Graphene field-effect transistor biosensor for detection of biotin with ultrahigh sensitivity and specificity.

Because avidin and biotin molecules exhibit the most specific and strongest non-covalent interaction, avidin-biotin technology is widely used in ELISA (enzyme-linked immunosorbent assay) kits for the detection of different bio-macromolecules linked to different diseases including cancer and influenza. Combining the outstanding electrical conductivity (200,000 cm2V-1s-1) of graphene with the unique avidin and biotin interaction, we demonstrate a novel graphene field-effect transistor (GFET) biosensor for the quantitative detection of bio-macromolecules. The GFET consists of six pairs of interdigital Cr/Au electrodes supported on Si/SiO2 substrate with an avidin immobilized single layer graphene channel as the sensing platform. By monitoring the real time current change upon the addition of biotin solution in bovine serum albumin (BSA) in the silicone pool preformed onto the GFET, the lowest detectable biotin concentration is estimated to be 90 fg/ml (0.37 pM). The specificity of the GFET is confirmed both by controlled and real sample measurements. From the magnitude of current change upon the addition of different concentrations of biotin solutions, the dissociation constant Kd is estimated to be 1.6 × 10-11 M. Since biotin is capable of conjugating with proteins, nucleotides and other bio-macromolecules without altering their properties, the present GFET sensor with its ultra-high sensitivity (0.37 pM) and specificity can be tailored to the rapid point-of-care detection of different types of desired biomolecules at very low concentration level through biotinylation as well as the exogenous biotin in blood serum.

Introduction

The strong interaction between avidin and biotin has been widely exploited in many applications such as protein and nucleic acid detection, immobilization, and purification methods (Guesdon et al., 1979; Hsu et al., 1981a, 1981b). Avidin is a tetrameric biotin-binding protein produced in the oviducts of birds, reptiles, and amphibians and is deposited in the whites of their eggs. Dimeric members of the avidin family are also found in some bacteria (Helppolainen et al., 2007). In chicken egg white, avidin makes up approximately 0.05% of total protein (approximately 1800 μg per egg). The tetrameric protein contains four identical subunits (homotetramer), each of which can bind to biotin (vitamin B7, vitamin H) with a high degree of affinity and specificity (Green, 1963). Biotin is involved in a wide range of metabolic processes, both in humans and other organisms, primarily related to the utilization of fats, carbohydrates, and amino acids (Zempleni et al., 2009). Because of its small size, biotin is a useful label for many proteins and nucleotides as it does not change the original properties of the proteins or nucleotides. The process of labeling a protein or nucleotide with biotin is known as biotinylation. It is also important that biotin protein ligases can attach biotin to specific lysine residues in vitro or in living cells (Chivers et al., 2011). The interaction between avidin and biotin is known to be the most specific and strongest non-covalent interaction (Kd = 10-15 M) between a protein and ligand (Green, 1963; Morag et al., 1996). The exceptionally strong affinity of avidin for biotin arises from the hydrophobic interactions of biotin and aromatic amino acids arranged in the binding pocket of avidin and multiple hydrogen bonding between heteroatoms in the ureido ring of biotin and asparagine, serine, tyrosine, and threonine residues in avidin (Chivers et al., 2011). Due to the strong interaction, the avidin-biotin complex is robust and stable against temperature, pH, harsh organic solvents and denaturing reagents. Given the unique properties of the avidin-biotin system, it is used often in enzyme-linked immunosorbent assay (ELISA) for many different types of medical applications such as cancer diagnosis (Ben Aissa et al., 2017; Gutiérrez-Zúñiga and Hernández-López, 2016; Lakshmipriya et al., 2016; Yang and Tseng, 2016). In addition, the avidin-biotin system is also used in other exciting applications such as labeled immunosensors (Yang et al., 2010), polymer-based detection (Ballesta-Claver et al., 2012).

