Introduction

With the advances of bioassay technology, early diagnosis of cancer has been rapidly developed, contributing to the decline of cancer mortality. In vitro diagnostics (IVD) has realized accurate, sensitive, and specific detection of biomarkers in blood and tissues, which has become indispensable in clinical practices [1,2,3,4]. The levels of cancer-related biomarkers in blood and tissue are important indicators for determining cancer development [5]. Folate receptor 1 (FOLR1) is a glycosylphosphatidylinositol-anchored epithelial cell surface glycoprotein [6,7,8,9]. A recent study reported that the serum FOLR1 level is significantly elevated in patients with ovarian cancer, compared to that of both healthy controls and patients with benign gynecological conditions [10]. In addition, levels of FOLR1 showed better accuracy for determining ovarian cancer than other biomarkers including CA125 [11, 12]. Hence, FOLR1 may be a new promising biomarker of liquid biopsy for cancer diagnosis.

Currently, most of FOLR1 bioassays have been realized based on enzyme-linked immunosorbent assay (ELISA) [13,14,15]. The involvement of antibodies leads to high cost [16], due to the complex purification process and low yield. [17] Another problem is the inherent instability of antibody [18]. The antibody is susceptible to external environment such as high temperature and low pH, which results in protein denaturation and limits the consistency of detection result [19]. Taken together, it is necessary to develop new strategy for biomarker capture. Recently, Zhang et al. reported a ligand-protein recognition manner, with hyaluronic acid ligand as the probe to capture CD44 [20]. Liu et al. designed ligand-protein sensor for the detection of transferrin receptor [21]. Compared with the antibody, the ligand-based specific recognition presented attractive advantages, such as superior capture efficiency, high stability, facile storage, and low cost for mass production [22,23,24]. Notably, many researchers have testified that FOLR1 could specifically bound to folic acid (FA) and its derivatives with a high affinity [8, 25]. As a result, ligand-protein interaction like FA-FOLR1 provides an opportunity to develop new analytical platform to detect FOLR1.

Over the past decade, label-free field-effect transistor (FET) bioassays have exhibited many prominent advantages as a promising point-of-care (POC) technique, including low cost, high sensitivity, high speed, small volume, intrinsic signal amplification, and circuit integration [26,27,28,29]. Many FET-based bioassays have been widely used in tumor biomarkers, such as PSA [30, 31] and MMP-9 [19]. FET-based bioassays can accumulate hole or electron effectively on the channel materials, which can be modified with surface functionalization. Therefore, it is able to achieve rapid response to different analytes with high selectivity and sensitivity [32,33,34].

In general, the configuration of FET bioassay is similar to that of metal-oxide-semiconductor FET. One interface of the channel is covered by insulating membrane. However, some reports suggested that the sensitivity and selectivity of FET bioassay would be hindered by the insulating membrane [30, 35, 36]. Because the thickness of insulting membrane enlarged the distance between analyte and semiconductor. In addition, insulator plays the role of capacitor and determines the sensing current. The sensitivity of the device can be dramatically affected with the defect and dielectric constant of it [37, 38]. Recently, many membraneless FET bioassays with high sensitivity have been reported [39, 40]. For example, Liu et al. applied bare MoS2 as the FET sensing channel to monitor DNA fragments, achieving an ultrahigh sensitivity with a limit of detection (LOD) below 100 aM [27].

