledge of the biotransformation pathways of the lead candidate to its metabolites is used to indicate the magnitude and duration of activity. Metabolite identi®cation is critical to many of these activities, and plays an important role in establishing the dose and toxicity levels. The identi®cation of metabolite structures with LC/ MS and LC/MS/MS techniques are an effective approach due to their ability to analyze trace mixtures from complex TABLE 13. Applications of LC/MS in preclinical development. Preclinical development activity Metabolism Process research Formulation 236 Analysis Metabolite identi®cation Impurity identi®cation Degradant identi®cation LC/MS application Standard methods and databases Natural products; standard methods and databases Predictive models for chemical degradants Predictive models for biomolecule degradants Selected References Kerns et al., 1997 Kerns et al., 1994 Rourick et al., 1997 Kleintop et al., 1998 LC/MS APPLICATIONS samples of urine, bile, and plasma. The key to structure identi®cation approaches is based on the fact that metabolites generally retain most of the core structure of the parent drug (Perchalski et al., 1982). Therefore, the parent drug and its corresponding metabolites would be expected to undergo similar fragmentations and to produce mass spectra that indicate major substructures. Recently, Kerns et al. demonstrated the application of LC/MS and LC/MS/MS standard method approaches in preclinical development for the metabolite identi®cation of buspirone, a widely used anxiolytic drug (Kerns et al., 1997). The success of this method relies on the performance of the LC/MS interface and the ability to generate abundant ions that correspond to the molecular weight of the drug and drug metabolites. The production of abundant molecular ions is an ideal situation for molecular weight con®rmation because virtually all the ion current is consolidated into an adduct of the molecular ion (i.e., [M ‡ H] ‡ , [M ‡ NH3] ‡ ). For example, full-scan mass spectra of buspirone contain an abundant [M ‡ H] ‡ ion signal with little detectable fragmentation. The product-ion spectrum reveals product ions and neutral losses that are associated with diagnostic substructures of buspirone [Fig. 28]. The product ion at m/z 122, for example, is indicative of the pyrimidine substructure. The presence of this ion in the product-ion spectrum of a metabolite indicates a structure that contains the pyrimidine substructure. Similarly, the m/z 180 product ion is diagnostic of the azaspirone decane & substructure, and the neutral loss of 164 (producing the m/z 222 product ion) is diagnostic of the butyl azaspirone decane dione substructure. To assist with the MS/MS structure identi®cation, the gross substructure of buspirone is categorized into pro®le groups (Kerns et al., 1995). Pro®le groups directly correlate speci®c product ions and neutral losses with the presence, absence, substitution, and molecular connectivity (Lee et al., 1996) of speci®c buspirone substructures and their modi®cations. The pro®le groups of buspirone are identi®ed with abbreviations that correspond to the three speci®c substructures: azaspirone decane dione (A), butyl piperazine (B), and pyrimidine (P). Substituted substructures are designated with a subscript (s), and a dash (±) denotes substructure connectivity. Thus, the buspirone molecule is represented by A±B±P. The As ±B±P designation refers to metabolite structures that contain the azaspirone decane dione, butyl piperazine, and pyrimidine substructures with substitution on the azaspirone decane dione substructure. The pro®le group categorization within a corresponding database allows the rapid visual recognition of primary substructures affected by metabolism. Table 14 illustrates a representative buspirone metabolite structure database. Information on the structure, molecular weight, UV characteristics, RRT, and product ions of metabolites obtained from rat bile, urine, and liver S9 samples are compiled. Using this format, the predominant buspirone metabolite pro®le groups, As ±B±P, A± FIGURE 28. Product-ion spectrum of the [M ‡ H] ‡ of buspirone, and the correspondence of ions to speci®c substructures that are diagnostic of the compound, and are used as a template for structure identi®cation of buspirone metabolites (Kerns et al., 1997). 237 & LEE AND KERNS B±Ps, and As ±B±Ps are easily recognized. These pro®le groups indicate azaspirone decane dione and pyrimidine as metabolically active sites of attack and the presence of multiple substitution sites on each of these substructures. In many cases, the pro®le groups indicate the occurrence of metabolic reactions on more than one substructure. Table 15 summarizes the product-ion spectra that are obtained for each buspirone metabolite structure. In combination with the data in Table 14, these results provide a comprehensive metabolite structure database that is indexed by speci®c analytical characteristics. This database is a useful source of information for accelerated preclinical development activities. It provides a useful tool for the rapid ``dereplication of metabolites'' observed in previous drug discovery studies, and serves as a future reference for the identi®cation of novel structures during the clinical stages of drug development. This protocol has also been found to be useful during drug discovery research in chemistry, metabolism, pharmacokinetics, activity, and safety studies (Lee et al., 1997). Extensions TABLE 14. Buspirone metabolite database of structures identi®ed with LC/MS and LC/MS/MS with a standard method (Kerns et al., 1997). MH ‡ b Pro®le groupc Sample sourcee ‡ OH ‡ OH ‡ OH ‡ OH ‡ OH ‡ OH N ‡ Glucoronyl ‡ Glucoronyl ‡ Glucoronyl ‡ Glucoronyl ‡ Glucoronyl ‡ Glucoronyl B B B B B B S9 B ‡ 2(OH) Ur, S9 ‡ Glucuronyl Ur, S9 tRRa Hydroxy buspirone glucoronide Hydroxy buspirone glucoronide Hydroxy buspirone glucoronide Hydroxy methoxy buspirone glucoronide Hydroxy methoxy buspirone glucoronide Hydroxy buspirone glucoronide 1-Pyrimidinyl piperazine Buspirone glucoronide Dihydroxy buspirone Methoxy buspirone glucuronide Dihydroxy buspirone Dihydroxy buspirone Despyrimidinyl peperazine buspirone 1 2 3 4 5 6 7 8 9 10 11 12 13 0.14 0.30 0.35 0.36 0.38 0.38 0.42 0.44 0.45 0.46 0.48 0.48 0.50 594 594 594 624 624 594 165 578 418 608 418 418 254 As-B-Ps As-B-Ps As-B-Ps As-B-Ps As-B-Ps As-B-Ps Bs-P A-B-Ps As-B-P A-B-Ps As-B-P As-B-Ps A-Bs ‡ OH ‡ OH ‡ OH ‡ OH, ‡ OH, ‡ OH U ‡ OH, U ‡ OH, U ‡ OH N Dihydroxy methoxy buspirone Hydroxy buspirone Dihydroxy buspirone Dihydroxy buspirone Dihydroxy buspirone Hydroxy buspirone Hydroxy buspirone Dihydroxy buspirone Despyrimidinyl buspirone Hydroxy buspirone Hydroxy buspirone Hydroxy buspirone Buspirone 14 15 16 17 18 19 20 21 22 23 24 25 26 0.51 0.53 0.55 0.55 0.56 0.59 0.69 0.70 0.74 0.74 0.80 0.87 1.00 448 402 418 418 418 402 402 418 308 402 402 402 386 As-B-Ps As-B-P As-B-P As-B-P As-B-P As-B-P As-B-P As-B-Ps A-Bs As-B-P As-B-P A-B-Ps A-B-P ‡ OH, ‡ OCH3 U U U U U U ‡ OH ‡ OH ‡ OH ‡ 2(OH) ‡ 2(OH) U U ‡ OH U ‡ OH ‡ OH U U a R1d R3d No. Proposed structure ‡ OCH3 ‡ OCH3 R2d ‡ Glucuronyl ‡ OCH3 ‡ 2(OH) ‡ OH U ÿPyrimidinyl piperazine, ‡ ------ O, ‡ OH ‡ 2(OH) ‡ OH ‡ OH ‡ OH ÿPyrimidine B Ur, S9 B, S9 Ur B, Ur, S9 Ur Ur B Ur, B, S9 Ur, B, S9 Ur, S9 B B B, Ur Ur, S9 B, Ur, S9 HPLC retention time relative to buspirone (tRR ˆ 1.00) using universal HPLC conditions. Buspirone retention time was 13.5 min. Molecular mass. c Refer to text for discussion of pro®le group categorization. A ˆ azaspirone decane dione, B ˆ butyl piperazine, P ˆ pyrimidine, subscript s refers to substitution in a particular substructure, NA ˆ not available from the data. d U ˆ unchanged substructure, N ˆ substructure not present. e Sample source: B ˆ rat bile in vivo, Ur ˆ rat urine, S9 ˆ rat liver S9 preparation in vitro. b 238 LC/MS APPLICATIONS & TABLE 15. Product-ion spectra of (M ‡ H) ‡ ions of buspirone metabolites and their structural interpretation based on buspirone as a structural template (Kerns et al., 1997). Fragmentation; m/z Proposed structure No. tRR MH ‡ Hydroxy buspirone 1 glucuronide Hydroxy buspirone 2 glucuronide Hydroxy buspirone 3 glucuronide Hydroxy methoxy 4 buspirone glucuronide Hydroxy methoxy 5 buspirone glucuronide Hydroxy buspirone 6 glucuronide 1-Pyrimidinyl piperazine 7 Buspirone glucuronide 8 0.42 165 0.44 578 Dihydroxy buspirone 9 0.45 418 Methoxy buspirone glucuronide Dihydroxy buspirone 10 0.46 608 11 0.48 418 Dihydroxy buspirone 12 Despyrimidinyl 13 piperazine buspirone 0.48 418 0.50 254 Dihydroxy methoxy buspirone Hydroxy buspirone 14 0.51 448 15 0.53 402 Dihydroxy buspirone 16 0.55 418 Dihydroxy buspirone Dihydroxy buspirone Hydroxy buspirone 17 18 19 0.55 418 0.56 418 0.59 402 Hydroxy buspirone 20 0.69 402 Dihydroxy buspirone 21 0.70 418 Despyrimidinyl buspirone 22 0.74 308 Hydroxy buspirone 23 0.74 402 Hydroxy buspirone 24 0.80 402 Hydroxy buspirone Buspirone 25 26 0.87 402 1.00 386 A B C D E F G 0.14 594 138 0.30 594 138 0.35 594 138 166 0.36 624 168 0.38 624 168 0.38 594 138 166 281 138 164 166 122 148 150 168 196 265 222 297 254 H I J Others 281 418 98 238 418 98 238 323 265 323 177 177 139 448 139 448 418 180 168 152 123 168 297 254 281 238 281 238 281 238 281 238 297 238 281 238 265 222 281 238 222 122 148 150 122 148 150 138 166 122 148 150 M 238 307 281 238 307 281 238 291 265 109 265 222 222 95 123 139 95 139 95 180 168 180 168 152 168 152 152 95 95 236, 208, 194, 109, 86, 81 383, 121, 100 98 139 400, 293, 98 98 98 219, 178, 109, 139 400 139 109, 100, 88, 74, 59, 43 168, 98 139 98, 95 139 152 95 180 168 180 168 402 109 432 100 151, 152 122 150 122 148 150 122 148 150 138 164 166 L 418 98 122 148 150 138 166 168 194 196 122 148 150 122 148 140 K 306, 267, 192, 98 108 109, 98 343, 109, 98 239 & LEE AND KERNS of this approach with microprobe NMR analysis of a 50 mg sample of isolated metabolite are routinely performed (Rourick et al., unpublished data, 1997). A ``chemical shift template'' characteristic of unique substructural sites is used to identify speci®c structural modi®cations and to distinguish isomers. 2. Impurity Identi®cation Synthetic impurities are of particular concern during process research and safety evaluation activities. Often, impurities are the result of synthetic by-products or starting materials of the scale-up process. Impurities provide a comprehensive indicator of the chemical process and are diagnostic of overall quality. The resulting information is used by process chemists to guide process optimization. Knowledge of the identity and relative amount of impurities is used to diagnose process reactions so that changes in reagents and reaction conditions lead to better yields and higher quality material. Although it is often dif®cult to assign an exact time period for the completion of chemical process research activities, it is usually the rate-determining step for preclinical development activities. With an increasing number of novel lead candidates that enter into preclinical development, considerable resources are needed to identify impurities. LC/MS-based approaches provide integrated sample clean-up and structure analysis procedures for the rapid analysis of impurities. This advantage was demonstrated during the preclinical development of TAXOL1. LC/MS played an important role for the identi®cation of impurities contained in extracts and process intermediates from Taxus brevifolia and T. baccata biomass. Because drugs derived from natural sources often have a very diverse set of structural analogs, it is important to determine which analogs are carried through the puri®cation process and ultimately appear as impurities. This task presents a unique challenge during the preclinical stage of drug development due to the highly complex nature of the samples. Similar to the approaches described previously for natural products dereplication during drug discovery, traditional strategies for impurity identi®cation rely on scale-up, extraction, isolation, and detailed structure analysis. Unfortunately, these methods are slow, timeconsuming, and problematic for accelerated preclinical development activities. Rapid structure identi®cation methods that use LC/MS and LC/MS/MS protocols are ideal for handling large numbers of drug candidates and are applicable to a diverse range of compound classes. Furthermore, the sensitivity of LC/MS-based methods are suf®cient to study impurities and active compounds at extremely low levels (0.2±2 nmol). 240 Kerns and coworkers developed a structure identi®cation strategy that incorporates LC/MS and LC/MS/MS techniques for rapid, sensitive, and high throughput impurity analysis (Kerns et al., 1994). This approach integrates traditional steps of sample preparation, separation, analysis, and data management into a single instrumental method. The resulting multidimensional data include retention time, molecular weight, UV, and substructure information. A structure database is developed for each candidate and is used to rapidly identify the same impurities in new samples. Structures are proposed based on using the drug candidate as a structural template and, with the use of a standard method approach, consistency for comparison of results throughout the preclinical development process is ensured. The full-scan mass spectrum of paclitaxel is shown in Fig. 29. The spectrum contains abundant molecular ions at m/z 854, 871, and 912, which correspond to [M ‡ H] ‡ , [M ‡ NH4] ‡ , and [M ‡ NH4 ‡ CH3CN] ‡ , respectively. This distinct molecular ion pattern is used to determine the molecular weight of the resulting impurities. The production spectrum of the [M ‡ NH4] ‡ ion of paclitaxel is shown in Fig. 30. Abundant product ions are present in the spectrum and correspond to unique substructures that are diagnostic of paclitaxel. Perhaps most strategic are the product ions associated with the paclitaxel core substructure (m/z 569) and side-chain (m/z 286). This unique pair of product ions serves as a diagnostic reference to structure. Once the product ions are assigned to their corresponding paclitaxel substructure, the spectrum is used as a substructural template for the identi®cation of impurities. A representative HPLC chromatogram with UV detection (230 nm) of a process penultimate lot of paclitaxel from T. brevifolia bark is shown in Fig. 31A. A comparable LC/MS±TIC chromatogram of the sample, obtained simultaneously with the LC/UV pro®le, is shown in Fig. 31B. The responses in each chromatogram compare favorably with each other. The molecular weight of each component is obtained on-line from the full-scan mass spectrum at a speci®c retention time. The sample is again subjected to HPLC separation and a product-ion spectrum is obtained for each compound. Structures of impurities are proposed based on a comparison of the molecular weight, product ions, and neutral losses that are associated with paclitaxel. Table 16 summarizes the novel taxane structures identi®ed with LC/MS. These impurities incorporate several consistent structural variations from the ®ve pro®le groups of paclitaxel [Fig. 32]. Nearly all of the compounds contain the characteristic paclitaxel core substructure as indicated by the product ion at m/z 509 with variations due to modi®cations. Many of these taxanes contain a side-chain similar to paclitaxel, with LC/MS APPLICATIONS & FIGURE 29. ESI full-scan mass spectrum of paclitaxel with a mobile phase that contains aqueous 2 mM ammonium acetate and acetonitrile (Kerns et al., 1994). FIGURE 30. Product-ion spectrum of the [M ‡ NH4] ‡ ion of paclitaxel and correspondence of ions to speci®c substructures diagnostic of the compound, used as a template for structure identi®cation of paclitaxel impurities (Kerns et al., 1994). 241 & LEE AND KERNS FIGURE 31. HPLC chromatograms of a process penultimate lot of paclitaxel from Taxus brevifolia bark. (A) LC/UV (230 nm) chromatogram. (B) LC/MS full scan TIC chromatogram (Kerns et al., 1994). variations occurring on the terminal amide of the sidechain. The product ions that differ from the characteristic side-chain ions of paclitaxel (m/z 286) by values indicative of speci®c substructures demonstrate these terminal amide variations. A comparison with the paclitaxel substructural template indicates structural differences beyond the position of the amide group in the side-chain substructure. When a new impurity is encountered during chemical process research, retention time and molecular weight information are compared to the database for rapid identi®cation. This approach is similar to the procedure described for natural product dereplication. If the compound is not contained in the structure database, then the corresponding LC/MS/MS analysis is performed to obtain substructural detail and the proposal of a new structure. Minimal sample preparation (dilution in HPLC mobile phase) is necessary. A standard reversed-phase HPLC method is used for all the samples that are associated with a drug candidate to reduce time-consuming method development/method re®nement procedures. Standard reversed-phase methods typically involve a 20± 242 30 min cycle time and provide information on a wide range of compounds. The incorporation of a standard method strategy allows the use of autosampling procedures and standard system software for data analysis. During the development of TAXOL1, 90 taxane impurities were rapidly identi®ed and added to the structure database. This MS/MS information is routinely obtained for impurities down to the 100 ng level (injected), and requires approximately 2±3 h for the analysis of each sample. The compounds are structurally categorized with pro®le group terminology. The LC/MSbased methods are signi®cantly faster than previous methods based on scale-up, isolation, fractionation, and individual structural analysis. 