To date, ELISA is one of the most widely used techniques for the detection of various types of biomarkers linked to various diseases such as influenza, cancer, AIDS (Carinelli et al., 2016; Safaei et al., 2015; Tauriello et al., 2018). Indeed, the detection of biomarkers by ELISA is based on the optical measurement of fluorescent markers, which requires integrated use of advanced spectral equipment. Biotinylated antibody and streptavidin-conjugated horseradish peroxide (streptavidin-HRP) often are used with ELISA to enhance the detection of various kinds of targets (Agnolon et al., 2020). A large range of clinical screening applications including those for infectious diseases such as SARS-CoV-2 (Wang et al., 2020) can be performed using ELISA diagnostic kits. However, from the sample collection to results output, the whole ELISA detection process takes at least a few hours and requires a specific analyzer (Rubio et al., 2017, Safaei et al., 2015). Moreover, because of the sensitivity of the ELISA kits, it cannot be used for the early stage diagnosis of a disease such as cancer (Fiala and Diamandis, 2018). Hence, the conventional ELISA kits cannot cope with urgent, quick and vigorous testing demands such as the recent COVID-19 pandemic. Fortunately, the graphene based FET sensor appears as a most promising approach to point-of-care medical diagnostics for rapid, sensitive, specific, low-cost detection and quantification of biomarkers (Afsahi et al., 2018), and thereby could be a complementary diagnostic tool to ELISA kits.

Graphene is a single atom thick, sp2-hybridized two-dimensional carbon material. Since its discovery, graphene's unique and extreme properties has triggered some revolutionary applications, such as batteries and super capacitor (Hu et al., 2016; Wen et al., 2016; Xiao et al., 2017; Zhang et al., 2018). Graphene also holds promise for energy storage (Yang et al., 2013), nano energy (Wu et al., 2012), catalysis (Machado and Serp, 2012), coating (Singh Raman et al., 2012) and biomedicine (Yang et al., 2012). Because of its extremely high electrical conductivity (200,000 cm2V-1s-1) and ambipolar transfer characteristics (Geim and Novoselov, 2009; Novoselov et al., 2004), graphene is considered as the most promising material for different types of ultra-sensitive graphene field-effect transistor (GFET) sensors such as chemical and biosensor. The extreme sensitivity of GFET relates to the Fermi energy shift induced by the adsorption of any molecules involving charge transfer between substrate and molecules, which ultimately changes the conductivity of graphene (Alberto and Pier, 2020). Even without any traceable change of Fermi level, any localized orbital distortion induced by the adsorbed molecule can also change the conductivity of graphene (Alberto and Pier, 2020). Hence, the carrier density of graphene is expected to be altered even by the non-covalent binding of biomolecules onto the surface, i.e., the electrical conductivity of graphene is expected to be altered by the non-covalent binding of a very small concentration of biomolecules onto the graphene surface.

The field-effect transistor (FET) is a type of transistor that uses an electric field to control the flow of current. In a typical FET, there are three terminals: source, gate, and drain. The flow of current between the drain and source through the conducting channel is controlled by the gate voltage (Frank, 2010). For the GFET sensor, not only the external gate voltage can regulate the current, but any change in graphene electronic states caused by the adsorption of molecules can also regulate the source-drain current (Deana et al., 2019). Note that graphene act as the conducting channel between source and drain in GFET. Thus, when a molecule is adsorbed onto the graphene surface in a working FET, the electronic states of the graphene channel are changed, and thereby changes the source-drain current. Hence the binding of any molecules onto the graphene channel of the GFET will trigger a real time change in source-drain current. Thus, the standard current vs concentration curve permits the quantitative analysis of target molecules. To date, a number of studies have been reported on the potential applications of GFET in biosensors (Ohno et al., 2009, 2010; Matsumoto, 2015; Afsahi et al., 2018; Cai et al., 2015; Pan et al., 2017; Xu et al., 2017; Yang et al., 2017; Yu et al., 2019; Zhou et al., 2017; Deana et al., 2019; Taniguchi et al., 2019; Nozaki et al., 2020). The detection of bio-macromolecules including cancer markers and RNAs by GFET biosensors has been reported using various types of acceptor/receptor design (Afsahi et al., 2018; Cai et al., 2015; Ohno et al., 2010; Pan et al., 2017; Wang et al., 2019; Xu et al., 2017; Yang et al., 2017; Yu et al., 2019; Zhou et al., 2017). Ohno et al. have reported the electrolyte gated GFET for pH and protein detection (Ohno et al., 2009). The aptamer modified GFET has been reported for label free detection of immunoglobulin (IgE) protein (Ohno et al., 2010). Recently, GFET has been shown to detect ethanol in ppb level (Nozaki et al., 2020). Deana et al. have reported the label free sensing of exosomes using functionalized graphene based FET (Deana et al., 2019).