Beside insulator, the dimension of the channel also plays a vital role in FET bioassay. The channel based on two-dimensional (2D) materials such as graphene and transition metal dichalcogenides have the high surface-to-volume ratio, which could supply more analyte binding sites and improve the sensitivity of the bioassay [41]. However, although graphene provides high mobility and sensing layer with atomic thickness, the sensitivity and response of graphene-based bioassay would be seriously restricted due to the absence of band gap and small on/off current ratio [42, 43]. Additionally, when utilizing 2D materials in bioassay, we must consider the toxicity of the materials. For example, some research results indicated that the 2D materials such as WSe2 [44], VTe2 [45], and NbTe2 [46] showed poor biocompatibility. Alternatively, MoS2 is one of the most extensively studied 2D semiconducting materials. The development in MoS2-based bioassays ascribed to the inherent advantages of MoS2, such as exceptional biocompatibility [47], easy modification [48], ultrafast saturable absorption [49], high on/off ratios, and high surface-to-volume ratio [50]. In particular, mechanically exfoliated MoS2 sheets with minimal defects can achieve relatively high mobility, on/off current ratio, and current level in FETs [51]. Until now, many ultrasensitive membraneless FET bioassays prepared with mechanically exfoliated MoS2 have been reported for the detection of PSA [52] and IgG [53]. However, few studies have been performed on the integration of membraneless MoS2 FET with ligand-protein interaction for the detection of FOLR1.

In this work, a membraneless MoS2 FET bioassay based on the ligand-protein interaction has been firstly prepared for FOLR1 detection. In this bioassay, BSA-FA is serving as the probe, that is, FA ligand as the capture component and BSA as the conjugate molecule to improve the biocompatibility and water solubility. The specific binding between BSA-FA and FOLR1 could mediate electrical signal and give ultrahigh sensitivity of the bioassay. A good linearity for threshold voltage (Vth) vs FOLR1 concentrations can be observed, with an extracted LOD of 0.057 pg/mL. In addition, a nonaqueous environment measurement has been applied in this study, which not only simplified the preparation process but also avoided the ionic screening effects.

Experimental section

Materials and reagents

Molybdenum disulfide (MoS2) was purchased from Resemi (Suzhou, China). 1-Pyrenebutanoic acid N-hydroxysuccinimidyl ester (PASE, 95%) was obtained from Sigma-Aldrich (China). Bovine serum albumin (BSA, 98%) was purchased from J&K (China). Bovine serum albumin-folic acid (BSA-FA) was purchased from Cusabio (Wuhan, China, www.cusabio.cn). Recombinant human FOLR1 protein (95%) and human FOLR1 antibody were obtained from R&D (USA).

The fabrication of MoS2 device

Few-layer MoS2 flakes were produced by mechanical exfoliation and then transferred onto p++Si substrate with 300 nm oxides via Scotch tape. That is, we first adhered adhesive tape to the surface of the bulk materials and peeled it off. Then, press the tape firmly yet gently onto the silicon wafer piece, peeling it off, and the silicon wafer washed with isopropanol to clean the organic residue. The wafer can now be viewed under the optical microscope. The thickness of MoS2 materials can be roughly estimated by the color of it, which will vary depending on the microscope. Finally, the few-layer MoS2 flakes were produced. TEM copper grid was then used to make patterns of electrodes on the assurance MoS2 flakes. Subsequently, the source and drain electrodes (In/Au 5/50 nm) were formed via vacuum coating machine. Thermal evaporation was carried out at the rate of 0.2 Å/s in the study. Removing the TEM copper grid template and the source and drain electrodes connected by the material were constructed into field-effect transistor device. The detailed preparation process is shown in Scheme S1 in the Electronic Supplementary Materials. The device was annealed at 200 °C for 2 h in mixed gases (Ar/H2, 100/10 sccm) to improve the contacts between MoS2 and electrodes, as well as removing residual organic contaminants during fabrication.

Immobilization of BSA-FA

To anchor BSA-FA on the channel of the MoS2 FET, the device was firstly immersed into 5 mMol 1-pyrenebutanoic acid succinimidyl ester (PASE) in DMF and incubated for 2 h. The PASE-coated device was washed sequentially in DMF, ethanol, and deionized (DI) water. Then, 10 μL of 100 μg/mL BSA-FA in 1× PBS (pH 7.2–7.4) was dropped on the chip and incubated overnight at 4 °C in a refrigerator for conducting the chemistry reaction between BSA-FA and PASE. The FET bioassay was rinsed with 1× PBS, DI water in order to remove the unbound BSA-FA. After that, the device was blocked with BSA solution (50 μg/mL, 10 μL) for 1 h to occupy nonspecific binding sites, followed by washing with 1× PBS, DI water, and drying with N2. Ultimately, the device was stored at 4 °C in a refrigerator for further use.

