Biosensors and Bioelectronics 66 (2015) 198–207

Contents lists available at ScienceDirect

Biosensors and Bioelectronics

journal homepage: www.elsevier.com/locate/bios

A sensitive and selective magnetic graphene composite-modified

polycrystalline-silicon nanowire field-effect transistor for bladder
cancer diagnosis
Hsiao-Chien Chen a,b,c,1, Yi-Ting Chen d,1, Rung-Ywan Tsai e, Min-Cheng Chen f,
Shi-Liang Chen a,b,g, Min-Cong Xiao a,b,g, Chien-Lun Chen h, Mu-Yi Hua a,b,g,n
a
Department of Chemical and Materials Engineering, Chang Gung University, Taoyuan 33302, Taiwan, ROC
b
Biosensor Group, Biomedical Engineering Research Center, Chang Gung University, Taoyuan 33302, Taiwan, ROC
c
Department of Biochemistry and Molecular Cell Biology, School of Medicine, College of Medicine, Taipei Medical University, Taipei 11031, Taiwan, ROC
d
Department of Biomedical Sciences, College of Medicine, Chang Gung University, Taoyuan 33302, Taiwan, ROC
e
Electronics and Optoelectronics Research Laboratories, Industrial Technology Research Institute, Hsinchu 310, Taiwan, ROC
f
National Nano Device Laboratories, Hsinchu Science Park, Hsinchu 31040, Taiwan, ROC
g
Green Technology Research Center, Chang Gung University, Taoyuan 33302, Taiwan, ROC
h
Department of Urology, Chang Gung University College of Medicine and Memorial Hospital, Taoyuan 33305, Taiwan, ROC

art ic l e i nf o a b s t r a c t

Article history: In this study, we describe the urinary quantification of apolipoprotein A II protein (APOA2 protein), a
Received 14 September 2014 biomarker for the diagnosis of bladder cancer, using an n-type polycrystalline silicon nanowire field-
Received in revised form effect transistor (poly-SiNW-FET). The modification of poly-SiNW-FET by magnetic graphene with long-
12 November 2014
chain acid groups (MGLA) synthesized via Friedel–Crafts acylation was compared with that obtained
Accepted 16 November 2014
Available online 18 November 2014
using short-chain acid groups (MGSA). Compared with MGSA, the MGLA showed a higher immobiliza-
tion degree and bioactivity to the anti-APOA2 antibody (Ab) due to its lower steric hindrance. In addition,
Keywords: the magnetic properties enabled rapid separation and purification during Ab immobilization, ultimately
Graphene preserving its bioactivity. The Ab-MGLA/poly-SiNW-FET exhibited a linear dependence of relative re-
FET
sponse to the logarithmical concentration in a range between 19.5 pg mL
1 and 1.95 mg mL
1, with a
Bladder cancer
limit of detection (LOD) of 6.7 pg mL
1. An additional washing step before measurement aimed at ex-
Biosensor
Apolipoprotein A II protein cluding the interfering biocomponents ensured the reliability of the assay. We conclude that our bio-
sensor efficiently distinguishes mean values of urinary APOA2 protein concentrations between patients
with bladder cancer (29–344 ng mL
1) and those with hernia (0.425–9.47 ng mL
1).
& 2014 Elsevier B.V. All rights reserved.

1. Introduction or cytopathologist. Diagnostic approaches include cystoscopy,

which is invasive and therefore painful, and urine cytology, which
Among men and women in the Western world, bladder cancer is not sufficiently sensitive for the detection of low-grade stages
is the fourth and eighth most common malignancy, respectively. (Araki et al., 2007).
Bladder cancer occurs in the epithelial lining of the bladder (Kir- A good biomarker reflects various conditions of a biological
kali et al., 2005). Non-muscle-invasive bladder cancer occurs in system and enables better disease diagnosis and/or treatment
approximately 70% of bladder cancer incidents. High-grade stages outcomes. The biomedical field is highly interested in the devel-
of muscle-invasive bladder cancer are associated with significant opment of disease biomarkers, where the biomarker development
tumor progression and, consequently, increased mortality. There- pipeline generally comprises three stages, which include bio-
fore, to determine the optimal treatment, it is critical to accurately marker discovery, verification, and clinical validation (Surinova
detect bladder cancer at an early stage. Traditionally, diagnosis is a et al., 2011). The process usually begins with biomarker discovery
complex process based on the experience of a urologist, radiologist based on a limited number of cases and uses deep profiling, which
can achieve the identification of thousands of proteins in a bio-
n
logical sample. Prior to the application of potential biomarker
Corresponding author at: Department of Chemical and Materials Engineering,
Chang Gung University, Taoyuan 33302, Taiwan, ROC. Fax: þ886 3 2118668.
candidates in translational medicine, the candidates must be
E-mail address: huamy@mail.cgu.edu.tw (M.-Y. Hua). evaluated in 100–1000 samples in the verification and validation
1
H.-C. Chen and Y.-T. Chen contributed equally to this work. phases using high-throughput, sensitive assays. A variety of

http://dx.doi.org/10.1016/j.bios.2014.11.019
0956-5663/& 2014 Elsevier B.V. All rights reserved.
 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207 199