3. Degradant Identi®cation The proactive characterization and identi®cation of degradants during the preclinical development phase of drug development offers advantages in the evaluation of drug candidates and their subsequent performance during clinical trials. The early knowledge of degradants provide LC/MS APPLICATIONS & TABLE 16. Novel taxane structures and related derivatives identi®ed by using paclitaxel as a structural template (Kerns et al., 1996). Pro®le groupc R1d R2d R3d R4d R5d 299 S1 C6H5 ± ± ± CH3 ix 0.18 313 S1 C6H5 ± ± ± C2H5 v 0.24 [1.61] 0.25 0.32 777 S3 ±Cs C3H7 H H C6H5 ± ii, vii 791 957 S3 ±C S3 ±Cs CH3 C6H5CH2 CH3CO H H b-Xylose C6H6 C6H5 ± ± i vii 0.36 0.48 805 825 S3 ±C S3 ±Cs C2H5 C6H5CH2 CH3CO H H H C6H5 C6H5 ± ± i viii 0.51 951 S3 ±Cs C6H13 H b-Xylose C6H5 ± vii 0.52 0.53 819 999 S3 ±C S3 ±Cs C3H7 C6H5CH2 CH3CO CH3CO H b-Xylose C6H5 C6H5 ± ± iii, i vi 0.56 [1.84] 0.60 805 S3 ±Cs C5H11 H H C6H5 ± ii, vii, viii 789 S3 ±Cs C4H7 H H C6H5 ± i 831 S3 ±Cs C6H5 CH3CO H CH3C ----- CH2CH3 ± iv 833 805 819 S3 ±C S3 ±Cs S3 ±Cs C4H9 C5H11 C3H7 CH3CO H CH3CO H H H C6H5 C6H5 C6H5 ± ± ± i i i 867 831 833 861 847 560 S3 ±C S3 ±Cs S3 ±Cs S3 ±C S3 ±Cs Cs C6H5CH2 C4H7 C4H9 C6H13 C5H11 ± CH3CO CH3CO CH3CO CH3CO CH3CO H H H H H H H C6H5 C6H5 C6H5 C6H5 C6H5 C6H5 ± ± ± ± ± ± i, iii i, iii ix i, iii i ii 522 Cs ± H H C4H7 ± ii 558 Cs ± H H C6H5CH2 ± ii 630 Cs ± H H C3H6OH ± ii 763 S3 ±Cs C2H5 H H C6H5 ± ii Proposed structure RRta Mol. Wt.b Paclitaxel side-chain methyl ester Paclitaxel side-chain ethyl esterf 10-Deacetyl propyl paclitaxel analog Methyl paclitaxel analog 7-Xylosyl-10-deacetyl benzyl paclitaxel analog Ethyl paclitaxel analogf 10-Deacetyl benzyl paclitaxel analog 7-Xylosyl-10-deacetyl hexyl paclitaxel analog n-Propyl paclitaxel analogf,g 7-Xylosyl benzyl paclitaxel analog 10-Deacetyl taxol C 0.17 10-Deacetyl-7-epicephalomannine Debenzoyl pentenoate 0.70 paclitaxel analog Butyl paclitaxel analog 0.75 10-Deacetyl-7-epi-taxol C 1.00 n-Propyl-7-epi-paclitaxel 1.01 analog Benzyl paclitaxel analog 1.05 7-epi-Cephalomanninef 1.10 7-epi-Butyl paclitaxel analog 1.12 Hexyl paclitaxel analog 1.15 7-epi-Taxol C 1.21 Hydroxy-10-deacetyl [0.75] baccatin III Debenzoyl-10-deacetyl [0.82] baccatin III pentenoate Debenzoyl-10-deacetyl [1.16] baccatin III phenylacetate 10-Deacetyl baccatin III [1.25] hydroxybutyrate 10-Deacetyl ethyl paclitaxel [1.51] Sample sourcee a HPLC retention time relative to paclitaxel (RRT ˆ 1.00) using HPLC condition I or condition II. Molecular weight. c Refer to text for categorization. S ˆ paclitaxel side-chain substructure; C ˆ paclitaxel core substructure; subscripts 1 and 3 refer to the position of substitution and subscript s refers to a general position of substitution on the respective substructure. d Refer to Fig. 1 for the general structures including R groups. For paclitaxel, R1 ˆ C6H5, R2 ˆ CH3CO, R3 ˆ H and R4 ˆ C6H5. e Sample source: i, Impurities in paclitaxel prepared from Taxus brevifolia. ii, Impurities in 10-deacetyl beccatin III puri®ed from T. baccata. iii, Impurities in paclitaxel from T. hicksii. iv, Impurities in semisynthetic paclitaxel. v, Paclitaxel heated with ethanol for 12 weeks at 50 C. vi, Impurities in preparations of 10-xylosyl paclitaxel puri®ed from T. brevifolia. vii, Impurities in preparations of 10-deacetyl xylosyl paclitaxel from T. brevifolia. viii, Impurities in preparations of 10-deacetyl paclitaxel from T. brevifolia. ix, Paclitaxel treated with Na2CO3 (0.05 M) for 10 min at pH 8 and diluted in methanol. f Con®rmed using authentic standard with the result of matching RRT, MW, and MS/MS product-ion spectrum. g Recently reported. b 243 & LEE AND KERNS FIGURE 32. General structures of the ®ve pro®le groups observed using LC/MS, based on the paclitaxel template (Kerns et al., 1995). insight into stability and toxicity. The value of early knowledge of degradants is to provide insights into critical issues and the development of corrective measures prior to clinical trials. During the course of drug development, the bulk drug and drug formulation are studied under a variety of stress conditions such as temperature, humidity, acidity, basicity, oxidization, and light. Qin et al. described the utilization of stressing conditions that may cause degradation (Qin et al., 1994). The resulting samples may be used to validate analytical monitoring methods and to serve as predictive tools for future formulation and packaging studies. A traditional approach to studying degradant formation involves similar time-consuming scale-up and preparation steps as described for metabolite and impurity analysis. Similarly, this area of pharmaceutical analysis 244 has experienced the issues associated with faster drug development cycles. Rourick and coworkers recently described proactive approaches to obtain degradant information with LC/MS methods during the preclinical development stage (Rourick et al., 1996). The procedure incorporates qualitative and quantitative process changes for analysis. The structural information necessary for successful drug development is emphasized. The corresponding structural information provides insight for decisions, based on which leads are to be developed for clinical testing. The early structural information on degradants of a drug candidate offers a unique capability for synthetic modi®cation to minimize degradation. Structural information can also facilitate planning of preclinical drug development in process research, formulation development, and safety assessment. The strategy for impurity and degradant identi®cation described by Rourick et al. (1996) subjects lead candidates to various development conditions, followed by LC/MS and LC/MS/MS analysis protocols. A structure database is constructed from the corresponding results and is used to reveal unstable regions within the drug structure as well as to ascertain which candidate or homologous series of drug candidates may be the most favorable for further development. High capacity and throughput speed are necessary so that many lead candidates may be evaluated. Applicability of the method to a wide range of compound classes is desirable. Once the drug candidate enters clinical development and manufacturing, the structure database is useful for the rapid identi®cation of impurities and degradants in samples generated during these stages of development. The method exposes drug candidates to forced degradation conditions, (e.g., acid, base, heat, and moisture) as a predictive pro®le. The coordinated use of LC/MS and LC/MS/MS provides structure identi®cation for speed, sensitivity, and high throughput. Standard methods, useful for 80% of the compounds, are applied. Various types of structural data are obtained for elucidation purposes (e.g., retention time, molecular weight, MS/ MS), and unknown compounds are elucidated with the candidate drug as a structural template. The LC/MS analysis provides retention time and molecular weight data, whereas LC/MS/MS provides substructural detail for structure identi®cation. Drug candidates are incubated under one of the following conditions: 0.3 N HCl, 0.01 N NaOH, 140 C, or 40 C in water. Automation via autosamplers and system software provides analytical speed and high throughput. Using this approach, ten degradants of cefadroxil, an orally effective semisynthetic cephalosporin antibiotic, were elucidated in a 2-day study. The conditions utilized [Table 17] are predictive of the conditions expected to occur under drug processing, storage, and physiological LC/MS APPLICATIONS & TABLE 17. Predictive pro®le conditions used to study cefadroxil and to rapidly produce a structure database (Rourick et al., 1996). Condition Reagent Acid Base HCl NaOH Heated solid Heated solution H2O Reagent concentration (N) 0.3 0.01 1 Time (H) 2 1.5 0.5 6 8 Temperature ( ) 24 Ambient Ambient 140 40 TABLE 18. Library of cefadroxil degradants and impurities obtained from predictive pro®les obtained by exposing the drug to the various development conditions shown in Table 17 (Rourick et al., 1996). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 0.15 0.20 0.26 0.32 0.40 0.62 0.64 0.77 1.0 1.0 1.98 2.12 2.15 2.17 2.20 2.25 2.35 2.38 379 379 379 317 363 363 233 381 363 363 363 377 398 329 512 363 Cefadroxil ‡ 16 Da Cefadroxil sulfoxide isomer Cefadroxil sulfoxide isomer Cefadroxil isomer
2-Cefadroxil isomer Cefadroxil with hydrolyzed lactam Cefadroxil 7-epi-Cefadroxil Piperazinedione cefadroxil isomer Methyl ester of cefadroxil (impurity) Additional side-chain (impurity) Isomer of piperazine dione of cefadroxil (
2 or 7-epi) conditions throughout drug lifetime. Standard LC/MS methods provide consistency from sample to sample throughout the development process, and allow for the construction and use of a structural database for the rapid identi®cation of impurities and degradants during development. The reversed-phase HPLC conditions provide a general measure of the polarity of each compound, useful for interpretation of substructural differences between related compounds. Due to the mass-resolving capability of the mass spectrometer, chromatographic resolution of coeluting or unresolved components is not required. Abundant protonated molecule ions, [M ‡ H] ‡ , provide reliable molecular weight information, and product-ion spectra generate valuable substructure information for each degradant. The product-ion spectrum of cefadroxil is used as a template for interpretation because speci®c product ions and neutral losses are compared to the spectra obtained from the unknown degradants. Product ions common to each spectra provide evidence of substructures unchanged by the degradation conditions and differences are indicative of structural variations. The product-ion spectrum of cefadroxil contains an abundant product ion at m/z 208, which is diagnostic of cleavage through the lactam substructure. The structural data for 18 observed impurities and degradants are shown in Table 18. The dimensions of chromatographic behavior and molecular weight are easily referenced within this ``LC/MS pro®le''. The mass chromatograms (EIC pro®les) corresponding to the [M ‡ H] ‡ ions of selected components are shown in Fig. 33. This particular representation is useful for distinguishing differences among degradants contained in the base-degraded sample. For example, the difference in molecular weight between cefadroxil and a proximateeluting degradant is 18 Da, which indicates a substructural difference that results from hydrolysis. The product-ion spectrum obtained from the degradant does not contain the diagnostic m/z 208 product ion. An abundant product ion at m/z 180 indicates a dihydrothiazine substructure consistent with a hydrolyzed lactam. The hydrolyzed lactam structure is further supported by the presence of a product ion at m/z 141, which results from the neutral loss of NH3. The resulting LC/MS structure library provides a predictive foundation for preclinical and clinical devel245 & LEE AND KERNS FIGURE 33. The mass chromatograms (EIC pro®les) of selected degradants at the m/z values corresponding to their [M ‡ H] ‡ ions in a base-degraded cefadroxil monohydrate sample (Rourick et al., 1996). opment stages that involve drug stability, synthesis, and process monitoring activities. Furthermore, this information aids the development of rugged stability-indicating methods for clinical drug release. Recently, this predictive degradation strategy has been demonstrated for biomolecules as a tool to accelerate drug development (Kleintop et al., 1998). Biomolecule samples are exposed to various accelerated stressed environments such as acid, base, heat, high intensity light, and humidity. Molecular weight information obtained from LC/MS indicated changes in the primary structure. Actual changes in protein sequence are determined with LC/ITMS methods, and data-dependent MS/MS spectra are generated. The resulting MS/MS spectra are interpreted with the SEQUEST database searching algorithm. Finally, a rapid LC/MS-based H/D exchange method is used to qualitatively detect changes in the tertiary structure. Similar to the examples described for small molecules, the resulting biomolecule degradant pro®le is used to predict potential problems before they appear during preclinical scale-up, formulation, and stability experiments. Furthermore, this information may serve as a diagnostic tool for structure identi®cation during the clinical development and manufacturing stages. 246 C. Clinical Development By the time the drug candidate reaches the clinical development stage, the scale of the chemical process has reached a full production batch size (kg). The information obtained on impurities and degradants during the preclinical development stage provides a useful historical database. Thus, when signi®cant process changes are made, the impurity pro®le is reviewed to determine whether the toxicological studies still support development. By far, the most intense use of LC/MS during the clinical development stage involves the quantitative bioanalysis of the drug and drug-related substances in support of pharmacokinetic activities. Sensitive assays are of paramount importance to success, particularly because lead compounds are more potent and require smaller doses than before. The LC/MSbased quantitative bioanalysis assay has emerged as the method of choice due to its high sensitivity and selectivity characteristics. Compatibility with conventional HPLC techniques has streamlined method development and analysis protocols during the transfer from the preclinical stage of development. Other analyses that involve LC/MS support activities deal with alternative dosing formulations (i.e., improved dissolution and LC/MS APPLICATIONS LTSS) and drug safety studies (i.e., toxic effects of metabolites, impurities, and degradants). Similar to the drug discovery stage, clinical development proceeds via three phases (I±III), where increasing levels of information are collected and increasing levels of criteria are required to proceed to the next phase. a. Analysis requirements. During the clinical development stage, the scale of process and sample volume is signi®cantly increased, with an intense focus on only a few drug candidates. For example, a multi-subject pharmacokinetics study can generate well over 1,000 samples in varying matrices such as plasma and urine. Also, attention shifts to the assurance of quality drug product for human use and the accumulation of information for NDA/MAA approval. Furthermore, clinical development studies often result in new impurity and degradant pro®les as the chemical scale-up process and formulation are improved. New degradants are generated from LTSS and require structure identi®cation. Also, human studies produce a new metabolite pro®le compared to those observed in animal models. In many ways, the analysis requirements for clinical development are similar to those required for drug discovery. Large numbers of samples must be analyzed, typically for a single analyte (e.g., pharmacokinetic studies), from clinical trials that involve 10,000 to 12,000 patients (Thompson, 1995) and a premium is placed on sensitivity (low ng/mL) and speed of analysis (<5 min). b. Analysis perspectives. A central issue of this phase is the completion of analyses under strict GLP and GMP guidelines. This requirement adds cost, but it assures reproducible results. Despite the diversity of sample matrices that are produced during clinical development, a large number can be analyzed with the same method. This standard approach produces greater ef®ciencies by providing: (1) reduced method development, (2) a reference database for a drug candidate, and (3) reduced repetitive analyses. Frequently, a compound that is initially identi®ed as a metabolite may also be observed as a degradant or impurity in pharmaceutics or process research samples. The early lock-in of methods leads to the ef®cient analysis of an enormous volume of samples throughout the lifetime of the project. In major studies, like pharmacokinetic or stability studies, large numbers of samples are produced; however, these studies typically feature a targeted analysis, where known compounds such as the parent drug or speci®c metabolite are quantitatively measured. Furthermore, the samples are usually presented within a consistent matrix (i.e., plasma, urine) that allows for ef®cient batch processing. The issues related to sample preparation are distinctly unique, and over the years, provide many important lessons to & consider for successful pharmaceutical analysis (Henion et al., 1998). c. LC/MS contributions. LC/MS analysis plays a major role in the success, ef®ciency, and timeliness of clinical development. The widespread acceptance of LC/ MS in the area of pharmacokinetics has led to major investments within pharmaceutical companies and contract analytical laboratories. LC/MS has demonstrated a clear advantage to HPLC for quantitative pharmacokinetic studies, in terms of method development time, cost, sample throughput, and turn-around. A recent review describes the powerful capabilities of LC/MS/MS approaches for quantitative bioanalysis studies (Brewer & Henion, 1998). Two LC/MS-based methods, SIM and SRM, are invaluable tools for quantitative analysis in clinical development and are featured in this section. The emphasis on highly automated chromatography-based sample preparations are highlighted with recent examples that involve off-line SPE and on-line extraction techniques. Structure identi®cation has also played a major role in accelerating clinical development. The rapid availability of structural information provides an improved understanding, in real-time, as processes and formulations are developed and metabolism is investigated. This information contributes to highly ef®cient clinical development activities that lead to successful NDA/MAA ®lings. LC/MS structural information is also used extensively as part of a multidisciplinary structure elucidation strategy for impurities, degradants, and metabolites; is information is also required for the NDA/ MAA. d. Overview. In this section, the application of LC/ MS and LC/MS/MS for quantitative bioanalysis during the clinical development stage is illustrated with examples that feature SIM (Fouda et al., 1991), SRM (Dear et al., 1998), automated off-line SPE extraction methods (Allanson et al., 1996; Simpson et al., 1998), and online extraction LC/MS methods (Needham et al., 1998). The SIM and SRM examples emphasize mass spectrometry-based analysis strategies, whereas extraction approaches emphasize the critical role of chromatographic separations, and to some extent, sample preparation. Also described are applications that deal with the identi®cation of metabolites during clinical studies (Lokiec et al., 1996) and degradants during LTSS (Volk et al., 1996). Table 19 provides a summary of the LC/MSbased applications featured in this section. The selection of a particular approach is discussed, with an emphasis on multiple stages of separation, chromatographic and mass, for providing unique capabilities and features for analysis. The applicability of these approaches to a wide variety of compounds of pharmaceutical interest is highlighted and 247 & LEE AND KERNS TABLE 19. Application of LC/MS in clinical development. Clinical development activity Pharmacokinetics Analysis Quantitative bioanalysis LC/MS application Selected ion monitoring Selected reaction monitoring Automated off-line solid-phase extraction Automated on-line extraction Metabolism Long-term stability Metabolite identi®cation Degradant identi®cation compared to traditional quantitative analysis methods that use HPLC and GC/MS. 1. Quantitative Bioanalysis±Selected Ion Monitoring The quantitative analysis of targeted components in physiological ¯uids is a major requirement in the clinical development stage. In 1991, Fouda et al. demonstrated the use of APCI±LC/MS for the quantitative determination of the renin inhibitor, CP-80,794, in human serum. Because the pharmacological action is below 200 pg/mL, a quantitative assay in the low pg/mL range is required to monitor the drug's pharmacokinetic and pharmacodynamic properties. The structure of CP-80,794 [Fig. 34] Template structure identi®cation Standard method LC/MS protocol Selected references Fouda et al., 1991; Wang-Iverson et al., 1992 Dear et al., 1998; Kaye et al., 1992 Allanson et al., 1996; Simpson et al., 1998 Needham et al., 1998; Ayerton et al., 1997 Lokiec et al., 1996 Volk et al., 1996 corresponds to a modi®ed peptide of molecular weight 620. The molecule lacks a signi®cant chromophore for UV detection with conventional HPLC methods. Also, the low volatility and thermal instability precluded analyses with GC/MS methods. Quantitative LC/MS assays generally involve four steps: sample preparation, assay calibration, sample analysis, and data management. In this method, the human serum samples are prepared with a liquid±liquid extraction procedure. The internal standard (1 ng) and carrier compound (50 ng), which is a structural analog of CP-80,794 [Fig. 34], are added to samples of human serum (1 mL aliquot). The addition of the carrier compound, a structural analog of CP-80,794, is respon- FIGURE 34. The structure of CP-80,794 and analogues used in the SIM LC/MS quantitative assay of CP80,794 in plasma (Fouda et al., 1991). 