In spite of enormous potential of avidin-biotin technology, the sensitivity and specificity of avidin-biotin interaction has not been investigated with GFET. Recently, Taniguchi et al. have studied the effect of pH on the interaction between avidin and iminobiotin using epitaxial graphene on SiC based FET and emphasized the need for concentration dependent measurements (Taniguchi et al., 2019). Here, we report the development of a novel GFET biosensor consisting of six pairs of interdigital electrodes covered by single layer graphene for detecting biotin molecules with ultrahigh sensitivity and specificity. Real time measurements are realized by monitoring the drain-source current changes upon the entry of biotin solution onto the avidin immobilized graphene channel. The detection limit of biotin is estimated to be 90 fg/ml (0.37 pM). Since the biotin can conjugate with a variety of proteins and nucleotides through biotinylation without changing their original properties, the proposed GFET, which integrates the graphene's extreme electrical conductivity with the unique avidin-biotin technology, is expected to be a breakthrough in the development of GFET biosensors for the rapid and point of care diagnosis of infectious diseases and different biomarkers with ultrahigh sensitivity and specificity.

Material and methods

Materials and instruments

A 4-inch 285 nm SiO2/Si wafer was purchased from ALLIANCE Biosystems (Osaka, Japan). The chemical vapor deposition (CVD) graphene on copper foil was purchased from Chongqing Graphene Technology Co., Ltd (Chongqing, China). Poly (methyl methacrylate) (PMMA) was purchased from Sigma-Aldrich (Tokyo, Japan). Ammonium peroxodisulfate was purchased from Kanto Chemical Co., Inc. (Tokyo, Japan). PBASE was synthesized. Neutravidin protein and phosphate-buffered saline (PBS) were purchased from Thermo Fisher (Massachusetts, USA). Bovine serum albumin (BSA), Lactoferrin, Vitamin C, Vitamin B3 were purchased from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan). Biotin was purchased from Tokyo Chemical Industry Co., Ltd (Tokyo, Japan). Ultra-pure water was obtained from water purifier WL220 (Tokyo, Japan).

3D height graphing of the interdigital electrode and substrate was carried out by 3D laser microscopy OLS-4000. A 405 nm laser source is used. Raman measurements were carried out on a Nicolet Almega XR Raman spectroscope with a 532 nm laser. The Raman spectra of the samples were acquired with a 100X objective, which reduced the spatial resolution to 1 μm. XPS measurements were performed on an AXIS-NOVA XPS system using an AlKα x-ray source. The incident and emission angles were 60 and 0° to the surface normal, respectively. The analyzer pass energies for the wide range and high-resolution measurements were set at 160 and 20 meV, respectively. The analyzer slit was set at 110 μm. A Keysight 4155B semiconductor parameter analyzer was used for the real-time GFET current measurements.

Device fabrication

The 4-inch 285 nm SiO2/Si wafer was cut into the typical size of 1 cm × 1 cm by a dicing saw (DISCO Corporation, Japan). Each 1 cm × 1 cm cut SiO2/Si was used as platform on which the interdigital electrode was fabricated. The interdigital shadow mask was carefully placed on the SiO2/Si surface. Metal chromium (Cr)/gold (Au) was used as the deposition material. The metal deposition with desired thickness was performed by the electron beam vacuum evaporation deposition (EIKO Engineering, Japan) technique. The thickness of the deposited Cr/Au was about 50 nm/50 nm.

