A Point-of-Care Immunosensor for Human Chorionic Gonadotropin in Clinical Urine Samples Using a Cuneated Polysilicon Nanogap Lab-on-Chip

Human chorionic gonadotropin (hCG), a glycoprotein hormone secreted from the placenta, is a key molecule that indicates pregnancy. Here, we have designed a cost-effective, label-free, in situ point-of-care (POC) immunosensor to estimate hCG using a cuneated 25 nm polysilicon nanogap electrode. A tiny chip with the dimensions of 20.5 × 12.5 mm was fabricated using conventional lithography and size expansion techniques. Furthermore, the sensing surface was functionalized by (3-aminopropyl)triethoxysilane and quantitatively measured the variations in hCG levels from clinically obtained human urine samples. The dielectric properties of the present sensor are shown with a capacitance above 40 nF for samples from pregnant women; it was lower with samples from non-pregnant women. Furthermore, it has been proven that our sensor has a wide linear range of detection, as a sensitivity of 835.88 μA mIU-1 ml-2 cm-2 was attained, and the detection limit was 0.28 mIU/ml (27.78 pg/ml). The dissociation constant Kd of the specific antigen binding to the anti-hCG was calculated as 2.23 ± 0.66 mIU, and the maximum number of binding sites per antigen was Bmax = 22.54 ± 1.46 mIU. The sensing system shown here, with a narrow nanogap, is suitable for high-throughput POC diagnosis, and a single injection can obtain triplicate data or parallel analyses of different targets.

In most instances, lab-dependent assays require a large working space and long durations and involve the use of chemicals, which slows down the result acquisition processes. The development of independent point-of-care (POC) electrochemical immunosensors will overcome these limitations and promises a number of significant advantages, including simplicity, fast response, suitability for onsite testing, high sensitivity and time and cost savings. Consequently, the frequency of hospital visits, travel expenses and lost productivity on the job could be reduced. The success of 'bedside analyzing systems', such as glucose meters, has promoted the development of similar quantitative biosensing devices for pregnancy, which rely on myriad detection schemes. One of the promising alternatives to existing biosensors is capacitive based and can be used for POC applications. Capacitive biosensors are favored due to their ability to detect the analyte directly in a sample with little or no preparation [33]. The capacitive behavior of this type of sensor is based on parameters such as changes in dielectric properties, charge distribution, permittivity and conductivity that originate from the complex matrix formation on the surface of the transducing electrodes. Hence, changes in such parameters could be measured when biomolecule binding occurs on the electrode surface. In the quest for higher sensitivities and more accurate measurements, the sensing electrode shown in the present study could be used as a capacitive sensor.
To develop the sensor, we used standard semiconducting materials, with silicon as the substrate and polysilicon (PS) as the sensing electrode. Polysilicon was chosen for its ability to withstand high temperatures and its high compatibility with standard semiconductor processing steps. The advantages of this material also include high sensitivity, a fast response time and high accuracy, as previously reported [34,35]. In the quest to reduce the use of expensive equipment, a conventional lithography microfabrication technique was implemented; this technique can be used to fabricate simple and non-complex nanostructures such as nanogaps and nanowires [36,37]. In this study, we developed a narrow-sized nanogap-based sensor with a single injection capability to obtain triplicate data or parallel analyses against different targets. Furthermore, because it is possible to block unused channel(s) to limit the sample usage, this system is more cost effective and easier to use. All abbreviation and symbols used in this study are described in Table 1.

