1135

American Journal of
EPIDEMIOLOGY
Copyright 2001 by the Johns Hopkins University
School of Hygiene and Public Health
Sponsored by the Society for Epidemiologic Research
Published by Oxford University Press
Volume 153
Number 12
June 15, 2001
Molecular Epidemiology of Infection Foxman and Riley
COMMENTARY
Molecular Epidemiology: Focus on Infection
Betsy Foxman
1
and Lee Riley
2
Molecular biology techniques have become increasingly integrated into the practice of infectious disease
epidemiology. The term molecular epidemiology routinely appears in the titles of articles that use molecular
strain-typing (fingerprinting) techniquesregardless of whether there is any epidemiologic application. What
distinguishes molecular epidemiology is both the molecular, the use of the techniques of molecular biology, and
the epidemiology, the study of the distribution and determinants of disease occurrence in human populations.
The authors review various definitions of molecular epidemiology. They then comment on the range of molecular
techniques available and present some examples of the benefits and challenges of applying these techniques
to infectious agents and their affected host using tuberculosis and urinary tract infection as examples. They close
with some thoughts about training future epidemiologists to best take advantage of the new opportunities that
arise from integrating epidemiologic methods with modern molecular biology. Am J Epidemiol 2001;153:
113541.
communicable diseases; epidemiology, molecular; tuberculosis; urinary tract infections
Received for publication June 21, 2000, and accepted for publi-
cation October 5, 2000.
Abbreviation: IS6110, insertion sequence 6110.
1
Department of Epidemiology and Center for Molecular and
Clinical Epidemiology of Infectious Diseases, School of Public
Health, University of Michigan, Ann Arbor, MI.
2
Divisions of Infectious Diseases and Epidemiology, School of
Public Health, University of California, Berkeley, Berkeley, CA.
Reprint requests to Dr. Betsy Foxman, Department of Epi-
demiology, School of Public Health, University of Michigan, 109 Obser-
vatory Street, Ann Arbor, MI 48109-2029 (bfoxman@umich.edu).
Over the past two decades, there has been a proliferation
of subspecialties among epidemiologists. Perhaps none of
these subspecialties has been received with more contro-
versy than molecular epidemiology, as the term molecu-
lar describes neither a disease category nor a substantive
area (1) but in jargonese refers to characteristics based on
nucleic acid- or amino acid-based content. The issue is fur-
ther confused by the independent emergence of the term
molecular epidemiology during the 1970s and early 1980s
in three separate substantive areas: cancer epidemiology,
environmental epidemiology, and infectious disease epi-
demiology. In many epidemiologic textbooks, molecular
epidemiology has been defined almost exclusively in terms
of biomarkers (2), ignoring the many applications in both
genetic and infectious disease epidemiology.
WHAT EXACTLY IS MOLECULAR EPIDEMIOLOGY?
Many different definitions of molecular epidemiology
have been published (table 1); all mention the use of mole-
cular tools, but not all explicitly mention epidemiology. This
is unfortunate, as molecular epidemiology is not just molec-
ular taxonomy, phylogeny, or population genetics but the
application of these techniques to epidemiologic problems.
Molecular taxonomy, phylogeny, population genetics, and
molecular epidemiology may use the same laboratory tech-
niques, but each follows distinct principles. In phylogeny/
taxonomy, the data are generated to describe properties and
characteristics of organisms. Population genetics often inter-
sects with epidemiology: both use population approaches to
describe the distribution of characteristics of interest and ana-
lyze data to identify the determinants of that distribution.
Epidemiology attempts to identify factors that determine dis-
ease distribution in time and place, as well as factors that
determine disease transmission, manifestation, and progres-
sion. Further, epidemiology is always motivated by an oppor-
tunity or possibility for intervention and prevention.
What distinguishes molecular epidemiology is both the
molecular, the use of the techniques of molecular biology
to characterize nucleic acid- or amino acid-based content,
and the epidemiology, the study of the distribution and
determinants of disease occurrence in human populations.