In the transfer process, the graphene on the copper foil was coated with a thin layer of PMMA using a spin coater (Mikasa Corporation, Japan) followed by 5 min of heating at 90 °C to evaporate the acetone solvent. The PMMA/graphene on the copper foil was cut into small pieces of 5 mm × 5 mm to suit the interdigital electrodes. Subsequent etching of the copper foil using ammonium peroxodisulfate aqueous solution left the PMMA/graphene film floating in the solution. After complete etching of copper foil, the PMMA/graphene film was transferred on the ultra-pure water, held for 30 min, then carefully transferred onto the surface of the interdigitated electrode substrate. The transferred PMMA/graphene film was then dried in air for 30 min at room temperature and heated at 60 °C for 30 min. Finally, the removal of PMMA with boiling acetone and isopropyl alcohol (IPA) left the graphene attached to the interdigital electrodes (Fei et al., 2017, Lin et al., 2012). The cleanliness of the transferred graphene was confirmed by optical images (Fig. S1 in supporting information, SI). Note that single and bilayer graphene transferred on SiO2 substrate can be clearly visualized by optical microscope. There might have some undetected small patches of PMMA on the graphene, which we think does not invalidate the principle of the present GFET biosensing platform utilizing the avidin-biotin technology.

Graphene modification

Initially, the transferred graphene on interdigital electrodes was modified with PBASE, which is used as the linker for the non-covalent binding of the biomolecule with graphene. The surface modification with PBASE is very significant in maximizing device performances. Hence to find the best modification condition, the surface modification with PBASE was optimized by incubating the graphene on different interdigital electrodes samples in a dry dimethylformamide (DMF) solution of 50 mM PBASE for 1, 2, 4, 6 and 8 h at room temperature individually. The PBASE modified graphene was washed with methanol three times and dried with rotary pump. Then a pre-designed liquid pool made with silicone sheet was fixed on the top of the PBASE modified graphene so that the bio-solution to be tested can be injected into this test pool.

To capture the avidin molecules on the surface of PBASE modified graphene, 100 μl of 1 mg/ml avidin-PBS solution is injected into the pool, and held for 1 h at room temperature. Then the device is washed with 100 μl PBS three times. To prevent the non-specific binding of the residual surface regions (regions unmodified by PBASE) to other molecules, which may affect the signal during actual testing of the device, the remaining unmodified surface regions of the graphene were blocked by injecting 100 μl of 0.01 mg/ml BSA-PBS solution into the pool and holding for 1 h at room temperature followed by washing with 100 μl PBS three times.

Detection of biotin

Biotin buffer solutions with different concentrations were injected into the liquid pool while the real time current was measured to achieve the quantitative detection of biotin in PBS solution at pH 7.4. The source-drain voltage was maintained at 0.1 V while the gate (Si/SiO2) was grounded. The Keysight 4155B semiconductor parameter analyzer was used in the real-time current measurements. The specificity of the graphene biosensor was investigated by two sets of experiments. In the first set, the same amount of PBS, BSA, biotin, BSA solutions were injected into the liquid pool separately while the current was monitored. Similarly, the same amount of Vitamin C, Vitamin B3, and lactoferrin were injected separately while monitoring the current.

Results and discussion

Characterization of interdigital electrodes

Fig. 1 a shows optical images of the interdigital electrodes fabricated on SiO2/Si substrate. The six pairs of the interdigital electrodes were fabricated on the 1 cm × 1 cm SiO2/Si substrate. Electrode width was estimated to be 200 μm on average and the gap between the adjacent electrodes was 200 μm on average. From the 3D height profile, as shown in Fig. 1b, the thickness of the Cr/Au electrode was estimated to be about 100 nm.

Fig. 1

(a) The full-scale optical image of the interdigital electrodes fabricated on a SiO2/Si substrate. (b) The 3D image showing the height of the interdigital electrode and substrate.