diagnostic tests have been investigated for the detection of blad- other charged species. Finally, individual urine samples from age-
der cancer. Valid, reliable and inexpensive tests that can be easily matched hernia patients and bladder cancer patients were ana-
and rapidly performed include Raman spectroscopy for diagnosing lyzed using the proposed biosensor; the results clearly indicated
urothelial carcinoma cell (Shapiro et al., 2011), elastic light scat- that the biosensor can distinguish among these two sample types.
tering for detecting papillary transitional cell carcinoma (Mourant These results demonstrate that bladder cancer diagnosis is feasible
et al., 1995), polymerase chain reaction for telomerase detection using a modified poly-SiNW-FET device.
(Kim et al., 2013), enzyme-linked immunosorbent assay (ELISA) for
MUC1 detection (Ferreira et al., 2008), silicon microring resonators
for monitoring fibroblast growth factor receptor 3 and Harvey RAS, 2. Experiments
nuclear magnetic resonance for measurements of leucine, tyr-
osine, lactate, glycine and citrate (Cao et al., 2012), electro- 2.1. Patients and materials
chemistry for H2O2 detection (Hua et al., 2011a; Roberts et al.,
2013), chromatography for detecting carnitine C9:1 (Kałuzna- Clinical specimens were collected using a previously described
Czaplinska and Jóźwik, 2014), and surface plasmon resonance for protocol (Chen et al., 2010, 2012). Briefly, first-morning urine
Her-2 determination (Tai et al., 2007). Additionally, a variety of samples were collected from hernia patient controls and bladder
candidate bladder cancer biomarkers such as RT112 cell CDH1, cancer patients into containers that contained a protease inhibitor
FHIT, LAMC2, RASSF1A, TIMP3, SFRP1, SOX9, PMF1 and RUNX3 cocktail tablet (one tablet per 50 mL of urine; Roche, Mannheim,
have been identified but require further validation (Roberts et al., Germany) and sodium azide (1 mM). As controls, first-morning
2013; Shin et al., 2013; Kandimalla et al., 2013). Apolipoprotein urine samples were collected from hernia patients of comparable
A-IIprotein (APOA2) was recently identified as a new biomarker age using an identical procedure after admission and before sur-
that occurs a highly elevated rates in pooled bladder cancer. The gical intervention. The collected urine samples were centrifuged at
relative difference in the level of APOA2 protein in urine is a highly 5000  g for 30 min at 4 °C within 5 h of collection to remove cells
significant factor (Chen et al., 2010, 2012). and debris, and the clarified supernatants were stored at
80 °C
The use of field-effect transistors (FETs) to diagnose bladder until further processing. All clinical samples were collected from
cancer from urine specimens has not been reported. Due to the the Department of Urology, Chang Gung Memorial Hospital,
associated advantages of miniaturization and superior sensitivity, Taoyuan, Taiwan. This study was approved by the Ethics Com-
significant effort has been expended toward the development of mittee of Chang Gung Memorial Hospital (IRB nos. 102-3232A3
FET biosensors in recent years (Kwon et al., 2012; Lee et al., 2009). and 99-2188B). All subjects received a description of the study and
High selectivity is achieved by immobilizing antibodies, peptides, provided written informed consent. APOA2 and anti-APOA2 were
and oligonucleotide-based aptamers to bind specific molecules. In obtained from Abcam, CB, UK. Maleic anhydride (MA) was pur-
this configuration, the binding analyte affects the channel con- chased from TCI. N-hydroxysulfo-succinimide sodium salt (sulfo-
ductivity in a manner that is similar to the effect of voltage ap- NHS) was purchased from Fluka. PBS, EDC, goat anti-human IgG
plication to a metallic gate electrode. However, earlier FETs with (Fc specific)-peroxidase antibody, BSA and MES buffer (pH 6.3)
planar or two-dimensional surfaces exhibited low sensitivity, were purchased from Sigma. Acetone, methanol, and 1-methyl-2-
limiting their application, particularly for trace analytes. One-di- pyrrolidone (NMP) were purchased from Tedia. Aluminum chlor-
mensional FET channels with nanostructures of nanowires (Li ide and TBO were purchased from Acros, and graphite was pur-
et al., 2013; Chen et al., 2011), graphene (Trung et al., 2014), na- chased from Alfa Aesar. NaOH, FeCl2 and FeCl3 were purchased
nobelts (Oh et al., 2013) or nanotubes (Chen et al., 2014) are ex- from Merck.
tremely attractive due to the depletion or accumulation of carriers
in the bulk of the nanoscale structures. 2.2. Syntheses of GLA and GSA
Magnetic and gold nanoparticles are loaded in the channels to
increase the surface area, thereby increasing the quantity of im- A 50 mg quantity of graphite as temporarily dispersed in 10 mL
mobilized biomolecules (Chen et al., 2014; Allen et al., 2007). of anhydrous NMP with sonication. One gram of MA was dissolved
Graphene sheets (Gs) are a potential candidate for the modifica- in 40 mL of anhydrous NMP solution under nitrogen, and AlCl3
tion of SiNW-FETs due to the inherently high surface area and was then added at a molar ratio of 1:1, 1:3, or 1:6 at 90 °C. After
superior conductivity (Sykes and Charles, 2009). However, Gs lack stirring for 3 h, the graphite solution was added to the three
functional groups for the immobilization of biomolecules, thus it is MA–AlCl3 solutions and reacted at 160 °C for 48 h. The samples
necessary to chemically modify Gs. The best-known functionali- were filtered through 0.1-μm PVDF membranes and washed with
zation of Gs is Hummers' chemical modification, in which treat- methanol and DI water 3 times. The yellow solutions were col-
ment with a strong acid and an oxidant is used to produce various lected after centrifugation at 3000 rpm. Finally, the products were
oxygen-containing groups, including short-chain carboxylic acids obtained from the filter cake by filtration through a 0.1-μm PVDF
(Liu et al., 2008). The carboxylic acid group can subsequently form membrane, and the products were labeled GLA11, GLA13, and
covalent bonds with amine-containing biomolecules via the EDC/ GLA16. GSA was synthesized from expandable graphite flake using
sulfo-NHS reaction. However, fewer investigations have focused on a method modified from Hummer (Liu et al., 2008).
the effects of chain length on the degree of biomolecule loading
and the associated role of steric hindrance. To clarify the effect of 2.3. Syntheses of MGLA and MGSA
the carboxylic acid chain length, Gs with a four-carbon-chain
carboxylic acid group were synthesized via Friedel–Crafts acyla- A 200-mg quantity of GLA was dispersed in 20 mL of DI water.
tion and compared with the product of Hummers' chemical FeCl3 (4.32 mmol) and FeCl2  4H2O (6.48 mmol) were dissolved in
modification in this study. Additionally, magnetic nanoparticles of 380 mL of DI water at room temperature and mixed with the GLA
Fe3O4 were synthesized in situ on the carboxylated Gs to enable solution under N2 gas. After heating this solution slowly to 50 °C,
rapid purification. After immobilizing anti-APOA2 on the magnetic and 30 mL of NaOH (0.576 N) was slowly added over 20 min, re-
carboxylated Gs and loading on the poly-SiNW-FET channel, a sulting in a final temperature of 80 °C. The reaction was then ra-
novel bladder cancer biosensor was developed based on a poly- pidly quenched on ice, and 0.1 N HCl was slowly added until a
SiNW-FET device. Performing an additional washing process en- neutral pH was obtained. MGLA was separated from the solution
sured the accuracy of the biosensor measurement by eliminating via the application of a magnetic field and then was washed with
200 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207

DI water several times. MGSA was prepared according to the 2.8. Bio-Plex assay for APOA2 protein quantification in individual
method described above. urine samples

