ARTICLES nature biotechnology Multiplexed electrical detection of cancer markers with nanowire sensor arrays gengfeng Zheng,, Fernando Patolsky 4, Yi Cui, Wayne U Wang 2& Charles M Lieber,3 describe highly sensitive, label-free, multiplexed electrical detection of cancer markers using silicon- nanowire field-effect vices in which distinct nanowires and surface receptors are incorporated into arrays. protein markers were routinely detected at femtomolar concentrations with high selectivity, and simultaneous incorporation of control nanowires enabled discrimination against false positives. Nanowire arrays allowed highly selective and sensitive multiplexed detection of prostate specific antigen E(PSA), PSA-ol-antichymotrypsin, carcinoembryonic antigen and mucin -l, including detection to at least 0. 9 pg/ml in undiluted serum samples. In addition, nucleic acid receptors enabled real-time assays of the binding, activity and small-molecule inhibition of telomerase using unamplified extracts from as few as ten tumor cells. the capability for multiplexed real-time o Genomics and proteomics research has elucidated many new biomar protein markers at femtomolar kers that have the potential to greatly improve disease diagnosis nents in undiluted serum samples The availability of multiple biomarkers is believed to be especially measurements of the binding, activity and small-I important in the diagnosis of complex diseases like cancer, for inhibition of telomerase from unamplified extracts of as few as ten a which disease heterogeneity makes tests of single markers, such as tumor cells using nucleic acid-modified nanowire devices. prostate specific antigen (PSA), inadequate. Patterns of multiple 巴3zo cancer markers might, however, provide the information necessary RESULTS or robust diagnosis of disease in any person within a population Array design and characteristics Moreover, detection of markers associated with different stages of The conversion of silicon-nanowire field-effect transistors into sensors disease pathogenesis could further facilitate early detection for cancer protein marker detection was carried out by attaching Widespread use of cancer markers in healthcare will ultimately monoclonal antibodies(mAbs) to the nanowire surfaces after device depend upon the development of techniques that allow rapid, fabrication. The basic linkage chemistry is similar to that used multiplexed detection of many markers with high selectivity and previously for protein microarrays and silicon-nanowire sensors sensitivity. This goal has not been attained with any existing method, for viruses and small molecules, 2, and involves three key steps. First, including the enzyme-linked immunosorbent assay(ELISA), a aldehyde propyltrimethoxysilane(APTMS)is coupled to oxygen common clinical approach for protein marker detection, surface plasma-cleaned silicon-nanowire surfaces to present terminal alde plasmon resonance(SPR)%, 0, nanoparticles-4 microcantilevers hyde groups at the nanowire surface. Second, the aldehyde groups are and carbon nanotubes 6,17 coupled to mAbs. Third, unreacted free aldehyde groups are blocked Here we present label-free, real-time multiplexed detection of by reaction with ethanolamine. In these studies, we show how this protein cancer markers with high selectivity and femtomolar sensitiv- surface chemistry affects the nanowire device sensitivity and selec- y using antibody-functionalized, silicon-nanowire field-effect sen- tivity, which are critical for development of a viable multiplexed sors. Previously, silicon-nanowirel8 and carbon-nanotubel6 devices detection technology have been used to detect binding and unbinding of proteins in The basic nanowire sensor chip(Fig. la; see Supplementary Fig. 1 aqueous solutions. However, this previous work did not demonstrate online)consists of integrated, electrically addressable silicon nanowires high sensitivity, nor did it show a capability for selective multiplexed with the potential for 200 individually addressable devices. Our detection of different proteins. To overcome previous limitations, we basic array design enables incorporation of different types of have developed integrated nanowire arrays in which distinct nano- addressable nanowires, for example, p-type and n-type doped silicon wires and surface receptors can be incorporated as individual device nanowires, during fabrication steps to form the addressable elements. We characterized the sensitivity and selectivity limits of these electrical contacts; that is, solutions of the different nanowires can devices and demonstrated simultaneous quantitative detection of be sequentially assembled in different regions of the device, and then Cambridge, Massachusetts 02138, USA. These authors contributed equally to the work Correspondence should be addressed to C.M. L(cn is harvard. edu) Received 4 April; accepted 26 July: published online 18 September 2005; doi: 10.1038/nbt1138 VOLUME 23 NUMBER 10 OCTOBER 2005 NATURE BIOTECHNOLOGY
Multiplexed electrical detection of cancer markers with nanowire sensor arrays Gengfeng Zheng1,4, Fernando Patolsky1,4, Yi Cui1, Wayne U Wang1,2 & Charles M Lieber1,3 We describe highly sensitive, label-free, multiplexed electrical detection of cancer markers using silicon-nanowire field-effect devices in which distinct nanowires and surface receptors are incorporated into arrays. Protein markers were routinely detected at femtomolar concentrations with high selectivity, and simultaneous incorporation of control nanowires enabled discrimination against false positives. Nanowire arrays allowed highly selective and sensitive multiplexed detection of prostate specific antigen (PSA), PSA-a1-antichymotrypsin, carcinoembryonic antigen and mucin-1, including detection to at least 0.9 pg/ml in undiluted serum samples. In addition, nucleic acid receptors enabled real-time assays of the binding, activity and small-molecule inhibition of telomerase using unamplified extracts from as few as ten tumor cells. The capability for multiplexed real-time monitoring of protein markers and telomerase activity with high sensitivity and selectivity in clinically relevant samples opens up substantial possibilities for diagnosis and treatment of cancer and other complex diseases. Genomics and proteomics research has elucidated many new biomarkers that have the potential to greatly improve disease diagnosis1–3. The availability of multiple biomarkers is believed to be especially important in the diagnosis of complex diseases like cancer4,5, for which disease heterogeneity makes tests of single