Design and fabrication of PSNG lab-on-chip (LOC)
The patterns for the PSNG electrode, contact pads and microfluidics were designed in Auto-CAD for pattern transfer via mask aligner (Fig 1a). The PSNG mask has a critical dimension of 1 μm at the electrode gap. The 20.5 mm × 12.5 mm LOC immunosensor is shown in Fig 1b. The LOC consists of three standalone PSNG electrodes as the base and is bonded with microfluidic channels that serve as the fluid delivery system to the specific electrodes (Fig 1c) [38][39][40][41][42][43][44][45][46][47]. A conventional lithography technique was used to fabricate the PSNG immunosensor, as reported elsewhere [37,43]. Fig 2a-2f explains the simplified fabrication process steps used to develop the PSNG electrode. Oxide was grown by wet oxidation, and polysilicon (PS) was deposited by the low-pressure chemical vapor deposition (LPCVD) technique. The deposited PS wafer was then annealed for 12 h at 600°C to allow the disordered PS molecules to rearrange after deposition. Al was first deposited as a sacrificial layer and patterned using the MIDAS Exposure System MDA-400M for the first electrode mask before PS etching, which used the reactive ion  etching (RIE) technique. To obtain a second mask pattern for Ti/Au contact padding, lift-off of Ti and Au was carried out after their deposition with a thermal evaporator (Auto 306 thermal evaporator; Edwards High Vacuum International, Wilmington, MA, USA). The cuneated PS electrode was first oxidized at 900°C in a dry O 2 atmosphere to form an oxide layer while simultaneously expanding the PS. Subsequent BOE etching revealed this expanded PS, and the entire oxidation and etching process was repeated until the desired nanogap (NG) was achieved (size expansion technique). The oxidation time varied from 5 to 15 min in each cycle of oxidation, and etching depended on the thickness of the polysilicon layer. The final fabricated PSNG electrode, bonded to the microfluidic channel via plasma oxidation, is shown in Fig 2g. The microfluidic layer on the LOC was fabricated using an SU-8 soft-lithography microstructure mold (Fig  2h-2j). A mixture of Sylgard 184 Silicone Elastomer Base and Sylgard 184 Silicone Elastomer Curing Agent was poured onto the SU-8 microstructure mold to form the PDMS polymer.

Surface modification on PSNG
PSNG surface hydroxylation. Because the PSNG structure is a non-biocompatible surface, it had to be modified to exhibit adhesion properties with biomolecules. The functionalization procedure is schematically shown in Fig 3. First, 10 ml of 0.1 M H 2 SO 4 and 12.5 ml of H 2 O 2 were diluted in water to a total volume of 100 ml. Then, 40 mg of EDTA were added, followed by 1.2 g of FeSO 4 incrementally added to the solution, which reacted and heated up during the addition process. When the solution became bubbly and brown, the PSNG electrode was immersed for 30 min for hydroxylation to occur, which is known as a Fenton reaction. This process produces a-OH terminated group onto PSNG (Fig 3a). hCGab immobilization on APTES. The amine-terminated PS surface was prepared by applying 1 μL of 10% APTES in water onto the PSNG electrode to establish a uniform layered surface (Fig 3b). The electrode was then incubated for 2 h and rinsed with ethanol to remove any unreacted APTES. hCGab (1 mg) was mixed with 1 ml DI water, 0.1 g of NaHCO 3 , the previously prepared mixture of 0.05 g tBOC and 1 ml chloroform. Then, 25 mM of EDAC and 50 mM of NHS were prepared in 0.01 M PBS solution. First, EDAC was added to the Ab solution and incubated for 20 min and then incubated in NHS for 10 min. Then, 1 μL of hCGab (8.6 mg/ ml) solution was applied onto the PSNG electrode and incubated for 3 h at room temperature and humidity of 85.5% (Fig 3c). Subsequently, the remaining free surface amine groups on the PSNG electrode after Ab immobilization were blocked with 5% BSA in PBS for 10 min to avoid non-specific binding of hCG on the electrode. Then, the-NH 2 protected by tBOC was removed by immersing the electrode in 10 mM TFA solution prior to detection with the hCG target [48].
hCG antigen binding. Clinical urine samples and hCG standard control solutions were reacted with the surface-modified PSNG electrode to analyze the signal transduction characteristics. Clinical samples consisting of pregnant and non-pregnant women's urine were first analyzed in a clinical laboratory to estimate the hCG concentration in the urine and the presence of impurities. Furthermore, hCG standard control samples of 0, 8, 16, 40, 100, and 240 mIU/ml from the ELISA Kit were also applied to the PSNG electrode as a reference to validate the immunosensor. Before electrical characterization was conducted, 1 μL of hCG sample concentration was applied to the PSNG electrode and incubated for 3 h at room temperature and 85.5% humidity (Fig 3d).