248 LC/MS APPLICATIONS sible for reducing any adsorption losses during sample preparation and LC/MS analysis (Self, 1979), and for enhancing the extraction ef®ciency and precision of the assay (Lee & Millard, 1975). Assay calibration involves the use of human serum samples forti®ed with CP-80,794 at 11 concentrations (six replicates per concentration), ranging from 0.05 to 10 ng/mL. In this particular case, due to a narrow linear dynamic range, two standard curves, ranging from 0.05 to 10 ng/mL, are constructed to provide the best accuracy. Serum blanks and an 11-point standard curve (two samples per concentration) are analyzed with each set of unknown samples. The LC/MS analysis involves the use of SIM to monitor the molecular ions [M-H]ÿ that correspond to the drug (m/z 619) and internal standard (m/z 633). In this LC/MS application, the negative ion mode is highly sensitive for this class of compound. Samples are loaded onto an HPLC autosampler and 80 mL aliquots are injected onto the column at 4-min intervals. The solvent front is observed at 1.0 min, and the elution times of the drug and internal standard are 3.1, and 3.4 min, respectively. The standard curves obtained from the CP-80,794 assay are shown in Fig. 35. Background subtraction routines are applied to obtain the best linear regression analyses and smallest y-intercept. The accuracy and precision of this assay are highlighted in Table 20, and representative HPLC chromatograms are shown in Fig. 36. At the time, this application provided a powerful benchmark for the use of LC/MS-based methods in the pharmaceutical industry. This particular assay successfully supported several clinical studies with sensitive and reliable results. This performance was benchmarked on more than 4,000 clinical samples, and led to a wider scope of application along with the development of routine, standard methods for quantitative bioanalysis. & TABLE 20. Analysis of serum samples that contain known amounts of CP-80,794; intra-assay parameters [n ˆ 6] (Fouda et al., 1991). Fort. conc.a (ng/mL) 0.05 0.10 0.20 0.30 0.40 0.50 0.75 1.00 2.50 5.00 10.00 Conc. found (ng/mL) C.V.b % accuracy 0.052 0.119 0.227 0.313 0.407 0.556 0.717 1.034 2.488 4.820 10.094 7.7 10.1 9.7 6.4 5.4 7.7 5.0 5.8 12.5 6.7 9.9 104 119 113 104 102 111 96 103 100 96 101 a Forti®ed concentration. Coef®cient of variation. b 2. Quantitative Bioanalysis±Selected Reaction Monitoring Since the work of Fouda et al. a tremendous growth in the LC/MS application for quantitative bioanalysis ensued in the clinical development stage of drug development. Similar LC/MS-based approaches that used SIM methods for quantitation (Wang-Iverson et al., 1992) to the highly selective SRM methods (Kaye et al., 1992) became popular. The strategic emphasis was on increased dimensions of analysis (chromatography and mass separations) to provide a higher selectivity and rapid cycle times. The use of SRM methods for quantitative bioanalysis represents increased dimensions of mass spectrometry analysis. SRM methods that use APCI±LC/MS/MS for the quantitative analysis of an antipsychotic agent, clozapine, in human plasma was recently described by FIGURE 35. Standard curves of the CP-80,794 SIM LC/MS assay for concentrations ranging from 0.05 to 10 ng/mL (Fouda et al., 1991). 249 & LEE AND KERNS FIGURE 36. Representative HPLC chromatograms of human serum samples for the CP-80,794 SIM LC/MS assay. Extracted ion current pro®les for: (A) Blank. (B) Blank plus internal standard. (C) 0.1 ng/mL CP-80,794. (D) 0.5 ng/mL CP-80,794 (Fouda et al., 1991). Dear et al. (Dear et al., 1998). Preclinical development studies of clozapine in rats and dogs used HPLC with ¯uorescence detection (FLD). With this method, a better limit of quantitation (LOQ) of 1 ng/mL was obtained. As the compound moved into the clinical stages of development, a more sensitive method of analysis was required to obtain rapid metabolic information in support of drug safety evaluation studies. A standard LC/MS/MS method is used for the quantitative analysis of clozapine (I) and four metabolites (II±V) in human plasma [Fig. 37]. FIGURE 37. Structures of clozapine (I) and four metabolites (II±V) assayed in human plasma with SRM LC/MS/MS (Dear et al., 1998). 250 The LC/MS/MS strategy is similar to previously described approaches for protein, natural products, metabolite, and impurity identi®cation. An ionization technique that generates abundant molecular ion species with very little fragmentation is desirable. The production spectrum is obtained to generate the substructural template of the molecule. Abundant and structurally unique transitions (molecular ion ! product ion) are identi®ed from the spectrum, and are used in the corresponding SRM experiment for quantitation. The SRM experiment provides a high degree of selectivity and better LOD than full-scan or SIM experiments for the analysis of complex mixtures (Johnson & Yost, 1985; Kusmierz et al., 1990). As highlighted in the previously described LC/MS/MS applications, the selectivity of MS/ MS reduces the requirements for complete chromatographic resolution of each component. Therefore, LC/MS/ MS experiments for quantitation typically emphasize short analytical run times to provide high sample throughput. The full-scan mass spectrum of clozapine contains an abundant [M ‡ H] ‡ ion at m/z 410 with little fragmentation [Fig. 38]. The 13C6-labeled clozapine internal standard exhibited similar performance in the full-scan mass spectrum; an abundant [M ‡ H] ‡ ion at m/z 416 is produced. Figure 38 also shows the product-ion spectra of the [M ‡ H] ‡ ion of clozapine and of the 13C6-labeled internal standard. This spectrum provides the structural template for identi®cation, and contains the abundant fragment at m/z 120 and 126, respectively, which corresponds to the benzamide substructure. Thus, the structurally unique transition of clozapine (410 ‡ ! 120 ‡ ) and the 13C6-labeled internal standard (416 ‡ ! 126 ‡ ) are selected for SRM quantitation LC/MS APPLICATIONS FIGURE 38. Positive APCI mass spectra: (A) Full-scan mass spectrum of clozapine. (B) product-ion spectrum of clozapine. (C) product-ion spectrum of 13C6-clozapine internal standard (Dear et al., 1998). experiments. Using this approach, the structurally unique transitions for the metabolites (II±V) are obtained. Figure 39 shows representative LC/MS/MS SRM chromatograms obtained from the analysis of an extract of a 50 pg/mL calibration standard. A mobile phase of 75% (v/v) acetonitrile in 20 nM ammonium formate at pH 4.0 provides good chromatographic peak shape, and the drug and internal standard elute <2min. The inter- and intra-assay precisions (% CV) of this method are less than 8% across the range of the limits of quanti®cation (0.05±10 ng/mL). The accuracy (% bias) for all spiked control concentrations does not exceed  4%. Same-day turnaround of results for over 100 samples is possible with this LC/MS/MS method. This method has been used to support an acute dose tolerance and pharmacokinetic study that involves the analysis of 1600 samples. 3. Quantitative Bioanalysis±Automated Solid-Phase Extraction As the trend toward obtaining more information earlier in the drug development cycle expanded, methodologies that involve sample preparation got exposed as the rate-limiting steps in work ¯ow during quantitative & FIGURE 39. Representative LC/MS/MS SRM chromatograms obtained from the analysis of an extract of a 50 pg/mL calibration standard. (A) Chromatogram for the m/z 410 precursor to m/z 120 product ion of clozapine. (B) Chromatogram of the m/z 416 precursor to m/z 126 product ion of 13C6-clozapine (internal standard) (Dear et al., 1998). bioanalysis activities. The LC/MS analysis was, indeed, the most ef®cient part of the overall process. Activities involved with sample tracking, sample preparation, data processing, and archiving presented the greatest challenge in a high-throughput environment. Opportunities with automated sample preparation using SPE in conjunction with LC/MS approaches attracted much attention due to their compatibility with robots and batch-mode processing (Kaye et al., 1996). The analysis emphasis focused on the increased dimensions of chromatographic separation. The work performed by Pleasance and coworkers highlighted the novel use of automated SPE in a ``96well'' format for high throughput bioanalysis in support of the migraine drug candidate 311C90 (Allanson et al., 1996). Initially, the clinical development of this drug candidate was supported with a 25-min HPLC analysis with ¯ourescence detection. Traditional packed-cartridge SPE was used for the extraction of human plasma in a sequential ``on-line'' mode (8-min extraction time). This method provides an LOQ of 2 ng/mL. However, as the clinical development of 311C90 progressed, the dose was 251 & LEE AND KERNS decreased, thus, a more sensitive method was needed. Also, cycle times of the automated on-line sample preparation and analysis were too slow for the LC/MS/ MS time-scale and the large number of samples from clinical studies. This situation required a method with increased sensitivity, shorter analysis time, and faster extraction cycle time. The analysis strategy uses LC/MS/MS in combination with a 96-well disk plate SPE. The activities associated with sample preparation and analysis are decoupled to simplify troubleshooting and to allow for greater ¯exibility. The bene®t of this off-line approach is the ability to simultaneously prepare 96 samples with SPE in 1 h, followed by the LC/MS/MS analysis on each sample. Previous approaches with individual SPE cartridges required 2±3 h. Sample preparation is automated for high-throughput batch processing with SPE, which is better suited for automation than other sample preparation methods (e.g., liquid±liquid extraction). The SPE cartridges are bundled together in a ``block'' with the same dimensions as a 96well plate (Kaye et al., 1996) to achieve increased throughput and the use with standardized robotics formats. Samples are collected into a deep well (1 mL) collection plate with a 300 mL elution volume. In this study, the LC/MS/MS analysis is <3 min, with injections occurring every 3.5 min. The previous HPLC method with ¯uorescence detection had a cycle time of 25 min. Thus, many opportunities exist with LC/MS/MS quantitative analyses to optimize chromatographic separations for speed. Similar to standard approaches described for open-access or metabolite identi®cation, standard or generic methods may be applicable with this application (Dear et al., 1998). The mass spectrometry performance of 311C90 and related compounds is illustrated in Fig. 40. The full-scan mass spectrum features an abundant [M ‡ H] ‡ ion at m/z 288, and essentially no fragmentation [Fig. 40A]. The corresponding product-ion spectrum of the [M ‡ H] ‡ ion of 311C90 indicates several useful product ions that correspond to unique substructures [Fig. 40B]. In this case, the product ion at m/z 58, corresponding to [(CH3)2N --- CH2]+, is selected with [M ‡ H] ‡ ion for the SRM experiment. For the desmethyl metabolite (183C91) and the deuterated internal standard (2H6 ± 311C91), the m/z 182 and 64 product ions, respectively, are selected with their corresponding [M ‡ H] ‡ ion for SRM. The LC/MS/MS SRM chromatograms obtained from the analysis of an extract of a 100 pg/mL calibration standard are shown in Fig. 41. Each analyte and internal standard elutes in less than 4 min. This strategy provides a 10-fold increase in sensitivity, which results in an LOQ of 0.1 ng/mL. All liquid handling (i.e., dilution, sample, internal standard addition) 252 FIGURE 40. Mass spectrometry of 311C90 and related compounds. (A) Full-scan positive ion electrospray spectrum of 311C90. Product-ion spectra of the: (B) [M ‡ H] ‡ ion of 311C90, (C) [M ‡ H] ‡ ion of 183C91 (desmethyl metabolite), and (D) [M ‡ H] ‡ ion of 2H6-311C90 (internal standard) (Allanson et al., 1996). is automated, and 96 samples are prepared in 1 h. This rate compares to 3 h via manual means or 5 h with a traditional on-line system approach. During the daytime, 2±3 blocks are extracted, and analyzed by LC/MS/MS during the evening. Chromatographic resolution of the drug, the metabolite, and the standard is not required, due to the structure-resolving capability of MS/MS. The instrument cycle time is reduced to 3.5 min from the former 25 min. In a period of 6 h, 96 samples plus standards and QC samples are analyzed with this method. Other assays have been performed with 2-min cycle times. Unattended analysis of six plates (576 injections) is common. Approximately 800 clinical samples have been analyzed with this method during 48 h of continuous operation. The work of Kaye et al. and Pleasance and coworkers provided the interest and motivation to extend sample LC/MS APPLICATIONS FIGURE 41. LC/MS/MS SRM chromatograms from the analysis of an extract of a 100 pg/mL calibration standard. (A) Chromatogram for the [M ‡ H] ‡ precursor ion to m/z 58 product ion of 311C90. (B) Chromatogram for the [M ‡ H] ‡ precursor ion to m/z 182 product ion of 183C91 (demethyl metabolite). (C) Chromatogram of the [M ‡ H] ‡ precursor ion to m/z 64 product ion of 2H6-311C90 (internal standard) (Allanson et al., 1996). preparation capabilities into an off-line batch mode process. This ``motivation'' was also stimulated by sample preparation bottlenecks, which would typically occur during on-line bioanalysis, where the limiting factor was associated with extraction and clean-up procedures. The rationale was to perform sample preparation tasks with an automated procedure followed by the transfer. A 96-well SPE system for the simultaneous extraction of drugs and metabolites in biological matrices developed by Wu et al. (Simpson et al., 1998) is shown in Fig. 42. In this approach, smaller elution volumes (75±200 mL) are used to improve SPE performance. This volume reduction allowed for the direct injection of samples without any evaporation and reconstitution. The collection plate that contains the elution fraction is loaded to an autosampler that is compatible with 96-well plates; therefore, elim- & FIGURE 42. Parallel extraction of 96 samples with an EmporeTM 96 individual disk SPE plate (A) and vacuum manifold system, including an acrylic vacuum manifold top (B), a polypropylene deep-well collection plate (C), two shims (D) for height adjustment of the collection plate, and a Delrin vacuum manifold (E) bottom (Simpson et al., 1998). inating the transfer to injection vials. This quantitative process improvement led to an improved analytical performance, considerable savings in time, and reduced cost. The performance of this LC/MS/MS quantitative analysis was demonstrated on a selective muscarinic M1 receptor partial agonist, SR 46559, which is under clinical development for the potential symptomatic treatment of the cognitive disorders that are associated with Alzheimer's disease. The full-scan mass spectrum and the corresponding product-ion spectrum are shown in Fig. 43. The m/z 268 product ion is identi®ed along with the [M ‡ H] ‡ ion at m/z 341 for the SRM experiment. Also, 253 & LEE AND KERNS intervention (adding the collection plate) and is signi®cantly faster than traditional extraction procedures (1 h per 96 samples versus 4±5 hrs.), the 96-well SPE format offers powerful advantages for high-throughput bioanalysis in combination with LC/MS/MS techniques. 4. Quantitative BioanalysisÐAutomated On-Line Extraction FIGURE 43. Positive electrospray mass spectra of SR 46559. (A) Fullscan mass spectrum. (B) Product-ion spectrum of the [M ‡ H] ‡ ion at m/z 341 (Simpson et al., 1998). the m/z 273 product ion and m/z 346 [M ‡ H] ‡ ion from the corresponding internal standard, 2H5 ±SR 46559, are identi®ed for SRM. Table 21 illustrates the cost analysis of a traditional disk cartridge SPE and the 96-well SPE system. Savings of $52 per 100 processed samples was calculated, which does not include the savings obtained from reduced solvent consumption and waste disposal. When combined with the fact that this system requires minimal user TABLE 21. Comparison in consumable costs between disk cartridge and 96-well plate for every 100 samples (Simpson et al., 1998). Disk cartridge $ 96-well plate $ Cartridge Receiving tube Injection vials Pipette tips 154 41 60 10 Plate Receiving plate 208 4.7 Total 265 Total 213 Based on invoices incurred from January to August 1997. 254 In response to off-line SPE assays in the 96-well plate format, LC/MS analyses that feature the direct injection of plasma with on-line extraction have recently been reported (Ayrton et al., 1997; Needham et al., 1998). This quantitative process approach eliminates the time-consuming sample extraction step, and provides an opportunity to perform an integrated on-line automated method for extraction and analysis. The use of turbulent ¯ow chromatography, which features high linear velocities and large particle size stationary phases, is described by Ayrton et al. (1997). The use of restricted access media (RAM) HPLC column, speci®cally designed to accommodate the direct injection of plasma and other biological ¯uids, is highlighted in the approach described by Needham et al. The utility of the RAM approach was demonstrated on a drug that was undergoing clinical evaluation as an anxiolytic agent, CP-93 393. The LC/MS analysis con®guration is shown in Fig. 44, and features a dual plasma extraction column set-up. This arrangement allows the extraction on one column to proceed while the other column is equilibrating and rinsing from the previous injection. The analysis procedure involves 2.5 min of extraction on column (2 mL/min), followed by back¯ushing with the analytical mobile phase to elute the analytes onto an analytical column (0.5 mL/min). With this on-line automated method, more than 250 samples have been analyzed unattended in a 24-h period. Good results are obtained over a 10±1000 ng/mL range. The LOQ and LOD are 190 pg and 58 pg, respectively. Accuracy and precision values are 9.0% or better over the entire range of the assay. Sample preparation issues inherent with on-line extraction techniques that deal with the addition of triethylamine to the mobile phase and the calculation of extraction recovery is likely to be addressed for future applications; however, this method appears to be widely applicable throughout drug development. Previous examples for the quantitative bioanalysis that is highlighted in this review feature single- and triple-quadrupole instruments. This example features the use of an ion trap mass spectrometer. The performance capabilities and potential implications of this mass spectrometer for quantitative bioanalysis have been recently described (Tiller et al., 1997b; Wieboldt et al., 1998). LC/MS APPLICATIONS FIGURE 44. Diagram of the instrument con®guration for integrated dual BioTrap column plasma extraction and LC/MS analysis. Extraction proceeds on one column while the other column is equilibrating and rinsing (Needham et al., 1998). 5. Metabolite Identi®cation During the course of clinical development, it is often important to identify the structures of metabolites. This information provides an opportunity to better understand interpatient variability in pharmacokinetics and toxicity. Clinical studies performed by Lokiec et al. on a semisynthetic derivative of 20(S)-camptothecin, CPT-11, demonstrate the use of LC/MS to investigate the in vivo metabolic pathways. CPT-11 is a potent inhibitor of topoisomerase II, which is an enzyme involved in DNA duplication, and exhibits signi®cant activity against various types of tumors in clinical studies. Here, the understanding and control of the main biotransformation pathways is particularly important for anticancer drugs because therapeutic doses are often close to the maximum tolerated dose. In the clinical studies involving CPT-11, bile and urine collected over a 48 h period was prepared with SPE procedures. The resulting extracts were pro®led with LC/ MS and LC/MS/MS. The analysis involves the use of the product-ion spectrum of CPT-11 [Fig. 45] as the structural & FIGURE 45. Product-ion spectrum of CPT-11, the structural template for the identi®cation of metabolic structures (Lokiec et al., 1996). template for the identi®cation of metabolite structures. The spectrum contains many diagnostic fragment ions that correspond to unique substructures of the molecule. Fragment ions at m/z 167 and 195 correspond to the bipiperidine and bipiperidine±carboxyl substructures, respectively. Abundant ions at m/z 375 and 331 correspond to the 7-ethyl-camptothecin core structure, and the ion at m/z 502 corresponds to the 7-ethyl-camptothecin core structure plus the proximal piperidine. As described in the previous sections that highlighted the MS/MS template approach for structure identi®cation, these ions serve as diagnostic markers for structural modi®cation. Thus, any ``shift'' observed in the product-ion spectra of CPT-11 metabolites serve to indicate a modi®ed CPT-11 substructure. The qualitative structural information derived by using this strategy is summarized in Table 22. Analysis of bile and urine extracts indicate that oxidation of the bipiperidine side chain is the major metabolic pathway. In addition, metabolism of CPT-11 occurs on three sites: (1) decarboxylation of the lactone ring; (2) oxidation of the camptothecin nucleus or of the 7ethyl group; and (3) multiple oxidations of the bipiperidine side chain. The information obtained from this study 255 & LEE AND KERNS TABLE 22. Identi®cation of CPT-11 metabolites in bile and urine (Lokiec et al., 1996). Peak intensity Compound Molecular mass 1 586 2 3 392 602 4 574 5 558 6 560 7 602 8 9 602 10 618 11 518 12 574 13 602 14 574 15 490 16 574 MH ‡ daughter spectrum m/z 543 (ÿCO2) m/z 502 (ÿ85 amu) m/z 84, 167, 195 ± m/z 559 (ÿCO2) m/z 502 [ÿ(85 ‡ 16) amu] m/z 183 (167 ‡ 16 amu) m/z 557 (ÿ18 amu) m/z 474 [ÿ(85 ‡ 16) amu] m/z 183 (167 ‡ 16 amu) m/z 541 (ÿ18 amu) m/z 474 (ÿ85 amu) m/z 84, 167, 195 m/z 543 (ÿ18 amu) m/z 476 (ÿ85 amu) m/z 84, 167, 195 m/z 559 (ÿCO2) m/z 518 (ÿ85 amu) m/z 84, 167, 195 m/z 559 (ÿCO2) m/z 502 [ÿ(85 ‡ 16) amu] m/z 183 (167 ‡ 16 amu) m/z 575 (ÿCO2) m/z 502 [ÿ(85 ‡ 32) amu] m/z 393 [(SN-38 ‡ H) ‡ ] m/z 349 (393-CO2) m/z 227 (195 ‡ 32 amu) m/z 393 [(SN-38 ‡ H) ‡ ] m/z 349 (393-CO2) m/z 99, 127 m/z 557 (ÿ18 amu) m/z [ÿ(85 ‡ 16) amu] m/z 183 (167 ‡ 16 amu) m/z 559 (ÿCO2) m/z 518 (ÿ85 amu) m/z 84, 167, 195 m/z 557 (ÿ18 amu) m/z 474 [ÿ(85 ‡ 16) amu] m/z 183 (167 ‡ 16 amu) No data obtained m/z 557 (ÿ18 amu) m/z 490 (ÿ85 amu) m/z 84, 167, 195 Comments Bile Urine CPT-11 High High SN-38 Oxidation of terminal piperidine High Low High Low Oxidation of terminal piperidine of metabolite 5 Low nda Decarboxylated: ÿ28 amu on camptothecin nucleus; modi®cation of the lactone ring High Low Decarboxylated: ÿ26 amu on camptothecin; modi®cation of the lactone ring High nd Oxidation of the camptothecin nucleus Low Low Oxidation of terminal piperidine High High Double oxidation of terminal piperidine High High Loss of terminal piperidine High High Oxidation of terminal piperidine of metabolite 5 High nd Oxidation of the camptothecin nucleus Low Low Oxidation of terminal piperidine of metabolite 5 Low nd Could correspond to metabolite 11 with the lactone ring of CPT-11 modi®ed as metabolite 5 Oxidation on camptothecin nucleus of metabolite 5 Low nd Low nd a ``nd, not detected. provides a clear indication that oxidative metabolism plays an important role in the elimination of CPT-11. 