markers, such as prostate specific antigen (PSA), inadequate. Patterns of multiple cancer markers might, however, provide the information necessary for robust diagnosis of disease in any person within a population6,7. Moreover, detection of markers associated with different stages of disease pathogenesis could further facilitate early detection. Widespread use of cancer markers in healthcare will ultimately depend upon the development of techniques that allow rapid, multiplexed detection of many markers with high selectivity and sensitivity. This goal has not been attained with any existing method, including the enzyme-linked immunosorbent assay (ELISA)8, a common clinical approach for protein marker detection, surface plasmon resonance (SPR)9,10, nanoparticles11–14, microcantilevers15 and carbon nanotubes16,17. Here we present label-free, real-time multiplexed detection of protein cancer markers with high selectivity and femtomolar sensitivity using antibody-functionalized, silicon-nanowire field-effect sensors. Previously, silicon-nanowire18 and carbon-nanotube16 devices have been used to detect binding and unbinding of proteins in aqueous solutions. However, this previous work did not demonstrate high sensitivity, nor did it show a capability for selective multiplexed detection of different proteins19. To overcome previous limitations, we have developed integrated nanowire arrays in which distinct nanowires and surface receptors can be incorporated as individual device elements. We characterized the sensitivity and selectivity limits of these devices and demonstrated simultaneous quantitative detection of multiple protein markers at femtomolar concentrations, including measurements in undiluted serum samples. In addition, we carried out direct measurements of the binding, activity and small-molecule inhibition of telomerase from unamplified extracts of as few as ten tumor cells using nucleic acid–modified nanowire devices. RESULTS Array design and characteristics The conversion of silicon-nanowire field-effect transistors into sensors for cancer protein marker detection was carried out by attaching monoclonal antibodies (mAbs) to the nanowire surfaces after device fabrication. The basic linkage chemistry is similar to that used previously for protein microarrays20,21 and silicon-nanowire sensors for viruses and small molecules22,23, and involves three key steps. First, aldehyde propyltrimethoxysilane (APTMS) is coupled to oxygen plasma–cleaned silicon-nanowire surfaces to present terminal aldehyde groups at the nanowire surface. Second, the aldehyde groups are coupled to mAbs. Third, unreacted free aldehyde groups are blocked by reaction with ethanolamine. In these studies, we show how this surface chemistry affects the nanowire device sensitivity and selectivity, which are critical for development of a viable multiplexed detection technology. The basic nanowire sensor chip (Fig. 1a; see Supplementary Fig. 1 online) consists of integrated, electrically addressable silicon nanowires with the potential for B200 individually addressable devices. Our basic array design enables incorporation of different types of addressable nanowires, for example, p-type and n-type doped silicon nanowires, during fabrication steps to form the addressable electrical contacts; that is, solutions of the different nanowires can be sequentially assembled in different regions of the device, and then Received 4 April; accepted 26 July; published online 18 September 2005; doi:10.1038/nbt1138 1Department of Chemistry and Chemical Biology, 2Biophysics Program and 3Division of Engineering and Applied Science, 12 Oxford Street, Harvard University, Cambridge, Massachusetts 02138, USA. 4These authors contributed equally to the work. Correspondence should be addressed to C.M.L. (cml@cmliris.harvard.edu). 1294 VOLUME 23 NUMBER 10 OCTOBER 2005 NATURE BIOTECHNOLOGY ARTICLES © 2005 Nature Publishing Group http://www.nature.com/naturebiotechnology
ARTICLES 8150 -log([PSAl(g/ml d 1,300 品 02000400060008,000 Time(s) Modification time(min) e Figure 1 Nanowire sensor arrays and detector properties. (a)Optical image(top)of a nanowire device array. the white lines correspond to the silicon tride passivated metal electrodes that connect to individual nanowire devices. the red rectangle highlights one of the repeated (vertical) regions where the nanowire devices are formed (see Supplementary Fig. l online for high resolution images of devices). The position of the microfluidic channel used to deliver has a total size of 6 m a The schematic(bottom)shows details of metal electrodes (golden lines)connecting nanowires (blue lines)in this region with orientation rotated 90%relative e to red rectangle. (b) Schematic showing two nanowire devices, I and 2, within an array, where the nanowires are modified with different (l, green; 2, red) antibody receptors. A cancer marker protein that binds specifically to its receptor (on nanowire- l)will produce a conductance change characteristic of he surface charge of the protein only on nanowire-l. (c)Change in conductance versus concentration of Psa for a p-type silicon nanowire modified with PSA-Abl receptor Inset: Conductance-versus-time data recorded after alternate delivery of Psa and pure buffer solutions; the psa concentrations were 0.9 ng/ml, 9 pg/ml, 0.9 pg/ml and 90 fg/ml, respectively. The buffer solutions used in all measurements were 1 uM phosphate(potassium salt)containing 2 HM KCl, pH=7.4.(d)Conductance-versus-time data recorded for a PSA-Abl-modified p-type silicon nanowire after alternate delivery of the following o protein and pure buffer solutions: (1)9 pg/ml PSA, (2)0.9 pg/ml PSA, (3)0.9 pg/ml PSA and 10 Hg/ml BSA, (4)10 ug/ml BSA and(5)9 PSA (e)Thickness dependence(red curve)of aldehyde silane layer on the sinw surfaces extracted from AFM measurements after different modification time the aldehyde propyltrimethoxysilane, and sensitivity dependence(blue curve) of detection of l ng/ml of PsA, after different modification time using a p-type SiNW device electrical contacts are formed in parallel by photolithography and The sensitivity limits of our silicon-nanowire devices were first metal deposition steps. In addition, different receptors can be printed determined by measuring conductance changes as the solution con- on the nanowire device array to allow selective multiplexed detection centration of PSa was varied, using devices modified with mAbs for (Fig. 1b). Selective binding of cancer marker proteins to surface-linked PSA(Abl). Representative time-dependent data(inset, Fig. lc)show