Graphene transfer and modification

Fig. 2 a shows the schematic of the top and side views of the transferred graphene on the interdigital electrodes (i.e. GFET). To ensure the quality of the transferred graphene, Raman measurements were done randomly over the entire regions. A typical Raman spectrum of the transferred graphene is shown in Fig. 2b. The two major peaks 2D and G characteristics of the clean graphene were observed. The absence of any defect induced peak around 1300 cm-1 (D band) indicates the cleanliness of the transferred graphene. The intensity ratio (I(2D)/I(G)) is 2.7, which is consistent with the intensity ratio of defect free monolayer graphene (Zhou et al., 2013). In addition, the 2D band of the bare graphene can be fitted into only one Lorentzian peak (R2 is above 0.99) with the full width at half-maximum (FWHM) 30.8 cm-1 (Fig. S2 in SI), which also indicates the monolayer graphene (Hao et al., 2010).

Fig. 2

(a) Schematic illustration of the graphene biosensor fabricated on SiO2/Si substrate. Graphene was carefully transferred to cover all six pairs of interdigital electrodes. (b) The typical Raman spectrum of the graphene transferred onto the interdigital electrodes.

The modification of graphene with the linker is an important part of designing a functional GFET biosensor system, because it realizes the immobilization of a receptor biomolecule (i.e., avidin) on the graphene which can specifically bind to a target ligand molecule (i.e., biotin). The modification steps are shown in Fig. 3 . A conventional non-covalent linker used in the case of graphene, and the carbon nanotube-based device is PBASE, which is normally used to handle molecules containing lysine residue (Chen et al., 2001). The van der Waals force between the graphene and the pyrene backbone of PBASE molecule ensure their tight binding, this binding is also called π stacking. To the best of our knowledge, in spite of its importance, no detailed study on the coverage dependent modification and its effect on graphene's properties has been reported. Hence, to optimize the parameters for optimum coverage, we performed the time-dependent measurements. Fig. 4 a shows the Raman spectra of transferred graphene after modification with PBASE for the different times indicated (1 h, 2 h, 4 h, 6 h, and 8 h). We can see that defect induced D (1350 cm-1) and D′ band (1620 cm-1) appear and the intensities of both bands increase with modification time. The D band is related to the disorder in the honeycomb structure of sp2-hybridized carbon and arises from double resonances involving the nearest neighbour sites. The mechanism that gives rise to the D′ band is similar to that of the D band. Since the PBASE molecule is expected to be immobilized onto the graphene through non-covalent interaction, it is unlikely that the defects in graphene arise from sp3 carbons in the graphene's structure. The most likely reason for the appearance of the D and D′ bands is that the localized vibrational modes of the PBASE interact with the extended phonon modes of graphene. These results suggest that the degree of disorder of the sp2 hybridized carbon system gradually increases as the modification time increases. Fig. 4b shows the ratios of 2D and G bands (2D/G) at different modification times. The 2D/G ratio gradually decreases and approaches 1 at around the modification time of 4 h. At 6 h, the 2D/G ratio still remains close to the value of 4 h but it further decreases at 8 h. Ratio of 1 is equivalent to that of the bilayer graphene. This indicates that the aggregation of pyrenyl groups onto the monolayer graphene forms a structure similar to that of bilayer graphene. Hence, to avoid any uncovered regions and excessive defects, the 4 h modification time seems optimal given the use of the 50 mM PBASE solution in dry dimethylformamide (DMF).

Fig. 3

The schematic illustration of the stepwise modification of graphene with avidin molecules followed by the binding of biotin molecules.

Fig. 4

(a) Raman spectra of the transferred graphene after modification with PBASE for the different times indicated (1 h, 2 h, 4 h, 6 h, 8 h). (b) The change of 2D/G ratio of the Raman spectra shown in (a). (c) The I–V curves of the bare graphene and after 4 h modification with PBASE. (d) The high-resolution N 1s XPS spectra of the clean, PBASE modified and avidin modified graphene.