PSNG Electrode Characterization
Optical and morphological characterization. Inspections of the PSNG during fabrication were performed using high-power microscopy (OLYMPUS-BX51) to make sure no contaminants were present. The fabricated nanogaps were also inspected and their gap size measured using scanning electron microscopy (SEM, JEOL JSM-6460LA). Chemical functionalization and Ag-Ab surface modification on 1 x 1 cm PS surface samples were inspected using field emission scanning electron microscopy (FESEM, Zeiss Ultra55) and Fourier transform infrared spectroscopy (FTIR, Spectrum 65 Perkin-Elmer). Then, for surface roughness and grain size measurement of PS, atomic force microscopy (AFM) (SPA400-SPI3800, Seiko Instruments Inc., Japan) was used. Furthermore, the luminescence properties of APTES were studied using photoluminescence (PL) (Horiba Fluorolog-3 for PL spectroscopy, HORIBA Jobin Yvon Inc., USA). The thin-film layer thickness and uniformity were observed using a Filmetrics F20-UV spectrometer.
Electrical characterization. Electrical measurements were conducted to analyze signals transduced by the Ag-Ab binding reaction from the PSNG electrode, which were probed via the Au contact pad on the LOC (Fig A in S1 File). AC measurements were conducted using an impedance spectrometer (Alpha-A High Performance Frequency Analyzer, Novocontrol Technologies, Hundsangen, Germany). A two-wire probe station was used with the impedance spectrometer to measure capacitance (C), permittivity (ε), loss tangent (tan δ) and conductivity (σ). Measurement was started at 0.1 V with a scaling factor of 1.4 and a frequency range of (1 x 10 0 to 1 x 10 6 Hz). The loss tangent (tan δ) value was continuously monitored during capacitance measurement to ensure it stayed below 100, indicating that the immunosensor was behaving as a capacitor. A Kiethley 6487 Picoammeter was used to measure current-voltage DC characteristics of the PSNG electrode, from which amperometric characteristics and sensitivity of the immunosensor could be analyzed. At each step during hCG detection, the PSNG electrodes were washed with DI water to ensure that measurement data were valid.

Results and Discussion
Several different sensing methods have been proposed to detect and discriminate hCG for pregnancy tests with various detection limits. Factors needed for a good sensor, such as sensitivity, selectivity, stability, reproducibility and simplicity, are critical and are achieved in the present study. Using the size expansion technique, a small nanogap size of 25 nm was achieved. Analyses were also conducted on the different gap sizes to measure their resistivity and the benefits of the small gap size. Analyses were performed using clinical urine samples and were correlated with hCG standard control samples for stability and selectivity. In addition, a single injection option facilitates triplicate data analyses, and the sensing surface can be regenerated.
Morphological analysis of the PSNG SEM, FESEM, FTIR and AFM analyses. SEM analysis of the fabricated PSNG electrodes showed that the smallest gap obtained was 25 nm from the size expansion process (Fig 4a-4c). Both SEM and FESEM (in red boxes) images were acquired to determine the gap sizes. Because the conventional lithography method was used, uneven nanogap geometrical shapes were the outcome. However, the reproducibility of the desired nanogaps is higher, at 49%, with nanogaps fabricated on a wafer, making this a great success (Fig B in S1 File). Surface modification and Ag-Ab binding on PSNG were examined under FTIR and FESEM. It was clearly observed that binding occurs on PS grains, which proves that good surface roughness is important for Ag-Ab attachments (Fig 4d). This finding was further supported by FTIR spectrometry, which could analyze PSNG/APTES, PSNG/APTES/hCGab and PSNG/APTES/hCGab/hCG complexes (Fig C in S1 File). The peaks of C-O stretching and N-H bending were observed at 1049-1201 cm -1 and 1641-1676 cm -1 , respectively, demonstrating low transmittance peaks after hCG target binding. Therefore, this spectrum confirms the presence of an additional layer over hCGab. The PSNG surface roughness properties were examined using AFM, as shown in Fig 4e, which showed an overall surface roughness (S a ) of 27.30 nm at a peak-valley (P-V) of 1.03 μm and an RMS value of 44.53 nm. However, with a small scan area, the surface roughness at the middle of the gap was S a = 6.68 nm, and the P-V was 161.20 nm. This indicates that the fabricated PSNG structure exhibits a good surface-to-volume ratio that enhances the biocompatibility of the immunosensor [49]. Fluorometry analysis of polysilicon. The photoluminescence properties of PS were examined using bare PS wafers via PL spectroscopy, with excitation from a 325 nm xenon lamp at both visible (VIS) and ultraviolet (UV) regions. Fig 5a shows the PL spectra of PS, which shows emission peaks at 221 and 465 nm. It is known that the emission band in the UV region is caused by a collisional recombination process and that the emission band in the VIS region is caused by electron-hole generation, depending on ionized vacancies. Therefore, the luminescence properties at both the UV and VIS regions reveal that the crystallinity of PS is not affected by a reduction in the annealing process [50].
Size expansion technique by oxide growth. The size expansion technique was used to reduce the gap size of PS. The initial patterned design on the silicon substrate had a 1 μm gap. During thermal oxidation, the patterned PS structure expanded as oxygen was being supplied (Fig 5b). After oxidation, the samples were immersed in BOE solution to remove excess oxide, leaving only the expanded underlying PS structure. Oxidation time reduced from 60 to 20 min (Fig 4a-4c) as the gap size reduced. The film thickness during each oxidation process was measured using Filmetrics Spectrometer F20-UV. Regression analysis of the oxidation process showed a linear regression, R 2 , of 0.8829 and linearity, y = 35.906x + 898.38, was achieved, where x is growth time and y is the film thickness. With the narrow nanogap obtained on our sensor, the effect of electrode polarization could be reduced, thus resulting in a negligible noise effect [37,51].