a mAbs should produce a conductance change in the corresponding well-defined conductance increase and subsequent return to baseline receptor-modified silicon-nanowire device but not in devices lacking when PSA solution and pure buffer, respectively, are alternately he specific antibody receptor. In the case of a p-type(boron-doped) delivered through a microfluidic channel to the devices. a plot of silicon nanowire, applying a positive gate voltage depletes carriers and these data(Fig. lc)shows that the conductance change is directly reduces the conductance, whereas applying a negative gate voltage proportional to the solution PSA concentration for values from leads to an accumulation of carriers and an increase in conductance 5 ng/ml down to 90 fg/ml (the opposite effect occurs in n-type semiconductors). The depen- There are several key features of these data. First, the reversibility of dence of the conductance on gate voltage makes field-effect transistors the conductance changes demonstrates that nonspecific, irreversible natural candidates for electrically based sensing since the electric field protein binding does not occur to a measurable extent on the devices resulting from binding of a charged species to the gate dielectric is Second, the increases in conductance with PSa binding to the analogous to applying a voltage using a gate electrode. Thus, the Abl-linked, p-type nanowire devices are consistent with binding of onductance of a p-silicon nanowire will increase(decrease) when a a protein with negative overall charge, as expected from the pl of PSA, protein with negative(positive)surface charge binds to the antibody 6.8(ref. 24), and the pH, 7.4, of our experiments. Third, the receptor, whereas the opposite response should be observed for an show that direct, label-free detection of PSA is routinely n-type(phosphorus-doped) silicon nanowire with a signal-to-noise ratio >3 for concentrations down to ATURE BIOTECHNOLOGY VOLUME 23 NUMBER 10 OCTOBER 2005 1295
electrical contacts are formed in parallel by photolithography and metal deposition steps. In addition, different receptors can be printed on the nanowire device array to allow selective multiplexed detection (Fig. 1b). Selective binding of cancer marker proteins to surface-linked mAbs should produce a conductance change in the corresponding receptor-modified silicon-nanowire device but not in devices lacking the specific antibody receptor. In the case of a p-type (boron-doped) silicon nanowire, applying a positive gate voltage depletes carriers and reduces the conductance, whereas applying a negative gate voltage leads to an accumulation of carriers and an increase in conductance (the opposite effect occurs in n-type semiconductors). The dependence of the conductance on gate voltage makes field-effect transistors natural candidates for electrically based sensing since the electric field resulting from binding of a charged species to the gate dielectric is analogous to applying a voltage using a gate electrode. Thus, the conductance of a p-silicon nanowire will increase (decrease) when a protein with negative (positive) surface charge binds to the antibody receptor, whereas the opposite response should be observed for an n-type (phosphorus-doped) silicon nanowire18. The sensitivity limits of our silicon-nanowire devices were first determined by measuring conductance changes as the solution concentration of PSA was varied, using devices modified with mAbs for PSA (Ab1). Representative time-dependent data (inset, Fig. 1c) show a well-defined conductance increase and subsequent return to baseline when PSA solution and pure buffer, respectively, are alternately delivered through a microfluidic channel to the devices. A plot of these data (Fig. 1c) shows that the conductance change is directly proportional to the solution PSA concentration for values from B5 ng/ml down to 90 fg/ml. There are several key features of these data. First, the reversibility of the conductance changes demonstrates that nonspecific, irreversible protein binding does not occur to a measurable extent on the devices. Second, the increases in conductance with PSA binding to the Ab1-linked, p-type nanowire devices are consistent with binding of a protein with negative overall charge, as expected from the pI of PSA, 6.8 (ref. 24), and the pH, 7.4, of our experiments. Third, these data show that direct, label-free detection of PSA is routinely achieved with a signal-to-noise ratio 43 for concentrations down to 75 fg/ml 2 2 1 1 200 1.35 1.25 1.15 1.05 0 2,000 4,000 6,000 Time (s) Time (s) Modification time (min) Layer thickness (nm) Conductance (µS) 8,000 150 100 1,300 1,250 1,200 200 150 100 50 20 40 Sensitivity (∆ nS) 60 80 100 120 0 2 4 6 0 0 0 1 2 3 4 5 2,000 4,000 6,000 8,000 50 ∆Conductance (nS) Conductance (nS) 0 9 10 11 –log ([PSA] (g/ml)) 12 13 14 a d e b c Figure 1 Nanowire sensor arrays and detector properties. (a) Optical image (top) of a nanowire device array. The white lines correspond to the silicon nitride passivated metal electrodes that connect to individual nanowire devices. The red rectangle highlights one of the repeated (vertical) regions where the nanowire devices are formed (see Supplementary Fig. 1 online for high resolution images of devices). The position of the microfluidic channel used to deliver sample is highlighted by the dashed white rectangle and has a total size of 6 mm 500 mm, length width. The image field is 8 mm 1.2 mm. The schematic (bottom) shows details of metal electrodes (golden lines) connecting nanowires (blue lines) in this region with orientation rotated 901 relative to red rectangle. (b) Schematic showing two nanowire devices, 1 and 2, within an array, where the nanowires are modified with different (1, green; 2, red) antibody receptors. A cancer marker protein that binds specifically to its receptor (on nanowire-1) will produce a conductance change characteristic of the surface charge of the protein only on nanowire-1. (c) Change in conductance versus concentration of PSA for a p-type silicon nanowire modified with PSA-Ab1 receptor. Inset: Conductance-versus-time data recorded after alternate delivery of PSA and pure buffer solutions; the PSA concentrations were 0.9 ng/ml, 9 pg/ml, 0.9 pg/ml and 90 fg/ml, respectively. The buffer solutions used in all measurements were 1 mM phosphate (potassium salt) containing 2 mM KCl, pH ¼ 7.4. (d) Conductance-versus-time data recorded for a PSA-Ab1-modified p-type silicon nanowire after alternate delivery of the following protein and pure buffer solutions: (1) 9 pg/ml PSA, (2) 0.9 pg/ml PSA, (3) 0.9 pg/ml PSA and 10 mg/ml BSA, (4) 10 mg/ml BSA and (5) 9 pg/ml PSA. (e) Thickness dependence (red curve) of aldehyde silane layer on the SiNW surfaces extracted from AFM measurements after different modification time of the aldehyde propyltrimethoxysilane, and sensitivity dependence (blue curve) of detection of 1 ng/ml of PSA, after different modification time using a p-type SiNW device. NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 10 OCTOBER 2005 1295 ARTICLES © 2005 Nature Publishing Group http://www.nature.com/naturebiotechnology