Detection of FOLR1 in PBS

The electrical response of the FET was measured with a semiconductor parameter analyzer (Keithley, 2643B) and a probe station (Cascade Microtech, MPS 150) at room temperature in air. In the measurements, 10 μL of FOLR1 in 1× PBS (pH 7.2–7.4) with a gradually increased concentration ranged from 100 fg/mL to 10 ng/mL was added on the device surface and incubated for 20 min at room temperature. Then, the chip was rinsed with 1× PBS, DI water and dried with N2. The MoS2 FET bioassay measurement was carried out after various concentrations of FOLR1 treatment. The transfer curves of the FET bioassay were measured with a silicon substrate as the back gate and a fixed source-drain voltage Vds = 0.5 V. The bioassay based on antibody-antigen mode for detecting FOLR1 was carried out in the same condition. The only difference is that BSA-FA is replaced by anti-FOLR1.

Detection of FOLR1 in human serum sample

Serum samples were volunteered by Qiyong Cai, the author of this article. He is 29 years old and in good health. The obtained serum samples were stored at − 80 °C until use. Human serum samples were treated by centrifugation at 2000×g for 10 min. The serum was diluted 10-fold. The recovery experiment was performed toward 10-fold diluted serum spiked with the different concentrations (50 pg/mL, 100 pg/mL, 200 pg/mL) of FOLR1. And then, six different concentrations of FOLR1 in 10-fold diluted serum from 100 fg/mL to 10 ng/mL were dissolved. All electrical measurement of MoS2 FET bioassay for detection of various FOLR1 in human serum remains the same as that described in PBS.

Selectivity and stability investigation

To determine the specificity of the bioassay for FOLR1, some common proteins were investigated, including carcinoembryonic antigen (CEA), squamous cell carcinoma antigen (SCCA), and P53, and the concentration is 10 ng/mL. The measurement procedure remains the same as the detection of FOLR1. The stability of the bioassay was conducted by monitoring the signal response to 10 pg/mL FOLR1 three times daily, and then the devices were stored at 4 °C in a refrigerator.

BSA-FA-based ELISA for the detection of FOLR1

The 1 mg/mL BSA-FA was diluted to 10 μg/mL with 50 mMol carbonate buffer. 100 μL of BSA-FA solution (10 μg/mL) was added in each well of 96 well polystyrene plate and coated at 4 °C overnight. The coating solution was removed and the plate was washed twice with 200 μL of 1× PBS per well. Then, 200 μL of blocking buffer (0.5% BSA solution) was added in each well to block the remaining protein-binding sites for 2 h at room temperature. The plate was washed twice with 1× PBS. Next, 100 μL of FOLR1 proteins (0 ng/mL, 1 ng/mL, 10 ng/mL, and 100 ng/mL) was added and then captured with the pre-coated FA. Because the recombination FOLR1 protein contained His tag, mouse anti-His monoclonal antibody was served as the detection antibody. Then, horseradish peroxidase–labeled goat anti-mouse secondary antibody was added and reacted with the detection antibody. Finally, the results were developed with tetramethylbenzidine (TMB) as colorimetric substrate. Absorbance was measured in a plate reader (SPARK, 10M, TECAN).

Characterization

To confirm the functionalization of all molecules onto the MoS2 surface, Raman spectroscopy was performed using invia-reflex system with a 532 nm excitation laser. The surface topography of MoS2 channel, functionalized with biomolecules in stepwise, was carried out using tapping-mode atomic force microscope (AFM) (Bruker Dimension Icon). To further investigate the immobilization of BSA-FA, X-ray photoelectron spectroscopy (XPS) spectra of pristine and BSA-FA-immobilized MoS2 were obtained by using AXIS SUPRA system with Al X-ray source (Kα 1486.6 keV).