The modification of graphene with PBASE was also characterized by I–V and XPS measurements. Typical I–V curves for clean and PBASE modified (modified for 4 h) graphene are shown in Fig. 4c. A large shift of the Dirac point towards higher voltages is observed. This shifts in Dirac point indicates that the PBASE modification gives the graphene P-doping characteristics. The XPS spectrum shown in Fig. 4d shows a small N 1s peak around 400 eV following the PBASE modification, which ensures the presence of PBASE molecules on the graphene surface. Avidin is immobilized on the PBASE modified graphene through the interaction between lysine residue of avidin and N-hydroxysuccinimide ester group of PBASE. After avidin modification, a strong N 1s peak is observed, which indicates the increased number of nitrogen atoms from the avidin protein on the surface. No significant difference in the survey and C 1s spectra acquired before and after the PBASE modification of the bare graphene is observed (Figs. S3 and S4 in SI), which is in agreement with previous reports (Zhou et al., 2017). Thus the Raman, XPS and I–V measurements ensure the perfect PBASE modification of the graphene and immobilization of avidin onto the graphene channel of the GFET device.

Quantitative detection of biotin

Quantitative detection of biotin was performed by adding the biotin solutions to the test pool as shown in Fig. 5 a. The 2-terminal measurement system is used while the gate-terminal was grounded. A blank test was run using 20 μl of PBS solution. Then 20 μl of each of the different concentrations of biotin-PBS solutions (180 fg/ml, 18 pg/ml, 1.8 ng/ml, 18 ng/ml, 36 ng/ml) were introduced to the test pool subsequently. No leakage of any liquid from the pool was observed during the analysis. Note that the concentration of the biotin solution in the test pool is lower than the original concentration because of the subsequent addition of the solution. Hence the actual concentrations of the test solutions are estimated to 90 fg/ml, 6.06 pg/ml, 0.45 ng/ml, 3.96 ng/ml, 9.33 ng/ml. Fig. 5b shows the Ids values while adding different concentrations of the biotin solutions in the real-time measurements. It is observed that as the concentration of biotin solution increases, the stable current value Ids gradually decreases. After the standard solution is added, the current Ids value stabilizes within 1 min. By monitoring the sharp current drops upon addition of the biotin-PBS solution and subsequent recovery to stable Ids, the biotin is detected down to the concentration level of 90 fg/ml. Detection of biotin with such ultra-sensitivity is achieved because of the extremely high conductivity of the graphene. The Ids current approaches the minimum value when the concentration of the biotin is close to 3.96 ng/ml. These results indicate that the present GFET biosensor can detect biotin levels as low as 90 fg/ml. When the concentration of the biotin is higher than 3.96 ng/ml, the graphene biosensor capability approaches saturated adsorption, i.e., the detection range of the present graphene biosensor is between 90 fg/ml and 3.96 ng/ml. Real-time Ids measurements were also performed using the biotin concentrations below the detection limit (i.e., 0.18 fg/ml, 1.8 fg/ml, 18 fg/ml), but no traceable change in stable Ids values were observed (Fig. S5 in SI). Note that reproducibility of biotin detection was confirmed in at least five independent measurements.

Fig. 5

(a) The 2-terminal real-time test system. A silicone pool is constructed onto the graphene transferred on interdigital electrodes fabricated on SiO2/Si substrate. (b) The real time current (Ids) upon addition of different concentrations of biotin molecules. (c) Langmuir fitting to the changes of Ids at different concentration of biotin. (d) The real-time current (Ids) upon the addition of PBS, BSA, biotin, and BSA at different times.

Estimation of dissociation constant of avidin-biotin interaction

Based on the real-time current change upon addition of the different concentrations of biotin solutions in PBS as shown in Fig. 5b, ΔI max (maximum current change) and K (the dissociation constant) can be estimated using the Langmuir adsorption model ΔI = (C*ΔI max)/(C + Kd), C represents the concentration of the biotin. By fitting the experimental results shown in Fig. 5c, ΔI max and K are estimated to -209 μA and 1.6 × 10-11 M, respectively. The correlation coefficient (R2) is above 0.99. This Langmuir adsorption model has the potential to be used as the calibration curve for the quantitative detection of biotin in a given sample. Note that the theoretical dissociation constant of the avidin-biotin interaction is ~ 1 × 10-15 M (Gitlin et al., 1987). This difference in Kd value, which is in agreement with previous reports, is thought to be caused by binding of the avidin to the substrate surface (Neish et al., 2002; Nguyen et al., 2012; Wayment and Harris, 2009). In addition, the affinity of avidin for biotin may also have been affected when avidin is combined with PBASE.