ARTICLES b 1.800 p-type nanowire(NWl, Fig 2a)and n-type nanowire (NW2, Fig. 2a) devices after M Nw1 introduction of 0.9 ng/ml of PSA showed a conductance increase in Nwl and a con- ductance decrease in Nw2, whereas the con- ductance returned to the baseline value of each device after introduction of buffer solu- tion without PSA. The magnitudes of the inductance changes in the two devices 1,00020003,0004,0005,000 were nearly the same and consistent with the concentration-dependent conductance Figure 2 Multiplexed detection with nanowire arrays (a)Complementary sensing of PSA using p-type measurements(Fig. 1). Similar behavior (NW1)and n-type(NW2)silicon- nanowire devices in the same array. The vertical solid lines observed with different concentrations o times at which PSA solutions of (1)0.9 ng/ml, (2)0.9 ng/ml, (3)9 pg/ml, (4)0.9 pg/ml and of PSA(Fig. 2a); that is, the p- and n-type E(5)5 ng/ml were connected to the microfluidic channel (b) Conductance-versus time data recorded devices showed ncentration E simultaneously from two silicon-nanowire devices in an array, where Nwl was functionalized increases and decreases, respectively, in con- s solutions of(1)9 pg/ml PSA, (2)1 pg/ml PSA, (3)10 Hg/ml BSA, (4) a mixture of i ng/ml PSA and ductance when the solution was alternated between psa and buffer a the points where the solution flow was switched from protein to pure buffer solutions e These experiments demonstrate key points bout multiplexed electrical de bserved for the or 2 fM. Similar ultrasensitive detection was achieved in studies and n-type elements are consistent with specific binding of PSa to e of carcinoembryonic antigen(CEA), 100 fg/ml or 0.55 fM,and field-effect devices, since the negatively charged protein will cause e mucin-1, 75 fg/ml or 0.49 fM,(see Supplementary Fig. 2 online) accumulation and depletion of charge carriers in the p-type and o using silicon-nanowire devices modified with mAbs for CEA and n-type nanowire elements, respectively. Second, the complementary tively electrical signals from p-and n-type devices provide a simple yet a in competitive binding experiments with bovine serum albumin noise or nonspecific binding of protein to one device; that is, real and 2(BSA)(Fig. Id). Conductance-versus-time measurements recorded selective binding events must show complementary responses in the on a silicon-nanowire device modified with Abl showed similar p-and n-type devices. The presence of correlated conductance signals onductance changes as above when 9 and 0.9 pg/ml solutions of in both devices(Fig. 2a), which occur at points when buffer and PSA were delivered to the device. These results show that reproducible PSA/buffer solutions are changed, illustrates clearly how this multi device-to-device sensitivity was achieved. Moreover, delivery of a plexing capability can be used to distinguish unambiguously nois solution containing 0.9 Pg/ml PSA and 10 ug/ml BSA showed the from protein-binding signals. conductance increase as a solution containing only PSA this concentration, whereas no conductance change was observed en the Bsa solution alone was delivered. These latter data demon- strate excellent selectivity and also show that high sensitivity was not Table I Summary of multiplexed detection experiments using mAb rec lost even with a 10-million-fold higher concentration of other proteins (f-PSA), Abl and cross-reactive for f-PSA and PSA-ACT complex, Ab2 We investigated the devices' modification chemistry to define their Conductance change, ns sensitivity limits. Atomic force microscopy measurements of the initial aldehyde-silane layer thickness on single nanowires(Fig. le)showed a Protein sample NWI-Abl NW2-Ab2 systematic increase with modification time. This thickness increase is onsistent with previous studies showing that similar silane reagents f-psA 154 can form multila ers25,26. Measurements of sensitivity showed that the sensor response decreases rapidly for initial reaction times >30 min The observed decrease in sensitivity is consistent with expectations for f-PSA a field-effect sensing device, and moreover, shows that the surface psa-Act modification chemistry must be controlled to achieve reproducible psa-act high-sensitivity devices. PSA-ACT PSA-ACT 55DDDDD Multiplexed detection with nanowire arrays f-PsA, PSA-ACT,Ab1850,3,200,1×10 For our initial studies of multiplexed detection, we used an array f-PSA, PSA-ACT, Abl 8.5, 320, 1 x 107 containing both p-type and n-type silicon-nanowire devices modified f-PSA, PSA-ACT, Abl 0.85, 3, 200, 1 x 107 138 with Abl. The incorporation of p-and n-type nanowires in a f-PSA, PSA-ACT, Abl 850, 0.32, 1 x 107 ingle sensor chip enables discrimination of possible electrical cross- f-PSA, Abl talk and/or false-positive signals by correlating the response versus ND me from the two types of device elements. Notably, simultaneous conductance changes are shown in Supplementary Figure 3onmesed to obtain VOLUME 23 NUMBER 10 OCTOBER 2005 NATURE BIOTECHNOLOGY
or B2 fM. Similar ultrasensitive detection was achieved in studies of carcinoembryonic antigen (CEA), 100 fg/ml or 0.55 fM, and mucin-1, 75 fg/ml or 0.49 fM, (see Supplementary Fig. 2 online) using silicon-nanowire devices modified with mAbs for CEA and mucin-1, respectively. We further investigated the devices’ reproducibility and selectivity in competitive binding experiments with bovine serum albumin (BSA) (Fig. 1d). Conductance-versus-time measurements recorded on a silicon-nanowire device modified with Ab1 showed similar conductance changes as above when 9 and 0.9 pg/ml solutions of PSA were delivered to the device. These results show that reproducible device-to-device sensitivity was achieved. Moreover, delivery of a solution containing 0.9 pg/ml PSA and 10 mg/ml BSA showed the same conductance increase as a solution containing only PSA at this concentration, whereas no conductance change was observed when the BSA solution alone was delivered. These latter data demonstrate excellent selectivity and also show that