a
t

I
n
s
t
i
t
u
t
e

o
f

T
r
o
p
i
c
a
l

M
e
d
i
c
i
n
e

-

T
h
e

L
i
b
r
a
r
y

o
n

D
e
c
e
m
b
e
r

1
3
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

1136 Foxman and Riley
Am J Epidemiol Vol. 153, No. 12, 2001
Molecular techniques may be applied to the measurement
of host or agent factors and of exposures. When applied to
studies of disease, the resulting enhanced measurement
increases our ability to more reliably detect associations.
Molecular techniques help to stratify and to refine data by
providing more sensitive and specific measurements, which
facilitate epidemiologic activities, including disease sur-
veillance, outbreak investigations, identifying transmission
patterns and risk factors among apparently disparate cases,
characterizing host-pathogen interactions, detecting uncul-
tivatable organisms, providing clues for possible infectious
causes of cancer and other chronic diseases, and providing
better understanding of disease pathogenesis at the molec-
ular level.
MOLECULAR TECHNIQUES
Molecular techniques do not substitute for conventional
methods. They address epidemiologic problems that cannot
be approached or would be more labor intensive, expen-
sive, and/or time consuming to address by conventional
techniques. Todays molecular technique can become
tomorrows conventional diagnostic tool or even consigned
to the wastebasket. For example, plasmid profile analysis
was a mainstay of molecular fingerprinting just a short
while ago and now has been almost entirely replaced by
other techniques.
Acknowledging that any list of molecular techniques will
be outdated from the time it is published, we present in table
2 techniques that have been applied in epidemiologic stud-
ies of infectious disease. They fall into two large categories:
identification and fingerprinting (strain typing). Rather than
describe the techniques themselves in detail, we describe
how the application of some of these techniques has
increased our understanding of the epidemiology of two
important infectious agents: Mycobacterium tuberculosis,
which causes tuberculosis, and uropathogenic Escherichia
coli, which causes urinary tract infection. Tuberculosis is the
most common infectious cause of deaths in adults world-
wide (3), and urinary tract infection is one of the most com-
mon bacterial infections, affecting half of all women (4) and
one seventh of all men at least once during their lifetime (5).
We will use these pathogens to illustrate the distinct
approaches and principles that must be considered when
conducting epidemiologic investigations using molecular
techniques.
TABLE 1. A snapshot of various definitions of molecular epidemiology
Higginson J (37)
Schulte PA (2)
Tompkins LS (38)
McMichael AJ (39)
Groopman JD, Kensler TW,
Links JM (40)
Hall A (41)
Shpilberg O, Dorman JS,
Ferrell RE, et al. (42)
Levin BR, Lipsitch M,
Bonhoeffer S (43)
Am J Pathol 1977;86:46084
In: Schulte PA, Perera FP, eds. San
Diego, CA: Academic Press,
1993:344
In: Miller VL, Kaper JB, Portnoy DA,
et al, eds. Washington, DC:
American Society for Microbiology,
1994:6373
Am J Epidemiol 1994;140:111
Toxicol Lett 1995;82-83:7639
Trop Med Int Health 1996;1:4078
J Clin Epidemiol 1997;50:6338
Science 1999;283:8069
Author(s) and ref. no. Definition Reference
the application of sophisticated techniques to the
epidemiologic study of biological material (p. 463)
molecular epidemiology is the use of biologic markers or
biologic measurements in epidemiologic research
(p. 13)
the application of molecular biology to the study of
infectious disease epidemiology (p. 65)
using molecular biomarkers in epidemiology (p. 5)
molecular epidemiologic research involves the identification
of relations between previous exposure to some
putative causative agent and subsequent biological
effects in a cluster of individuals in populations (p. 763)
the analysis of nucleic acids and proteins in the study of
health and disease determinants in human
populations (p. 407)
molecular epidemiology uses molecular techniques to define
disease and its pre-clinical states, to quantify exposure
and its early biological effect, and to identify the presence
of susceptibility genes (p. 633)
the practical goals of molecular epidemiology are to identify
the microparasites responsible for infectious diseases
and determine their physical sources, their biological
relationships, and their route of transmission and those
of the genes responsible for their virulence, vaccine-
relevant antigens and drug resistance (p. 806)