Results and discussion

Fabrication of MoS2 FET bioassay

The fabrication procedure of the MoS2 FET bioassay is illustrated in Fig. 1. Prior to surface modification, back gate MoS2 FET was prepared. In brief, mechanically exfoliated few-layer MoS2 was transferred onto SiO2/Si substrate, where SiO2 with the thickness of 300 nm functioned as the gate dielectric, and the p-doped Si worked as the back gate. The Raman spectra with the E12g and A1g modes were observed at 380 and 404 cm−1, respectively, which suggests that the exfoliated material is indeed few-layer MoS2, as shown in Fig. S1. Source/drain electrodes (In/Au 5/50 nm) were patterned on MoS2 by thermal evaporation. Aiming to achieve superior sensing performance and high field-effect mobility (μ = 10–20 cm2V−1 s−1) [54], the large surface area of the channel with a width of 12 μm and the thickness of MoS2 film ranged from 5 to 15 nm was chosen (Fig. S2). AFM image clearly showed the exact thickness of MoS2 is 8 nm. Functionalization was performed with layer-by-layer assembling. Firstly, PASE molecules were attached onto MoS2 surface as the linker [55]. Then, the amino group of BSA-FA was reacted with carbonyl group of PASE via the N-hydroxy succinimide (NHS) chemistry for immobilizing BSA-FA [40, 43]. Next, BSA was applied as the blocking agent to prevent nonspecific binding between FOLR1 and substrate material, as well as blocking the excess carbonyl groups. Finally, the assay was applied in detecting different concentrations of FOLR1, which is specifically bound to BSA-FA. The variation of electrical conductance was recorded synchronously.

Fig. 1
figure 1

Schematic illustration of the preparation process of MoS2 FET bioassay for FOLR1 detection

Full size image

Characterization of MoS2 FET bioassay

To validate the functionalization process, the MoS2 flake was characterized by AFM in a stepwise manner. Compared to the pristine MoS2 film surface topography, the root-mean-squared (RMS) roughness of the immobilized BSA-FA increased from 0.217 to 1.54 nm, suggesting that BSA-FA is successfully bonded to PASE (Fig. 2a, b). The RMS roughness of the sample slightly increased to 1.58 nm after being blocked by BSA, indicating that nonspecific binding sites are blocked well (Fig. 2c). Lastly, the FOLR1 protein was introduced into the MoS2 channel (Fig. 2d). The RMS roughness increased to 1.78 nm. Here, the value of RMS roughness of each step was estimated from the full size of the image. Moreover, as shown in Fig. 2d, in contrast to previous steps, the number of bright spots is significantly increased, indicating an increased average height of the device surface. The topographic images of stepwise modification are shown in Fig. S3. From AFM images, the average height increased by 1.2 nm after the device was incubated in FOLR1 solution (Fig. S3d), which is lower than that of FOLR1. Because the molecular weight of FOLR1 is roughly 25.4 kDa in this work. This may be due to the diversity of binding directions of FOLR1 and BSA-FA and measurement errors. These results clearly identified that the FOLR1 proteins bond to BSA-FA immobilized on the FET device.