high sensitivity was not lost even with a 10-million-fold higher concentration of other proteins in solution. We investigated the devices’ modification chemistry to define their sensitivity limits. Atomic force microscopy measurements of the initial aldehyde-silane layer thickness on single nanowires (Fig. 1e) showed a systematic increase with modification time. This thickness increase is consistent with previous studies showing that similar silane reagents can form multilayers25,26. Measurements of sensitivity showed that the sensor response decreases rapidly for initial reaction times 430 min. The observed decrease in sensitivity is consistent with expectations for a field-effect sensing device27, and moreover, shows that the surface modification chemistry must be controlled to achieve reproducible high-sensitivity devices. Multiplexed detection with nanowire arrays For our initial studies of multiplexed detection, we used an array containing both p-type and n-type silicon-nanowire devices modified with Ab1. The incorporation of p- and n-type nanowires in a single sensor chip enables discrimination of possible electrical crosstalk and/or false-positive signals by correlating the response versus time from the two types of device elements. Notably, simultaneous conductance-versus-time data recorded from p-type nanowire (NW1, Fig. 2a) and n-type nanowire (NW2, Fig. 2a) devices after introduction of 0.9 ng/ml of PSA showed a conductance increase in NW1 and a conductance decrease in NW2, whereas the conductance returned to the baseline value of each device after introduction of buffer solution without PSA. The magnitudes of the conductance changes in the two devices were nearly the same and consistent with the concentration-dependent conductance measurements (Fig. 1). Similar behavior was observed with different concentrations of PSA (Fig. 2a); that is, the p- and n-type devices showed concentration-dependent increases and decreases, respectively, in conductance when the solution was alternated between PSA and buffer. These experiments demonstrate key points about multiplexed electrical detection with nanowire devices. First, the complementary conductance changes observed for the p-type and n-type elements are consistent with specific binding of PSA to field-effect devices, since the negatively charged protein will cause accumulation and depletion of charge carriers in the p-type and n-type nanowire elements, respectively. Second, the complementary electrical signals from p- and n-type devices provide a simple yet robust means for detecting false-positive signals from either electrical noise or nonspecific binding of protein to one device; that is, real and selective binding events must show complementary responses in the p- and n-type devices. The presence of correlated conductance signals in both devices (Fig. 2a), which occur at points when buffer and PSA/buffer solutions are changed, illustrates clearly how this multiplexing capability can be used to distinguish unambiguously noise from protein-binding signals. 2,000 1,800 1,800 1,600 1,400 1,200 0 900 1,800 2,700 3,600 1,600 1 2 3 45 1 2 34 1,400 1,200 1,000 Conductance (nS) Conductance (nS) 1,000 2,000 Time (s) Time (s) 3,000 4,000 5,000 NW2 NW1 NW2 NW1 0 a b Figure 2 Multiplexed detection with nanowire arrays. (a) Complementary sensing of PSA using p-type (NW1) and n-type (NW2) silicon-nanowire devices in the same array. The vertical solid lines correspond to times at which PSA solutions of (1) 0.9 ng/ml, (2) 0.9 ng/ml, (3) 9 pg/ml, (4) 0.9 pg/ml and (5) 5 ng/ml were connected to the microfluidic channel. (b) Conductance-versus-time data recorded simultaneously from two p-type silicon-nanowire devices in an array, where NW1 was functionalized with PSA Ab1, and NW2 was modified with ethanolamine. The vertical lines correspond to times when solutions of (1) 9 pg/ml PSA, (2) 1 pg/ml PSA, (3) 10 mg/ml BSA, (4) a mixture of 1 ng/ml PSA and 10 mg/ml PSA Ab1 were connected to the microfluidic channel. Black arrows in a and b correspond to the points where the solution flow was switched from protein to pure buffer solutions. Table 1 Summary of multiplexed detection experiments using nanowire devices modified with mAb receptors specific for free PSA (f-PSA), Ab1 and cross-reactive for f-PSA and PSA-ACT complex, Ab2 Conductance change, nS Protein sample [Protein] pg/ml NW1-Ab1 NW2-Ab2 f-PSA 1,700 192 154 f-PSA 850 185 132 f-PSA 8.5 98 81 f-PSA 0.85 45 50 f-PSA 0.085 15 10 PSA-ACT 3,200 ND 143 PSA-ACT 320 ND 124 PSA-ACT 3.2 ND 67 PSA-ACT 0.32 ND 19 f-PSA, PSA-ACT, Ab1 850, 3,200, 1 107 ND 140 f-PSA, PSA-ACT, Ab1 8.5, 320, 1 107 ND 118 f-PSA, PSA-ACT, Ab1 0.85, 3,200, 1 107 ND 138 f-PSA, PSA-ACT, Ab1 850, 0.32, 1 107 ND 15 f-PSA, Ab1 850, 1 107 ND ND ND corresponds to no detected conductance change. Sensor data used to obtain conductance changes are shown in Supplementary Figure 3 online. 1296 VOLUME 23 NUMBER 10 OCTOBER 2005 NATURE BIOTECHNOLOGY ARTICLES © 2005 Nature Publishing Group http://www.nature.com/naturebiotechnology
ARTICLES Figure 3 Multiplexed detection of cancer marker protein detection by three silicon-nanowire devices in an array Devices 1, 2 and 3 fabricated from similar nanowires, and then differentiated with distinct mAb receptors 1, red; 2, green; 3, blue)specific to three different cancer markers.(b) Conductance. L detection of PSA, CEA and mucin-l on p-type silicon-nanowire array in which Nwl, Nw2 and NW3 were functionalized with mAbs for psa CEA and mucin-l, respectively. The solutions 020004,0006,0008000 0 2,000 4.000 6.000 ere delivered to the nanowire array sequentially d SA, (3)0.2 ng/ml CEA, (4)2 pg/ml CEA, ml mucin-l, (6)5 pg/ml mucin-1 Buffer solutions were injected following each protein solution at points indicated by black es8g药88o arrows.(c)Conductance-.time data 1600 o recorded for the detection of PSA-containing 1,000 1,500 1500 ,0004,5006000 e donkey serum samples on a p-type silicon- Time(s) g nanowire array in which Nwl was functionalized 3 with Abl and NW2 was passivated with ethanolamine. The solutions were delivered to the nanowire array sequentially as follows: (1)1 uM phosphate buffer 3i+2 HM KCL, pH=7.4,(2)donkey serum, (3) a mixture of donkey serum and 90 pe/ml of PSA, (4) a mixture of donkey serum and 0.9 ng/ml of PSA. The onkey serum was injected at points indicated by the black arrows. (d)Conductance-versus-time data recorded for the same two p-type devices after addition of (1)a mixture of donkey serum and 0. 