2.4. Preparation of Ab-MGLA and Ab-MGSA The urine level of APOA2 apolipoprotein was determined with
the MILLIPLEX MAP Human Apolipoprotein Panel kit (Millipore,
EDC (50 mg) and sulfo-NHS (60 mg) were dissolved in 5 mL of MA, USA) using the Bio-Plex system (Bio-Rad Laboratories), as
MES in the dark. A 0.2-mL aliquot of this solution was mixed with previously reported (Chen et al., 2013a). The assay procedure was a
0.1 mL of MGLA (10 mg mL
1) at 25 °C with shaking for 30 min in modification of the blood sample-suitable protocol provided by
the dark. After separation by the application of a magnetic field, Millipore. The immunobeads were analyzed using the Bio-Plex
the MGLA was washed with 0.8 mL of MES, re-suspended in 200 system (Bio-Rad Laboratories). Standard curves and analyte
0.2 mL of MES, and mixed with 10, 25, 50, 100, 150 or 200 ng of concentrations were obtained using Bio-Plex Manager software
anti-APOA2 at 25 °C for 3 h. Ab-MGLA was separated from the version 4.2 (Bio-Rad Laboratories). In total, 20 urine samples (10
solution and washed with PBS to remove free anti-APOA2. Ab- from hernia patients, 10 from patients with bladder cancer) were
MGSA was prepared as described above, except that 0.1 mL of analyzed using the Human Apolipoprotein Kit, which is commer-
MGSA (21.3 mg mL
1) was used. cially available as a 96-well plate immunoassay. The details of
measurements using the Bio-Plex assay are provided in Supporting
2.5. ELISA assay for anti-APOA2 quantification information.

300, 150, 75, 37.5, 18.8, 9.4, 4.7 and 2.3 ng mL
1 of anti-APOA2 2.9. Fabrication of poly-SiNW-FET
(150 μL) were coated on Microlite 2 multiwell plates (Thermo
Labsystems, Franklin, MA) for 1.5 h. After washing with PBS, 150 μL The sample nanowire devices were manufactured on standard
six-inch p-type wafers (Fig. 1A). Initially, a buried oxide/nitride
of BSA (5 mg mL
1) were used as block for 1.5 h before hy-
layers were deposited onto the substrate surface as the gate di-
bridization with 1.95 μg mL
1 of APOA2 protein (150 μL) for 1.5 h.
electric of the nanowire FETs to avoid surface reaction species
A total of 150 μL of BSA (5 mg mL
1) were utilized for further
penetrating to substrate gate. A 50-nm polysilicon layer was then
blocking. 100 μL of biotin-antibody was added to each well and
deposited via chemical vapor deposition. The poly-Si wire was
incubated for 1 h at 37 °C. After removing the solution, 90 μL of
subsequently patterned using a standard I-line stepper from the
TMB substrate was added to each well and incubated for 30 min.
CMOS semiconducting process. The photoresist was trimmed
Finally, 50 μL of stop solution was added to each well and gently
using reactive plasma etching followed by Si etching, and the
tapped the plate to ensure thorough mixing. The optical density at
nanowire dimensions were reduced to approximately 0.3 μm. A
450 nm of each well was measured within 5 min. All processes
channel protection photoresist pattern was then formed via I-line
were performed at 37 °C in a dark room.
lithography. Channel protection patterning was performed both to
prevent the channel from intrinsically implanting the n þ source/
2.6. ELISA assay for APOA2 protein quantification
drain (S/D) and to increase the field sensitivity of the nanowire.
The n þ S/D was subsequently implanted using a 1015 cm
2 P31 þ
10 μg mL
1 Anti-APOA2 (150 μL) were coated on Microlite
ion beam at 10 keV to reduce the parasitic resistance of the na-
2 multiwell plates (Thermo Labsystems, Franklin, MA) for 1.5 h. nowire. Then, the channel protection photoresist was removed.
After washing with PBS, 150 μL of BSA (5 mg mL
1) were used as Finally, the S/D dopant was activated by annealing at 600 °C for
block for 1.5 h before hybridization with 55.7, 24.8, 12.9, 6.5, 3.2, 30 min under a N2 atmosphere. A thick SiNx passivation layer was
1.6 and 0.8 ng mL
1 of APOA2 protein (150 μL) for 1.5 h. A total of deposited onto the wafer to protect the Si substrate gate from
150 μL of BSA (5 mg mL
1) were utilized for further blocking. damage by the solutions used during pH testing. This n-type poly-
100 μL of biotin-antibody was added to each well and incubated SiNW-FET fabrication requires only two additional masks and can
for 1 h at 37 °C. After removing the solution, 90 μL of TMB sub- be integrated with a conventional 0.35 μm CMOS process. The
strate was added to each well and incubated for 30 min. Finally, morphology of poly-SiNW channel is shown in Fig. 1B.
50 μL of stop solution was added to each well and gently tapped
the plate to ensure thorough mixing. The optical density at 450 nm 2.10. Modification of poly-SiNW-FET and fabrication of the biochip
of each well was measured within 5 min. All processes were per-
formed at 37 °C in a dark room. The functional modification of the poly-SiNW-FET with term-
inal amine groups was performed as follows. A 5- μL solution of
2.7. ELISA assay for anti-APOA2 activity using Ab-MGLA and Ab- 2 wt% APTES in ethanol was placed on the poly-SiNW for 1 h to
MGSA form a self-assembled monolayer. The modified chip was then
washed with ethanol and dried in an oven at 100 °C for 1 h. To add
100 μL of Ab-MGLA or Ab-MGSA with solid containing the Ab-MGLA or Ab-MGSA, the chip was further treated with 5%
0.5 mg mL
1 were coated on Microlite 2 multiwell plates (Thermo glutaraldehyde at room temperature for 1 h, resulting in the for-
Labsystems, Franklin, MA) for 1.5 h. After the liquid solution was mation of terminal aldehyde groups for direct immobilization after
removed without washing, 150 μL of BSA (5 mg mL
1) were used the addition of Ab-MGLA or Ab-MGSA to poly-SiNW-FET for 1 h.
as block for 1.5 h before hybridization with 1.95 μg mL
1 of APOA2 Finally, 5 μL of BSA (5 mg mL
1) in PBS was added at 4 °C for 1 h in
protein (150 μL) for 1.5 h. A total of 150 μL of BSA (5 mg mL
1) the dark to block non-specific binding, followed by washing with
were utilized for further blocking. 100 μL of biotin-antibody was PBS three times before use for further detection.
added to each well and incubated for 1 h at 37 °C. After removing
the solution, 90 μL of TMB substrate was added to each well and 2.11. APOA2 protein detection on Ab-MGLA/poly-SiNW-FET and Ab-
incubated for 30 min. Finally, 50 μL of stop solution was added to MGSA/poly-SiNW-FET
each well and gently tapped the plate to ensure thorough mixing.
The optical density at 450 nm of each well was measured within We used the response current of the FET in 0.5 mM PBS as the
5 min. All processes were performed at 37 °C in a dark room and baseline current. The APOA2 protein standard was injected into
under an additional magnetic field. the microfluid channel and reacted for 10 min in the dark at room
 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207 201