6. Degradant Identi®cation Volk et al. recently described the use of LC/MS approaches for the identi®cation of butorphanol degradants that are present in LTSS samples during the clinical 256 development stage (Volk et al., 1996). Butorphanol, the active ingredient in Stadol1 NS, is formulated as an intranasal analgesic product and is currently used in the treatment of the pain that is associated with post-surgical situations, dental intervention, and migraine (Lippman et al., 1977; Zeedick, 1979). Careful monitoring of degradant formation is an important aspect in determining the stability of the drug. Identi®cation of the degradants is LC/MS APPLICATIONS useful to determine potency and to provide insight into improved formulations. In this study, several low-level degradants are observed in the HPLC/UV chromatogram. Using a standard method LC/MS-pro®ling protocol, structural and substructural data for trace, butorphanol-related components were obtained rapidly and systematically in the formulated drug without prior fractionation. Similar to the analysis strategies described for metabolite and impurity identi®cation, chromatographic resolution of coeluting or unresolved components is not required to obtain product-ion data for structural analysis, due to the mass-resolving capability of mass spectrometry. As a result, this analysis strategy permits the development of an LC/MS-based butorphanol degradant database that includes relative retention time, molecular weight, and product-ion spectra. The degradation pathway and LC/MS pro®le data are summarized in Fig. 46. These results indicate the degradative processes that are observed for the LTSS samples of butorphanol tartrate: oxidative products proposed as 9-hydroxy and 9-keto-butorphanol, norbutorphanol, a ring-contraction degradant, and 1,10a butorphanol. In less than one day, detailed structural information regarding trace level components in the stability samples is obtained. When this information was correlated with known or predicted chemically labile portions of the molecule, structures were rapidly proposed. As a result, synthesis of the proposed degradants and subsequent con®rmation occurred much sooner than with the traditional structure identi®cation strategies. Degradant structures are identi®ed with the production spectrum of butorphanol as a structural template. The subsequent product-ion spectra obtained for each degradant represents a unique ®ngerprint of each compound, and can be used for structure identi®cation purposes as well as for con®rmation of the presence of a suspected component. A database of degradant structures is & constructed from the resulting data [Table 23] or referenced to an existing database assembled during the preclinical development stage. Thus, the comparison of degradant product-ion spectra with the butorphanol structural template allows for a rapid and systematic approach for identi®cation. D. Manufacturing Once the NDA is approved, attention shifts to issues that relate to the reliable production, packaging, storage, and distribution of the drug. Thus, quality assurance of the drug product for human use is of highest concern. Regulatory agencies provide intense oversight of the manufacturing and patient side-effects. When new impurities or degradants appear, a major effort is invested in identifying the source and in generating an appropriate remedy to the situation. The development and maintenance of good relationships with regulatory agencies is of great importance. a. Analysis requirements. Clinical supplies must pass rigid speci®cations that were established during earlier stages. When a product is no longer within speci®cations (i.e., failure), a tremendous amount of analytical work goes into characterizing the impurity. These situations may arise from many events, such as an unexpected process reaction, failure of a cleanup step, or contamination from a packaging material. Competitive analysis can involve the assessment of advantageous characteristics, such as stability. It can also involve the accumulation of information regarding patent infringement. b. Analysis perspectives. The majority of manufacturing analyses focus on the routine monitoring of highly characterized products. Quality assurance analysis of a drug can involve the monitoring of 5±20 impurities, and FIGURE 46. Degradation pathway for butorphanol tartrate in long-term stability samples (LTSS). The dashed area indicates the substructure involved in the proposed ring contraction (Volk et al., 1996). 257 & LEE AND KERNS TABLE 23. Butorphanol degradants from long-term stability storage of an aqueous formulation identi®ed with LC/MS and template structure analysis (Volk et al., 1996). Proposed structure RRT Molecular weight Norbutorphanol Hydroxy-butorphanol Ring-contracted butorphanol Butorphanol Keto-butorphanol
1, 10a-butorphanol 0.58 0.78 0.90 1.0 1.3 1.5 259 343 313 327 341 309 R1 H ±CH2(CH7) ±CH2(C4H7) ±CH2(C4H7) ±CH2(C4H7) ±CH2(C4H7) R2 R3 H OH H H ----- O H OH OH OH OH OH H TABLE 24. Applications of LC/MS in manufacturing. Manufacturing activity Production Quality Control Analysis LC/MS application Impurity identi®cation Protein characterization Data-dependent analysis Peptide mapping the monitoring of a biologically derived drug can involve amino acid sequence veri®cation and the monitoring of variant structural forms. When troubleshooting situations arise via the introduction of a new impurity in the drug product, intense efforts are invested in identi®cation and quantitation procedures so that the cause can be rapidly addressed and eliminated from the manufacturing process. Prompt action can maintain an excellent relationship with regulatory agencies and can lead to signi®cant savings. c. LC/MS contributions. HPLC is used throughout the pharmaceutical industry, and is often the technique by which manufacturing issues are ®rst identi®ed. The transfer of this analysis to the LC/MS is, therefore, useful for organizational collaboration. During this development stage, structural information is readily correlated with HPLC peaks observed in the QC laboratory. Thus, LC/MS is particularly suitable for troubleshooting studies. The inherent sensitivity, selectivity, and speed of LC/MS allows the rapid production of information. The structural information from LC/MS is often suf®cient for the rapid assessment of the source and extent of the problem. The sensitivity and selectivity of LC/MS allows for the facile application to trace impurities in complex samples. d. Overview. In this Section, the application of LC/ MS, featuring data-dependent analysis techniques for the identi®cation and quantitation of contaminants from drug product packaging, is illustrated (Tiller et al., 1997a). In 258 Selected references Tiller et al., 1997a; Wu et al., 1997 Chang et al., 1997 the past, leachables from packaging materials have appeared as impurities in drug product QC monitoring. Identi®cation and quantitation of these leachables with LC/MS can provide information for the selection of suitable materials. Also described is the application of LC/ MS for the monitoring of biologic drugs (Chang et al., 1997). The amino acid sequence can be con®rmed, and impurities due to other sequences or post-translational modi®cations may be monitored. Each example highlights the unique capabilities of LC/MS during this ®nal stage of drug development [Table 24]. It should be mentioned here that the use of LC/MS methods for the identi®cation of impurities and degradants in support of patents is also an important aspect of the manufacturing stage. Similar strategies as described for preclinical development support are used, and the resulting LC/MS structure pro®les that are obtained from samples in the ®eld are compared to in-house reference standards. Due to the proprietary nature of this application, few results and details are published. 1. Impurity Identi®cation Using Data-Dependent Analysis Regulatory authorities strictly scrutinize the leachables (e.g., plasticizers, impurities) that may come from medical devices and drugs. It is the responsibility of the drug or medical device company to identify the leachables and to provide adequate testing of their toxicity. Monitor- LC/MS APPLICATIONS ing methods must be developed and validated to effectively control toxic leachables during the manufacture of high quality pharmaceuticals. As a result, materials for medical devices and drug products must be tested for leachable components. Once a known toxic compound is discovered, it must be identi®ed for the assessment of toxicity, followed by the monitoring of levels using validated methods as required by the FDA. This identi®cation procedure could be a time-consuming process with traditional methods that are based on fractionation and individual component analysis. Tiller et al. have demonstrated an analytical strategy with on-line LC/UV/MS and LC/MS/MS to rapidly obtain structural information for leachables from a drug-delivery device (Tiller et al., 1997a). The analysis strategy makes use of ``data-dependent'' analysis, wherein the mass spectrometer ®rst obtains molecular ions using full-scan techniques, and makes real-time decisions about MS/MS product-ion spectra that must be obtained. In this way, molecular weight and substructural information are both obtained for many components during a single HPLC run. In this study, adhesive was applied to a glass bottle and cured. Highly puri®ed water was placed over the adhesive, heated at 50 C for three days, and analyzed with gradient reversed-phase HPLC. A LCQ ion trap mass spectrometer with ESI was used to pro®le the polyesters in the adhesive extracts with full-scan mass spectra and corresponding product-ion spectra triggered by an ion abundance that surpassed a threshold. Many components are readily observed in the ESI± LC/MS chromatogram, and several polyester leachables were identi®ed [Fig. 47]. The ESI±LC/MS chromatogram reveals 15 components, compared to the three components that were observed in the 220 nm UV chromatogram. This difference illustrates the capability of ESI±LC/MS to provide a more universal detection when the analytes do not contain strongly UV-absorbing substructures (e.g., aromatic). The method is highly ef®cient because molecular weight and substructural information via the full-scan and product-ion experiments, respectively, are both obtained for the sample components. Figure 48 shows the full-scan mass spectrum and the product-ion spectrum obtained from the peak at 25.16 min. A molecular weight of 472 is easily con®rmed due to the presence of the [M ‡ H] ‡ ion at m/z 473 and the ammonium adduct [M ‡ NH4] ‡ at m/z 490. With this information and knowledge of the starting materials, a reaction product of 2-isophthalic acid molecules with two diethylene glycol molecules was proposed [Fig. 48C]. The product-ion spectrum contains the ions at m/z 429 and 385, which correspond to two successive losses of either C2H4O or CO2 from the protonated molecule ion to yield the m/z 429 and 385 product ions, respectively. & Similar to previous structure identi®cation methods described for metabolites, impurities, and degradants, the knowledge of the physiological or chemical process, in this case the adhesive synthesis process, helped in the rapid interpretation of the MS/MS spectra of the unknown components. No user input about the sample composition is needed for the ``data-dependent'' analysis scheme; thus, these experiments are simple and rapid to perform. The result is a fairly routine approach to structural screening of unknown mixtures during the manufacturing stage. A recent report by Wu et al. highlighted the importance of similar LC/MS-based procedures to identify organic extractables from rubber stoppers in biotech products (Wu et al., 1997). Here, the capability of analyzing the extractables from mixtures of a peptide drug along with its formulated excipients allowed for the detection of butylated hydroxytolulene to the 0.1 ppm level. Furthermore, the presence of a coeluting polymer highlighted the advantages of on-line LC/MS-based approaches. 2. Peptide Mapping in Quality Control Quality control involves a carefully designed series of analysis and protocols. The purpose of this activity during the manufacturing stage of drug development is to ensure the production of safe, high quality drug products. These measures are helpful for the producers of the product as well as the regulators (i.e., FDA). In this way, adherence to protocols and procedure are carefully monitored on a routine basis. When any uncertainty in the manufacturing process occurs, procedures are referenced and data are analyzed to determine the speci®c stage of manufacturing to begin examining. Thus, the responsibility of drug manufacturers and regulating agencies are to determine when and how a process went awry. The ability to do so in an ef®cient, straight-forward manner is helpful to both parties, and ultimately, the consumer. Thus, the ability to provide this information is highly dependent on the manufacturer's ability to control the process. LC/MS-based approaches provide the same analytical bene®ts during the manufacturing stage of drug development as described for earlier stages. Information relating to the process, i.e., retention time, molecular weight, and structure, obtained during manufacturing activities help to accelerate drug development. LC/MS analysis of protein digests provides a powerful tool for mapping peptides and for assuring quality during the manufacturing process. Furthermore, the characterization of minor components contained in the digest may provide information on degradative processes (deamidation, oxidation, proteolysis), incorrect folding and disul®de rearrangement, and errors in translation by the host cell. 259 & LEE AND KERNS FIGURE 47. Chromatograms from the LC/UV/ion trap MS analysis of leachables that were extracted from a test adhesive with deionized water at 50 C for 3 days. (A) UV chromatogram at 220 nm. (B) TIC full-scan MS chromatogram (Tiller et al., 1997a). Somidobove is a recombinant bovine growth hormone (rbST) with a MW 22,818 that is composed of 199 amino acids. It is used to treat lactating dairy cows for increased milk production. Chang et al. recently described the use of LC/MS as a routine peptide-mapping tool to examine the amino acid sequence of the protein molecule during manufacturing (Chang et al., 1997). The peptides that are generated from the tryptic digest of somidobove provide a diagnostic con®rmation of the primary structure, and an important aspect of quality control. The use of LC/MS in a quality control environment with biologicals involves three steps. First, the expected cleavage sites (in this case, trypsin) within the amino acid sequence of the protein somidobove are indicated in Fig. 260 49. This tryptic map serves as a template for the expected peptide map. Second, an analytical method, using chromatography columns and conditions that provide the best resolution and reproducibility, is developed. An opportunity exists to optimize the analysis based on chromatography, digestion, and LC/MS performance. Finally, the resulting LC/MS data are pro®led, according to amino acid sequence, peak number, and [M ‡ H] ‡ , as shown in Fig. 50. Other properties such as relative retention time can also be added in this format. This LC/MS approach is analogous to the strategies that are employed during earlier phases of drug development. Rigid protocols result in a tight control of the process and an adherence to desired speci®cations. LC/MS APPLICATIONS & FIGURE 48. Mass spectra of the leachate component eluting at 25.16 min. (A) Full scan mass spectrum of component. (B) Product-ion spectrum of m/z 473. The proposed structure is the reaction product between two isophthalic acid molecules with two diethylene glycol molecules to form the compound shown in (C) (Tiller et al., 1997a). Similarly, the LC/MS screening approaches provide a measure of control as well, to ensure the ef®cient harvesting of discovery and the detection of serendipity. VII. FUTURE APPLICATIONS AND PROSPECTS With a perspective on recent LC/MS applications that illustrated the contributions to the acceleration of drug development, new insights may be derived to speculate on future prospects. Certainly, the meaning of pharmaceutical analysis has evolved, and has resulted in the continual development of LC/MS-based technologies and applications. In fact, these improvements have led to novel applications that constitute a major factor for industry acceptance. Pharmaceutical industry applications seem to follow an iterative cycle that emphasizes high throughput, sensitivity, and structural detail. Current 261 & LEE AND KERNS FIGURE 49. Amino acid sequence of somidobove, which serves as a template for the expected peptide map (Chang et al., 1997). FIGURE 50. Somidobove LC/MS tryptic peptide mapping (Chang et al., 1997). 