9 pg/ml of Psa, (2) donkey serum. (e)Conductance-versus time dat detection of PSA-containing human serum samples on a p-type silicon- nanowire array in which Nwl was functionalized with Abl and Nw2 was passivated 9 with ethanolamine. The solutions were delivered to the nanowire array sequentially as follows: (1)1uM phosphate buffer 2 uM KCl, PH=7.4,(2)a o mixture of human serum preblocked with 10 ug/ml Abl, (3)human serum and (4)same as(2) a device array consisting of p-type silicon-nanowire elements with dependent conductance changes only in NW2. In addition, control C either PSA Abl receptors(NWI, Fig. 2b) or surfaces passivated experiments in which solutions of f-PSA, PSA-ACT and Abl were with ethanolamine(NW2, Fig. 2b). Simultaneous measurements delivered showed concentration-dependent conductance changes in of the conductance of Nwl and NW2 show that well-defined, NW2 but not NWl, because f-PSA was blocked by Abl present in oncentration-dependent conductance increases were only observed the solution. These multiplexing results demonstrate selective, high- in Nwl delivery of PSA solutions (9 and 1 pg/ml), sensitivity detection of both markers and show that nanowire sensor although small conductance spikes were observed in both devices at arrays measured the f-PSA and PSA-ACT concentrations in a single he points where PSA and buffer solutions were changed. Delivery of real-time assay. BSA at 10 ug/ml showed no response in either NWl or NW2, and We also investigated multiplexed detection of three cancer marker subsequent delivery of a solution of Psa(I ng/ml) and Abl proteins, f-PSA, CEA and mucin-1, using silicon-nanowire devices 10 ug/ml), which complexes the free PSA, did not cause measurable functionalized with mAbs for f-PSA (NWI), CEA(NW2)and mucin-1 conductance changes in either device. Together, these multiplexing (NW3)(Fig. 3a). Conductance-versus-time measurements were experiments demonstrate that the electronic signals measured can recorded simultaneously from NWl, NW2 and Nw3 as different be readily attributed to selective marker-protein binding, show that protein solutions were sequentially delivered(Fig. 3b). First, intro our surface passivation chemistry effectively prevents nonspecific duction of f-PSA and buffer solutions led to concentration-dependent protein binding, and provide a robust means for discriminating conductance increases only when NwI was exposed to PSA solution; against false-positive signals arising from no conductance changes were observed in NW2 or NW3. Similarly, introduction of CEA solutions yielded concentration-dependent con- ductance changes only in NW2, whereas subsequent delivery of Multiplexed detection of cancer markers mucin-l solutions resulted in concentration-dependent conductance vire arrays for multiplexed detec- changes only in Nw3. These results demonstrate multiplexed real- tion of marker proteins relevant to cancer, we first focused on free PSA time, label-free marker-protein detection with sensitivity to the (f-PSA)and PSA-al-antichymotrypsin(PSA-ACT)complex, which femtomolar level and essentially complete selectivity are generally measured in the diagnosis of prostate cancer->. we fabricated a device array from p-type silicon-nanowire elements that Protein detection in serum samples were then modified either with mAbs for f-PSA, Abl, or mAbs that Cancer diagnosis will require rapid analysis of clinically relevant cross-react with f-PSA and the PSA-ACT complex, Ab2. Simultaneous samples, such as blood serum. To this end, we investigated detection conductance measurements of Nwl, which was modified with Abl, of PSA in undiluted donkey and human serum samples that were and NW2, which was modified with Ab2, were carried out for a wide desalted in a rapid and simple purification step. The measurements range of conditions(see Supplementary Fig. 3 online)and are were made using p-type silicon-nanowire elements with either PSA summarized in Table 1. The data show that delivery of f-PSA resulted Abl receptors(NWl, Fig. 3c-e)or surfaces passivated with ethanol in concentration-dependent conductance changes in both NWl and amine(NW2, Fig 3c-e)in the same sensor array. ATURE BIOTECHNOLOGY VOLUME 23 NUMBER 10 OCTOBER 2005
We carried out a second test of multiplexing capabilities using a device array consisting of p-type silicon-nanowire elements with either PSA Ab1 receptors (NW1, Fig. 2b) or surfaces passivated with ethanolamine (NW2, Fig. 2b). Simultaneous measurements of the conductance of NW1 and NW2 show that well-defined, concentration-dependent conductance increases were only observed in NW1 upon delivery of PSA solutions (9 and 1 pg/ml), although small conductance spikes were observed in both devices at the points where PSA and buffer solutions were changed. Delivery of BSA at 10 mg/ml showed no response in either NW1 or NW2, and subsequent delivery of a solution of PSA (1 ng/ml) and Ab1 (10 mg/ml), which complexes the free PSA, did not cause measurable conductance changes in either device. Together, these multiplexing experiments demonstrate that the electronic signals measured can be readily attributed to selective marker-protein binding, show that our surface passivation chemistry effectively prevents nonspecific protein binding, and provide a robust means for discriminating against false-positive signals arising from either electronic noise or nonspecific binding. Multiplexed detection of cancer markers To test the capabilities of the nanowire arrays for multiplexed detection of marker proteins relevant to cancer, we first focused on free PSA (f-PSA) and PSA-a1-antichymotrypsin (PSA-ACT) complex, which are generally measured in the diagnosis of prostate cancer28–30. We fabricated a device array from p-type silicon-nanowire elements that were then modified either with mAbs for f-PSA, Ab1, or mAbs that cross-react with f-PSA and the PSA-ACT complex, Ab2. Simultaneous conductance measurements of NW1, which was modified with Ab1, and NW2, which was modified with Ab2, were carried out for a wide range of conditions (see Supplementary Fig. 3 online) and are summarized in Table 1. The data show that delivery of f-PSA resulted in concentration-dependent conductance changes in both NW1 and NW2, whereas the introduction of PSA-ACT yielded concentrationdependent conductance changes only in NW2. In addition, control experiments in which solutions of f-PSA, PSA-ACT and Ab1 were delivered showed concentration-dependent conductance changes in NW2 but not NW1, because f-PSA was blocked by Ab1 present in the solution. These multiplexing results demonstrate selective, highsensitivity detection of both markers and show that nanowire sensor arrays