Effect of gap size on PSNG sensitivity
The DC electrical characteristics of the PSNG electrode were examined during chemical functionalization and attachment of APTES and hCGab on the fabricated PSNG. Fig 5c illustrates the investigation of l g = 75, 60 and 25 nm gap sizes, which were applied with 1 μL of prepared 8.6 mg/ml hCGab on APTES-modified PSNG. The graph shows that the resistance drops gradually when the gap size decreases from 75 to 60 to 25 nm, and the measured resistances were 65.7, 55.7, and 22.6 μΩ, respectively, as proven by Eqs 1 and 2. Therefore, to enhance the sensitivity of the electrode, a gap size of lg = 25 nm was selected for hCG measurement because it exhibits low resistance (S1 File).

PSNG current; I PSNG ¼
Av 2r e l e þ r air l g ð1Þ It is also known that a small gap size, l g , could reduce the electrode polarization effect, which is a major source of error in nanogap electrodes. Therefore, it was suggested that l g less than 100 nm could reduce the effect because the accumulation of ions known as the Debye layer at the smallest gap showed negligible error. Hence, reduction by l g causes the Debye layer to overlap, and the potential drop could be reduced [37,52].

DC and AC measurements on PSNG electrodes-clinical sample analyses
Current-voltage characteristics. The prepared PSNG electrode functionalized with APTES followed by hCGab immobilization was tested with pregnant and non-pregnant women's urine samples, followed by hCG standard control samples. Each modified layer on the PSNG surface exhibits ohmic characteristics that influence the electrode resistivity according to R T = (ρ TlT )/A. A total of six samples were tested (n = 6) for hCG, with three samples each from pregnant and non-pregnant women. In the pregnant women's urine samples, the gestational ages of pregnancy were 4, 9 and 12 weeks, measured as 970, 16,200 and 230,000 mIU/ml, respectively, as obtained from a clinical laboratory. These results clearly indicated that there were no non-specific interferences by albumin, epithelial or glucose in the samples (Table 2). Therefore, by testing with our PSNG immunosensor, it was observed that an increase in hCG concentrations caused a reduction in current with both urine and hCG standard control samples (Fig 5d). hCG Current; I hCG ¼ Av 2ðr e l e þ r SM l SM Þ þ r hCG l hCG ð3Þ Therefore, changes in hCG current, ΔI hCG , were due to the changes in resistivity of the hCG concentration, Δρ hCG , as proven in Eq 3, while other parameters remained constant (S1 File). It can thus be inferred that there is a binding event between hCGab and hCG on the surface of the PSNG electrode, which can transduce current based on hCG concentration [53].
Capacitance-frequency characteristics. During AC detection, each modification layer on the PSNG surface was signified as a multilayer dielectric material, and both PSNG electrodes were signified as parallel plate capacitors, thus behaving as a capacitive immunosensor based on capacitance, C = ε r ε o A/d. This distinctive property of the PSNG electrode was used in this study and was clearly reflected in the results obtained from capacitance, permittivity (Fig D in S1 File) and conductivity (Fig E in S1 File) signals upon hCG target binding on the electrodes. The capacitance and permittivity properties of PSNG increased when urine samples with higher hCG concentrations were applied, as shown in Eq 4. The changes in hCG concentration cause the change in hCG permittivity, Δε hCG , which eventually increases the hCG capacitance, C hCG .
Therefore, those samples in which hCG has reacted with hCGab, forming a peptide bond, correspond to an increment in C hCG of the PSNG electrode observed in the range 80 to 100 nF with the pregnant women's urine samples [54]. However, samples without hCG do not exhibit this characteristic and exhibit C hCG in the range of 30 to 40 nF. The sensor ε hCG values ranged from 160 F/m to 185 F/m for pregnant samples. In addition, the measured capacitance is not only due to the sample capacitance but also influenced by the sample conductance, as shown in Eq 5. The sensor conductivity values ranged from 80 nS/cm to 100 nS/cm with increasing hCG levels in the pregnant urine samples. Lonappan et al. [55] and Sánchez et al. [56] have reported that an increment in hCG levels leads to an increased level of conductivity. Hence, in the sensor operation, it can be deduced that the measured capacitance values were due to the dielectric term and were also related to sample conductivity [57].
