Fabrication of Nano-BSA Graphene FET
Titanium and gold were patterned on the SiO2/Si substrate as source and drain electrodes by photolithography, metal deposition, and a liftoff process, as shown in Fig. 1a, b. Graphene grown by chemical vapor deposition was transferred to the metal electrodes on the substrate, and the poly(methyl methacrylate) film on the graphene was removed by acetone (Fig. 1c). Based on denatured BSA-doped graphene [39], native BSA dissolved in deionized water at a concentration of 1.5 mM was then dropped on the graphene (Fig. 1d) and the native BSA was denatured on the graphene surface at 80 °C, as shown in Fig. 1e. Obvious thin dBSA films were formed on graphene via noncovalent interactions, consistent with previous observations [44]. The dBSA functionalized graphene was defined by photolithography (Fig. 1f) and simultaneously etched using O2 plasma. The photoresist was removed by acetone (Fig. 1g). The dBSA functionalized graphene channel was 60 µm in width and 30 µm in length. The thickness of this channel measured optically was approximately 26.4 nm (Fig. S1). This multifunctional nano-dBSA film acted to prevent surface contamination and destruction of the graphene, and also functioned as a cross-linker between the graphene FET biosensor and bioconjugate receptor. Finally, to protect the gold contacts from the electrolyte and obviate the leakage current from these metal contacts to the electrolyte, a layer of chemically stable SU-8 photoresist was coated on the gold electrodes (Fig. 1h).
Functionalization and Characterization of Bioprobes Based on Nano-dBSA Film
To achieve sensitive CEA recognition, the functionalized dBSA film enriched with chemical groups was used as a cross-linker of graphene. The film interacted with the graphene by π–π stacking. Anti-CEA mAb antibodies were conjugated onto the dBSA films via an immobilization procedure involving EDC and sulfo-NHS. EDC reacted with anti-CEA mAb to create an o-acylisourea intermediate, and a sulfo-NHS ester intermediate was formed by adding the sulfo-NHS, which could couple with amine-containing dBSA film on graphene. The resulting anti-CEA-dBSA functionalized graphene FETs (GFETs) acted as sensitive bio-interfaces to specifically recognize CEA. After immobilizing anti-CEA mAb and rinsing with PBS, native BSA solution was added to the channel of the dBSA functionalized GFET to block the excess reactive groups remaining on the dBSA surface. Finally, anti-CEA-dBSA functionalized GFETs were rinsed with deionized water and prepared for subsequent detection of target molecules. The entire process of the modification for anti-CEA-dBSA functionalized GFETs is shown in Fig. 2a.
Sandwich fluorescent immunoassay is a commonly used approach in biotechnology [45]. It was used to characterize the immobilization of anti-CEA mAb on dBSA film in this study. Secondary anti-CEA mAb conjugated with QDs was mixed with CEA solution and incubated with anti-CEA-dBSA functionalized graphene and bare dBSA functionalized graphene. Compared with the control group, the fluorescent images shown in Fig. S2 revealed that anti-CEA mAb was successfully immobilized on the dBSA functionalized graphene surface by the activation of EDC and sulfo-NHS. The results indicated that this novel method based on the dBSA film could be effective in the design of graphene biosensors.
Construction of Electrolyte-Gated Anti-CEA-dBSA Functionalized GFET
The reaction chamber made of polydimethylsiloxane was anchored on the substrate using silicone. The miniaturized Ag/AgCl electrochemical reference electrode was immersed in the reactive chamber as the gate of the anti-CEA-dBSA functionalized GFET. Drain–source voltage (Vds) and gate–source voltage (Vgs) were applied to force the operation of the devices. One terminal of the miniaturized Ag/AgCl electrochemical reference electrode was fixed on the shelf, and another terminal was immersed in the reactive chamber as the gate. Considering the sensitivity of anti-CEA-dBSA functionalized GFETs, 0.1 mM PBS was added to the reactive chamber as the electrolyte to maintain an appropriate Debye length [46]. A schematic diagram of electrolyte-gated anti-CEA-dBSA functionalized GFET is shown in Fig. 2b. In addition, a representative optical micrograph of the dBSA functionalized graphene channel with an SU-8 insulating layer is shown in Fig. 2c.