a
t

I
n
s
t
i
t
u
t
e

o
f

T
r
o
p
i
c
a
l

M
e
d
i
c
i
n
e

-

T
h
e

L
i
b
r
a
r
y

o
n

D
e
c
e
m
b
e
r

1
3
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Molecular Epidemiology of Infection 1137
Am J Epidemiol Vol. 153, No. 12, 2001
MOLECULAR EPIDEMIOLOGY OF TUBERCULOSIS
Tuberculosis is one infectious disease in which molecular
biology techniques have yielded novel information that
would have been difficult if not impossible to obtain by con-
ventional laboratory methods. Several molecular tech-
niques are currently used to subtype M. tuberculosis, the
etiologic agent of tuberculosis. The method that has come to
be accepted as standard is based on a repetitive DNA ele-
ment called insertion sequence 6110 (IS6110) (6). The
strains are typed according to electrophoretic banding pat-
terns generated by strain differences in location in the chro-
mosome and copy numbers of this repetitive DNA element.
In strains with fewer than six copies of IS6110, one or more
of several secondary typing methods are used. The discrim-
inatory power and reproducibility of these secondary typing
methods have been compared and recently reported (7).
The IS6110-based strain-typing method has been used for
a variety of tuberculosis epidemiologic investigations,
including the following: 1) confirmation of an outbreak in
institutional settings, 2) identifying an outbreak in what
appears to be sporadic cases of tuberculosis, 3) identifying
risk factors for recent infections or rapidly progressive dis-
ease, 4) tracking geographic spread of M. tuberculosis
clones of public health importance, and 5) evaluating labo-
ratory cross-contamination with M. tuberculosis.
Confirmation of an outbreak in institutional settings
One of the first epidemiologic applications of the IS6110-
based typing method was performed in a study of a tubercu-
losis outbreak in a housing facility for human immunodefi-
ciency virus-infected persons in San Francisco, California
(8). In this situation, it was clear that there was an outbreak
of tuberculosis even before the M. tuberculosis isolates were
typed. The importance of this study was that the IS6110-
based typing method itself was validated in a recognized
outbreak setting. All patients suspected to be part of an insti-
tutional outbreak were infected with strains that had similar
TABLE 2. Applications of molecular techniques in epidemiologic studies and available techniques as of this writing
Identification
Fingerprinting
Conventional
Nucleic acid based
PCR* based
Protein based
Conventional
Nucleic acid based
PCR based
Protein based
Gene expression
Applications Technique Method
Culture
Enzyme-linked immunosorbent assay (ELISA)
Enzyme immunosorbent assay (EIA)
Monoclonal antibodies
DNA hybridization for known genes
Direct sequencing of one or more regions
Multilocus sequence typing (MLST)
Amplification of a single target specific to a pathogen
Ligase chain reaction (LCR)
Western blot or immunoblotting
Serotype
Antibiotic susceptibilities
Plasmid profiles
Restriction fragment length polymorphism (RFLP)
Pulsed field gel electrophoresis (PFGE)
Segmented RNA gel electrophoresis
Ribosomal RNA gel electrophoresis
Direct sequencing of one or more regions
Multilocus sequence typing (MLST)
Amplification of a single target specific to a pathogen
Targeting known repetitive sequences (enterobacterial repetitive intergenic
consensus sequences (ERIC), repetitive extragenic palindromic sequences
(REP), double repetitive element (DRE), BOX, insertional sequence (IS),
polymorphic guanine/cytosine-rich repetitive sequences (PGRS))
Random primers (randomly amplified polymorphic DNA (RAPD), arbitrary primed
PCR (AP-PCR))
Restriction endonuclease of a single amplified product
Amplified fragment length polymorphism (AFLP)
Multilocus enzyme electrophoresis (MLEE)
Reverse transcriptase PCR
Microarray technologies
* PCR, polymerase chain reaction.