Fig. 1. (A) Schematic representation of poly-SiNW-FET device fabrication and (B) SEM image of a typical poly-SiNW-FET device.

temperature. The unreacted APOA2 protein was removed by Friedel–Crafts chemical acylation using AlCl3 as the catalyst to
washing three times, and PBS was re-injected into the microfluid perform maleic anhydride (MA) ring opening. After 48 h of reac-
channel prior to measurement. Similarly, APOA2 detection in in- tion, carboxylated graphene with short-chain acid groups (GSA)
dividual urine from patients suffering from bladder cancer was obtained by either Hummers' method or GLA was well dispersed
performed according to the above reacting, washing and measur- in aqueous solution compared to graphite (Fig. S2, inset), in-
ing processes. The measured volumes of the samples were 5 μL. All dicating the presence of a hydrophilic group (acid group). To de-
measurements were obtained using a 50-mV bias voltage, VB. We termine the optimal reaction conditions, various molar ratios of
chose a current at a constant gate voltage (VG ¼3.0 V) as the cri- MA and AlCl3 were examined. As shown in Fig. S2, GLA13
terion to quantify the APOA2 protein. (MA/AlCl3 is 1/3) exhibited absorption peaks at 226 nm, corre-
sponding to the π–πn transition of the aromatic C ¼C band and the
2.12. Characterization methods n–πn transition of the C¼ O band (Chen et al., 2013b). Additionally,
the absorbance of the broad band extending to 900 nm was higher
To characterize GLA, GSA, MGLA and MGSA, FT-IR spectra were than that observed for the other two conditions. Inadequate cat-
obtained using a Bruker-Tensor 27 spectrometer at a spectral re- alyst led to a lower degree of modification, while excess catalyst
solution of 8 cm
1. The presence of exfoliated Gs was confirmed increased the viscosity of the solution, hindering reactive colli-
by XRD patterns, obtained using a Rigaku D/Max-2B with nickel- sions and thereby also decreasing the degree of modification.
filtered Cu Kα radiation at a scanning rate of 1° min
1 and scan- Additionally, the absorption intensity of GLA at wavelengths
ning range from 5° to 90°. TEM images were obtained using a greater than 300 nm increased more significantly than for GSA,
Hitachi H-7500 system. Field-emission scanning electron micro- indicating that GLA has a greater degree of conjugation than GSA.
scopy (FE-SEM) was performed using a Hitachi S-5000 system. In Hummers' method, the addition of a strong oxidizing agent
Absorption spectra were recorded on a Perkin-Elmer Lambda 800/ destroys the plane and edge structure and forms acid groups, in-
900 spectrometer. XPS measurements were obtained using a VG cluding other oxygen groups; however, in Friedel–Crafts acylation,
Scientific ESCALAB 250 system. electrophilic aromatic substitution occurs on the structure with
the highest electron density (edge site). Thus, GLA retains a na-
tively conjugated structure.
3. Results and discussion The degree of destruction of the conjugated structure was ex-
amined by X-ray photoelectron spectroscopy (XPS). The C1s
3.1. Characterization of GLA spectrum of pristine graphite was deconvoluted into two peaks,
which correspond to the aromatic ring (C ¼C/C–C) at 284.6 eV and
Fig. S1 shows the synthetic scheme of carboxylated graphene the original defective structures of hydroxyl and epoxy/ether
with long-chain acid groups (GLA) from pristine graphite via groups (C–O–C) at 286.1 eV (Fig. S3) (Roldán et al., 2012). The
202 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207