262 LC/MS APPLICATIONS trends highlight the growing importance of structural information and a waning perception of LC/MS as a specialized or dif®cult analysis. A perspective on future applications of LC/MS in drug development is presented with an emphasis on integrated sample generation and analysis activities. A. Workstations Workstations are perhaps a more ¯exible complement to a robotics method. These systems are capable of inseries or parallel analysis. Methods are developed with very speci®c and specialized functions that allow higher throughput and operation in a batch mode. These dedicated approaches would seem to be a popular choice in the drug discovery and preclinical development stages. Workstations could be con®gured according to specialized needs and easily interchangeable with changes in workload or priorities. Since the testing in these areas of drug development are less rigorous (i.e., method validation, QC) than in clinical development and manufacturing, the overall analysis speed is enhanced. A batch-mode process allows for relatively simple correction without catastrophic errors or delays. In any case, each approach seems to focus on opportunities to optimize the human interface during analysis. The robotics approach emphasizes integrated analyses and provides a mechanism to eliminate tedious processes manually performed by a scientist. A workstation approach highlights rapid, diverse analyses where a corrective action can be taken immediately if the LC/MS method is not performing as expected. This approach would favor situations where constantly changing priorities create demands on sample analysis and instrument ¯exibility. & Automated approaches to LC/MS will play a major role in the future development of dedicated workstations. Robotic systems allow for the integration of many processes (i.e., sample preparation, separation, analysis) in virtually any format. Once set up and properly con®gured, these systems can perform the analysis with little or no operator intervention. Analyses are either performed in series as samples are individually and sequentially passed though the con®guration or off-line in a batch mode. These approaches would appear to be wellsuited for clinical development and manufacturing activitiesÐapplications where the analysis is fairly uniform and repetitive. The most critical oversight is provided during the method development and method validation stages. It can be expected that similar approaches will be adopted in the early stages of drug development as sample-generating technologies continue to mature and standard formats become established. The prospects for the application of open-access LC/ MS support strategies beyond combinatorial and medicinal chemistry seem good. Areas such as process chemistry optimization could bene®t tremendously and provide accelerated development for chemical scale-up campaigns (Storer, 1996); therefore, it is likely that similar LC/MS methods will be required (Cepa & Searle, 1998). Extensions of these approaches can be envisioned to include selection of salt form as well as toxicology, solubility, and lipophilicity applications. Once accepted, routine and highly integrated quantitation approaches [Fig. 51] as described by Cole et al. (Cole et al., 1998) are likely to become more widespread. Also, instrument formats that feature high resolution chromatography and/ or mass spectrometry may provide the required structural detail in some instances. Similar to small molecule FIGURE 51. An automated quantitation approach that performs all of the necessary tasks required in a quantitative analysis. The user supplies only the molecular ions of the analytes and internal standards (Cole et al., 1998). 263 & LEE AND KERNS analysis, the widespread use of open-access LC/MS systems for protein synthesis activities (Burdick & Stults, 1997) appear to be imminent. Screening applications appear to be logical and necessary. LC/MS-based approaches are likely to continue to be the ®rst choice analysis to overcome important bottlenecks for several reasons. First, the quantities of compounds (drug and target) required for drug development will continue to decrease. For example, the amount of compound contained in a single combinatorial bead is usually ca. several hundred picomoles. Second, miniaturization and scale-down methods will become popular strategies to screen compounds, gather data, and improve product and processes in a cost-effective manner. Smallscale experiments reduce the amount of compound/ reagents as well as provide opportunities to explore new avenues to development. Finally, new screening approaches will continue to be desirable in areas that feature trace mixture analysis. The use of LC/MS to directly gather data from mixtures will reduce sample preparation steps for fractionation and puri®cation, and will help alleviate the stress brought about by the increased number of targets and compounds that require analysis. B. Multidimensional Analysis Certainly, screening-based strategies can be extended to include multidimensional separations or multiple stages of mass analysis (i.e., LC/LC/MS/MS, LC/MSn). The integration of expanded capabilities to increase performance will continue to impact chromatography and mass spectrometry formats. A recent example that highlights the bene®ts of multidimensional chromatography approaches (LC/LC) may provide uniquely integrated capabilities for target-based screening in drug discovery (Hsieh et al., 1997), on-line sample extraction (Needham et al., 1998), and integrated approaches for sample cleanup (Sharma et al., 1997). New applications that feature mass spectrometry also have an exciting potential. The recent studies performed by Hopfgartner et al. illustrate the use of a tandem quadrupole TOF instrument to obtain exact mass measurements on drug metabolites (Hopfgartner et al., 1998). The highly automated MS/MS-based approaches for real-time structure identi®cation described by Yu et al. (Yu et al., 1999) are a signi®cant step toward an integrated sample analysis with virtually no operator involvement. Furthermore, the possibilities for routine and/or data-dependent MSn protocols for structure identi®cation are an exciting development for the detailed analysis of unknown compounds such as natural products (Gilbert & Lewer, 1998) and metabolites (Lopez et al., 1998). It should be mentioned that, although multidimensional LC/MS approaches will continue to expand with 264 powerful and unique capabilities for pharmaceutical analysis, emerging areas of drug development are likely to bene®t from benchmarked applications. For example, the ``in-process'' monitoring and testing of raw materials in bacterial or mammalian cell culture would appear to be logical extension of current applications for chemical process scale-up. In this case, LC/MS can be used to determine the optimum timepoints to harvest the product. Further QC applications of LC/MS are envisioned that deal with the characterization of product lots for consistency and speci®cations. An important aspect is the assessment of purity in the presence of inherent microheterogeneity, which is characteristic of biologics that are derived from bacterial or mammalian cell culture. Performance aspects of chromatographic separations, particularly for the mapping of post-translational modi®cations, will be important considerations. C. Miniaturization A 96-well plate format has been the standard for drug screening over the past ten years (Bubiak, 1997), and miniaturized formats using a 384-well plate and still smaller ones have begun to appear in the pharmaceutical industry (Kolb & Neumann, 1997). It is likely that these formats, or a derivative, in combination with the LC/MS assay will provide the necessary throughput for liquid handling of samples. Approaches to miniaturize or scale-down the pharmaceutical analysis will be popular. For example, pooling strategies are an effective method to alleviate the problems associated with high-volume liquid handling and analysis. This approach has been used to test closely related compounds such as a series of analogs in drug discovery (Hop et al., 1998). Recent compound-pooling strategies developed for testing combinatorial libraries (Devlin et al., 1996; Wilson-Lingardo et al., 1996) would appear to be well-suited for many LC/MS-screening approaches. A matrix-pooling strategy involves a 10  10 matrix of separate compounds pooled in the x- and ydirections. The result is 20-sets that contain ten compounds each. The library size is reduced by a factor of ®ve, and each compound can be tested in duplicate because each compound resides in a x- and y-axis pool. Scaledown approaches that feature predictive models for metabolic (Rourick et al., 1998) and chemical (Fink et al., 1997) stability are likely to be highly utilized in early development stages. These strategies feature integrated methodologies and parallel processing of complex mixtures that result in multidimensional structural data [Fig. 52]. Scale-down approaches are also likely to be extensively used to provide comparative pro®les for metabolism (Gillespie et al., 1998), process impurities (Rourick et al., 1998), degradants (Rourick et al., 1996), LC/MS APPLICATIONS & FIGURE 52. Multidimensional structural data produced by predictive models, featuring integrated methods and parallel processing of drug candidates for metabolic (Rourick et al., 1998) and chemical (Fink et al., 1997) stability. cellular uptake (Kerns et al., 1998), and interaction (Lee et al., 1995b). LC/MS interfaces that accommodate miniaturized formats for biomolecule analysis such as nanoelectrospray (Wilm & Mann, 1996) or microelectrospray (Figeys et al., 1996) with a variety of mass-detection devices, ranging from triple quadrupole (Swiderek et al., 1998), TOF (Medzihradszky et al., 1998), and quadrupole TOF (Morris et al., 1996; Hanisch et al., 1998) to ion traps (Figeys & Aebersold, 1997; Arnott et al., 1998), appear to be headed for tremendous growth. Future developments with instrumentation and improvements in performance will drive this growth, which will permit the facile conversion to automated approaches (Ducret et al., 1998) and routine procedures for isotopic labeling of peptides for sequence analysis (Shevchenko et al., 1997). Furthermore, it may be reasonable to presume that these advances will result in the more frequent investigations of intact membrane proteins (Whitelegge et al., 1998), tRNA (Taniguchi & Hayashi, 1998), and preparative approaches (Immler et al., 1998). D. Information Management A key area that will continue to emerge with LC/MS applications is data management. With the variety of systems available from benchtops to tandem mass spectrometers, the ability to generate vast amounts of data quickly and easily has created a dilemma for accelerated development activities. Researchers will quickly realize the consequence of data generation and data management. There are three factors that will help to shape this important area of focus. First, consistency of data will be of signi®cant importance. A consistent data format could involve the separate treatment of raw, processed, and interpreted data. Standard, de®ned database ®elds for data entry are essential. Database searching techniques for protein identi®cation (Yates et al., 1998b; Yates, 1998c) are likely to expand in scope throughout the drug development cycle. Investigations into similar procedures for the identi®cation of small molecules (i.e., metabolites, impurities, degradants) will be important such as automatic deconvolution (Higgs, 1998) and correlation analysis (Fernandez-Metzler et al., 1998). Second, a relational database to access, compare, and rank data is desirable. Simultaneous comparison of structure, purity, chemical and metabolic stability, lipophilicity, and bioavailability can be made to assist the drug candidate selection process (Rourick et al., 1998). Finally, the continued simpli®cation of data will be a critical step for the continued use and acceptance of LC/MS-based techniques in the pharmaceutical industry. Chromatograms and mass spectra will still have tremendous value, and will likely evolve into highly visual and intuitive representations such as contour maps (Li et al., 1998). However, the interpreted result is a more powerful currency. The visualization of metabolic stability by using structures, tables, and graphs is already a standard practice. A picture or number (or groups of each) is still the most the easily understood data for decision-makers regardless of how it was obtained or derived. Visualization approaches are likely to progress from the actual operating system of the instrument to the drug development process. E. Strategic Outsourcing Historically, the pharmaceutical industry relied almost entirely on internal resources to support drug development. The surge in new technologies has made it dif®cult to keep current in all areas of responsibility and to maintain adequately trained staff. Due to the availability of high quality research services outside large pharmaceutical companies, many organizations have taken advantage of outsourced analytical services. Similar calculations for FTE described earlier in the accelerated 265 & LEE AND KERNS development section indicate that outsourcing is a highly ef®cient approach for LC/MS analysis. The cost that is involved with instrument selection, purchase, and installation combined with investments in personnel, training, and maintenance required to generate reliable data have made outsourcing a justi®able, and perhaps, desirable, even necessary, solution. Outsourcing strategies offer a highly ¯exible and adaptable strategy for clinical development support. This approach to drug development generally involves the use of contract research organizations (CRO's) and contract analytical laboratories, which perform a variety of specialized clinical development-based functions. This practice has recently spread throughout the drug development cycle, and it is expected that 20% of pharmaceutical R & D budgets will be spent externally by the year 2000 (Taaffe, 1996). F. Summary Analysis will continue to respond to advances in pharmaceutical technologies. Similar to the pharmaceutical industry's response to molecular biology technology in the 1970's, where quantity and puri®cation were no longer the bottlenecks for the discovery of new drugs, continued advances in combinatorial chemistry, genomics, and recently, proteomics are likely to lead to signi®cant changes. These changes will necessitate the development of analytical technologies that address the criteria of a dynamic drug development environment. The repetitive yet diverse nature of drug development suggests that high throughput, automated analysis tools will play a major role. Growth in product pipelines and limitations in analytical capabilities are likely to continue to drive the ®eld toward miniaturization. Meanwhile, the traditional methods for analysis continue to lose meaning in many aspects of drug development. For example, the classical approaches of NMR or IR are no longer feasible for the analysis of combinatorially derived compounds. Certainly, NMR or IR spectra of a multicomponent library mixture can be obtained, but they are not routinely diagnostic of structure. The loss of these powerful tools creates both burdens and opportunities. Regardless of the applications, the successful application of LC/MS techniques to drug development will remain closely linked to their integration with distinct drug development activities. The ultimate success is likely to depend on continued collaboration and relationship. VIII. PERSPECTIVES ON THE FUTURE GROWTH OF LC/MS There appear to be some common motivating factors by which many organizations have responded to LC/MS 266 accepted it. These factors provide a perspective on why LC/MS has become so prominent throughout the drug development cycle and provide unique insights into the future prospects for growth. Accelerated drug development schemes have shifted the need for analytical instrumentation to include criteria for high throughput (quantitative process approaches) and the capability to contribute to an application that produces information for accelerated decision-making (qualitative process approaches). The value of sensitivity, selectivity, and detail are still signi®cant; however, new parameters dealing with ef®ciency, productivity, and information content have become the new watermarks for analytical instrumentation in the pharmaceutical industry. These factors have been affected by shorter timelines and the pharmaceutically relevant information required for decision-making. Ongoing technological advances have caused relationship-building to be a key aspect of growth. Scientists are compelled to seek more depth of knowledge in their primary speciality, yet they are constantly put in an environment that demands broader knowledge, ¯exibility, relationship skills, and the ability to effectively integrate new technologies. And, as responsibilities within the pharmaceutical industry have become broader, interpersonal skills such as the ability to understand needs, create relationships, and develop alliances have become more important. Today's analytical end-users are made up of: (1) researchers who specialize in analysis, and (2) researchers from other pharmaceutical specialities. As in the past, researchers who specialize in analysis (i.e., analytical chemists) with formal training in the fundamentals of instrumentation, method development, and the treatment of data are continually challenged with relearning the fundamentals of chemistry (Laitinen, 1980) and other pharmaceutical sciences, such as molecular biology. They must keep current with advances in the ®eld as well as the pharmaceutical business in general. Recent trends have made scientists more dependent on analytical information, and consequently, they are more apt to want to control or perform analysis activities themselves. No doubt, the human factor will inevitably arise and create reluctance, and perhaps, barriers to adopt new approaches. For example, the LC/MS applications that deal with open-access support for combinatorial and medicinal chemistry, made the various aspects of acceptance quite clear. A chemist preferred simple answers to simple questions. Certainly, scientists within the pharmaceutical industry will continue to be immensely challenged by the introduction of new technologies such as LC/MS and their subsequent application to achieve accelerated development goals. In an effort to make these new technologies LC/MS APPLICATIONS & TABLE 25. The elements of the information-gathering process during drug development, based on relationships between the analytical data and decision. Drug development need Analysis method Analysis strategy Analysis solution Analysis activity Method Method/process re®nement Method/process validation ``user friendly'' to researchers in non-analytical specialities, speci®c tasks are being incorporated into the operating system of the analytical instrument. Still, productivity and ef®ciency are of paramount importance to success. Future growth will continue to depend on strong relationships with instrument manufacturers and academia. The pharmaceutical industry will continue to be dependent on innovative products and highly trained scientists to provide the data to support breakthroughs in medicine. Technologies such as combinatorial chemistry and molecular biology are capable of providing unique information at unprecedented levels. These technologies have been adapted and integrated within the pharmaceutical environment to provide quantitative and qualitative process solutions for drug development. For the academicians and instrument manufacturers, there has been no such breakthrough. Speci®cally, there has not been an equivalent process-related technology that would signi®cantly accelerate the rate of invention and graduation or the rate of instrument development and manufacturing. Yet the burden of keeping pace with technology and the diverse demands of the end-user increases. In general, progress in drug development is dependent on data obtained from analytical instruments. Thus, how this information is derived and provided to decisionmakers is critical to the success of the pharmaceutical industry. LC/MS technologies have effectively and uniquely supported the productivity (quantitative processes) and ef®ciency (qualitative processes) needs of drug development. Future issues appear to be directed toward enhancement of the information-gathering process that focuses on the relationship between data and decision [Table 25]. There is no single approach to the selection and implementation of analytical technologies. There will always exist diversity in how different organizations proceed to accomplish their goals. This approach relates to the organization's history and culture as well as to the people's areas of expertise and philosophies. IX. CONCLUSIONS An impressive variety of LC/MS-based solutions, incorporating quantitative and qualitative process Data Currency for drug development decision Raw Information Processed Information/knowledge Interpreted Integrated information/knowledge approaches, are now routinely applied to accelerate drug development. The results are signi®cant and led to the successful development of innovative therapies and numerous novel drugs. The use of LC/MS with other technologies for sample preparation, analysis, and data management is now an inextricably