Fig. 2
figure 2

Surface roughness of AFM images of the same device with pristine MoS2 (a) and sequentially immobilizing with BSA-FA (b), blocking with BSA (c) and treating with FOLR1 (d)

Full size image

The electrical characteristics in all steps were also measured to further testify the functionalization of the MoS2 FET. After the binding of charged biomolecules on bioassay channel, the signal transduction can be altered by electrostatic gating and/or charge transfer, inducing the changes of sensing current. During the experiment, all biomolecules were in 1× PBS (pH 7.2~7.4). The isoelectric point (pI) of BSA-FA, BSA, and FOLR1 is ~ 4.7, ~ 4.7, and ~ 8.2, respectively, indicating that BSA-FA and BSA are negatively charged, and FOLR1 is positively charged in the PBS. Notably, both the theory of “pH-memory” and many studies prove that the biomolecules retained ionization state in a nonaqueous environment [31]. Therefore, as depicted in Fig. 3a, the stepwise transfer characteristics of MoS2 FET bioassay were observed under nonaqueous environment. The source-drain current (Ids) was measured as a function of back gate voltage (Vgs) from − 60 to 60 V with a set drain voltage (Vds) of 0.5 V.

Fig. 3
figure 3

Transfer curves (a) and Vth changes (b) of the MoS2 FET bioassay functionalization process in a stepwise manner: (1) pristine, (2) pristine + PASE, (3) pristine + PASE + BSA-FA, (4) pristine + PASE + BSA-FA + BSA, (5) pristine + PASE + BSA-FA + BSA + FOLR1 (100 pg/mL). Transfer characteristics of the bioassay were obtained with Vgs various from − 60 to 60 V and Vds = 0.5 V (in order to show the current change more clearly, the data with Vgs various from − 60 to − 20 V was omitted)

Full size image

As shown in Fig. 3b, step 1, the pristine MoS2 FET showed typical n-type conductance, consistent with previously reported results [19]. In addition, the linear output curve of the device is presented in Fig. S4, demonstrating the presence of Ohmic contact between In and MoS2. After dropping PASE solution onto the device, a positive shift of threshold voltage (Vth) from 8.18 to 10.3 V was obtained (Fig. 3b, step 2), suggesting the p-type doping of PASE. Furthermore, Raman spectroscopy was performed before and after PASE coupled to MoS2. As shown in Fig. S5, the clear blue-shift of Raman signatures was obtained as compared with pristine MoS2, which confirms the p-type doping effect of PASE [56]. The result is similar to the PASE-doped graphene in other studies [57]. When negatively charged BSA-FA was reacted with PASE, there is a considerable left-shifted Vth (Fig. 3b, step 3). This can be interpreted by improved surface negative charge density of MoS2 and increased electron injection to the MoS2 channel.

Notably, in our nonaqueous system, electrostatic gating effect induced by the weakly charged BSA-FA can be ignored [58]. Additionally, a previous study reported that electrons of protein could be directly transferred into graphene or carbon nanotube via the amine group [39]. For MoS2, we postulate that the immobilization of BSA-FA could generate n-type doping effect on MoS2, and, consequently, increase the on-current. In other words, negatively charged biomolecules act as electron donors causing left-shifted Vth, and increasing Ids. This change is consistent with previously reported results [40, 57, 58]. To further prove the functionalization of BSA-FA, we performed XPS measurement because the N-content of BSA-FA can induce the changes of N 1s peaks. As shown in Fig. S6, in contrast to the pristine MoS2, an additional peak area and increased peak intensity was observed in N 1s core level, which is assigned to the immobilization of BSA-FA on the device. After the device was blocked in BSA (50 μg/mL) solution, Vth showed a relatively small negative shift (Fig. 3b, step 4), implying that the amount of the immobilized BSA is extremely lower than that of BSA-FA. Finally, the positively charged FOLR1 (100 pg/mL) specifically combined with BSA-FA, and the value of Vth is positively shifted to 10.8 V (Fig. 3b, step 5). It may be resulted from the opposite charge between BSA-FA and FOLR1, in which FOLR1 decreased negative charge density on MoS2. As a result, the electrons transfer in MoS2 is significantly hindered by FOLR1, and consequently, the n-type doping effect of BSA-FA is weakened. Thereby, FOLR1 functioned as p-type dopant and decreased the on-current. To further identify the immobilization of BSA-FA, BSA, and FOLR1, the Raman spectroscopy was performed (Fig. S7). After the stepwise BSA-FA and BSA treatment, as shown in Fig. S7b, the frequency of the E12g and A1g mode gradually decreases, indicating that BSA-FA and BSA fixed on the device exhibits the n-type doping effect [59]. When the device was incubated in FORL1 solution, the enhancement of Raman mode intensity proves the p-type doping effect [60]. The above results certified that the MoS2 FET bioassay was successfully fabricated in a stepwise manner, in good agreement with AFM measurement result.