degree of defective structure was determined by calculating the process; next, the GLA was magnetized through the in situ
ratio of the area of defective structure to the total area, yielding a synthesis of Fe3O4 nanoparticles on its surface. Fig. 2A shows that
value of 14.0%. Similarly, GLA was deconvoluted into two peaks at MGLA was miniaturized to approximately 40 nm and was covered
287.7 eV (C ¼O of ketone) and 288.9 eV (O–C ¼O) with area ratios with nanoparticles, which were 5 nm in size. Moreover, no free
of 8.7% and 8.7%, respectively (Fig. S4) (Hua et al., 2012). This result nanoparticles were observed either within or dispersed around
is consistent with the theoretical value for 1-one-butenoic acid, the composite. These results indicate that the nanoparticles were
which is the chain derived from MA ring opening. Compared with adsorbed on GLA, while free nanoparticles were displaced during
graphite, the degree of defective structure did not significantly purification.
increase, indicating that the conjugated structure remained nearly Compared with the absorption bands of GLA and GSA, a new
complete. shoulder, corresponding to the Fe3O4 nanoparticles, was observed
To further confirm the structural modification, Fourier-trans- at approximately 400 nm in MGLA and MGSA (Fig. 2B) (Yang et al.,
form infrared (FT-IR) spectroscopy was performed on the synthe- 2013). This result demonstrates that the nanoparticles on the GLA
sized GLA (Fig. S5). The typical band characteristic of graphite at surface are Fe3O4, confirming the synthesis of MGLA. The photo
1530 cm
1 was assigned to the stretch vibration of C ¼C (νC ¼ C). also confirms that the MGLA composite is magnetic (Fig. 2B inset).
After Friedel–Crafts acylation, GLA13exhibited several new peaks MGLA was further characterized by XRD. Six new diffraction peaks
at 1703 cm
1 (νC ¼ O), 1613 cm
1 (νC ¼ C of graphene overlapping at 30.4°, 35.8°, 43.5°, 53.7°, 57.3° and 63.1° were attributed to the
the νC ¼ O of the ketone), 1445 and 1264 cm
1 (coupling between crystalline Fe3O4 (Fig. 2C). Additionally, FT-IR spectroscopy re-
the in-plane O–H bending and the C–O stretching of the dimer), vealed the characteristic peak of Fe–O at 584 cm
1 in MGLA,
and 1375 cm
1 (–CH2 bending) (Hua et al., 2011b). Peaks char- which is a slight red shift on the order of 2 cm
1compared to
acteristic of the asymmetric and symmetric νC ¼ O of MA were not Fe3O4 NPs. This result reveals that interactions between GLA and
observed at 1869 or 1777 cm
1, confirming MA ring opening and Fe3O4 increase adsorption (Fig. S10).
1-one-butenoic acid formation at the edge site. The presence of The magnetization of GLA to MGLA increased from 0 to
the acid chain facilitates the dispersion of GLA in aqueous solution 51.0 emu g
1 (Fig. 2D). However, MGSA exhibited higher magne-
due to the negative charge repulsion of the anion generated by tization (61.0 emu g
1) than MGLA. The increased magnetization
dissociation. of MGSA is most likely due to the destruction of the GSA structure
Toluidine blue O (TBO) was used as a probe to quantify the and associated oxygen groups, favoring the adsorption of Fe3O4
carboxylic group density based on the absorption spectrum at NPs. Additionally, the carboxylic group densities of MGLA and
633 nm (Pistillo et al., 2011). As shown in Fig. S6, GLA13 exhibited MGSA after magnetization decreased significantly to
the highest carboxylic group density (1.33  10–6 mol mg
1), 0.17  10–6 mol mg
1 and 0.08  10–6 mol mg
1; this finding
confirming the optimal reaction conditions and the presence of suggests that the highly alkaline solution used during the mag-
carboxylic groups. This value is also higher than that of the gra- netizing process cleaved the acid chain (Fig. S11).
phene oxide (0.38  10–6 mol mg
1) obtained via Hummers'
method. 3.3. Quantification of urinary APOA2 and clinical importance for
The stability of the carboxylic side chain was confirmed by bladder cancer
thermal gravimetric analysis (TGA) from 100 °C to 690 °C (Fig. S7).
Due to the high strength of the graphite structure, little weight We previously demonstrated that the average urinary con-
loss was observed. However, both GLA and GSA exhibited sig- centration of the APOA2 protein was approximately 3.0–78.4-fold
nificant weight loss due to the destruction of the structure during higher in bladder cancer patients at different stages than in non-
chemical modification and the instability of the side chain. GSA tumor controls, as determined using a mass spectrometry (MS)-
initially exhibited weight loss due to the thermal degradation of based platform and a Western blot assay (Chen et al., 2010, 2012).
the side chain at approximately 100 °C. Compared to GSA, GLA The quantification of a protein marker using a MS platform is
exhibited higher thermal stability, with initial weight loss occur- specific and multiplexable but requires costly instrumentation that
ring at 240 °C, indicating that enhanced preservation of the gra- may not be affordable for common laboratories and clinics. As part
phene structure improved side chain stability. Additionally, the of the biomarker development pipeline, a potential biomarker
total weight loss of GLA was greater than that of GSA, revealing a discovered or verified in academia should be further translated to
high molecular weight and high degree of grafting to the 1-one- the clinic using a high-throughput, sensitive immunoassay to
butenoic acid group. quantify the target protein in a large number of clinical samples
Further, the conformation of GLA was analyzed by XRD in the (Surinova et al., 2011; Solier and Langen, 2014). Fig. 3A shows that
presence of nickel powder (15% wt/wt), which yielded diffraction the higher urinary concentration of APOA2 in bladder cancer pa-
peaks at 44.8°, 52.2° and 76.7°, and was used as a normalization tients was confirmed by an immune-based Bio-Plex system in a
standard (Fig. S8) (Chen et al., 2013b). Graphite exhibits two strong larger number of samples with statistical significance (54.5-fold
diffraction peaks at 26.9° and 55.0° due to crystalline heavy higher, n¼ 111, p o0.001, area under the curve (AUC)¼ 0.864)
stacking. In contrast, for GLA, the former peak decreases in size (Chen et al., 2013a). To further address the problems of tedious
and shifts to a lower angle, while the latter peak disappears. These workflow for urine sample preparation using the Bio-Plex system
results indicate that the heavy stacking sheet is destroyed due to while improving the sensitivity of detection, we explored the
exfoliation into graphene, which increases the associated d-space. quantification performance of urine APOA2 protein using the Ab-
Additionally, TEM analysis revealed a transparent thin sheet with a MGLA/poly-SiNW-FET-based biosensor.
wrinkled appearance (Fig. S9), which is in agreement with the
structural features of a graphene sheet. This result demonstrates 3.4. Immobilization and characterization of Ab-MGLA
that graphite can be functionalized and exfoliated into grapheme
via Friedel–Crafts chemical acylation. The immobilization of ananti-APOA2 antibody (Ab) on MGLA
was accomplished using EDC/sulfo-NHS as the linker to form a
3.2. Characterization of MGLA bioconjugate bond between Ab and MGLA. Ab is specific for the
AOPA2 protein. The immobilized Ab and its bioactivity were
For size compatibility with the poly-SiNW-FET, the relatively quantified using an enzyme-linked immunosorbent assay (ELISA).
large synthesized GLA was miniaturized using an ultrasonic The optimal loading on 1 μg of MGLA was examined by reaction
 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207 203

Fig. 2. (A) TEM analysis of MGLA. (B) The absorption spectra of (▲) Fe3O4 NPs, (●) GLA, and (■) MGLA. Inset: Photos of GLA and MGLA obtained under an applied magnetic
field. (C) The XRD spectra of GLA and MGLA. (D) The magnetization curves of GLA, MGLA, and MGSA.