a
t

I
n
s
t
i
t
u
t
e

o
f

T
r
o
p
i
c
a
l

M
e
d
i
c
i
n
e

-

T
h
e

L
i
b
r
a
r
y

o
n

D
e
c
e
m
b
e
r

1
3
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

1138 Foxman and Riley
Am J Epidemiol Vol. 153, No. 12, 2001
IS6110 patterns, while those unrelated to this cluster of
tuberculosis cases did not have these patterns. Therefore, the
relatedness of the M. tuberculosis isolates in an outbreak
setting was not determined by the similarity of the elec-
trophoretic banding patterns generated by the IS6110 typing
method but by the knowledge that the patients with tubercu-
losis were part of a recognized outbreak. Outbreaks provide
one of the best opportunities to validate a new molecular
strain-typing method for epidemiologic application. Once
validated, the technique can then be applied to other epi-
demiologic investigations, as outlined below.
Identifying an outbreak in what appears to be sporadic
cases of tuberculosis
Tuberculosis in most communities occurs in a typical
endemic pattern, without obvious clustering in time or
place. In New York City in the early 1990s, there were sev-
eral institutional outbreaks of multidrug-resistant tuberculo-
sis that were traced to a single strain with a characteristic
IS6110 banding pattern designated the W strain (9). This
multidrug-resistant strain had an atypical antimicrobial
resistance pattern, one of which was resistance to
kanamycin. Hence, resistance to kanamycin served as a rel-
atively easy identifiable and tractable marker for multidrug
resistance for this particular strain of M. tuberculosis in New
York City. However, the relation of drug-susceptible tuber-
culosis cases is more difficult to establish epidemiologically.
Analysis of M. tuberculosis isolates by the IS6110 typing
method in a number of studies helped to identify cases of
tuberculosis belonging to clusters that were not found using
conventional contact-tracing methods (1014).
Identifying risk factors for recent infection or rapidly
progressive disease
Tuberculosis results from either reactivation of an infec-
tion acquired in the remote past or from rapid progression
from an infection acquired recently. It has come to be gen-
erally accepted that cases of tuberculosis caused by M.
tuberculosis strains with identical IS6110 banding patterns
isolated from two or more persons (cluster patterns) repre-
sent recent exogenous infection, while those infected with
strains with patterns that are not observed among any other
clinical isolate (unique patterns) from that community are
considered to represent endogenous reactivation disease
(10, 12, 15). It should be cautioned, however, that the inter-
pretation of IS6110 cluster patterns as representing cases of
tuberculosis arising from recent infection may not always be
appropriate. In highly stable populations, this assumption
may not be valid, as was shown in a study in Arkansas, a
state with largely rural populations (16). A large proportion
of newly diagnosed tuberculosis cases were found to have
isolates with similar IS6110 patterns, and many of these
cases were not epidemiologically linked. Therefore, addi-
tional factors must be taken into consideration to make the
assumption that cluster patterns represent recent infections.
These include information such as the mobility of the study
population, time of arrival and duration of residence of those
who develop tuberculosis in the geographic place of study,
differences in the mean age of cases of tuberculosis infected
with cluster pattern strains versus those infected with unique
pattern strains, and other epidemiologically plausible char-
acteristics supportive of this assumption. Again, it is the epi-
demiologic information that validates the technique, as well
as the interpretation based on the information provided by
the technique.
The ability to differentiate the proportion of tuberculosis
cases in a community due to recent infection versus reacti-
vation has major implications in the assessment of a tuber-
culosis control program in that community. New cases of
tuberculosis that arise from recent infections reflect rates of
current active transmissions in that community: the higher
the incidence, the poorer the tuberculosis control program.
Hence, obtaining information about the incidence of tuber-
culosis due to recent infection, regardless of whether this
occurs in human immunodeficiency virus-infected or unin-
fected populations, is an important component of tuberculo-
sis control efforts. The IS6110-based strain-typing method
helps to obtain such information.
Once this type of information is obtained, conventional
analytical epidemiologic methods are used to identify risk
factors of tuberculosis due to recent infection. Such studies
are usually performed using a case-control design, where
cases are defined as those who develop tuberculosis from
recent infection and controls are those patients who develop
reactivation tuberculosis. The ability to further stratify
tuberculosis patients by their M. tuberculosis isolates based
on the IS6110 patterns provides an opportunity to refine
case-control analyses. The conventional laboratory methods
are often not discriminating enough to provide the data strat-
ification needed to conduct such an analysis. Case-control
studies made possible by the IS6110 typing analysis of M.
tuberculosis isolates have identified a variety of risk factors
associated with recent infection. Many of these factors are
shared among tuberculosis patients in different communities
(young age, ethnic minorities, homelessness, acquired
immunodeficiency syndrome), but some are specific to a
particular community (10, 12, 14, 15, 1719) .
Tracking geographic spread of M. tuberculosis clones
of public health importance
The comparison of the standardized IS6110 pattern data-
bases from different geographic areas makes it possible to
track M. tuberculosis strains considered to be of major pub-
lic health importance. It also helps to evaluate the ability of
certain strains of M. tuberculosis to spread in a community.
For example, a search of databases was performed to track
the spread of the highly multidrug-resistant strain of M.
tuberculosis from New York City designated as the W
strain (9). This strain has been identified from tuberculosis
patients in Florida, Nevada, Georgia, Colorado, and France.
The patients in Denver and France were previous residents
of New York City. In Denver, possible secondary transmis-
sions were documented, where the index case spread the
infection to at least two household contacts and several
health care facility workers. The so-called W strains in
New York City belong to a clade of M. tuberculosis called