Feasibility study

Many researchers have reported that FOLR1 presented high affinity with FA and its derivatives [8, 61]. We have tried to verify whether the increased roughness and shifted Vth after the FOLR1 combination are ascribed to the specific binding of FOLR1 to BSA-FA. Conventional ELISA was performed with various concentrations of FOLR1 ranged from 1 to 100 ng/mL. As shown in Fig. 4, optical density (OD) value was increased with the improved concentration of FOLR1, and the color of the solution was changed from colorless to yellow correspondingly (Fig. 4, inset). These results prove that BSA-FA can specifically bind to FOLR1, instead of simple charge interaction. They also demonstrated that the small molecular FA can be alternative of antibody for the detection of FOLR1.

Fig. 4
figure 4

Various concentrations of FOLR1 (1 ng/mL, 10 ng/mL, and 100 ng/mL) were monitored by BSA-FA-based ELISA

Full size image

Electrical response of various FOLR1 concentrations in PBS

In order to obtain the optimal incubation time of the target molecules, the specific binding curve was plotted. Fig. S8 presents the variation of Vth with binding time, at a fixed FOLR1 concentration of 1 ng/mL. The dynamic equilibrium of binding reaction was completed after 20 min. Thus, the incubation time is set as 20 min for various concentrations of FOLR1. In addition, the buffer is also a critical consideration for the specific binding, and 1× PBS serving as the optimal buffer. The result is related to Debye screening length, namely the charged molecules on a certain length scale (Debye length λD) would be screened by counterions. Many studies reported that increased buffer ionic strengths yield a reduced Debye length, according to the formula λD = 0.32 (I)−0.5, where I is the ionic strength of the buffer solution [62]. Thus, the Debye length of 1× PBS (~ 0.7 nm) buffer solution is shorter compared to that of 0.1× PBS (~ 2.3 nm) and 0.01× PBS (~ 7.3 nm). The shorter Debye length may allow the screening of more charges [63], thus promoting the specific binding and improving the sensitivity of bioassay.

The detection of different concentrations of FOLR1 was performed by MoS2 FET bioassay in a nonaqueous environment. As shown in Fig. 5a, in the range between 100 fg/mL and 10 ng/mL, the Ids-Vgs curves were obtained with Vgs various from − 60 to 60 V and Vds = 0.5 V. The bioassay blocked with BSA and then immersed into PBS to establish a baseline current. After incubating with positively charged FOLR1, the Ids was decreased and the Vth was gradually shifted to the positive gate voltage with increased concentrations (Table S1), representing p-type doping effects on MoS2. Moreover, as the concentration further increased to 10 ng/mL, the electrical signal became stable, suggesting binding sites become saturated. The same results were obtained independently from three different FET bioassays (Fig. S9).

Fig. 5
figure 5

a Transfer curves of the MoS2 FET bioassay in response to various FOLR1 concentrations in 1× PBS. b Calibrated response ΔVth versus the FOLR1 concentration for the MoS2-based bioassay. The inset showed a linear relationship between ΔVth and various concentrations of FOLR1. c A least-squares fit to Hill-Langmuir equation, yielding the KD of 12 pg/mL