measured the f-PSA and PSA-ACT concentrations in a single real-time assay. We also investigated multiplexed detection of three cancer marker proteins, f-PSA, CEA and mucin-1, using silicon-nanowire devices functionalized with mAbs for f-PSA (NW1), CEA (NW2) and mucin-1 (NW3) (Fig. 3a). Conductance-versus-time measurements were recorded simultaneously from NW1, NW2 and NW3 as different protein solutions were sequentially delivered (Fig. 3b). First, introduction of f-PSA and buffer solutions led to concentration-dependent conductance increases only when NW1 was exposed to PSA solution; no conductance changes were observed in NW2 or NW3. Similarly, introduction of CEA solutions yielded concentration-dependent conductance changes only in NW2, whereas subsequent delivery of mucin-1 solutions resulted in concentration-dependent conductance changes only in NW3. These results demonstrate multiplexed realtime, label-free marker-protein detection with sensitivity to the femtomolar level and essentially complete selectivity. Protein detection in serum samples Cancer diagnosis will require rapid analysis of clinically relevant samples, such as blood serum. To this end, we investigated detection of PSA in undiluted donkey and human serum samples that were desalted in a rapid and simple purification step. The measurements were made using p-type silicon-nanowire elements with either PSA Ab1 receptors (NW1, Fig. 3c–e) or surfaces passivated with ethanolamine (NW2, Fig. 3c–e) in the same sensor array. Figure 3 Multiplexed detection of cancer marker proteins. (a) Schematic illustrating multiplexed protein detection by three silicon-nanowire devices in an array. Devices 1, 2 and 3 are fabricated from similar nanowires, and then differentiated with distinct mAb receptors (1, red; 2, green; 3, blue) specific to three different cancer markers. (b) Conductanceversus-time data recorded for the simultaneous detection of PSA, CEA and mucin-1 on p-type silicon-nanowire array in which NW1, NW2 and NW3 were functionalized with mAbs for PSA, CEA and mucin-1, respectively. The solutions were delivered to the nanowire array sequentially as follows: (1) 0.9 ng/ml PSA, (2) 1.4 pg/ml PSA, (3) 0.2 ng/ml CEA, (4) 2 pg/ml CEA, (5) 0.5 ng/ml mucin-1, (6) 5 pg/ml mucin-1. Buffer solutions were injected following each protein solution at points indicated by black arrows. (c) Conductance-versus-time data recorded for the detection of PSA-containing donkey serum samples on a p-type siliconnanowire array in which NW1 was functionalized with Ab1 and NW2 was passivated with ethanolamine. The solutions were delivered to the nanowire array sequentially as follows: (1) 1 mM phosphate buffer + 2 mM KCl, pH ¼ 7.4, (2) donkey serum, (3) a mixture of donkey serum and 90 pg/ml of PSA, (4) a mixture of donkey serum and 0.9 ng/ml of PSA. The donkey serum was injected at points indicated by the black arrows. (d) Conductance-versus-time data recorded for the same two p-type silicon-nanowire devices after addition of (1) a mixture of donkey serum and 0.9 pg/ml of PSA, (2) donkey serum. (e) Conductance-versus-time data recorded for the detection of PSA-containing human serum samples on a p-type silicon-nanowire array in which NW1 was functionalized with Ab1 and NW2 was passivated with ethanolamine. The solutions were delivered to the nanowire array sequentially as follows: (1) 1 mM phosphate buffer + 2 mM KCl, pH ¼ 7.4, (2) a mixture of human serum preblocked with 10 mg/ml Ab1, (3) human serum and (4) same as (2). 123 1 23 2,250 2,100 1,950 1,800 1,650 1,500 1,800 1,700 1,600 0 2,000 4,000 Conductance (nS) Conductance (nS) Conductance (nS) Conductance (nS) 6,000 8,000 0 2,000 4,000 6,000 8,000 NW3 NW2 NW2 NW1 NW1 NW1 NW2 NW2 NW1 1 1 2 23 4 345 6 0 1 2 34 0 1 2 1,740 1,720 1,620 1,800 1,700 1,600 500 1,500 3,000 4,500 Time (s) Time (s) 1,000 1,500 6,000 a d e b c NATURE BIOTECHNOLOGY VOLUME 23 NUMBER 10 OCTOBER 2005 1297 ARTICLES © 2005 Nature Publishing Group http://www.nature.com/naturebiotechnology
ARTICLES Figure 4 Sch (Top)Silicon-nanowire devices modified with oligonucleotide primer. (Middle)A solution containing telomerase is delivered to the device array. and telomerase binds in a concentration-dependent process. Bottom) Subsequent addition of dNTPs leads to telomerase-catalyzed primer extension/elongation. showed little change in baseline for either NWl or NW2, although subsequent addition of undiluted human serum showed a well defined conductance increase in NWl. These results clearly demon strate the detection of cancer markers with high sensitivity and selectivity in human serum. Telomerase detection and activity ● dNTPs To define further the potential of our silicon-nanowire arrays as cancer diagnostic tools, we investigated a nucleic acid-based marker assay involving detection of telomerase. Telomerase is a eukaryotic ribonu- cleoprotein complex,4 that catalyzes the addition of the telomeric repeat sequence TTAGGG to the ends of chromosomes using its intrinsic RNA as a template for reverse transcription334.Telomerase known human cancers >,o, but no activity was found in most of the adjacent somatic tissues, and telomerase has thus been proposed as a marker for cancer g Conductance-versus-time measurements recorded simultaneously detection and as a therapeutic target. e from NWl and NW2 as different donkey serum solutions were The telomerase assay is illustrated in Figure 4. First, silicon o sequentially delivered to the devices are shown in Figure 3c,d Delivery nanowire device elements within an array are functionalized with E of donkey serum containing 59 mg/ml total protein did not lead to an oligonucleotide primers complementary to the telomerase bindi appreciable conductance change relative to the standard assay buffer. site. Second, the presence or absence of telomerase is then detected by Donkey serum solutions containing f-PSA led to concentration- monitoring the nanowire conductance after delivery of a sample cell a dependent conductance increases only for NWl; no conductance extract to the device array. In the case of p-type nanowire device g changes were observed in NW2(Figs. 3c, d). Well-defined conductance elements, binding should reduce the conductance, as telomerase changes were observed for PSA concentrations as low as 0.9 pg/ml, (pl 10)7 is positively charged at physiological pH. Third,addition hich corresponds to a concentration 100-billion times lower than of deoxynucleotide triphosphates(dNTPs) leads, in the presence of hat of the background serum proteins. Similar results were obtained telomerase, to elongation, which should produce an increase in for human serum samples( Fig. 3e). Specifically, addition of undiluted conductance owing to the incorporation of negatively charge o human serum, which contains f-PSA, blocked with an excess of Abl, nucleotides near the nanowire surface a b 880 1.020 Time(s) Figure 5 Detection of telomerase (a)Conductance-versus-time data recorded for oligonucleotide-modified p-type silicon-nanowire devices, after the introduction of (l) a solution containing extract from 100 HeLa cells and 0. 