with various concentrations of Ab. The loaded quantity of Ab was negatively charged proteins are subject to charge repulsion, in-
calculated by ELISA of the un-immobilized Ab at a monitoring creasing the difficult of targeting and decreasing the bioactivity of
wavelength of 450 nm (Fig. 3B and C). The optimal amount of Ab the material. In contrast, the long chain can extend and bend at
immobilized on 1 μg of MGLA was approximately 42 ng. The de- random angles to produce a large space for protein targeting, re-
gree of immobilization was 1.4%, in which MGLA provided sulting in higher bioactivity.
1.7  10
10 mol of carboxylic groups to load 2.4  10
12 mol of Ab
(17.4 kDa) (Fig. 3D). To investigate the effect of chain length, MGSA 3.5. Preparation and characterization of Ab-MGLA/poly-SiNW-FET
(2.13 μg) containing the same quantity of carboxylic groups was biosensors
also loaded with Ab. However, the optimal amount of Ab im-
mobilized on 2.13 μg of MGSA was approximately 29 ng. The de- The development of a poly-SiNW-FET biosensor depends on
gree of immobilization was 0.98%, such that MGSA provided the measurement of the variation in conductance, which is in-
1.7  10
10 mol of carboxylic groups to load 1.67  10
12 mol of duced by disturbing the surface charge. Therefore, poly-SiNW-FET
Ab. The quantity of immobilized Ab was 1.43-fold higher on MGLA with a wire length of 2 μm and a width of 0.15 μm was used as the
than on MGSA. This result indicates that the longer chain, which biochip to fabricate the Ab-MGLA/poly-SiNW-FET biosensor
has more degrees of freedom, can overcome the steric hindrance (Fig. 1B). Considering the limitation of the Debye length, the direct
within the large biomolecule during immobilization. The activity deposition of Ab-MGLA on poly-SiNW was excluded. To fabricate
of the immobilized Ab, which is the dominant factor in biosensor the thinnest possible layer of Ab-MGLA on the poly-SiNW, the
performance, was also investigated. The retained activity of the surface of poly-SiNW first must be modified by (3-aminopropyl)
immobilized Ab on MGLA was approximately 71% as determined triethoxysilane (APTES) and glutaraldehyde (GA) to provide
by ELISA (Fig. 3C), while 62% activity was observed on MGSA. The terminal aldehyde groups that can form covalent bonds with Ab-
immobilized Ab fraction on MGSA was likely confined to the sur- MGLA (Fig. 4A). Unmodified sites are blocked by bovine serum
face in clusters. As proteins are introduced, the very large proteins albumin (BSA) to prevent non-specific reactions or binding. The
(37 kDa) are subject to steric hindrance among themselves. The efficient detection range of a FET biosensor is within the Debye
204 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207

Fig. 3. (A) Validation of APOA2 using a multiplexed Bio-Plex assay. The concentration of urinary APOA2 was elevated in bladder cancer patients compared to hernia patients
(total n ¼111). The optical density (O.D.) calibration curve for (B) APOA2 protein and (C) anti-APOA2 concentrations based on ELISA (n¼ 3). (D) The optimization of anti-
APOA2 immobilization (n¼3).

length, which is associated with the ionic strength of the solution. the Debye length. The change is also affected by other charged
Although a higher ionic strength results in a shorter Debye length, species, particularly smaller components, that are nearly physi-
the FET exhibits higher sensitivity. TEM analysis revealed that cally adsorbed on the FET surface during the measurement of
MGLA was approximately 40 nm wide and 11 nm thick, assuming unknown samples, resulting in decreased selectivity. Moreover,
that the thickness of GLA is 1 nm and that two Fe3O4 NPs are lo- the presence of unknown species dissolved in solution also
cated on the top and bottom of the GLA plane. Therefore, the use changes the ionic strength and Debye length, varying the current
of 0.5 mM PBS with a Debye length of approximately 15.2 nm is response. Fig. 5A shows the results for Ab-MGLA/poly-SiNW-FET
suitable. The optimal conditions for Ab-MGLA loading were in- after injection with immunoglobulin G (IgG, 1000 ng dL
1), im-
vestigated by reacting with various concentrations of Ab-MGLA, munoglobulin M (IgM, 192.5 mg dL
1), glucose (5 mM), ascorbic
ranging from 0.025 to 1 mg mL
1 and corresponding to 1 ng mL
1 acid (AA, 4.3 μg mL
1), and uric acid (UA, 0.295 mM) at con-
protein (Fig. 4B). The current of the Ab-MGLA/poly-SiNW-FET centrations typically found in human plasma. These species were
biochip decreased with increasing concentration of Ab-MGLA in
also studied at concentrations that were twice as large as the va-
the reaction. The relative current change (
ΔI/I0, %) at each con-
lues listed above. The injection of either normal or excess con-
centration of Ab-MGLA in the reaction was 6.3% at 0.025 mg mL
1,
centrations of species changed the current. However, the major
11.3% at 0.125 mg mL
1, 22.6% at 0.625 mg mL
1and 23.6% at 1.0 m
variations were reduced when the solutions were removed, fol-
g mL
1. The relative current increases slightly after 0.625 μ
lowed by washing and the injection of pure buffer. Although the
g mL
1, indicating that the modification was saturated. These re-
currents differed slightly from those in the presence of PBS, due to
sults also indicate that the protein was negatively charged. The
either the limited adsorption of species or the interference of Ab-
protein attracted carriers from the substrate to the channel and
decreased the conductance between the source and drain elec- protein binding as a result of the species present, washing the FETs
trodes, resulting in decreased conductance. before carrying out the measurement eliminated most of the in-
terference. To evaluate the effect of protein binding in the pre-
3.6. Performance of Ab-MGLA/poly-SiNW-FET biosensors for the sence of key interfering species, 1 ng of protein was hybridized in
detection of protein the presence of each of the species identified above (Fig. 5B). In
addition to species adsorption, the slight differences observed
Recent reports of biosensors based on FETs have been devel- relative to the pure protein solution demonstrates the presence of
oped for real-time detection after the injection of specific analytes, interfering species and indicates that interactions between the
such as proteins or viruses (Oh et al., 2010; Huang et al., 2010). protein and interfering species affects protein binding. The chan-
However, FET sensitivity is charge-dependent, such that the cur- ges in the current are dependent on the quantity of bound protein,
rent or conductance of FET changes when the analytes are within which can be exploited to improve selectivity.
 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207 205

Fig. 4. (A) A schematic of the process of poly-SiNW-FET device surface modification and Ab-MGLA/poly-SiNW-FET biosensor preparation. (B) ISD–VG curves of the reaction of
[Ab-MGLA] with a, b, c, and d and the corresponding ISD–VG of a′, b′, c′ and d′ under 1 ng mL
1 protein binding.