PSNG electrode analytical validation for specificity, sensitivity and stability The obtained experimental results are highly valuable in calculating the performance of the sensor and its capabilities. In the present investigation, various analyses, such as Ag-Ab specific binding, sensitivity, selectivity, limit of detection (LOD), reproducibility and stability were carried out and are shown in Fig 6. During these tests, a non-specific blocking agent, BSA, was used to ensure that only hCG molecules bind to the hCGab probe. Therefore, one-site specific binding was analyzed, the equilibrium dissociation constant (K d ) was calculated to be 2.23 ± 0.66 mIU, and the maximum number of binding sites per antigen would be B max = 22.54 ± 1.46 mIU, as indicated in Fig 6a and Eq 6, where x is reaction time and y is hCG concentration.
These results indicate a good affinity for hCG detection and the importance of sensitive PSNG electrodes in biosensing [58][59][60]. When a wide range of hCG concentrations were tested, sensitivity analysis determined the sensor's ability to transduce significant currents. The linearity achieved for hCG standards was with a correlation coefficient R 2 = 0.9943, while for urine samples it was with R 2 = 0.9279 (Fig 6b). Therefore, our sensor exhibits a sensitivity of 835.88 μA mIU -1 ml -2 cm -2 at the active sensing area, with 'A' of 0.00262 cm 2 and a gradient of 2.19 μA mIU -1 ml -2 , as calculated from Eq 9. The proposed sensitivity is higher than the reported values of 0.255 μA mIU -1 ml -2 by Tan et al. [26] and 0.1032 μA mIU -1 ml -2 by Li et al. [27]. Here, the gradients of the urine sample plots are larger than standard reference samples, which could be due to a non-linear concentration range in the urine samples. Therefore, the gradients indicate normal behavior of urine samples in that they do not contain only hCG, unlike the ideal reference hCG samples. This proves that, even though our immunosensor is highly selective, the linearity depends on the type of sample. Therefore, this sensor could also measure hCG that exists in a different medium. Because the hCG concentration detected in urine samples is larger than that in the hCG standards, linearity proves that both slopes decrease as the concentration increases. This result has been further confirmed by plotting a curve that theoretically fits both the urine and standard samples.
Selectivity tests were carried out by using interfering species, such as CEA, AA, UA and glucose (GLUC), with 8 mIU/ml of hCG (Fig 6c). It can be observed that the response of interfering species on our PSNG electrode is less than 10 nS/cm at the lowest concentration of hCG. In addition, the LOD of the sensor was measured at 0.28 mIU/ml (27.78 pg/ml) at 3σ from PSNG electrode conductivity at a frequency of 1 MHz. To test this, a smaller concentration of the hCG target (0.1 mIU/ml) was prepared and its minimum conductivity saturation point was observed at 5.13 nS/cm, as shown in the top inset of (Fig 6c(i)). Hence, our LOD is proven to be considerably better than that of the nanoporous gold and graphene sensor reported by Li et al., at 0.34 mIU/ml [14] and polyethylene glycol microbeads reported by Zhu et al at 0.38 mIU/mL [61], as shown in Table 3. The Fig 6c(ii) inset shows that the stability of the present LOC decreases to 98.5% after 22 days of testing with 8 mIU/ml of hCG. This is acceptable for a sensor and comparable with the 7.1% decrease after 1 month using a MWCNT modified immunoassay, as reported by Na li et al. [27], and a 14% decrease after 3 weeks using a nanoporous gold and graphene sensor, as reported by Ru Li et al. [14].
Reproducibility between LOCs was also tested to ensure intersensor signal accuracy. Fig 6d  shows that there is little variation in capacitance with measured values of replica samples. However, it does not affect the electrode sensitivity, in which the relative standard deviations (RSDs) of 4.3, 3.3 and 4.9% changes in capacitance at 4, 9 and 12 weeks of pregnancy, respectively, indicate good reproducibility. This could be because the surface and geometrical factor of the PSNG electrode provide a good platform for biosensing [62]. Furthermore, for regeneration purposes, the PSNG sensor may also be reusable after washing it in urea to remove the bonded Ag target on immobilized APTES/hCGab [30]. This strategy enables the reuse of the sensing surface for multiple samples with tripartite sensing.