Full size image

To effectively assess the quantification performance of this bioassay, assay response is defined as ΔVth and calibrated as ΔVth = Vth (C) – Vth (n = 0), indicating the variation of Vth between the concentration of FOLR1 and the baseline. As shown in Fig. 5b, the relationship between the calibrated response ΔVth and FOLR1 concentration was summarized (the data in Fig. 5a converted to 5b). When the concentration of FOLR1 is above 1 ng/mL, ΔVth became stable, indicating that all binding sites had theoretically reacted with it and the assay signal reached saturated. Moreover, a great detection linearity for FOLR1 with the concentration spanned from 100 fg/mL to 1 ng/mL was observed (Fig. 5b, Inset). The regression equation is ΔVth = 1.17 × lgCFOLR1–1.26, and the correlation coefficient is 0.9988. The standard deviation (RSD) was calculated from measurements of three devices (n = 3), which is not significantly deviated from the mean value (< 6%) at different levels. Above result suggested high reproducibility for MoS2 bioassay. Furthermore, the LOD was estimated to be 0.057 pg/mL (S/N = 3), representing high sensitivity of FET bioassay.

Compared with some presented detection methods (Table S2), the developed MoS2-based FET bioassay exhibits lower LOD, higher sensitivity, and faster response time. These results supported the applications of FET bioassay for the ultrasensitive and quantitative detection of FOLR1. In addition, the predominant detection performance is also attributed to the strong binding affinity between FA and FOLR1. Figure 5c shows CFOLR1/ΔVth plotted as a function of various FOLR1 concentrations, which can be described by Hill-Langmuir equation [33, 40, 57]:

$$ \frac{C_{\mathrm{FOLR}1}}{\varDelta {V}_{\mathrm{th}}}=\frac{C_{\mathrm{FOLR}1}}{\varDelta {V}_{\mathrm{th},\max }}+\frac{K_{\mathrm{D}}}{\varDelta {V}_{\mathrm{th},\max }} $$
(1)

where ΔVth, max is the maximum change of the Vth, CFOLR1 is the concentration of FOLR1, and KD is the binding affinity dissociation constant for BSA-FA and FOLR1. The KD value was estimated to be 12 ± 6 pg/mL, lower than that of currently reported antibody-antigen bioassay [64], indicating the relatively high affinity between the small molecular ligand FA and FOLR1. To further prove FA ligand gives ultrahigh sensitivity and quantification of the device, the FET bioassay in antibody-antigen mode for detecting FOLR1 was performed. Firstly, the reaction of dynamic equilibrium between anti-FOLR1 and FOLR1 was carried out. The binding stability was completed after 50 min (Fig. S10), which is longer than that of ligand-antigen mode. Then, the transfer characteristics of the device were obtained, as shown in Fig. S11a, and the current decreases with increasing FOLR1 concentration. A great detection linearity for FOLR1 with the concentration ranged from 100 fg/mL to 1 ng/mL was observed, and the KD value was also estimated to be 14 ± 7 pg/mL, as shown in Fig. S11b and S11c. By contrast, the KD value is lower than that of antibody-antigen FET bioassay, which may be due to multiple binding sites between FA and FOLR1 [65]. This result further validates the high sensitivity of the FET bioassay. Beside lower KD value, the ultrasensitive and fast detection for FOLR1 is also attributed to nonaqueous measurement environment. Because Debye screening length made few effects on detection, and avoided the leakage (less than 0.1 nA) induced by the ions in buffer solution.

For verifying that the obtained Vth shift can be indeed mediated by FOLR1, several control experiments were set and conducted. As depicted in Fig. S12a, the electrical response of bioassay without BSA-FA was recorded to demonstrate nonspecific reaction between BSA and FOLR1. The addition of various concentrations of FOLR1 to the device presented negligible change of transfer curves, suggesting no noticeable nonspecific binding between them. In addition, owing to the nonaqueous measurement condition, bioassay was immersed into 1× PBS before the immobilization of FOLR1, and then dried with N2. As shown in Fig. S12b, the current signal only slightly changed, revealing the buffer made no significant interfering effects during the stepwise modification. The above results supported the reliability of our detection system.