Enhanced Performance of Anti-CEA-dBSA Functionalized GFET
The performances of electrolyte-gated anti-CEA-dBSA functionalized GFETs were evaluated by the fundamental measurements of GFETs. The transfer characteristics were shown in Fig. 3a, which depicted the successful functionalization of nano-dBSA films on graphene and anti-CEAs mAb with the retention of the intrinsic property of graphene. The ambipolar curves indicating the Dirac points (denoted VD) at Vgs were between 0.1 and 0.25 V, while Vds was below 0.2 V, suggesting that the anti-CEA-dBSA functionalized Gr can be classified as the p-type. A greater difference was observed between two neighboring drain–source currents (Ids) in the hole regime (the gate–source voltage was denoted Vgs, Vgs < VD) than in the electron regime (Vgs < VD), indicating that the gate voltage in the hole regime could be a better choice for the detection of target molecules. To maintain the performance of the electrodes and graphene channel, a low Vds at 0.1 V was applied to drive the anti-CEA-dBSA functionalized GFET [47].
The transconductance parameter gm for a transistor device is widely used to describe FET devices. This parameter represents the amplification capability of GFETs [26, 48], where a higher gm enables a greater conductivity response per unit of biomolecule charge excitation. Therefore, this parameter is positively correlated with the device sensitivity and is valuable for sensing applications. The transconductance gm of anti-CEA-dBSA functionalized GFETs under different drain–sources voltages is defined as the derivative of Ids with respect to Vgs in Fig. 3b. While the Vds was set at 0.1 V, gm = –577.78 μS approached the maximum (denoted gmax−) in the hole regime at a special gate voltage Vgs = 0.07 V (denoted Vmax-). Similarly, at Vgs = 0.24 V (denoted Vmax+), gm = 434 μS approached another maximum (denoted gmax+) in the electron regime. The average transconductance value of several anti-CEA-dBSA modified GFETs in Table S1 was higher than that of the anti-CEA mAb PYR-NHS modified GFETs in Table S2 and many other reported electrolyte-gated GFET devices [26, 49, 50], which revealed the high sensitivity of this device for biomolecule detection.
For detailed investigation of the transconductance enhancement mechanism, the hole and electron mobility parameters of the anti-CEA-dBSA functionalized GFETs were calculated according to the transconductance gm using Eq. (1) [51]:
$$\mu = g_{\text{m}} L/WC_{\text{tot}} V_{\text{ds}}$$
(1)
where L is the channel length, W is the channel width, Ctot is the gating capacitance per unit channel area (F cm−2), Vds is the source–drain voltage (V), and gm is the differential transconductance. For the interfacial capacitance of the graphene–water interface, the quantum capacitance CQ of graphene and the double-layer capacitance Cdl of the electrolyte are in series connection to construct the gate capacitance. Subsequently, according to the capacitive equivalent circuit model of the graphene conducting channel shown in the insert of Fig. 3a, the total gating capacitance per unit area is calculated as: Ctot = CQCdl/(CQ + Cdl). The double-layer capacitance Cdl acts as a parallel-plate capacitor, which could be calculated using equation: \({C_{\text{dl}}} = {\epsilon_0}{\epsilon_r}/{d_{\text{dl}}}\), where \({\epsilon_0}\) is the permittivity of free space, \({\epsilon_r}\) is the dielectric constant of the electrolyte (~ 78), and ddl is the Debye length on the bio-interface. According to the buffer ionic strength of the electrolyte, the Debye length is estimated to be approximately 23 nm, and the corresponding double-layer capacitance Cdl is approximately 2.97 μF cm−2. While the graphene channel potential is Vch, the quantum capacitance CQ of Gr is defined as [26, 52]:
$$C_{Q} = \frac{{8\pi q^{2} k_{B} T}}{{\left( {hv_{\text{F}} } \right)^{2} }}\ln \left[ {2\left( {1 + \cosh \frac{{qV_{\text{ch}} }}{{k_{\text{B}} T}}} \right)} \right],$$
(2)
where q = − 1.602 × 10−19 C is the electron charge, KB = 1.381 × 10−23 J K−1 is the Boltzmann constant, h = 6.626 × 10−34 JS is the Planck constant, vF = 1.1 × 106 m s−1 is the Fermi velocity of Dirac fermions, and T = 300 K at room temperature. The potential distribution in the electrolyte-gated anti-CEA-dBSA modified GFET device is described as Eq. (3) [53]:
$$C_{Q} /C_{\text{dl}} = \left( {V_{\text{gs}} - V_{\text{ch}} } \right)/V_{\text{ch}} ,$$
(3)
where Vgs is the gate–source voltage. Thus, the CQ value at an arbitrarily given Vgs can be analytically determined by substituting Eq. (3) into Eq. (2) and solving for CQ.