Commercial potential of POC PSNG immunosensor
Our lab-on-chip immunosensor with electrical readouts offers several advantages over immunoassays due to its reduced cost, reduced form factor and ease of signal acquisition. The costs associated with laboratory equipment, chemicals and labeling procedures for hCG diagnosis via immunoassay methods could be reduced using POC immunosensors [63][64][65]. Because our POC PSNG is fabricated with multiple electrodes, it could facilitate multiplex detection from a single sample, which reduces cost and diagnosis time, as demonstrated in other sensing systems [33,66,67].
Recently, a study on the sensitivity and limitations of commercially available qualitative POC pregnancy test kits was reported by Greene et al. Two commonly used hCG POC tests, OSOM and QuickVue+, were examined, and the authors concluded that both devices have a detection range of 20-300 IU/ml with urine samples, which is insufficient for early pregnancy diagnosis. Furthermore, the mean gestational age for specimens corresponding to the false negative results was 4 weeks [68]. Siemens Immulite 1000 hCG Assay shows detection limit at 2.0 mIU/mL [69]. With our method, comparable sensitivity and specificity could be achieved from electrical transductions of the binding activity on PSNG, as shown in Table 4. Moreover, the gap size of 25 nm on the proposed lab-on-chip sensor is minimal compared to other commonly available lab-onchip systems. In addition, the tripartite flow-channel minimizes the sample usage with uniform flow for parallel analyses. Overall, the PSNG electrode immunosensor used in the current study exhibited higher performance and a more facile fabrication process than other available systems, and the sensor can be readily fit with commercially available detection systems.

Conclusion
The human chorionic gonadotrophin (hCG) level in pregnant women is usually low during early pregnancy and increases drastically during the second trimester. As detection of hCG in the clinical urine sample is critical during pregnancy, we have demonstrated the detection of hCG with a surface functionalized PSNG electrode immunosensor fabricated by a conventional lithography and size expansion technique. Our lab-on-chip sensor exhibited good stability, high sensitivity (835.88 μA mIU -1 ml -2 cm -2 ), good reproducibility, high selectivity and a low detection limit (0.28 mIU/ml, or 27.78 pg/ml). Due to the narrow nanogap (25 nm) fabricated on this immunosensor, it shows high performance with a good signal-to-noise ratio, and it can be applied for multiplexed POC diagnosis, further reducing expensive equipment usage. Moreover, the dielectric analyses on the PSNG electrodes with clinical human urine samples provide reliable measurements and stability against regeneration reagents, which favors multiple sample analyses on the tripartite channel with the option of parallel runs.
Supporting Information S1 File. This file contains. Fig A. (a) Electrical characterization setup for AC measurements using a frequency analyzer and (b) DC measurements using the Kiethley 6487 Picoammeter.