a
t

I
n
s
t
i
t
u
t
e

o
f

T
r
o
p
i
c
a
l

M
e
d
i
c
i
n
e

-

T
h
e

L
i
b
r
a
r
y

o
n

D
e
c
e
m
b
e
r

1
3
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Molecular Epidemiology of Infection 1139
Am J Epidemiol Vol. 153, No. 12, 2001
the Beijing family (20). Members of this clade have been
recently shown to be responsible for large epidemics of mul-
tidrug-resistant tuberculosis in Russia.
Evaluating laboratory cross-contamination with M.
tuberculosis
One of the common problems that hospital epidemiolo-
gists must deal with is to assess whether a cluster of infec-
tions that arises in a nosocomial setting represents a true out-
break or a pseudo outbreak due to laboratory contamination.
This is an epidemiologic issue that is not easily addressed by
conventional epidemiologic methods. The application of the
IS6110-based analysis has proven to be quite useful in con-
firming cross-contamination in clinical mycobacteriology
laboratories. In two reports, investigations of clusters of M.
tuberculosis cultures associated with patients without clinical
suspicion of tuberculosis led to a conclusion that laboratory
contamination had occurred (21, 22).
Summary
The molecular epidemiologic approach to studying tuber-
culosis epidemiology has identified several new observa-
tions that could not have been obtained by conventional
epidemiologic or laboratory approaches. We now know that,
in places such as the United States, new cases of tuberculo-
sis resulting from recent infections are more common than
previously believed, and that in persons with acquired
immunodeficiency syndrome, greater than 50 percent of
them develop the disease from new infections. The ability to
distinguish tuberculosis due to recent infection from reacti-
vation tuberculosis helps to assess the current rates of active
transmission of tuberculosis in a community and may guide
appropriate tuberculosis control efforts in such a commu-
nity. The application of molecular epidemiologic methods to
the study of tuberculosis can be generalized to other infec-
tious diseases that clearly occur as outbreaks.
MOLECULAR EPIDEMIOLOGY OF E. COLI URINARY
TRACT INFECTION
Urinary tract infection is a common, frequently recurring
condition that affects almost 11 million women annually (4)
and is a major cause of nosocomial infection (23). In con-
trast with tuberculosis, urinary tract infection is not usually
thought of as a disease that occurs in outbreaks. We are
aware of only two reports of E. coli urinary tract infection
outbreaks outside hospital settings (24, 25). Both outbreaks
were detected because of the unique phenotype of the agent
combined with a particularly severe clinical presentation.
Probably many such outbreaks occur, but they are not easily
detected. Detection is difficult, because the background rate
of urinary tract infection is so high, and because many dif-
ferent organisms can cause urinary tract infection. Further,
the bacteria that most commonly cause urinary tract infec-
tion, E. coli, a usual bowel inhabitant, are extremely hetero-
geneous. For multiagent syndromes such as urinary tract
infection or diseases caused by heterogeneous species like
E. coli, molecular methods are essential for understanding
the epidemiology.
The types of questions that are addressed by molecular
techniques for diseases like urinary tract infection are dif-
ferent from those for tuberculosis. Questions pertinent for
urinary tract infection include the following. 1) Can any E.
coli cause urinary tract infection (i.e., is urinary tract infec-
tion a result of fecal contamination)? 2) In the absence of
obvious outbreaks, how are uropathogenic E. coli transmit-
ted between persons? 3) Does the great variety of uropatho-
genic E. coli reflect multiple pathogenic mechanisms, the
horizontal gene transfer into different strains of E. coli, or
both?
Can any E. coli cause urinary tract infection?
E. coli (a normal bowel inhabitant) cause a number of
human diseases and are the most common cause of urinary
tract infection (26). Although a single species when classi-
fied by biochemical tests, E. coli are quite heterogeneous.
The genome ranges from 4.5 to 5.5 megabases (27, 28), with
sequence homology as low as 70 percent. In comparison,
human and chimpanzee genomes have 98 percent sequence
homology. E. coli cause the vast majority of urinary tract
infections. Given the heterogeneity of E. coli, and that E.
coli are a normal bowel inhabitant, it is reasonable to won-
der if any bacteria can cause urinary tract infection in an oth-
erwise healthy individual. This question cannot be answered
without molecular techniques.
To address this question, fecal E. coli from normal,
healthy individuals have been compared with E. coli iso-
lated from persons with kidney and bladder infections, by a
variety of different molecular methods (29, 30). All com-
parisons showed that uropathogenic E. coli share certain
characteristics and that these characteristics differentiate
them from fecal E. coli, implying that urinary tract infection
in normal, healthy individuals is not merely the result of
fecal contamination.
Transmission of uropathogenic E. coli
Because E. coli are found in the bowel flora, it has been
presumed that uropathogenic E. coli are transmitted by the
fecal-oral route. Epidemiologic evidence suggests that, at
least in sexually active young women, both the frequency of
vaginal intercourse and the duration of sexual partnership are
important predictors of urinary tract infection acquisition.
While this does not preclude fecal-oral transmission, it does
suggest that sexual transmission may be possible. This obser-
vation led us to search for urethral colonization with E. coli in
the male sex partners of women with urinary tract infection
(31). Urinary, vaginal, and fecal E. coli isolates from 19
women with urinary tract infection were compared with E.
coli found in random initial voids from their most recent male
sex partner. E. coli were isolated from four of 19 male sex
partners. In each case, the E. coli isolated from the male were
identical by pulsed-field gel electrophoresis and bacterial vir-
ulence profile to the urinary E. coli from his sex partner, sug-
gesting that sexual activity may lead to transmission. Without