4 mM dcTP, (2)a mixture all four dntPs(dATP, dgtP, duTP and dcTP)each at O 1 mM, (3)a solution containing extract from 100 HeLa cells and 0. 4 mM dCTP and (4)0.4 mM dCTP only Points()and (4)were recorded using a second device. ( b)Conductance-versus time data recorded for oligonucleotide-modified nanowire device after delivery of (1) a solution containing extract from 100,000 normal human fibroblast cells and 0. 4 mM dcTP, (2)a mixture of all four dntPs each at o 1 mM, (3)a solution containing extract from 10,000 HeLa cells, 0.4 mM dCTP, and 5 uM oligonucleotide(sequence: 5-TTTTITAATCCGTCGAGCAGAGTT-3), (4)a mixture of all four dNTPs each at 0.1 mM, (5)a solution containing extract from 10, 000 heat-deactivated HeLa cells(90C, 10 min) and 0.4 mM dCTP and(6)a mixture of all four dNTPs at 0. 1 mM. (c)Conductance-versustime data recorded on a p-type silicon- nanowire device after the introduction of (1)a solution containing extract 100 HeLa cells and 0. 4 mM dCTP, and (2)a mixture of all four dNTPs each at o1 mM and 20 uM AZTTP. Inset: Plot of the inhibition of elongation AZTTP concentration, where 100% corresponds to conductance change associated with elongation in the absence of AZTTP VOLUME 23 NUMBER 10 OCTOBER 2005 NATURE BIOTECHNOLOGY
Conductance-versus-time measurements recorded simultaneously from NW1 and NW2 as different donkey serum solutions were sequentially delivered to the devices are shown in Figure 3c,d. Delivery of donkey serum containing 59 mg/ml total protein did not lead to an appreciable conductance change relative to the standard assay buffer. Donkey serum solutions containing f-PSA led to concentrationdependent conductance increases only for NW1; no conductance changes were observed in NW2 (Figs. 3c,d). Well-defined conductance changes were observed for PSA concentrations as low as 0.9 pg/ml, which corresponds to a concentration B100-billion times lower than that of the background serum proteins. Similar results were obtained for human serum samples (Fig. 3e). Specifically, addition of undiluted human serum, which contains f-PSA, blocked with an excess of Ab1, showed little change in baseline for either NW1 or NW2, although subsequent addition of undiluted human serum showed a welldefined conductance increase in NW1. These results clearly demonstrate the detection of cancer markers with high sensitivity and selectivity in human serum. Telomerase detection and activity To define further the potential of our silicon-nanowire arrays as cancer diagnostic tools, we investigated a nucleic acid–based marker assay involving detection of telomerase. Telomerase is a eukaryotic ribonucleoprotein complex31,32 that catalyzes the addition of the telomeric repeat sequence TTAGGG to the ends of chromosomes using its intrinsic RNA as a template for reverse transcription33,34. Telomerase activity has been found in at least 80% of all known human cancers35,36, but no activity was found in most of the adjacent somatic tissues, and telomerase has thus been proposed as a marker for cancer detection and as a therapeutic target. The telomerase assay is illustrated in Figure 4. First, siliconnanowire device elements within an array are functionalized with oligonucleotide primers complementary to the telomerase binding site. Second, the presence or absence of telomerase is then detected by monitoring the nanowire conductance after delivery of a sample cell extract to the device array. In the case of p-type nanowire device elements, binding should reduce the conductance, as telomerase (pI B 10)37 is positively charged at physiological pH. Third, addition of deoxynucleotide triphosphates (dNTPs) leads, in the presence of telomerase, to elongation, which should produce an increase in conductance owing to the incorporation of negatively charged nucleotides near the nanowire surface. Telomerase dNTPs Figure 4 Schematic of the telomerase binding and activity assay. (Top) Silicon-nanowire devices modified with oligonucleotide primer. (Middle) A solution containing telomerase is delivered to the device array, and telomerase binds in a concentration-dependent process. (Bottom) Subsequent addition of dNTPs leads to telomerase-catalyzed primer extension/elongation. 960 950 1,060 100 80 60 40 % Elongation 20 0 0 5 10 [AZTTP] (µM) 15 20 1,040 1,020 1,000 980 940 930 920 910 900 0 200 400 600 800 1,000 0 200 400 600 800 1,000 1,200 880 1 1 1 3 4 5 6 2 2 2 3 4 800 720 0 400 Conductance (nS) 800 Time (s) Time (s) Time (s) 1,200 1,600 ab c Figure 5 Detection of telomerase (a) Conductance-versus-time data recorded for oligonucleotide-modified p-type silicon-nanowire devices, after the introduction of (1) a solution containing extract from 100 HeLa cells and 0.4 mM dCTP, (2) a mixture all four dNTPs (dATP, dGTP, dUTP and dCTP) each at 0.1 mM, (3) a solution containing extract from 100 HeLa cells and 0.4 mM dCTP and (4) 0.4 mM dCTP only. Points (3) and (4) were recorded using a second device. (b) Conductance-versus-time data recorded for oligonucleotide-modified nanowire device after delivery of (1) a solution containing extract from 100,000 normal human fibroblast cells and 0.4 mM dCTP, (2) a mixture of all four dNTPs each at 0.1 mM, (3) a solution containing extract from 10,000 HeLa cells, 0.4 mM dCTP, and 5 mM oligonucleotide (sequence: 5¢-TTTTTTAATCCGTCGAGCAGAGTT-3¢), (4) a mixture of all four dNTPs each at 0.1 mM, (5) a solution containing extract from 10,000 heat-deactivated HeLa cells (90 1C, 10 min) and 0.4 mM dCTP and (6) a mixture of all four dNTPs each at 0.1 mM. (c) Conductance-versus-time data recorded on a p-type silicon-nanowire device after the introduction of (1) a solution containing extract from 100 HeLa cells and 0.4 mM dCTP, and (2) a mixture of all four dNTPs each at 0.1 mM and 20 mM AZTTP. Inset: Plot of the inhibition of elongation versus AZTTP concentration, where 100% corresponds to conductance change associated with elongation in the absence of AZTTP. 1298 VOLUME 23 NUMBER 10 OCTOBER 2005 NATURE BIOTECHNOLOGY ARTICLES © 2005 Nature Publishing Group http://www.nature.com/naturebiotechnology