linked element of drug development. The LC/MS applications highlighted in this review indicate that this unique technology is now fundamentally established as a valuable tool for analysis in the pharmaceutical industry. The illustration of a widened scope of application in the pharmaceutical industry, the ®fth stage of an analytical method, suggests that LC/MS is establishing itself as a widely accepted and routine tool (sixth stage) for pharmaceutical analysis in all stages of drug development. It appears likely that the balance between research and development needs along with partnership between academia, instrument manufacturers and industry researchers will allow for the continued acceptance and advancement of LC/MS technologies in the pharmaceutical industry. However, the pharmaceutical industry is hardly analyst-driven. Yet, it is analyst-dependent. And, whereas new technologies continue to advance, the need for information-rich tools such as LC/MS will increase. A continual change in the top-line business combined with tremendous focus on the bottom-line will indeed challenge many. And, as the pressures also increase with increasing responsibilities of generating the data, managing the laboratory, and utilizing the information, decision-makers will need to understand technology and its applications much faster. Of course, relating this understanding to business goals represents an equally signi®cant challenge. Though many of the developments illustrated in this review are truly unprecedented, we should realize that history does have a tendency of repeating itself. And like the seven stages of an analytical method, there are signi®cant social events which impact drug development. The insightful and powerful statement made by Izaak Kolthoff in 1973 (Kolthoff, 1973) is an eloquent reminder, ``We analytical chemists are no longer the maidservants of other chemists, but together we contribute to further progress in the whole ®eld of chemistry.'' Hopefully, as the bases of understanding application, strategy and 267 & LEE AND KERNS analysis become more complicated and more integrated with drug development, the depth of understanding new technology, process and collaboration will continue to improve. The recent applications of LC/MS for accelerated drug development suggest an exciting future in the pharmaceutical industry. It will be ®lled with many needs, anxieties and triumphs. Perhaps the next ten years will be equally exciting as the last. X. GLOSSARY OF TERMS The glossary provides a list of the terminologies and de®nitions that are encountered during analysis activities in drug development. The objective is to assist the understanding of principles, concepts, and analysis strategies. The glossary has been compiled in part from the well-accepted de®nitions obtained from the literature (Pure Appl Chem, 1992; Pure Appl Chem, 1993). Many of the meanings are derived from an analyst's perspective; speci®cally, activities that involve LC/MS-based methods. ADMEÐThe abbreviation for absorption, distribution, metabolism, and excretion. (see Pharmacokinetics.) Af®nityÐThe tendency of one molecule to associate with another. AnalogÐA drug whose structure is related to that of another drug, but whose chemical and biological properties may be quite different. Analytical ChemistryÐThe science of chemical characterization and measurement. ApplicationÐAn analytical procedure developed for the production of speci®c types of data from certain samples. BioanalysisÐThe measurement of a drug and related structures that are contained in complex matrices, typically at the part per million to part per trillion level. BioassayÐA procedure for determining the concentration, purity, and/or biological activity of a drug by measuring its effect on an organism, tissue, cell, enzyme, or receptor as compared to a standard preparation. BiotransformationÐThe chemical conversion of substances by living organisms or enzyme preparations. Chemical Manufacture and Control (CMC)ÐThe section in the IND or NDA that contains information on the chemical characteristics of the lead candidate. 268 ChemometricsÐThe application of statistics to the analysis of chemical data and design of chemical experiments and simulations. Clinical DevelopmentÐThe stage of drug development that is responsible for registering the drug candidate and ®ling the NDA. Clinical Trial Application (CTA)ÐThe equivalent of the IND document ®led in Europe. Collisionally Induced Dissociation (CID)ÐThe process where a selected precursor ion undergoes collisions with neutral gas molecules in a collision region to yield product ions. Combinatorial ChemistryÐThe process of preparing large sets of organic compounds by combining sets of building blocks. Combinatorial LibraryÐA set of compounds prepared by combinatorial chemistry. De Novo DesignÐThe design of bioactive compounds by the incremental construction of a ligand model within the receptor or enzyme active site. DegradantÐA drug analog formed from the drug as a result of the action of chemical and/or physical conditions. DereplicationÐA process of rapidly identifying known compounds contained in a natural products extract. Double Blind StudyÐA clinical study of potential and marketed drugs where neither the investigators nor the subjects know which subjects will be treated with the active principle and which ones receive a placebo. DrugÐAny substance presented for treating, curing, or preventing disease. Drug DiscoveryÐThe stage of drug development responsible for generating and optimizing the lead compounds. Drug DispositionÐRefers to all the processes involved in absorption, distribution, metabolism, and excretion of drugs in a living organism. Drug ProductÐThe manufactured formulated drug appropriate for clinical studies. Drug SubstanceÐThe manufactured drug compound appropriate for formulation into drug product. Ef®cacyÐThe success of the drug in alleviating a targeted disease in humans. Electrospray Ionization (ESI)ÐThe liberation of ions from electrically charged droplets in an atmospheric pressure ionization source region. Extracted Ion Current Pro®le (EICP)ÐThe mass spectrometry signal that corresponds to an ion LC/MS APPLICATIONS plotted against time. Also referred to as a mass chromatogram. Food and Drug Administration (FDA)ÐThe government agency responsible for instituting policies and monitoring science to assure safe, high quality pharmaceutical products. Full-Time EquivalentÐOne full-time employee with skills and productivity appropriate to the designated responsibilities. Gas Chromatography/Mass Spectrometry (GC/ MS) ÐAn analytical instrument that integrates a gas chromatograph with a mass spectrometer. GenomeÐThe complete set of chromosomal and extrachromosomal genes of an organism, a cell, an organelle, or a virus. The complete DNA component of an organism. GenomicsÐThe study of the relationship of speci®c genes to the expression of certain proteins or biological functions (e.g., disease). Good Laboratory Practice (GLP)ÐThe regulations and requirements for the operation of laboratories that produce data that are used in regulatory ®lings. Good Manufacturing Practice (GMP)ÐThe regulations and requirements for validation, personnel, and record-keeping of facilities that are associated with making drug products that are administered to humans. High Performance Liquid Chromatography (HPLC) ÐA technology in which compounds are partitioned between a stationary media and a ¯owing liquid phase in a tubular format under high pressure, resulting in the separation of a mixture of components and the reproducible time of elution of each speci®c compound. HomologÐA compound that belongs to a series of compounds that differ from each other by a repeating unit, such as a methylene group, a peptide residue, etc. ImpurityÐAn extraneous compound present in a drug substance or drug product. Investigational New Drug (IND)ÐThe registration document ®led to begin testing in humans. Lead OptimizationÐThe synthetic modi®cation of a biologically active compound, to improve stereoelectronic, physicochemical, pharmacokinetic, and toxicological properties. Limit of Detection (LOD)ÐThe amount of analyte that yields a signal-to-noise ratio adequate to provide the desired con®dence for detection of an analyte. A signal-to-noise (S/N) ratio of 3 is typical. & Limit of Quanti®cation (LOQ)ÐThe amount of analyte that yields a signal-to-noise ratio adequate to provide the desired con®dence for quantitation of an analyte. A signal-to-noise (S/N) ratio of 10 is typical. LipophilicityÐThe af®nity of a molecule for a lipophilic environment. ManufacturingÐThe stage of drug development responsible for the manufacture of marketed drug products. Marketing Authorization Application (MAA)ÐThe equivalent to the NDA ®led in Europe. Mass ChromatogramÐsee Extracted Ion Current Pro®le. Matrix-Assisted Laser Desorption/Ionization (MALDI)ÐThe production of ions from an analyte in the solid state by irradiation with a laser, and the facilitation of a co-precipitated compound that readily absorbs the laser light. Medicinal ChemistryÐThe design and synthesis of drug candidates. MetabolismÐThe physical and chemical processes involved in the maintenance and reproduction of life in which nutrients are broken down to generate energy and to give simpler molecules (catabolism), which by themselves may be used to form more complex molecules (anabolism). MetaboliteÐAny intermediate or product that results from metabolism. Molecular BiologyÐThe study of genetic materials and their expression at the molecular level. New Chemical Entity (NCE)ÐA compound not previously described in the literature. New Drug Application (NDA)ÐThe registration document ®led to rigorously document an NCE's properties, manufacturing, safety, and detailed evaluation in human clinical studies. Neutral Loss MS/MS ScanÐThe operation of a tandem mass spectrometer in which the two mass analyzers are scanned at the same rate, with the second mass analyzer offset (lower) from the ®rst by a constant m/z ratio. Nuclear Magnetic Resonance (NMR)ÐA technique for detailed structural analysis that involves observation of the absorption of discrete radio frequencies by speci®c atoms in molecules exposed to a magnetic ®eld. Parallel AnalysisÐSimultaneous processing of more than one sample or compound during one or more steps of an analytical protocol. 269 & LEE AND KERNS Pattern RecognitionÐThe identi®cation of patterns in large data sets, using appropriate mathematical methodologies. Perfusive ChromatographyÐHPLC with stationary media that have pores through the particles of the media, resulting in reduced back-pressure. PharmacokineticsÐRefers to the study of absorption, distribution, metabolism, and excretion (ADME) of bioactive compounds in higher organisms. PharmacophoreÐThe ensemble of steric and electronic features that are necessary to ensure the optimal supramolecular interactions with a speci®c biological target structure and to trigger (or to block) its biological response. PlaceboÐAn inert dosage form that is identical in appearance, ¯avor, and odor to the drug product dosage form, but contains no drug. MotencyÐThe amount of drug substance in a speci®c lot of drug product. Preclinical DevelopmentÐThe stage of drug development responsible for evaluating lead compounds and ®ling the IND. Precursor-Ion ScanÐThe operation of a tandem mass spectrometer in which the ®rst mass analyzer is scanned while the second is held at a speci®c m/z ratio. Process ChemistryÐDevelopment of reagents and conditions for the synthesis of bulk drug substance. ProdrugÐAny compound that undergoes biotransformation before exhibiting its pharmacological effects. Product-Ion ScanÐThe operation of a tandem mass spectrometer in which the ®rst mass analyzer is held at a speci®c m/z ratio, while the second is scanned. Also, the operation of an ion trap mass spectrometer in which ions of a speci®c m/z ratio are isolated and collisionally activated, followed by scanning the product ions. ProteomicsÐDetermination of the proteins expressed under speci®c biological states (e.g., disease) so as to determine their role. ReceptorÐA molecule or a polymeric structure in or on a cell that speci®cally recognizes and binds a compound that acts as a molecular messenger. Selected Ion Monitoring (SIM)ÐA mass spectrometry-based screening approach for quantitative analysis where a single m/z value, which corresponds to the drug molecule, is detected. Selected Reaction Monitoring (SRM)ÐA mass spectrometry-based screening approach for quan- 270 titative analysis where two dimensions of mass analysis are employed to detect a unique product ion corresponding to the drug molecule. SensitivityÐRefers to a detectable signal for analysis. De®ned by the slope of the calibration curve [d(signal)/d(amount)]. Serial AnalysisÐSeparate and sequential processing of samples or compounds during one or more steps of an analytical protocol. Soft SpotsÐRefers to the labile substructures of a molecule that indicate chemical and/or metabolic instability. Structure-Activity Relationship (SAR)ÐThe relationships between chemical structure and pharmacological activity for a series of compounds. Solid-Phase Extraction (SPE)ÐUse of solid chromatographic media to extract compounds from a liquid sample. Standard Operating ProcedureÐA document stipulating a speci®c protocol for analytical or organizational activities, for the purpose of satisfying regulatory or organizational guidelines. Substructure ConnectivityÐThe use of MS/MS or MSn to delineate substructural relationships. Analogous to two-dimensional NMR techniques. Thin Layer ChromatographyÐA technology in which compounds are partitioned between stationary media on a solid support (e.g., glass) and a liquid phase at atmospheric pressure. Total Ion Current (TIC) ChromatogramÐThe mass spectrometry signal that corresponds to the sum of all ions detected in a scan plotted against time. Thermospray Ionization (TSI)ÐThe liberation of ions from heated droplets occurring in a vacuum region. XenobioticÐA compound foreign to an organism. REFERENCES Ackermann BL, Regg BT, Colombo L, Stella S, Coutant JE. 1996a. Rapid analysis of antibiotic-containing mixtures from fermentation broths by using liquid chromatography±electrospray ionization±mass spectrometry and matrix-assisted laser desorption ionization±time-of¯ight±mass spectrometry. J Am Soc Mass Spectrom 7:1227±1237. Ackermann BL, Gillespie TA, Regg BT, Austin KF, Coutant JE. 1996b. Application of packed capillary liquid chromatography/mass spectrometry with electrospray ionization to the study of the human biotransformation of the antiemetic drug dolasetron. J Mass Spectrom 31:681±689. LC/MS APPLICATIONS Ackermann BL, Ruterbories KJ, Hanssen BR, Lindstrom TD. 1998. Increasing the throughput of microsomal stability screening using fast gradient elution LC/MS. Proc 46th ASMS Conf Mass Spectrom and Allied Topics, Orlando, Florida, p 16. Aebersold R, Bures EJ, Namchuk M, Goghari MH, Shushan B, Covey TC. 1992. Design, synthesis, and characterization of a protein sequencing reagent yielding amino acid derivatives with enhanced detectability by mass spectrometry. Protein Sci 1:494±503. Afeyan NB, Fulton SP, Regnier FE. 1991. Perfusion chromatography packing materials for proteins and peptides. J Chromatogr 544:267±279. Allanson JP, Biddlecombe RA, Jones AE, Pleasance S. 1996. The use of automated solid phase extraction in the `96 well' format for high throughput bioanalysis using liquid chromatography coupled to tandem mass spectrometry. Rapid Commun Mass Spectrom 10:811±816. Allen MC, Shah TS, Day WW. 1998. Rapid determination of oral pharmacokinetics and plasma free fraction using cocktail approaches: methods and application. Pharm Res 15:93±97. Anderegg RJ, Wagner DS, Blackbrun RK, Opiteck GJ, Jorgenson JW. 1997. A multidimensional approach to protein characterization. J Protein Chem 16:523±526. Andren PE, Emmett MR, Caprioli RM. 1995. Micro-electrospray: zeptomole/attomole per microliter sensitivity for peptides. J Am Soc Mass Spectrom 5:867±869. Apffel A, Chakel J, Udiavar S, Hancock WS, Souders C, Pungor, E Jr. 1995. Application of capillary electrophoresis, high-performance liquid chromatography, on-line electrospray mass spectrometry and matrix-assisted laser desorption ionization±time of ¯ight mass spectrometry to the characterization of single-chain plasminogen activator. J Chromatogr A 717:41±60. Apffel A, Chakel JA, Hancock WS, Souders C, M' Timkulu T, Pungor E Jr. 1996. Application of multidimensional af®nity high-performance liquid chromatography and electrospray ionization liquid chromatography±mass spectrometry to the characterization of glycosylation in singlechain plasminogen activator. Initial results. J Chromatogr A 750:35±42. Arnott DP, Henzel WJ, Stults JT. 1995. Identi®cation of proteins from two-dimensional electrophoresis gels by peptide mass ®ngerprinting. In: Snyder AP, editor. Biochemical and biotechnological applications of electrospray ionization mass spectrometry. Washington, DC: American Chemical Society. 619:226±243. Arnott D, O'Connell KL, King KL, Stults JT. 1998. An integrated approach to proteome analysis: identi®cation of proteins associated with cardiac hypertrophy. Anal Biochem 258:1±18. Arpino PJ, Dawkins BG, McClafferty FW. 1974. A liquid chromatography/mass spectrometry system providing continuous monitoring with nanogram sensitivity. J Chromatogr Sci 12:574±578. & Arpino PJ, Guiochon G. Krien P, Devant G. 1979. Optimization of the instrumental parameters of a combined liquid chromatograph±mass spectrometer, coupled by an interface for direct liquid introduction. J Chromatogr 185: 529±547. Arpino PJ. 1982. On-line liquid chromatography/mass spectrometry? An odd couple! Trends Anal Chem 1: 154±158. Ayrton J, Dear GJ, Leavens WJ, Mallett DN, Plumb RS. 1997. The use of turbulent ¯ow chromatography/mass spectrometry for the rapid, direct analysis of a novel pharmaceutical compound in plasma. Rapid Commun Mass Spectrom 11:1953±1958. Ayrton J, Plumb R, Leavens WJ, Mallett D, Dickins M, Dear GJ. 1998a. Application of a generic fast gradient liquid chromatography tandem mass spectrometry method for the analysis of cytochrome p450 probe substrates. Rapid Commun Mass Spectrom 12:217±224. Ayrton J, Dear GJ, Leavens WJ, Mallett DN, Plumb RS. 1998b. Use of generic fast gradient liquid chromatography± tandem mass spectroscopy in quantitative bioanalysis. J Chromatogr B Biomed Sci Appl 709:243±254. Babiak J. 1997. Transforming your robotics into an infrastructure for the future. J Biomol Screening 2:139±143. Baillie TA. 1992. Advances in the application of mass spectrometry to studies of drug metabolism, pharmacokinetics and toxicology. Int J Mass Spectrom Ion Proc 118/ 119:289±314. Banerjee PK, Rosofsky M. 1997. Drug discovery: the quest for innovation and productivity. Scrip Magazine 35±38. Banks JF, Gulcicek EE. 1997. Rapid peptide mapping by reversed-phase chromatography on nonporous silica with on-line electropsray time-of-¯ight mass spectrometry. Anal Chem 69:3973±3978. Bansal S, Liang Z. 1998. Rapid, uniform and rational bioanalytical methods development using LC/MS/MS. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 716. Barbuch RJ, Coutant JE, Welsh MB, Setchell KD. 1989. The use of thermospray liquid chromatography/tandem mass spectrometry for the class identi®cation and structural veri®cation of phytoestrogens in soy protein preparations. Biomed Environ Mass Spectrom 18:973±977. Bates JH. 1995. Tuberculosis chemotherapy. The need for new antituberculosis drugs is urgent. Am J Respir Crit Care Med 151:942±943. Baxter AD. 1997. Meeting client needs in combinatorial chemistry: a service company perspective. Drug Discovery Today 2:307±308. Bean KA, Henion JD. 1997. Direct determination of anabolic steroid conjugates in human urine by combined highperformance liquid chromatography and tandem mass spectrometry. J Chromatogr B Biomed Sci Appl 690: 65±75. Beattie IG, Blake TJ. 1989. The structural identi®cation of drug metabolites using thermospray liquid chromatography/ 271 & LEE AND KERNS mass spectrometry. Biomed Environ Mass Spectrom 18:872±877. Beaudry F, Yves Le Blanc JC, Coutu M, Brown NK. 1998. In vivo pharmacokinetic screening in cassette dosing experiments: the use of on-line prospekt1 liquid chromatography/atmospheric pressure chemical ionization tandem mass spectrometry technology in drug discovery. Rapid Commun Mass Spectrom 12:1216±1222. Blakely CR, Carmody JJ, Vestal ML. 1980. Liquid chromatograph±mass spectrometer for analysis of non-volatile samples. Anal Chem 52:1636±1640. Blakely CR, Vestal ML. 1983. Thermospray interface for liquid chromatography/mass spectrometry. Anal Chem 55: 750±754. Brewer E, Henion JD. 1998. Atmospheric pressure ionization LC/MS/MS techniques for drug disposition studies. J Pharm Sci 87:395±402. Brownsill R, Combal J-P, Taylor A, Bertrand M, Luijten W, Walther B. 1994. The application of electrospray and neutral-loss mass spectrometry to the identi®cation of the metabolites of S12813 in rat liver slices. Rapid Commun Mass Spectrom 8:361±365. Bruins AP, Covey TR, Henion JD. 1987. Ion spray interface for combined liquid chromatography/atmospheric pressure ionization mass spectrometry. Anal Chem 59:2642±2646. Bruins AP. 1991. Liquid chromatography±mass spectrometry with ionspray and electrospray interfaces in pharmaceutical and biomedical research. J Chromatogr 554:39±46. Bruins AP. Mechanistic aspects of electrospray ionization. J Chromatogr A 794:345±357. Bryant MS, Korfmacher WA, Wang S, Nardo C, Nomeir AA, Lin CC. 1997. Pharmacokinetic screening for the selection of new drug discovery candidates is greatly enhanced through the use of liquid chromatography±atmospheric pressure ionization tandem mass spectrometry. J Chromatogr A 777:61±66. Burdick DJ, Stults JT. 1997. Analysis of peptide synthesis products by electrospray ionization mass spectrometry. Methods Enzymol 289:499±519. Burlingame AL, Carr SA, editors. 1996. Mass spectrometry in the biological sciences. Totowa, NJ: Humana Press. Burlingame AL, Boyd RK, Gaskell SJ. 1998. Mass spectrometry. Anal Chem 70:647R±716R. Caldwell GW, Easlick SM, Gunnet J, Masucci JA, Demarest K. 1998. In vitro permeability of eight beta-blockers through caco-2 monolayers utilizing liquid chromatography/electrospray ionization mass spectrometry. J Mass Spectrom 33:607±614. Carr SA, Huddleston MJ, Bean MF. 1993. Selective identi®cation and differentiation of N- and O-linked oligosaccharides in glycoproteins by liquid chromatography±mass spectrometry. Protein Science 2:183±196. Carroll DI, Dzidic I, Stillwell RN, Haegele KD, Horning EC. 1975. Atmospheric pressure ionization mass spectrometry: corona discharge ion source for use in liquid chromato- 272 graph±mass spectrometer±computer analytical system. Anal Chem 47:2369±2373. Celis JE, Rasmussen HH, Leffers H, Madsen P, Honore B, Gesser B, Dejgaard K, Vandekerckhove J. 1991. Human cellular protein patterns and their link to genome DNA sequence data: usefulness of two-dimensional gel electrophoresis and microsequencing. FASEB J 5: 2200±2208. Cepa S, Searle P. 1998. Establishing an open access LC/MS lab for pharmaceutical discovery and development. Proc 46th ASMS Conf Mass Spectrom and Allied Topics, Orlando, Florida. p 283. Chakel JA, Pungor E Jr, Hancock WS, Swedberg SA. 1997. Analysis of recombinant DNA-derived glycoproteins via high-performance capillary electrophoresis coupled with off-line matrix-assisted laser desorption ionization time-of¯ight mass spectrometry. J Chromatogr B Biomed Sci Appl 689:215±220. Chang JP, Kiehl DE, Kennington A. 1997. Separation and characterization of the tryptic peptide mapping of recombinant bovine growth hormone by reversed-phase high-performance liquid chromatography electrospray mass spectrometry. Rapid Commun Mass Spectrom 11: 1266±1270. Chen TM, Chan KY, Coutant JE, Okerholm RA. 1991. Determination of the metabolites of terfenadine in human urine by thermospray liquid chromatography±mass spectrometry. J Pharm Biomed Anal 9:929±933. Clauser K, Baker P. 1997. Protein prospector, V.3.0, University of California, San Francisco. Cole MJ, Cunniff J, Drexler D, Hemenway E, Land A. 1998. Large scale quantitative analysis of pharmaceuticals via LC±MS and LC±MS/MS ion trap mass spectrometry. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 12. Cooks RG, Busch KL. 1982. Counting molecules by desorption ionization and mass spectrometry/mass spectrometry. J Chem Educ 59:926±933. Covey TR, Lee ED, Henion JD. 1986. High-speed liquid chromatography/tandem mass spectrometry for the determination of drugs in biological samples. Anal Chem 58: 2453±2460. Covey TR, Huang EC, Henion JD. 1991. Structural characterization of protein tryptic peptides via liquid chromatography/mass spectrometry and collision-induced dissociation of their doubly charged molecular ions. Anal Chem 63:1193±1200. Covey T. 1995. Analytical characteristics of the electrospray ionization process. In: Snyder AP, editor. Biochemical & biotechnological applications of electrospray ionization mass spectrometry. Washington, DC: American Chemical Society. 619:21±59. Davies J. 1994. Inactivation of antibiotics and the dissemination of resistance genes. Science 264:375±382. Davis MT, Stahl DC, Hefta SA, Lee TD. 1995. A microscale electrospray interface for on-line, capillary liquid chroma- LC/MS APPLICATIONS tography/tandem mass spectrometry of complex peptide mixtures. Anal Chem 67:4549±4556. Davis MT, Lee TD. 1998. Rapid protein identi®cation using a microscale electrospray LC/MS system on an ion trap mass spectrometer. J Am Soc Mass Spectrom 9: 194±201. Davis RG, Anderegg RJ, Blanchard SG. Iterative size-exclusion chromatography coupled with liquid chromatographic mass spectrometry to enrich and identify tightbinding ligands from complex mixtures. Tetrahedron, in press. Dear GJ, Fraser TJ, Patel DK, Long J, Pleasance S. 1998. Use of liquid chromatography±tandem mass spectrometry for the quantitative and qualitative analysis of an antipsychotic agent and its metabolites in human plasma and urine. J Chromatogr A 794:27±36. Desai MC, Zuckermann RN, Moos WH. 1994. Recent advances in the generation of chemical diversity libraries. Drug Dev Res 33:174±188. Devlin JJ, Lian A, Trinh L, Plolkoff MA, Senator D, Sheng W, Kondracki J, Kretschmer PJ, Morser J, Lipson SE, Spann R, Loughlin JA, Dunn KV, Morrissey MM. 1996. High capacity screening of pooled compounds: identi®cation of active compound without re-assay of pool members. Drug Dev Res 37:80±85. Ding J, Burkhart W, Kassel DB. 1994. Identi®cation of phosphorylated peptides from complex mixtures using negative-ion ori®ce-potential stepping and capillary liquid chromatography/electrospray ionization mass spectrometry. Rapid Commun Mass Spectrom 8:94±98. Dower WJ, Fodor SPA. 1991. The search for molecular diversity (II): recombinant and synthetic randomized peptide libraries. Annu Rep Med Chem 26:271±280. Drews J, Ryser S. 1997. Pharmaceutical innovation between scienti®c opportunities and economic constraints. Drug Discovery Today 2:365±372. Ducret A, Van Oostveen I, Eng JK, Yates JR 3rd, Aebersold R. 1998. High throughput protein characterization by automated reverse-phase chromatography/electrospray tandem mass spectrometry. Protein Sci 7:706±719. Ehlhardt WJ, Woodland JM, Baughman TM, Vandenbranden M, Wrighton SA, Kroin JS, Norman BH, Maple SR. 1998. Liquid chromatography/nuclear magnetic resonance spectroscopy and liquid chromatography/mass spectrometry identi®cation of novel metabolites of the multidrug resistance modulator LY335979 in rat bile and human liver microsomal incubations. Drug Metab Dispos 26: 42±51. Emmett MR, Caprioli RM. 1994. Micro-electrospray mass spectrometry: ultra-high-sensitivity analysis of peptides and proteins. J Am Soc Mass Spectrom 5: 605±613. Eng JK, McCormack AL, Yates JR. 1994. An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J Am Soc Mass Spectrom 5:976±989. & Fenn JB, Mann M, Meng CK, Wong SF, Whitehouse CM. 1989. Electrospray ionization for mass spectrometry of large biomolecules. Science 246:64±71. Fernandez-Metzler CL, Ellis JD, King RC, Merkle KM, Olah TV, Owens KG. 1998. Application of correlation analysis to the identi®cation of drug metabolites in complex biological matrices. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 712. Figeys D, Ducret A, Yates JR 3rd, Aebersold R. 1996. Protein identi®cation by solid phase microextraction-capillary zone electrophoresis±microelectrospray±tandem mass spectrometry. Nat Biotechnol 14:1579±1583. Figeys D, Aebersold R. 1997. High sensitivity identi®cation of proteins by electrospray ionization tandem mass spectrometry: initial comparison between an ion trap mass spectrometer and a triple quadrupole mass spectrometer. Electrophoresis 18:360±368. Figeys D, Aebersold R. 1998. High sensitivity analysis of proteins and peptides by capillary electrophoresis±tandem mass spectrometry: recent developments in technology and applications. Electrophoresis 19:885±892. Fink SW, Rourick RA, Whitney JL, Volk KJ, Klohr SE, DiDonato GC, Kerns EH, Lee MS. 1997. Accelerating drug development using automated predictive stability pro®ling methodologies. Annual Meeting of the American Association of Pharmaceutical Scientists, Boston, Massachusetts, p 590. Foltz RL. 1978. Quantitative analysis of abused drugs in physiological ¯uids by gas chromatography/chemical ionization mass spectrometry. In: De Leenheer AP, Roncucci, RR, Van Peteghem C, editors. Quantitative Mass Spectrometry in Life Sciences II. Elsevier: Amsterdam 39±62. Fouda H, Nocerini M, Schneider R, Gedutis C. 1991. Quantitative analysis by high-performance liquid chromatography atmospheric pressure chemical ionization mass spectrometry: the determination of the renin inhibitor CP80,794 in human serum. J Am Soc Mass Spectrom 2: 164±167. Gallop MA, Barrett RW, Dower WJ, Fodor SPA, Gordon EM 1994. Applications of combinatorial technologies to drug discovery. 1. Background and peptide combinatorial libraries. J Med Chem 37:1233±1251. Games DE, McDowell MA, Levsen K, Schafer KH, Dobberstein P, Gower JL. 1984. A comparison of moving belt interfaces for liquid chromatography mass spectrometry. Biomed Mass Spectrom 11:87±95. Ganem B, Li Y, Henion JD. 1991. Detection of non covalent receptor±ligand complexes by mass spectrometry. J Am Chem Soc 113:6294±6296. Ganguly AK, Praminik BN, Tsarbopoulos A, Covey TR, Huang E, Fuhrman SA. 1992. Mass spectrometric detection of the non covalent GDP-bound conformational state of the human H-ras protein. J Am Chem Soc 114:6559±6560. Garland WA, Powell ML. 1981. Quantitative selected ion monitoring (QSIM) of drugs and/or drug metabolites in biological matrices. J Chromatogr Sci 19:392±434. 273 & LEE AND KERNS Georgopapadakou NH, Walsh TJ. 1994. Human mycoses: drugs and targets for emerging pathogens. Science 264:371±373. Gilbert JR, Lewer P. 1998. LC/MSn as a tool for the identi®cation of biologically active compounds from natural product mixtures. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 21. Gillespie T, Barbuch R, Berna M, Bowsher R, Jensen C. LC/ MS/MS identi®cation and pro®ling of pergolide mesylate metabolites in human, rat, and dog liver slices. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 706. Gordon EM, Barrett RW, Dower WJ, Fodor SPA, Gallop MA 1994. Applications of combinatorial technologies to drug discovery. 2. Combinatorial organic synthesis, library screening strategies, and future directions. J Med Chem 37:1385±1401. Hancock WS, Apffel A, Chakel J, Souders C, M'Timkulu T, Pungor E, Guzzetta AW. 1996. Reversed-phase peptide mapping of glycoproteins using liquid chromatography/ electrospray ionization±mass spectrometry. Methods Enzymol 271:403±427. Hanisch FG, Green BN, Bateman R, Peter-Katalinic J. 1998. Localization of O-glycosylation sites of MUC1 tandem repeats by QTOF ESI mass spectrometry. J Mass Spectrom 33:358±362. Hayes MJ, Lankmayer EP, Vouros P, Karger BL, McGuire JM. 1983. Moving belt interface with spray deposition for liquid chromatography/mass spectrometry. Anal Chem 55:1745±1752. Heath TG, Giordani AB. 1993. Reversed-phase capillary highperformance liquid chromatography with on-line UV, ¯uorescence and electrospray ionization mass spectrometric detection in the analysis of peptides and proteins. J Chromatogr 638:9±19. Heath TG, Mooney JP, Broersma R. 1997. Narrow-bore liquid chromatography±tandem mass spectrometry with simultaneous radioactivity monitoring for partially characterizing the biliary metabolites of an arginine ¯uoroalkyl ketone analog of D-MePhe-Pro-Arg, a potent thrombin inhibitor. J Chromatgr B Biomed Sci Appl 688:281±289. Heller R Hindle T. 1998. New York: DK Publishing, Inc. Essential manager's manual. p 168. Henion JD, Covey TR. 1986. The determination of drugs in urine by LC±MS and LC±MS/MS. In: Gaskel SJ, editor. Mass spectrometry in biomedical research. New York: John Wiley & Sons 477±486. Henion J, Brewer E, Rule G. 1998. Sample preparation for LC/MS/MS: analyzing biological and environmental samples. Anal Chem 70:650A±656A. Henzel WJ, Billeci TM, Stults JT, Wong SC, Grimley CG, Wantanabe CW. 1993. Identifying proteins from twodimensional gels by molecular mass searching of peptide fragments in protein sequence databases. Proc Natl Acad Sci 90:5011±5015. Herron WJ, Eadie J, Penman AD. 1995. Estimation of ranolazine and eleven phase I metabolites in human 274 plasma by liquid chromatography±atmospheric pressure chemical ionization mass spectrometry with selected-ion monitoring. J Chromatogr A 712:55±60. Higgs RE. 1998. Automatic deconvolution of complex mixture components from HPLC±MS data. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 816. Hirschman R. 1991. Medicinal chemistry in the golden age of biology: lessons from steroid and peptide research. Angew Chem 30:1278±1301. Hop CE, Wang Z, Chen Q, Kwei G. 1998. Plasma-pooling methods to increase throughput for in vivo pharmacokinetic screening. J Pharm Sci 87:901±903. Hopfgartner G, Bean K, Henion J, Henry R. 1993. Ion spray mass spectrometric detection for liquid chromatography: a concentration- or a mass-¯ow-sensitive device? J Chromatogr 647:51±61. Hopfgartner G, Chernushevich I, Covey T, Bonner R. 1998. Exact mass measurements on product ions for structural elucidation of drug metabolites with a tandem quadrupole TOF instrument. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando Florida, p 713. Houthaeve T, Gausepohl H, Mann M, Ashman K. 1995. Automation of micro-preparation and enzymatic cleavage of gel electrophoretically separated proteins. FEBS Lett 376:91±94. Houthaeve T, Gausepohl H, Ashman K, Nillson T, Mann M. 1997. Automated protein preparation techniques using a digest robot. J Protein Chem 16:343±348. Hsieh YF, Gordon N, Regnier F, Afeyan N, Martin SA, Vella GJ. 1997. Multidimensional chromatography coupled with mass spectrometry for target-based screening. Mol Divers 2:189±196. Hung DT, Jamison TF, Schreiber SL. 1996. Understanding and controlling the cell cycle with natural products. Chem Biol 3:623±639. Hunt DF, Henderson RA, Shabanowitz J, Sakaguchi K, Michel H, Sevilir N, Cox AL, Appella E, Engelhard VH. 1992. Characterization of peptides bound to the class I MHC molecule HLA-A2.1 by mass spectrometry. Science 255: 1261±1263. Immler D, Gremm D, Kirsch D, Spengler B, Presek P, Meyer HE. 1998. Identi®cation of phosphorylated proteins from thrombin-activated human platelets isolated by twodimensional gel electrophoresis by electrospray ionization±tandem mass spectrometry (ESI±MS/MS) and liquid chromatography±electrospray ionization±mass pectrometry (LC±ESI±MS). Electrophoresis 19:1015±1023. Irabarne JV, Dziedzic PJ, Thomson BA. 1983. Atmospheric pressure ion evaporation±mass spectrometry. Int J Mass Spectrom Ion Phys 50:331±347. Janiszewski J, Schneider P, Hoffmaster K, Swyden M, Wells D, Fouda H. 1997. Automated sample preparation using membrane microliter extraction for bioanalytical mass spectrometry. Rapid Commun Mass Spectrom 11: 1033±1037. LC/MS APPLICATIONS Janota K, Carter GT. 1998. Natural products library screening using LC/MS and LC/MS/MS. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, 557. Johnson JV, Yost RA. 1985. Tandem mass spectrometry for trace analysis. Anal Chem 57:758A±768A. Josephs JL. 1996. Detection and characterization of fumonisin myotoxins by liquid chromatography/electrospray ionization using ion trap and triple quadrupole mass spectrometry. Rapid Commun Mass Spectrom 10:1333±1344. Joyce KB, Jones AE, Scott RJ, Biddlecombe RA, Pleasance S. 1998. Determination of the enantiomers of salbutamol and its 4-O-sulphate metabolites in biological matrices by chiral liquid chromatography tandem mass spectrometry. Rapid Commun Mass Spectrom 12:1899±1910. Kanazawa H, Konishi Y, Matsushima Y, Takahashi T. 1998. Determination of sedatives and anesthetics in plasma by liquid chromatography±mass spectrometry with a desalting system. J Chromatogr A 797:227±236. Kassel DB, Shushan B, Sakuma T, Salzmann JP. 1994. Evaluation of packed capillary perfusion column HPLC/ MS/MS for the rapid mapping and sequencing of enzymatic digests. Anal Chem 66:236±243. Kaur S, McGuire L, Tang D, Dollinger G, Huebner V. 1997. Af®nity selection and mass spectrometry-based strategies to identify lead compounds in combinatorial libraries. J Prot Chem 16:505±511. Kaye B, Clark MW, Cussans NJ, Macrae PV, Stopher DA. 1992. The sensitive determination of abanoquil in blood by high-performance liquid chromatography/atmospheric pressure ionization mass spectrometry. Biol Mass Spectrom 21:585±589. Kaye B, Herron WJ, Macrae PV, Robinson S, Stopher DA, Venn RF, Wild W. 1996. Rapid, solid phase extraction technique for the high-throughput assay of darifenacin in human plasma. Anal Chem 68:1658±1660. Kennedy RT, Jorgenson JW. 1989. Preparation and evaluation of packed capillary liquid chromatography columns with inner diameters from 20 to 50 mm. Anal Chem 61: 1128±1135. Kerns EH, Volk KJ, Hill SE, Lee MS. 1994. Pro®ling taxanes in taxus extracts using LC/MS and LC/MS/MS techniques. J Nat Prod 57:1391±1403. Kerns EH, Volk KJ, Hill SE, Lee MS. 1995. Pro®ling new taxanes using LC/MS and LC/MS/MS substructural analysis techniques. J Rapid Commun Mass Spectrom 9: 1539±1545. Kerns EH, Rourick RA, Volk KJ, Lee MS. 1997. Buspirone metabolite structure pro®le using a standard liquid chromatographic-mass spectrometric protocol. J Chromatogr B 698:133±145. Kerns EH, Hill SE, Detlefsen DJ, Long BH, Carboni J, Lee MS. 1998. Cellular uptake pro®le of paclitaxel using liquid chromatography tandem mass spectrometry. J Rapid Commun Mass Spectrom 12:620±624. Kleintop B, Volk K, Hail M, Sanders M, Reddick C, Bolgar M, Warrack B, Liu J, Detlefsen D, Kerns E, Lee M. 1998. & Predictive degradation pro®ling of biomolecules using LC/ MS methodologies. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 240. Kolb AJ, Neumann K. 1997. Beyond the 96-well microplate: instruments and assay methods for the 384-well format. J Biomol Screening 2:103±109. Kolthoff IM. 1973. Development of analytical chemistry as a science. Anal Chem 45:25A±37A. Kusmierz JJ, Sumrada R, Desiderio DM. 1990. Fast atom bombardment mass spectrometric quantitative analysis of methionine-enkephalin in human pituitary tissues. Anal Chem 62:2395±2400. Laitinen HA. 1973. The seven ages of an analytical method. Anal Chem 45:2305. Laitinen HA. 1980. Analytical chemistry in a changing world. Anal Chem 52:605A±612A. Lant MS, Martin LE, Oxford J. 1985. Qualitative and quantitative analysis of ranitidine and its metabolites by high-performance liquid chromatography±mass spectrometry. J Chromatogr 323:143±152. Lee ED, Henion JD. 1985. Micro-liquid chromatography/mass spectrometry with direct liquid introduction. J Chromatogr Sci 23:253±264. Lee MG, Millard BJ. 1975. A comparison of unlabelled and labelled internal standards for quanti®cation by single and multiple ion monitoring. J Biomed Mass Spectrom 2: 78±81. Lee MS, Yost RA, Perchalski RJ. 1986. Tandem mass spectrometry and drug metabolism. Annu Rep Med Chem 21:313±321. Lee MS, Yost RA. 1988. Rapid identi®cation of drug metabolites with tandem mass spectrometry. Biomed Env Mass Spectrom 15:193±204. Lee MS, Kerns EH, DiDonato GC, Gedamke R, Hail ME, Klohr SE, Rourick R, Starrett AM, Taylor SC, Thompson WL, Volk KJ, Warrack BM, Whitney JL. 1995a. Diversity pro®ling: rapid pharmaceutical development based on ESI/ MS. Proc 43rd ASMS Conf Mass Spectrom and Allied Topics Atlanta, Georgia, 148. Lee MS, Volk KJ, Liu J, Pucci MJ, Handwerger S. 1995b. Identi®cation of a cytoplasmic peptidoglycan precursor in antibiotic-resistant bacteria. In: Snyder AP, editor. Biochemical and biotechnological applications of electrospray ionization mass spectrometry. Washington, DC: American Chemical Society 619:106±129. Lee MS, Klohr SE, Kerns EH, Volk KJ, Leet JE, Schroeder DR, Rosenberg IE. 1996. The Coordinated Use of Tandem Mass Spectrometry and High Resolution Mass Spectrometry for the Structure Elucidation of the Kedarcidin Chromophore. J Mass Spectrom 31:1253±1260. Lee MS, Kerns EH, Hail ME, Liu J, Volk KJ. 1997. Recent applications of LC±MS techniques for the structure identi®cation of drug metabolites and related compounds. LC GC 15:542±558. Lee MS, Kerns EH. 1998. High throughput drug metabolite analysis using LC/MS screening methods. British Mass 275 & LEE AND KERNS Spectrometry Society 23rd Annual Meeting, University of Warrick, UK. Li C, Chauret N, Ducharme Y, Trimble LA, Nicoll-Grif®th DA, Yergey JA. 1995. Integrated application of capillary HPLC/continuous-¯ow liquid secondary ion mass spectrometry to discovery stage metabolism studies. Anal Chem 67:2931±2936. Li MX, Wu JT, Parus S, Lubman DM. 1998. Development of a three-dimensional topographic map display for capillary electrophoresis/mass spectrometry with an ion trap/re¯ectron time-of-¯ight mass spectrometer detector: applications to tryptic digests of isoforms of myelin basic protein. J Am Soc Mass Spectrom 9:701±709. Ling V, Guzetta AW, Canova-Davis E, Stults JT, Hancock WS, Covey TR, Shushan BI. 1991. Characterization of the tryptic map of recombinant DNA derived tissue plasminogen activator by high-performance liquid chromatography±electrospray ionization mass spectrometry. Anal Chem 63:2909±2915. Lippman M, Mok MS, Steen SN, Lane AZ, Caruso FS. 1977. Analgesic onset time of intravenous butorphanol in post surgical patients: a placebo-controlled study. Curr Ther Res 22:276±283. Liu J, Volk KJ, Kerns EH, Klohr SE, Lee MS, Rosenberg IE. 1993. Structural characterization of glycoprotein digest by microcolumn liquid chromatography±ionspray tandem mass spectrometry. J Chromatogr 632:45±56. Liu J, Zhao H, Volk KJ, Klohr SE, Kerns EH, Lee MS. 1996. Analysis of monoclonal antibody and immunoconjugate digests by capillary electrophoresis and capillary liquid chromatography. J Chromatogr A 735:357±366. Liu J, Volk KJ, Mata MJ, Kerns EH, Lee MS. 1997. Analysis of amino acid enantiomers derived from antitumor antibiotics using chiral capillary electrophoresis. J Pharm Biomed Anal 15:1729±1739. Liu JL, Lu KV, Eris T, Katta V, Westcott KR, Narhi LO, Lu HS. In vitro methionine oxidation of recombinant human leptin. Pharm Res 15:632±640. Lokiec F, du Sorbier BM, Sanderink GJ. 1996. Irinotacan (CPT11) metabolites in human bile and urine. Clin Cancer Res 2:1943±1949. Lopez LL, Yu X, Cui D, Davis MR. 1998. Identi®cation of drug metabolites in biological matrices by intelligent automated liquid chromatography/tandem mass spectrometry. Rapid Commun Mass Spectrom 12:1756±1760. Malcolm SL, Madigan MJ, Taylor NL. 1990. Thermospray mass spectrometer as a quantitative speci®c, sensitive, detector for liquid chromatography. Its application to the analysis of pyridostigmine in human plasma. J Pharm Biomed Anal 8:771±776. Manini P, Andreoli R, Mutti A, Bergamaschi E, Niessen WM. 1998. Determination of n-hexane metabolites by liquid chromatography/mass spectrometry. 2. Glucuronide-conjugated metabolites in untreated urine samples by electrospray ionization. Rapid Commun Mass Spectrom 12:1615±1624. 276 Maurer HH, Kraemer T, Ledvinka O, Schmitt CJ, Weber AA. 1997. Gas chromatography±mass spectrometry (GC±MS) and liquid chromatography±mass spectrometry (LC±MS) in toxicological analysis. Studies on the detection of clobenzorex and its metabolites within a systematic toxicological analysis procedure by GC±MS and by immunoassay and studies on the detection of alpha- and beta-amanitin in urine by atmospheric pressure ionization electrospray LC±MS. J Chromatogr B Biomed Sci Appl 689:81±89. Mazsaroff I, Yu W, Kelley BD, Vath JE. 1997. Quantitative comparison of global carbohydrate structures of glycoproteins using LC±MS and in-source fragmentation. Anal Chem 69:2517±2524. McClafferty FW, editor. 1983. Tandem mass spectrometry, New York: John Wiley and Sons. McCormack AL, Schieltz DM, Goode B, Yang S, Barnes G, Drubin D, Yates JR 3rd. 1997. Direct analysis and identi®cation of proteins in mixtures by LC/MS/MS and database searching at the low-femtomole level. Anal Chem 69:767±776. Medzihradszky KF, Maltby DA, Hall SC, Settineri CA, Burlingame AL. 1994. Characterization of protein Nglycosylation by reversed-phase microbore liquid chromatography/electrospray mass spectrometry, complementary mobile phases, and sequential exoglycosidase diegestion. J Am Soc Mass Spectrom 5:350±358. Medzihradszky KF, Besman MJ, Burlingame AL. 1998. Reverse-phase capillary high performance liquid chromatography/high performance electrospray ionization mass spectrometry: an essential tool for the characterization of complex glycoprotein digests. Rapid Commun Mass Spectrom 12:472±478. Moos WH, Green GD, Pavia MR. 1993. Recent advances in the generation of molecular diversity. Annu Rep Med Chem 28:315±324. Morris HR, Paxton T, Dell A, Langhorne J, Berg M, Bordoli RS, Hoyes J, Bateman RH. 1996. High sensitivity collisionally-activated decomposition tandem mass spectrometry on a novel quadrupole/orthogonal-acceleration time-of-¯ight mass spectrometer. Rapid Commun Mass Spectrom 10:889±896. Moseley MA, Deterding LJ, Tomer KB, Jorgenson JW. 1991. Nanoscale packed-capillary liquid chromatography coupled with mass spectrometry using a coaxial continuous-¯ow fast atom bombardment interface. Anal Chem 63:1467±1473. Nadin A, Nicolaou KC. 1996. Chemistry and biology of the zaragoic acids (squalestatins). Angew Chem Int Ed Engl 35:1622±1656. Nedved ML, Habibi-Goudarzi S, Ganem B, Henion JD. 1996. Characterization of benzodiazepine `combinatorial' chemical libraries by on-line immunoaf®nity extraction, coupled column HPLC±ion spray mass spectrometry±tandem mass spectrometry. Anal Chem 68: 4228±4236. LC/MS APPLICATIONS Needham SR, Cole MJ, Fouda HG. 1998. Direct plasma injection for high-performance liquid chromatographic± mass spectrometric quantitation of the anxiolytic agent CP-93 393. J Chromatogr B 718:87±94. Nicolas EC, Scholz TH. 1998. Active drug substance impurity pro®ling part II. LC/MS/MS ®ngerprinting. J Pharm Biomed Anal 16:825±836. Niessen WM. 1998. Advances in instrumentation in liquid chromatography±mass spectrometry and related liquidintroduction techniques. J Chromatogr A 794:407±435. Olah TV, McLoughlin DA, Gilbert JD. 1997. The simultaneous determination of mixtures of drug candidates by liquid chromatography/atmospheric pressure chemical ionization mass spectrometry as an in vivo drug screening procedure. Rapid Commun Mass Spectrom 11:17±23. Oxford J, Lant MS. 1989. Development and validation of a liquid chromatographic±mass spectrometric assay for the determination of sumatriptan in plasma. J Chromatogr 496: 137±146. Patchett AA. 1993. Excursions in drug discovery. J Med Chem 36:2051±2058. Perchalski RJ, Yost RA, Wilder BJ. 1982. Structural elucidation of drug metabolites by triple-quadrupole mass spectrometry. Anal Chem 54:1466±1471. Perico N, Remuzzi G. 1996. New antirejection drugs. Exp Opin Ther Patents 6:871±891. Pettit GR. 1994. Marine animal and terrestrial plant anticancer constituents. Pure Appl Chem 66:2271±2281. Poon GK, Wade J, Bloomer J, Clarke SE, Maltas J. 1996. Rapid screening of taxol metabolites in human microsomes by liquid chromatography/electrospray ionizationmass spectrometry. Rapid Commun Mass Spectrom 10: 1165±1168. Pullen FS, Kerkins GL, Burton KI, Ware RS, Teague MS, Kiplinger JP. 1995. Putting mass spectrometry in the hands of the end user. J Am Soc Mass Spectrom 6:394±399. Qin XZ, Ip DP, Chang KH, Dradransky PM, Brooks MA, Sakuma T. 1994. Pharmaceutical application of LC±MS. 1ÐCharacterization of a famotidine degradate in package screening study by LC±APCI MS. J Pharm Biomed Anal 12:221±233. Regnier FE. 1991. Perfusion chromatography. Nature 350: 634±635. Reinhold VN, Reinhold BB, Costello CE. 1995. Carbohydrate molecular weight pro®ling, sequence, linkage, and branching data: ES±MS and CID. Anal Chem 67: 1772±1784. Roepstorff P, Fohlman J. 1984. Proposal for a common nomenclature for sequence ions in mass spectra of peptides. Biomed Mass Spectrom 11:601±602. Rourick RA, Volk KJ, Kerns EK, Lee MS. 1996. Predictive strategy for the rapid structure elucidation of drug degradants. J Pharm Biomed Anal 14:1743±1752. Rourick RA, Fink SW, Volk KJ, Whitney JL, Klohr SE, Kerns EH, Lee MS. 1998. The impact of LC/MS based structure & pro®le libraries on the drug discovery±development continuum. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 17. Scott DO, Heath TG. 1998. Investigation of the CNS penetration of a potent 5-HT2a receptor antagonist (MDL 100,907) and an active metabolite (MDL 105,725) using in vivo microdialysis sampling in the rat. J Pharm Biomed Anal 17:17±25. Self R. 1979. A polemic on the uses and functions of deuterated analogues in quantitative mass spectrometry. Biomed Mass Spectrom 6:315±316. Selway CN, Terrett NK. 1996. Parallel-compound synthesis: methodology for accelerating drug discovery. Bioorg Med Chem 4:645±654. Sharma J, Panico M, Barber J, Morris HR. 1997. Puri®cation and determination of intact molecular mass by electrospray ionization mass spectrometry of the photosystem II reaction center subunits. J Biol Chem 272:33153±33157. Shevchenko A, Chernushevich I, Ens W, Standing KG, Thomson B, Wilm M, Mann M. 1997. Rapid `de novo' peptide sequencing by a combination of nanoelectrospray, isotopic labeling and a quadrupole/time-of-¯ight mass spectrometer. Rapid Commun Mass Spectrom 11: 1015±1024. Simpson H, Berthemy A, Buhrman D, Burton R, Newton J, Kealy M, Wells D, Wu D. 1998. High throughput liquid chromatography/mass spectrometry bioanalysis using 96well disk solid phase extraction plate for the sample preparation. Rapid Commun Mass Spectrom 12:75±82. Smith RD, Johnson AL. 1981. Deposition method for moving ribbon liquid chromatograph±mass spectrometer interfaces. Anal Chem 53:739±740. Smith RD, Light-Wahl KJ. 1993. The observation of noncovalent interactions in solution by electrospray ionization mass spectrometry: promise, pitfalls and prognosis. Biol Mass Spectrom 22:493±501. Snyder AP. 1995. Electrospray: a popular ionization technique for mass spectrometry. In: Snyder AP, editor. Biochemical & biotechnological applicatons of ESI MS. Washington, DC: Americana Chemical Society 619:1±20. Snyder LR, Kirkland JJ. 1979. Introduction to modern liquid chromatography, New York: John Wiley and Sons. Stevens GJ, Hitchcock K, Wang YK, Coppola GM, Versace RW, Chin JA, Shapiro M, Suwanrumpha S, Mangold BL. 1997. In vitro metabolism of N-(5-chloro-2-methylphenyl)N'-(2-methylpropyl)thiourea: species comparison and identi®cation of a novel thiocarbamide-glutathione adduct. Chem Res Toxicol 10:733±741. Stevenson TI, Loo JA, Greis KD. 1998. Coupling capillary high-performance liquid chromatography to matrixassisted laser desorption/ionization mass spectrometry and N-terminal sequencing of peptides via automated microblotting onto membrane substrates. Anal Biochem 262:99±109. Stoeckli M, Farmer TB, Caprioli RM. 1999. Automated mass spectrometry imaging with a matrix-assisted laser deso- 277 & LEE AND KERNS rption ionization time-of-¯ight instrument. J Am Soc Mass Spectrom 10:67±71. Storer R. 1996. Solution-phase synthesis in combinatorial chemistry: applications in drug discovery. Drug Discovery Today 1:248±254. Straub KM, Rudewicz P, Garvie C. 1987. `Metabolic mapping' of drugs: rapid screening techniques for xenobiotic metabolites with MS/MS techniques. Xenobiotica 17:413±422. Swiderek KM, Davis MT, Lee TD. 1998. The identi®cation of peptide modi®cations derived from gel-separated proteins using electrospray triple quadrupole and ion trap analyses. Electrophoresis 19:989±997. Taaffe P. 1996. Outsourcing in the pharmaceutical industry. Financial Times Management Report. Takach E, Minkoff M, Peltier J, Verentchikov A, Vestal M, Martin S. 1998. Rapid analysis and identi®cation of small molecules using automated ESI time-of-¯ight mass spectrometry. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 775. Taniguchi H, Hayashi N. 1998. A liquid chromatography/ electrospray mass spectrometric study on the posttranscriptional modi®cation of tRNA. Nucleic Acids Res 26:1481±1486. Taylor LCE, Johnson RL, Raso R. 1995. Open access atmospheric pressure chemical ionization mass spectrometry for routine sample analysis. J Am Soc Mass Spectrom 6: 387±393. Terrett NK, Bell AS, Brown D, Ellis P. 1996. Sildena®l (Viagra), a potent and selective inhibitor of type 5 cGMP phosphodiesterase with utility for the treatment of male erectile dysfunction. Bioorg Med Chem Lett 6:1819±1824. Thompson WL. 1995. Research investments in a managed care, cost-containment oligopsony. Annu Rep Med Chem 30: 339±345. Thomson BA. 1998. Atmospheric pressure ionization and liquid chromatography/mass spectrometryÐtogether at last. J Am Soc Mass Spectrom 9:187±193. Thomson BA, Iribarne JV. 1979. Field induced ion evaporation from liquid surfaces at atmospheric pressure. J Chem Phys 71:4451±4463. Tiller PR, Mutton IM, Lane SJ, Bevan CD. 1995. Immobilized human serum albumin: liquid chromatography/mass spectrometry as a method of determining drug-protein binding. Rapid Commun Mass Spectrom 9:261±263. Tiller PR, El Fallah Z, Wilson V, Juysman J, Patel D. 1997a. Qualitative assessment of leachables using data-dependent liquid chromatography/mass spectrometry and liquid chromatography/tandem mass spectrometry. Rapid Commun Mass Spectrom 11:1570±1574. Tiller PR, Cunniff J, Land AP, Schwartz J, Jardine I, Wake®eld M, Lopez L, Newton JF, Burton RD, Folk BM, Buhrman DL, Price P, Wu D. 1997b. Drug quantitation on a benchtop liquid chromatograpy±tandem mass spectrometry system. J Chromatogr A 771:119±125. Tiller PR, Land AP, Jardine I, Murphy DM, Sozio R, Ayrton A, Schaefer WH. 1998. Application of liquid chromatogra- 278 phy±mass spectrometry analyses to the characterization of novel glyburide metabolites formed in vitro. J Chromatogr A 794:15±25. Tomoda H, Kim YK, Nishida H, Musuma R, Omura S. 1994. Pyripyropenes, novel inhibitors of acyl-CoA:Cholesterol acyltransferase produced by Aspergillus fumigatus. I. Production, isolation, and biological properties. J Antibiotics 47:148±153. Tong H, Bell D, Siegel MM. 1998. Modules for automated data massaging, interpretation and e-mailing in open access mass spectrometry. Proc 46th ASMS Conf Mass Spectrom and Allied Topics Orlando, Florida, p 110. Turner WW, Rodriguez MJ. 1996. Recent advances in the medicinal chemistry of antifungal agents. Curr Pharm Des 2:209±224. Van Breemen RB, Bartlett MG, Tsou YH, Culver C, Swaisgood H, Unger SE. 1991. Degradation of peptide drugs by immobilized digestive proteases. Drug Metab Dispos 19: 683±690. Van Breemen RB, Huang C-R, Nikolic D, Woodbury CP, Zhao Y-Z, Venton DL. 1997. Pulsed ultra®ltration mass spectrometry: a new method for screening combinatorial libraries. Anal Chem 69:2159±2164. Van Breemen RB, Nikolic D, Bolton JL. 1998. Metabolic screening using on-line ultra®ltration mass spectrometry. Drug Metab Dispos 26:85±90. Vogler B, Klaiber I, Roos G, Walter CU, Hiller W, Sandor P, Kraus W. 1998. Combination of LC±MS and LC±NMR as a tool for the structure determination of natural products. J Nat Prod 61:175±178. Volk KJ, Klohr SE, Rourick RA, Kerns EH, Lee MS. 1996. Pro®ling impurities and degradants of butorphanol tartrate using liquid chromatography/mass spectrometry and liquid chromatography/tandem mass spectrometry substructural techniques. J Pharm Biomed Anal 14: 1663±1674. Volk KJ, Hill SE, Kerns EH, Lee MS. Pro®ling degradants of paclitaxel using liquid chromatography±mass spectrometry and liquid chromatography±tandem mass spectrometry substructural techniques. J Chromatogr B Biomed Sci Appl 696:99±115. Voyksner RD, Bursey JT, Hines JW. 1985. Analysis of candidate anticancer drugs by thermospray high-performance liquid chromatography±mass spectrometry. J Chromatogr 323:383±394. Wang-Iverson D, Arnold ME, Jemal M, Cohen AI. 1992. Determination of SQ 33,600, a phosphinic acid containing HMG CoA reductase inhibitor, in human serum by highperformance liquid chromatography combined with ionspray mass spectrometry. Biol Mass Spectrom 21: 189±194. Weidolf L, Covey TR. 1992. Studies on the metabolism of omeprazole in the rat using liquid chromatography/ ionspray mass spectrometry and the isotope cluster technique with [34S]omeprazole. Rapid Commun Mass Spectrom 6:192±196. LC/MS APPLICATIONS Weintraub ST, Pinckard RN, Hail M. 1991. Electrospray ionization for analysis of platelet-activating factor. Rapid Commun Mass Spectrom 5:309±311. Whitehouse CM, Dreyer RN, Yamashita M, Fenn JB. 1985. Electrospray interface for liquid chromatographs and mass spectrometers. Anal Chem 57:675±679. Whitelegge JP, Gundersen CB, Faull KF. 1998. Electrosprayionization mass spectrometry of intact intrinsic membrane proteins. Protein Sci 7:1423±1430. Wieboldt R, Campbell DA, Henion J. Quantitative liquid chromatographic-tandem mass spectrometric determination of orlistat in plasma with a quadrupole ion trap. J Chromatogr B Biomed Sci Appl 708:121±129. Williamson BL, Tomlinson AJ, Mishra PK, Gleich GJ, Naylor S. 1998. Structural characterization of contaminants found in commercial preparations of melatonin: similarities to case-related compounds from L-tryptophan associated with eosinophilia±myalgia syndrome. Chem Res Toxicol 11:234±240. Wilm M, Mann M. 1996. Analytical properties of the nanoelectrospray ion source. Anal Chem 68:1±8. Wilson-Lingardo L, Davis PW, Ecker DJ, Hebert N, Acevedo O, Sprankle K, Brennan T, Schwarcz L, Freier SM, Wyatt JR. 1996. Deconvolution of combinatorial libraries for drug discovery: experimental comparison of pooling strategies. J Med Chem 39:2720±2726. Wong SF, Meng CK, Fenn JB. 1988. Multiple charging in electrospray ionization of poly(ethylene glycols). J Phys Chem 92:546±550. Wong PSH, Yoshioka K, Xie F, Kissinger PT. In vivo microdialysis/liquid chromatography/tandem mass spectrometry for the on-line monitoring of melatonin in rat. Rapid Commun Mass Spectrom 13:1±5. Wood SA, Parton AH, Simmonds RJ, Stevenson D. 1996. Stereoselective metabolism of two keto-pyrrol analogues & of 8-hydroxy-2-dipropyl-aminotetralin by freshly isolated rat hepatocytes. Chirality 8:264±270. Wu SL, Wang YJ, Hu J, Leung D. 1997. The detection of the organic extractables in a biotech product by liquid chromatography on-line with electrospray mass spectrometry. PDA J Pharm Sci Technol 51:229±237. Yates JR, Speicher S, Grif®n PR, Hunkapiller T. 1993. Peptide mass maps: A highly informative approach to protein identi®cation. Anal Biochem 214:397±408. Yates JR. 1998a. Mass spectrometry and the age of the proteome. J Mass Spectrom 33:1±19. Yates JR 3rd, Morgan SF, Gatlin CL, Grif®n PR, Eng JK. 1998b. Method to compare collision-induced dissociation spectra of peptides: potential for library searching and subtractive analysis. Anal Chem 70:3557±3565. Yates JR 3rd. 1998c. Database searching using mass spectrometry data. Electrophoresis 19:893±900. Yinon J, Hwang DG. 1985. Metabolic studies of explosives. II. High-performance liquid chromatography±mass spectrometry of metabolites of 2,4,6-trinitrotoluene. J Chromatogr 339:127±137. Yost RA, Boyd RK. 1990. Tandem mass spectrometry: quadrupole and hybrid instruments. Methods Enzymol 193:154±200. Yu X, Cui D, Davis MR. 1999. Identi®cation of in vitro metabolites of indinavir by intelligent automated LC±MS/MS (INTAMS) utilizing triple quadrupole tandem mass spectrometry. J Am Soc Mass Spectrom 10:175±183. Zavitsanos AP, Alebic-Kolbah T. 1998. Enantioselective determination of terazosin in human plasma by normal phase high-performance liquid chromatography±electrospray mass spectrometry. J Chromatogr A 794:45±56. Zeedick JF. 1979. A double-blind comparison of oral butorphanol and codeine. Curr Med Res Opinion 6:178±182. 279