a
t

I
n
s
t
i
t
u
t
e

o
f

T
r
o
p
i
c
a
l

M
e
d
i
c
i
n
e

-

T
h
e

L
i
b
r
a
r
y

o
n

D
e
c
e
m
b
e
r

1
3
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

1140 Foxman and Riley
Am J Epidemiol Vol. 153, No. 12, 2001
molecular techniques, in this case pulsed-field gel elec-
trophoresis, we would have been unable to determine whether
the E. coli isolated from males were identical to those causing
urinary tract infection in their sex partner.
Diversity of uropathogenic E. coli
The genomes of bacteria can be quite plastic. In addition
to adding and deleting genes on the chromosome, bacteria
can change phenotype through the gain or loss of plasmids.
Antibiotic resistance, for example, can be transferred via
plasmid and has been demonstrated to be transmitted across
bacterial species. It may be that at least some uropathogenic
characteristics are acquired in this fashion. In this scenario
there could be heterogeneity in the organism structure
despite using a common method of pathogenesis. An out-
break would follow the path of the plasmid (horizontal and
vertical gene transfer) rather than the path of a particular
bacterial clone. The identification of uropathogenic factors
and their mode of transmission between pathogens would
greatly assist our understanding of urinary tract infection
epidemiology and pathogenesis and our ability to prevent
disease via vaccination or other strategies. These types of
studies require epidemiologic methods to collect appropriate
sample isolates from well-defined populations and to make
appropriate inferences about the findings based on labora-
tory analyses.
As an example, we studied a collection of first-episode
urinary tract infection isolates from epidemiologically well-
characterized women and grouped them by the presence or
absence of nine putative virulence genes. Although the num-
bers were small, we detected a time-space cluster in one
such grouping. Pulsed-field gel electrophoresis analysis
showed an apparent clonal group. Only a small proportion
of fecal E coli with the same combination of virulence genes
had the same pulsed-field gel electrophoresis pattern. We
reasoned that, for genomic subtraction, choosing one isolate
from the apparent cluster and the second isolate from the
fecal isolates would increase our possibility of detecting uri-
nary tract infection virulence factors (32). This experiment
resulted in the identification of 37 DNA sequences physi-
cally located all over the genome of the chosen urinary tract
infection isolate.
Summary
The application of molecular techniques to the study of
heterogeneous organisms enhances epidemiologic studies
by improving our ability to subclassify these organisms
into meaningful groups. This facilitates the detection of
disease outbreaks that may otherwise be undetected and
allows the epidemiologist to identify risk factors of out-
breaks, sporadic cases, or both. In the case of tuberculosis,
epidemiologic information helped to validate the molecu-
lar techniques, which were, in turn, applied to further char-
acterize the epidemiology of tuberculosis. In the case of E.
coli, both epidemiologic information and molecular labo-
ratory information have to be analyzed simultaneously to
characterize the epidemiology of urinary tract infection.
With E. coli, collections resulting from population-based
epidemiologic studies can assist the molecular biologist in
identifying groups for studying pathogenesis and in mak-
ing inferences about the potential role of newly identified
genes. These results, in turn, can be used to further char-
acterize the epidemiology.
A LOOK TO THE FUTURE
We touched here upon the applications of molecular tech-
niques to the study of the epidemiology of infectious agents,
focusing on the pathogens themselves. We did not touch on
the other side of the host-pathogen relation, the role of the
host in disease susceptibility and resistance. Recent studies
have revealed an association of a number of candidate genes
with tuberculosis, although such genes comprise a small
attributable fraction of all those who develop tuberculosis
(33). There is also some suggestion that chronic recurring
urinary tract infection (four or more episodes in a 12-month
period) may have a genetic component: women who are
nonsecretors of blood group antigens are more likely to have
recurring urinary tract infections (34), and they are also
more susceptible than secretors to colonization with E. coli
with uropathogenic potential (35). Studies of host genetic
susceptibility to infection are indeed part of the molecular
epidemiology discipline and are reviewed in more detail
elsewhere (36).
There is a clear need to train individuals who are capable
of developing new theories and methods to address the
questions that will arise from the interface of molecular
biology and epidemiology. Most of the molecular tech-
niques listed in table 2 can be mastered in a short period of
time. The more time-consuming part of this type of training
is the epidemiology. By definition, molecular epidemiology
training requires practical application of both the laboratory
and epidemiologic techniques to address a real-world infec-
tious disease problem. The molecular epidemiologist will
need to interface with clinicians, statisticians, epidemiolo-
gists, molecular biologists, computer scientists, engineers,
and practitioners in the new field of bioinformatics and
computational biology. Although we will never conquer
infectious diseases, we can certainly learn to live in greater
harmony with them. This may perhaps be our most effective
intervention strategy. Discovering how to do so will be the
great challenge for molecular epidemiologists of the future.
REFERENCES
1. Ambrose CB, Kadlubar FF. Toward an integrated approach to
molecular epidemiology. Am J Epidemiol 1997;146:91218.
2. Schulte PA, Perera FP, eds. Molecular epidemiology: princi-
ples and practices. San Diego, CA: Academic Press, 1993.
3. Raviglione MC, Snider DE, Kochi A. Global epidemiology of
tuberculosis. JAMA 1995;273:2206.
4. Foxman B, Barlow R, dArcy H, et al. Urinary tract infection:
self-reported incidence and associated costs. Ann Epidemol
2000;10:50915.
5. Zielske JV, Lohr KN, Brook RH, et al. Conceptualization and
measurement of physiologic health for adults. Vol 16. Urinary