Using this model for the interfacial capacitance, the field-effect mobility of charge carriers in the device can be obtained. The mobility values extracted at the transconductance maximum points (gmax− for holes, gmax+ for electrons) were used as the mobility parameters of anti-CEA mAb modified GFET devices. Average values of hole mobility μave-h1 and electron mobility μave-e1 for seven anti-CEA-dBSA GFET devices were estimated to be approximately 2763.9 and 1169.6 cm2 V−1 s−1, respectively. As shown in Fig. 3c, d, the average mobility parameters of several anti-CEA mAb functionalized GFET devices based on a noncovalent functionalized linker (PYR-NHS) were lower than those of anti-CEA-dBSA GFET devices. These results indicated that GFET biosensors based on this multifunctional and self-protecting dBSA film could improve the performance of GFET biosensors.
Performance of Anti-CEA-dBSA Functionalized GFET
The output characteristic curves were obtained by recording the Ids versus Vds under different Vgs, as shown in Fig. 4a. The dependence of the Ids with Vds variation (− 0.5–0.1 V) verified the good electrical contact between the graphene and gold electrode. The leakage currents were recorded under different top gate voltages (Fig. 4b). Compared with the values of the net change in drain currents, the absolute values of leakage currents were always below 80 nA, which could be considered negligible. To preclude false signals, especially those arising from nonspecific binding, several control groups were used to assess the utility of the anti-CEA-dBSA functionalized GFET. The responses of the drain–source current (Ids) after adding the same value (10 ng mL−1) of control, cytokeratin-19-fragment (CYFRA21-1), SCC, and CEA are shown in Fig. 4c. The general serum diluent was used to dilute the biomarkers, which served as the control group at the same time. As shown in the specific detection curves of the anti-CEA mAb functionalized GFET in Fig. 4c, when the control group, CYFRA21-1, and SCC were added to the buffer solution of the anti-CEA mAb functionalized GFET, no obvious increase was shown in Ids. Upon the addition of the CEA protein, a large increase in drain current caused by the binding of CEA was observed. To accelerate the reaction between the anti-CEA mAb and CEA protein, the solution was stirred for several seconds after the addition of each protein. The isoelectric point of CEA was approximately 4.4–4.7 [54], indicating that these target molecules were negatively charged in the nearly neutral pH buffer solution. These results demonstrated that the negatively charged CEA protein was avidly bound by the anti-CEA-dBSA functionalized GFET, resulting in an increase in the drain–source current upon addition of CEA. Interestingly, the addition of control proteins SCC and CYFRA21-1 did not induce a similar increase in drain–source current, which indicated that the nonspecific binding of dBSA functionalized graphene with nontarget proteins could be negligible. Taken together, the findings indicated that anti-CEA mAb functionalized GFETs exhibited good specificity for the detection of CEA.
The drain–source current of the anti-CEA-dBSA functionalized GFET was monitored at various CEA protein concentrations to evaluate its sensing characteristics. The target CEA proteins at concentrations of 10 pg mL−1, 100 pg mL−1, 1 ng mL−1, 10 ng mL−1, and 100 ng mL−1 were introduced into the channel of the anti-CEA-dBSA functionalized GFET as the time-dependent response of the drain current was recorded (Fig. 4d). The mechanism of detection for anti-CEA-dBSA functionalized GFETs involved the adsorption of negative CEA proteins on the surface of the graphene. These proteins acted as electron donors, resulting in conductance changes. For this reason, the drain–source current increased gradually after injection of the target CEA at each concentration (Fig. 4d). According to the response of the control group, the limit of detection was less than 56 fM.