The specificity of detection

In real applications, the sample to be tested contains not only the target molecule but also other molecules. Therefore, it is necessary to investigate sensor specificity. To confirm the specificity of the present GFET sensor, real time measurements with both the controlled and real samples were done. In the first set of controlled measurement, 20 μl of PBS, BSA, biotin, BSA solutions was poured sequentially into the test pool and observed the change in current. The concentrations of the BSA and biotin solutions were 0.01 mg/ml and 180 fg/ml, respectively. The results are shown in Fig. 5d. When the PBS and BSA solutions were added to the pool, the current signal remained almost unchanged, but upon addition of biotin solution, the current signal is dropped and remained unchanged even after introducing the BSA solution in the final step. In the second set of the experiment, 20 μl of 0.01 mg/ml of vitamin C, vitamin B3, and lactoferrin solutions in PBS was poured into the test pool sequentially and change in current was observed. The results are shown in Fig. S6 in SI. The results suggest that the current signal remains almost unchanged before and after the addition of the solution. These results indicate that the immobilized avidin does not interact with PBS, BSA, vitamin C, vitamin B3 and lactoferrin but readily interacted with biotin, i.e., the present GFET is very specific to the biotin molecules. For testing the specificity with a real sample, a preliminary experiment with chicken egg whites and yolk was performed (egg was purchased from supermarket). As expected, injection of 50 μl of egg yolk solution in PBS into the test pool of GFET induces the large drop of stable Ids current in real time measurements, which indicates the presence of free biotin or biotinylated compound. In contrast, injection of egg white solution in PBS does not induce any significant change in stable Ids current, which implies that no chemical components of egg white can non-specifically bind to the immobilized avidin on GFET. It is well known that the chicken egg yolk contains significant amount of biotin (13–25 μg/yolk), which is normally conjugated with specific biotin binding protein (BBP) (Meslar et al., 1978). These observations further support the specificity and sensing ability of the present GFET sensor in the case of a real sample. The detailed results are shown Fig. S7 in SI.

Exogenous biotin interferences

One of the main disadvantages of the avidin-biotin technology is the interference of exogenous biotin present in the blood of the subject which can range from 0.12 to 0.36 nM (Zempleni et al., 2009; Colon and Greene, 2018; Luong et al., 2019). Since the sensitivity of the present GFET lies in the pM range, the lowering or removal of biotin content in the target sample should be considered before charging the sample in the test pool. Different methods have been suggested for lowering or pre-removing the biotin from the blood sample. For example, biotin can be removed from blood samples by interacting the sample with avidin/streptavidin immobilized on insoluble matrices such as magnetic particles, polymers (e.g. agarose beads), silica, etc (Colon and Greene, 2018). Because of the short life time of biotin (2 h), the biotin content of the blood sample should be cleared within 4–5 h (Luong et al., 2019). Hence, for the analysis of blood sample using the present GFET sensor, the pre-removal of biotin content or the pre-estimation of the exogenous biotin interference should be considered. If the target sample is other than blood such as swab or exhaust gas, then the pre-removal of biotin will not be needed.

Comparative sensitivity and practical applicability

Recently, a number of studies have reported the utilization of avidin-biotin interaction focusing on sensing technology (Lin et al., 2019; Buzid et al., 2018, 2019; Kergaravat et al., 2012; Ho and Hung, 2008). The reported detection limits ranged from 2 pM to 84 nM using various detection methods such as competitive immunoassay, electrochemical, cyclic voltammetry and immunoaffinity chromatography as tabulated in Table 1 . The GFET biosensor presented here shows the highest sensitivity and the minimum time requirement for real time detection.

Table 1

A comparison detection limits as regards biotin concentrations taken from recent publications.