Selectivity and stability

The selectivity and stability of MoS2 FET bioassay in complex media was investigated to further evaluate the practicality of the method. In selectivity test, the sensing signal of potential interference factors was recorded, such as carcinoembryonic antigen (CEA), squamous cell carcinoma antigen (SCCA), and P53. Negligible Vth shifts (ΔVth) were observed when the device was exposed to the above interferences. The concentration of CEA, SCCA, and P53 (10 ng/mL) is ten times that of FOLR1 (1 ng/mL), while a significant ΔVth was observed after the introduction of only 1 ng/mL FOLR1. The selectivity of the device is further tested with the mixed solutions, including CEA (10 ng/mL) + FOLR1 (1 ng/mL), SCC (10 ng/mL) + FOLR1 (1 ng/mL), and P53 (10 ng/mL) + FOLR1 (1 ng/mL). For above solutions, acceptable difference of ΔVth was observed (Fig. 6a), showing the excellent selectivity of the MoS2 FET bioassay for FOLR1 detection. In addition, the stability of the bioassay was examined (Fig. 6b). The BSA-FA functionalized device was preserved at 4 °C in refrigerator for different days and the ΔVth response to 10 pg/mL FOLR1 was recorded. After the storage for 7 days, the ΔVth response still retained 90% of its initial response (ΔVth, 0). It indicated acceptable storage stability of this FET bioassay. However, bioassay platform with better stabilization still poses challenges to mass production.

Fig. 6
figure 6

a Selectivity of MoS2 FET bioassay for FOLR1 from a series of interference factors. b Stability test of the MoS2 FET bioassay after the storage for 7 days

Full size image

Electrical response of various FOLR1 concentrations in serum

The potential applicability of the bioassay as POC device was validated in human serum. The FOLR1 with different concentrations (50 pg/mL, 100 pg/mL, 200 pg/mL) were added into 10-fold diluted serum, and the recoveries of FOLR1 were estimated to be 98%, 102%, and 95.5%, respectively (Table S3). These results suggest that the developed bioassay is reliable for detecting FOLR1. The electrical response of the bioassay corresponding to the detection of different concentrations of FOLR1 in 10-fold diluted human serum is depicted in Fig. S13a. Treatment process and measurement condition are maintained in the PBS buffer. The sensing linearity of the FET device between calibrated response ΔVth and FOLR1 concentration (from 100 fg/mL to 1 ng/mL) was observed (Fig. S13b). The linear regression equation of bioassay was ΔVth = 1.49 × lgCFOLR1–0.46 (inset in Fig. S13b), and the LOD was extracted to 0.086 pg/mL (S/N = 3). The error bar was calculated from three devices. The sensing response of different FET devices to various FOLR1 concentrations is similar, as displayed in Fig. S14. Importantly, the detection performance in human serum is comparable to that of in PBS buffer. The bioassay exhibited ultrahigh sensitivity and satisfactory selectivity in the real human serum. This satisfactory result provides the potential to cell capture for disease diagnosis in the future [66,67,68]. Therefore, the MoS2 FET bioassay showed promising application prospects in future clinical diagnosis.

Conclusion

We have successfully realized the detection of FOLR1 by an MoS2 FET bioassay for the first time in a nonaqueous environment. Based on the charge doping, the specific binding between small molecular FA and FOLR1 with a high affinity produced significant conductance variation on the membraneless MoS2 channel. The shifted Vth was obtained with various concentrations of FOLR1, in a linear relationship. The high electivity and reliability were observed based on the results of selectivity and control experiments. We have also demonstrated the label-free MoS2 bioassay platform can be applicable to the detection of FOLR1 in serum. In view of the importance of the detection of biomarkers for determining cancer development, the FET bioassay based on ligand-protein interaction provides the possibility of IVD testing. However, the poor stability of this method can be a challenge for mass production. Thus, further improving the performance of the device will be conducive to the real-life application of it.