The process of protein binding, washing and measuring was for Ab-MGLA/poly-SiNW-FET and 19.5% for Ab-MGSA/poly-SiNW-
used to assess the performance of the Ab-MGLA/poly-SiNW-FET FET after one week of storage, compared to the values obtained for
biosensor under different concentrations of protein. The current fresh biosensors. The immobilized fractions of Ab on MGLA/poly-
decreased with increasing protein concentration due to the de- SiNW-FET and MGSA/poly-SiNW-FET perform the stable property.
pletion of charge carriers in the poly-SiNW-FET, while the nega-
tively charged protein remained bound to the Ab-MGLA surface
(Fig. 5C).The relationship between ISD and the protein level reveals 3.7. Determination of APOA2 protein in human urine
that the relative current change increased proportionally with
increasing protein fraction from 19.5 pg mL
1 to 1.95 mg mL
1 The APOA2 protein is a bladder cancer biomarker that is ex-
(Fig. 5D). Compared with Ab-MGLA/poly-SiNW-FET, the biosensor hibited in higher concentrations in biofluids (Chen et al., 2010,
based on Ab-MGSA/poly-SiNW-FET exhibited a linear dependence 2012). Therefore, the performance of the Ab-MGLA/poly-SiNW-
of relative response to the logarithmical concentration in a range FET biosensor developed for clinical analysis was evaluated by
of from 195 pg mL
1 to 19.5 mg mL
1, with a lower slope that examining urine samples collected from bladder cancer patients.
corresponded to lower sensitivity. The limit of detection (LOD) of In hernia patients, the serum level of protein is
Ab-MGLA/poly-SiNW-FET and Ab-MGSA/poly-SiNW-FET were 0.425–9.47 ng mL
1, which is significantly lower than the con-
calculated using 3sS
1 as previously described (Palaniappana centration of 29–344 ng mL
1 that is typically observed in pa-
et al., 2010), where s is the standard deviation of device in pure tients suffering from late- and advanced-stage bladder cancer
buffer solution and S indicates the sensitivity (slope) of the device (Fig. 6). These results demonstrate that the developed Ab-MGLA/
at lower concentrations. The standard deviations of Ab-MGLA/ poly-SiNW-FET biosensor can be used to distinguish protein levels
poly-SiNW-FET and Ab-MGSA/poly-SiNW-FET in the absence of
in hernia and bladder cancer patients. The results of the biosensor
APOA2 were 0.88% and 0.71%, respectively. The LOD of Ab-MGLA/
are also in agreement with those measured by Bio-Plex, indicating
poly-SiNW-FET and Ab-MGSA/poly-SiNW-FET can be estimated
that the diagnosis of bladder cancer based on the developed Ab-
through 3sS
1 and were 6.7 pg mL
1 and 95.9 pg mL
1, respec-
MGLA/poly-SiNW-FET is feasible and accurate. Our assay is less
tively. In addition, the time of analysis is dependent on the scan
rate. In this study, the scan range was between 0 and 3 V with a complex and time-consuming than the Bio-Plex technology. Dif-
scan rate of 0.5 V s
1, resulting in a response time of 6 s. The su- ferently from cystoscopy, the Ab-MGLA/poly-SiNW-FET biosensor
perior performance of the Ab-MGLA/poly-SiNW-FET biosensor can is non-invasive and can limit patient pain and discomfort. Com-
be attributed to the higher amount of bioactive sites and lower pared with urine cytology, the proposed methodology is operator-
steric hindrance with respect to protein binding. independent and can offer a higher sensitivity and diagnostic ac-
Additionally, the stability of each biosensor type was in- curacy. Although subject to future validation, our novel APOA2
vestigated after dry storage at 4 °C for one week. The relative biosensor based on Ab-MGLA/poly-SiNW-FET may open new
currents for 1 ng mL
1 protein decreased by approximately 20.2% perspectives in bladder cancer diagnostics.
206 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207

Fig. 5. (A) The effects of interfering species (red bar: high concentration; blue bar: normal concentration) on the current variations in Ab-MGLA/poly-SiNW-FET biosensors
before (red and blue bars) and after washing (green and cyan bars) (n¼ 3). (B) The effects of interfering species for 1 ng mL
1 protein binding. (C) ISD–VG curves of the Ab-
MGLA/poly-SiNW-FET biosensor at various protein concentrations. (D) The calibration curves of the current responses to various protein concentrations for the Ab-MGLA/
poly-SiNW-FET biosensor and Ab-MGSA/poly-SiNW-FET biosensor (n ¼5). (For interpretation of the references to color in this figure legend, the reader is referred to the web
version of this article.)

provided more freedom, thereby preventing steric hindrance

during the immobilization of anti-APOA2 compared to a shorter
chain. ELISA results indicated that the immobilized anti-APOA2 on
MGLA retained most of its bioactivity. In summary, the MGLA/
poly-SiNW-FET biochip exhibited superior performance compared
to the MGSA/poly-SiNW-FET biochip. Further, the analysis of urine
samples from bladder cancer patients demonstrated the feasibility
of the use of the proposed biosensor to diagnose bladder cancer.
Thus, this biosensor exhibits substantial potential for use in clin-
ical diagnosis.

Acknowledgments

We thank the National Science Council of the Republic of China,

the Chang Gung Memorial Hospital and the Ministry of Economic
Affairs for their financial assistance: NSC 101-2221-E-182-013-
Fig. 6. APOA2 protein concentrations in urine samples from hernia and bladder MY3, CMRPD2D0081, CMRPD2A0043, NSC 102-2113-M-182-001-
cancer patients based on Ab-MGLA/poly-SiNW-FET (red bars) and Bio-Plex (green
MY2, 102-D0607 (GERPD2C0141) and D301AR1H10. We also thank
bars) (n¼ 3). (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.) the Chang Gung Memorial Hospital Microscopy Core Laboratory
for assistance.
4. Conclusions

A label-free electrical method of APOA2 protein detection was

developed based on an Ab-MGLA/poly-SiNW-FET biochip. The Appendix A. Supplementary material
magnetized MGLA accelerated purification during the im-
mobilization of anti-APOA2, thus preventing the denaturation of Supplementary data associated with this article can be found in
anti-APOA2. Additionally, the longer bioactive chain of MGLA the online version at http://dx.doi.org/10.1016/j.bios.2014.11.019
 H.-C. Chen et al. / Biosensors and Bioelectronics 66 (2015) 198–207 207