Platform	Material	Method	Limit of detection	Time	Reference
Beads and gold nanoparticles	anti-biotin antibody	competitive immunoassay	2pM	15min	Lin et al. (2019)
Boron-doped diamond electrode	captavidin	electrochemical sensing	1 nM	10min	Buzid et al. (2019)
Boron-doped diamond electrode	streptavidin	electrochemical sensing	5 nM	10min	Buzid et al. (2018)
Electrochemical magneto biosensor	streptavidin	cyclic voltammetry	84 nM	40min	Kergaravat et al. (2012)
Sepharose beads	Anti-biotin antibody	immunoaffinity chromatographic	41pM	30min	(Ho and Hung, 2008)
Graphene FET biosensor	streptavidin (neutravidin)	current Ids real-time monitor	0.37 pM	1min	This work

Similar to the biotin-avidin system used in the conventional ELISA kits (Carinelli et al., 2016; Safaei et al., 2015; Tauriello et al., 2018), the present GFET biosensor can be utilized for detection of antigen-antibody, hormone-receptor, nucleic acid system in body fluid, tissue and cell, and other bioactive macromolecules through the judicious choice of biotinylation. Because of its small size, biotin conjugation with protein, nucleic acid and other macromolecules does not appear to change the original bioactivity of these molecules (Wilchek and Bayer, 1990). The widely used biotinylation methods utilize enzymatic biotinylation, primary amine biotinylation, sulfhydryl biotinylation, carboxyl biotinylation, glycoprotein biotinylation, and oligonucleotide biotinylation (Diamandis and Christopoulos, 1991). In the clinical detection applications, it can be used for the detection of biomarkers via anti-biotin antibodies or avidins/streptavidin-tagged detection strategies such as enzyme reporters (e.g., horseradish peroxidase, alkaline phosphatase) or fluorescent probes. For example, by monitoring the real time Ids of GFET upon the addition of biotinyled M1 protein (the universal biomarker for influenza virus) on the avidin or streptavidin immobilized graphene channel, influenza can be diagnosed with ultrahigh sensitivity even before any symptom appears. Thus, because of the wide scope of avidin-biotin system, the present GFET is expected to be a breakthrough in the development of various ultrasensitive GFET biosensors for rapid and point-of-care low cost medical diagnosis. In addition, the ultra-high sensitivity of the present GFET is expected to allow quantitative detection of the exogenous biotin level in blood serum.

Conclusions

By combining the extreme electrical conductivity of graphene with the unique interaction between avidin and biotin molecules, we have realized a novel two terminal GFET biosensor for the detection of biotin with ultrahigh sensitivity and specificity. The GFET consists of six pairs of Cr/Au electrodes and an avidin immobilized single layer graphene channel supported on the Si/SiO2 substrate. By monitoring the source-drain current change upon addition of biotin solution in BSA, biotin can be detected to levels as low as 90 fg/ml (0.37 pM), and so is the most sensitive GFET for biomolecule detection. The detection range of the present GFET biosensor is estimated to lie between 90 fg/ml and 3.96 ng/ml. The high specificity of the GFET was confirmed both by controlled and real samples. Based on the Langmuir adsorption model, ΔI max and K are estimated to be -209 μA and 1.6 × 10-11 M. The correlation coefficient (R2) is about 0.99. Because of the wide range of conjugation capability of biotin molecules through biotinylation, the present GFET can be employed for the detection of various biomarkers and bio-macromolecules. In addition, it may also be used for the rapid quantitative detection of exogenous biotin. Because of its real-time rapid detection, ultra-low detection limit, and high specificity, the present GFET biosensor is expected to be a breakthrough in medical devices offering real-time point-of-care clinical diagnosis.

CRediT authorship contribution statement

Shiyu Wang: Conceptualization, Writing - original draft, Methodology. Md Zakir Hossain: Conceptualization, Writing - review & editing. Kazuo Shinozuka: Supervision. Natsuhiko Shimizu: Methodology. Shunya Kitada: Methodology. Takaaki Suzuki: Supervision. Ryo Ichige: Supervision. Anna Kuwana: Supervision. Haruo Kobayashi: Conceptualization, Funding acquisition, Supervision, Writing - review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.