a
t

I
n
s
t
i
t
u
t
e

o
f

T
r
o
p
i
c
a
l

M
e
d
i
c
i
n
e

-

T
h
e

L
i
b
r
a
r
y

o
n

D
e
c
e
m
b
e
r

1
3
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Molecular Epidemiology of Infection 1141
Am J Epidemiol Vol. 153, No. 12, 2001
tract infection. Arlington, VA: The Rand Corporation, 1981.
(Document no. R2262/16-HHS).
6. van Embden JD, Cave MD, Crawford JT, et al. Strain identifi-
cation of Mycobacterium tuberculosis by DNA fingerprinting:
recommendations for a standardized methodology. J Clin
Microbiol 1993;31:4069.
7. Kremer K, van Sooligen D, Frothingham R, et al. Comparison
of methods based on different molecular epidemiological
markers for typing of M. tuberculosis complex strains: inter-
laboratory study of discriminatory power and reproducibility. J
Clin Microbiol 1999;37:260718.
8. Daley CL, Small PM, Schecter GF, et al. An outbreak of tuber-
culosis with accelerated progression among persons infected
with the human immunodeficiency virus. N Engl J Med 1992;
326:2315.
9. Bifani PJ, Plikaytis BB, Kapur V, et al. Origin and interstate
spread of a New York City multidrug-resistant Mycobacterium
tuberculosis clone family. JAMA 1996;275:4527.
10. Small PM, Hopewell PC, Singh SP, et al. The epidemiology of
tuberculosis in San Francisco: a population-based study using
conventional and molecular methods. N Engl J Med 1994;330:
17039.
11. Tabet SR, Goldbaum GM, Hooton TM, et al. Restriction frag-
ment length polymorphism analysis detecting a community-
based tuberculosis outbreak among persons infected with human
immunodeficiency virus. J Infect Dis 1994;169:18992.
12. Friedman CR, Stoeckle MY, Kreiswirth BN, et al.
Transmission of multidrug-resistant tuberculosis in a large
urban setting. Am J Respir Crit Care Med 1995;152:3559.
13. Barnes PF, Yang Z, Preston-Martin S, et al. Patterns of tuber-
culosis transmission in Central Los Angeles. JAMA1997;278:
115963.
14. Van Deutekom H, Gerritsen JJ, van Soolingen D, et al. Amol-
ecular epidemiological approach to studying the transmision of
tuberculosis in Amsterdam. Clin Infect Dis 1997;25:10717.
15. Alland D, Kalkut GE, Moss AR, et al. Transmission of tuber-
culosis in New York City: an analysis by DNA fingerprinting
and conventional epidemiologic methods. N Engl J Med 1994;
330:171016.
16. Braden CR, Templeton GL, Cave DM, et al. Interpretation of
restriction fragment length polymorphism analysis of
Mycobacterium tuberculosis isolates from a state with a large
rural population. J Infect Dis 1997;175:144652.
17. Friedman CR, Quinn GC, Kreiswirth BN, et al. Widespread
dissemination of a drug-susceptible strain of Mycobacterium
tuberculosis. J Infect Dis 1997;176:47884.
18. Tornieporth NG, Ptachewich Y, Poltoratskaia N, et al.
Tuberculosis among foreign-born persons in New York City,
199294: implications for tuberculosis control. Int J Tuberc
Lung Dis 1997;1:52835.
19. Bauer J, Yang Z, Poulsen S, et al. Results from 5 years of
nationwide DNA fingerprinting of Mycobacterium tuberculo-
sis complex isolates in a country with a low incidence of M.
tuberculosis infection. J Clin Microbiol 1998;36:3058.
20. Van Soolingen D, Qian L, De Haas PE, et al. Predominance of
a single genotype of Mycobacterium tuberculosis in countries
of east Asia. J Clin Microbiol 1995;33:32348.
21. Small PM, McClenny NB, Singh SP, et al. Molecular strain
typing of Mycobacterium tuberculosis to confirm cross-
contamination in the mycobacteriology laboratory and modifi-
cation of procedures to minimize occurrence of false-positive
cultures. J Clin Microbiol 1993;31:167782.
22. Dunlap NE, Harris RH, Benjamin WH, et al. Laboratory con-