References Kirkali, Z., Chan, T., Manoharan, M., Algaba, F., Busch, C., Cheng, L., Kiemeney, L.,
Kriegmair, M., Montironi, R., Murphy, W.M., Sesterhenn, I.A., Tachibana, M.,
Weider, J., 2005. Urology 66, 4–34.
Allen, B.L., Kichambare, P.D., Star, A., 2007. Adv. Mater. 19, 1439–1451. Kwon, O.S., Park, S.J., Hong, J.T., Han, A.R., Lee, J.S., Lee, J.S., Oh, J.H., Jang, J., 2012. ACS
Araki, M., Nieder, A.M., Manoharan, M., Yang, Y., Soloway, M.S., 2007. Urology 69, Nano 6, 1486–1493.
270–274. Lee, H.S., Kim, K.S., Kim, C.J., Hahn, S.K., Jo, M.H., 2009. Biosens. Bioelectron. 24,
Cao, M., Zhao, L., Chen, H., Xue, W., Lin, D., 2012. Anal. Sci. 28, 451–456. 1801–1805.
Chen, Y.T., Chen, C.L., Chen, H.W., Chung, T., Wu, C.C., Chen, C.D., Hsu, C.W., Chen, M. Li, B.R., Chen, C.W., Yang, W.L., Lin, T.Y., Pan, C.Y., Chen, Y.T., 2013. Biosens. Bioe-
C., Tsui, K.H., Chang, P.L., Chang, Y.S., Yu, J.S., 2010. J. Proteome Res. 9, lectron. 45, 252–259.
5803–5815. Liu, Z., Robinson, J.T., Sun, X.M., Dai, H.J., 2008. J. Am. Chem. Soc. 130, 10876–10877.
Chen, K.I., Lia, B.R., Chen, Y.T., 2011. Nanotoday 6, 131–154. Mourant, J.R., Bigio, I.J., Boyer, J., Conn, R.L., Johnson, T., Shimada, T., Laser, 1995.
Chen, Y.T., Chen, H.W., Domanski, D., Smith, D.S., Liang, K.H., Wu, C.C., Chen, C.L., Surg. Med 17, 350–357.
Chung, T., Chen, M.C., Chang, Y.S., Parker, C.E., Borchers, C.H., Yu, J.S., 2012. J. Oh, J., Chang, Y.W., Kim, H.J., Yoo, S., Kim, D.J., Im, S., Park, Y.J., Kim, D., Yoo, K.H.,
Proteomics 75, 3529–3545. 2010. Nano Lett. 10, 2755–2760.
Chen, H.C., Chen, Y.H., Chen, S.L., Chern, Y.T., Tsai, R.Y., Hua, M.Y., 2013a. Biosens. Oh, J., Yoo, G., Chang, Y.W., Kim, H.J., Jose, J., Kim, E., Pyun, J.C., Yoo, K.H., 2013.
Bioelectron. 46, 84–90. Biosens. Bioelectron. 50, 345–350.
Chen, Y.T., Parker, C.E., Chen, H.W., Chen, C.L., Domanski, D., Smith, D.S., Wu, C.C., Palaniappan, Al, Goh, W.H., Tey, J.N., Wijaya, I.P.M., Moochhala, S.M., Liedberg, B.,
Chung, T., Liang, K.H., Chen, M.C., Chang, Y.S., Borchers, C.H., Yu, J.S., 2013b. R. Mhaisalkar, S.G., 2010. Biosens. Bioelectron. 25, 1989–1993.
Soc. Chem. 0, 271–314. Pistillo, B.R., Perrotta, A., Gristina, R., Ceccone, G., Nardulli, M., d'Agostino, R., Favia,
Chen, H.C., Qiu, J.T., Yang, F.L., Liu, Y.C., Chen, M.C., Tsai, R.Y., Yang, H.W., Lin, C.Y., Lin, P., 2011. Surf. Coat. Technol. 205, S534–S536.
C.C., Wu, T.S., Tu, Y.M., Xiao, M.C., Ho, C.H., Huang, C.C., Lai, C.S., Hua, M.Y., 2014. Roberts, W.S., Davis, F., Holmes, J.L., Collye, S.D., Larcombe, L.D., Morgan, S.L., Hig-
Anal. Chem. 86, 9443–9450. son, S.P.J., 2013. Biosens. Bioelectron. 41, 282–288.
Ferreira, C.S.M., Papamichael, K., Guilbault, G., Schwarzacher, T., Gariepy, J., Mis- Roldán, L., Santos, I., Armenise, S., Fraile, J.M., Garcĭa-Bordejé, E., 2012. Carbon 50,
sailidis, S., 2008. Anal. Bioanal. Chem. 390, 1039–1050. 1363–1372.
Hua, M.Y., Lin, Y.C., Tsai, R.Y., Chen, H.C., 2012. J. Mater. Chem. 22, 2566–2574. Shapiro, A., Gofrit, O.N., Pizov, G., Cohen, J.K., Maier, J., 2011. Eur. Urol. 59, 106–112.
Hua, M.Y., Chen, H.C., Chuang, C.K., Tsai, R.Y., Jeng, J.L., Yang, H.W., Chern, Y.T., 2011a. Shin, Y., Perera, A.P., Park, M.K., 2013. Sensor. Actuators B-Chem. 178, 200–206.
Solier, C., Langen, H., 2014. Proteomics 14, 774–783.
Biomaterials 32, 4885–4895.
Surinova, S., Schiess, R., Hüttenhain, R., Cerciello, F., Wollscheid, B., Aebersold, R.,
Hua, M.Y., Lin, Y.C., Tsai, R.Y., Leu, Y.L., Liu, Y.C., Lai, J.T., 2011b. J. Mater. Chem. 21,
2011. J. Proteome Res. 10, 5–16.
7254–7262.
Sykes, E., Charles, H., 2009. Nat. Chem. 1, 175–176.
Huang, S.C., Artyukhin, A., Misra, M., Martinez, J., Stroeve, P., Grigoropoulos, C., Ju, J.,
Tai, S.P., Wu, Y., Shieh, D.B., Chen, L.J., Lin, K.J., Yu, C.H., Chu, S.W., Chang, C.H., Shi, X.
Noy, A., 2010. Nano Lett. 10, 1812–1816.
Y., Wen, Y.C., Lin, K.H., Liu, T.M., Sun, C.K., 2007. Adv. Mater. 19, 4520–4523.
Kałuzna-Czaplinska, J., Jóźwik, J., 2014. TrAC-Trends Anal. Chem. 56, 1–90.
Trung, T.Q., Tien, N.T., Kim, D., Jang, M., Yoon, O.J., Lee, N.E., 2014. Adv. Funct. Mater.
Kandimalla, K., van Tilborg, A.A., Zwarthoff, E.C., 2013. Nat. Rev. Cancer 10, 327–335.
24, 117–124.
Kim, K.W., Shin, Y., Perera, A.P., Liu, Q., Kee, J.S., Han, K., Yoon, Y.J., Park, M.K., 2013.
Yang, H.W., Hua, M.Y., Chen, S.L., Tsai, R.Y., 2013. Biosens. Bioelectron. 41, 172–179.
Biosens. Bioelectron. 45, 152–157.