Target Detection in Diluted Serum Samples
Analysis of clinically relevant samples, such as blood serum, could be very important in the clinical diagnosis of cancer. To verify target detection in serum samples using the anti-CEA-dBSA functionalized GFET, the target CEA proteins in diluted blood serum at concentrations of 10 pg mL−1, 100 pg mL−1, 1 ng mL−1, 5 ng mL−1, and 45 ng mL−1 were added to the reactive chamber, and the drain–source current was recorded at the same time. As shown in Fig. 5a, the drain–source currents increased with the target molecule concentrations. The gradually increasing drain–source current response with increasing CEA concentration in blood serum was consistent with the results in Fig. 4d. The general serum diluent was used to dilute the CEA-containing serum sample, which also worked as a control group.
The average net drain–source currents of the anti-CEA-dBSA functionalized GFET caused by the control, 10 pg mL−1, 100 pg mL−1, 1 ng mL−1, 5 ng mL−1, and 45 ng mL−1 groups were 0.0747, 0.661, 1.01, 1.77, 2.73, and 4.99 μA, respectively. The dissociation constant (Kd) for the interaction between the anti-CEA mAb and CEA could be estimated by measuring the drain current (Ids) of the anti-CEA-dBSA functionalized GFET at different CEA concentrations. The quantity of net drain–source current (∆Ids) was calculated as a function of CEA protein concentrations, as shown in Fig. 5b. The plot of the data yielded a nonlinear curve, indicating that the relationship between ∆Ids and the binding CEA could fit the Hill adsorption model [55, 56] calculated as Eq. (4):
$$\Delta I =\Delta I_{\hbox{max} } C_{\text{cea}}^{n} /\left( {K_{d}^{n} + C_{\text{cea}}^{n} } \right),$$
(4)
where Kd is the dissociation constant of the interaction between CEA and anti-CEA mAb, ΔImax is the saturated net drain–source current, Ccea is the protein concentration, and n is the Hill cooperativity coefficient of the binding interaction.
According to the fitted red curve shown in Fig. 5b, ΔImax, Kd, and n were estimated to be 12.1 µA, 122.8 ng mL−1, and 0.35, respectively. The calculated value of n was less than 1, which indicated the negative cooperativity in binding interaction between CEA and anti-CEA mAb. The molecular weight of CEA of approximately 180 kD [57] resulted in a dissociation constant of 6.82 × 10−10 M. The dissociation constant between CEA and anti-CEA mAb had been investigated previously [58, 59], and it was determined to vary from 4 × 10−12 to 1 × 10−7 M. Therefore, the value of the resulting dissociation constant evaluated in this study using anti-CEA-dBSA functionalized GFETs was in accordance with previously reported results, indicating a high affinity between CEA and anti-CEA mAb. From Eq. (4) and the definition of the dissociation constant (Kd) [60], while the ligand concentration was equal to the dissociation constant (Kd), the percentage of bound receptors at equilibrium was 50%. According to the calculated value (122.8 ng mL−1) of Kd, the available receptors on the dBSA functionalized GFETs bio-interface were sufficient for the detection of CEA molecules under different concentrations in this study. According to the fitting results, the limit of detection was estimated to be approximately 337.58 fg mL−1, which was lower than for other graphene FET biosensors [41, 61, 62]. Well-defined drain–source current changes were observed for low CEA concentrations (337.58 fg mL−1) in diluted serum, which were much smaller than the cutoff value (5 ng mL−1) used in clinical diagnosis. In addition, compared with other nanomaterial-based CEA immunosensors in Table S3, the sensitivity of multifunctional dBSA functionalized GFETs showed obvious superiority. These results clearly demonstrated the promising potential of anti-CEA-dBSA functionalized GFETs in clinical applications.