tamination with Mycobacterium tuberculosis cultures. Am J
Crit Care Med 1995;152:17024.
23. Sedor J, Mulholland SG. Hospital-acquired urinary tract infec-
tions associated with the indwelling catheter. Urol Clin North
Am 1999;26:8218.
24. Riley PA, Threlfall EJ, Cheasty T, et al. Occurrence of Fime
plasmids in multiply antimicrobial-resistant Escherichia coli
isolated from urinary tract infection. Epidemiol Infect 1993;
110:45968.
25. Olesen B, Komos HJ, Orskov F, et al. Cluster of multiresistant
Escherichia coli O78:H10 in Greater Copenhagen. Scand J
Infect Dis 1994;26:40610.
26. Kunin CM. Detection, prevention and management of urinary
tract infections. 5th ed. Baltimore, MD: Williams & Wilkins,
1997.
27. Arthur M, Johnson CE, Rubin RH, et al. Distribution of chro-
mosome length variation in natural isolates of Escherichia
coli. Mol Biol Evol 1998;15:616.
28. Blattner FR, Plunkett G, Bloch CA, et al. The complete
genome sequence of Escherichia coli K-12. Science 1997;277:
145374.
29. Johnson JR. Virulence factors in Escherichia coli urinary tract
infection. Clin Microbiol Rev 1991;4:80128.
30. Caugnant DA, Levin BR, Lidin-Janson G, et al. Genetic diver-
sity and relationships among strains of Escherichia coli in the
intestine and those causing urinary tract infections. Prog
Allergy 1983;33:20327.
31. Foxman B, Zhang L, Tallman P, et al. Transmission of
uropathogens between sex partners. J Infect Dis 1997;175:
98992.
32. Zhang L, Foxman B, Manning SD, et al. Molecular epidemio-
logic approaches to urinary tract infection gene discovery in
uropathogenic Escherichia coli. Infect Immun 2000;68:200915.
33. Abel L, Casanova JL. Genetic predisposition to clinical tuber-
culosis: bridging the gap between simple and complex inheri-
tance. Am J Hum Genet 2000;67:2747.
34. Hooton TM, Roberts PL, Stamm WE. Effects of recent sexual
activity and use of a diaphragm on the vaginal microflora. Clin
Infect Dis 1994;19:2748.
35. Stapleton A, Hooton TM, Fennell C, et al. Effect of secretor
status on vaginal and rectal colonization with fimbriated
Escherichia coli in women with and without recurrent urinary
tract infection. J Infect Dis 1995;171:71720.
36. Harrison LH, Griffin DE. Infectious diseases. In: Schulte PA,
Perera FP, eds. Molecular epidemiology: principles and prac-
tices. San Diego, CA: Academic Press, 1993:30139.
37. Higginson J. The role of the pathologist in environmental med-
icine and public health. Am J Pathol 1977;86:46084.
38. Tompkins LS. Molecular epidemiology: development and
application of molecular methods to solve infectious disease
mysteries. In: Miller VL, Kaper JB, Portnoy DA, et al, eds.
Molecular genetics of bacterial pathogenesis: a tribute to
Stanley Falkow. Part 1. Retrospective look at early advances.
Washington, DC: American Society for Microbiology, 1994:
6373.
39. McMichael AJ. Invited commentarymolecular epidemiol-
ogy: new pathway or new travelling companion? Am J
Epidemiol 1994;140:111.
40. Groopman JD, Kensler TW, Links JM. Molecular epidemiology
and human risk monitoring. Toxicol Lett 1995;82-83:7639.
41. Hall A. What is molecular epidemiology? (Editorial). Trop
Med Int Health 1996;1:4078.
42. Shpilberg O, Dorman JS, Ferrell RE, et al. The next stage:
molecular epidemiology? J Clin Epidemiol 1997;50:6338.
43. Levin BR, Lipsitch M, Bonhoeffer S. Population biology, evo-
lution, and infectious disease: convergence and synthesis.
Science 1999;283:8069.

a
t

I
n
s
t
i
t
u
t
e

o
f

T
r
o
p
i
c
a
l

M
e
d
i
c
i
n
e

-

T
h
e

L
i
b
r
a
r
y

o
n

D
e
c
e
m
b
e
r

1
3
,

2
0
1
2
h
t
t
p
:
/
/
a
j
e
.
o
x
f
o
r
d
j
o
u
r
n
a
l
s
.
o
r
g
/
D
o
w
n
l
o
a
d
e
d

f
r
o
m