SARS-CoV-2 detection by using graphene FET arrays with a portable microfluidic measurement system

We developed graphene FET (G-FET) arrays combined with a portable microfluidic measurement system for SARS-CoV-2 detection. Multiple G-FETs modified with SARS-CoV-2 spike antibodies and those not modified were integrated onto the same chip. By calculating the difference in the FET-responses, we aimed to minimize noise including virus physisorption and baseline drifts. The microfluidic system was used to change ionic strengths of buffers without manual pipetting. The virus was incubated in a high ionic strength solution, followed by electrical measurements in a low ionic strength solution, leading to effective binding and electrical detection. Upon introducing the virus at a concentration of 108 virus ml−1, a response of 7.9 mV was obtained. To confirm whether the response was attributed to the virus, we employed a scanning electron microscope (SEM). SEM observation indicates that the virus was much adsorbed on the antibody-modified surface compared to the non-modified surface, which agrees with the G-FET response.


Introduction
Graphene possesses exceedingly high electron/hole mobilities when compared to conventional semiconductors. 1,2)n addition, as its carbon atoms are exposed to the surface, it is easy to immobilize various receptors via non-covalent bonding such as π-π interactions. 3)These distinctive properties render graphene a highly sensitive sensing platform.We have developed a variety of graphene-based sensors [4][5][6] and technologies, 7,8) and demonstrated detection of biomolecules including proteins, [9][10][11] lectins, 12) and bacteria.15][16][17][18] For example, E. Piccinini et al. detected spike protein using antibodymodified G-FETs, with a limit of detection (LOD) of 6.7 nM.16) A. Silvestri et al. also detected spike protein using angiotensin-converting enzyme 2 (ACE2) modified G-FETs, with a LOD of 2.4 aM. 17) Wile many studies have employed spike proteins as the sample, only a few have demonstrated the detection of whole viruses. 18)Furthermore, in most of the studies, a single G-FET sensor was used for virus detection, despite the fact that the electric characteristics of G-FETs vary significantly due to the process variation, 19) thereby rendering results based on single G-FET measurements potentially controversial. Morover, baseline drift is a well-known drawback of G-FET biosensors because the drift makes it difficult to distinguish between signal and noise.20,21) Nonetheless, the issue is often dismissed and not discussed.
Here, we measured the whole SARS-CoV-2 using a G-FET array integrated with a portable microfluidic measurement system.The array was composed of multiple G-FETs with SARS-CoV-2 spike antibodies.Multiple non-modified G-FETs were also integrated onto the same chip, and used as reference sensors.As the G-FET characteristics were diverse due to the fabrication process variation, we calculated their averages to cancel out the variation and improve the measurement reliability.We also calculated the difference in G-FET responses between the antibody-modified and nonmodified G-FETs to suppress noise from the physisorption of the virus and baseline drifts.Furthermore, we implemented a microfluidic channel over the G-FET array to change buffer ionic strengths.The virus is effectively bound to receptors in a physiological ionic strength, 22) whereas the G-FET response becomes stronger in a lower ionic strength, 23) indicating a trade-off between binding and detection.Therefore, we incubated the virus in a higher ionic strength solution and subsequently measured the virus in a lower ionic strength solution.The portable microfluidic measurement system facilitated buffer exchange without manual pipetting, reducing perturbations by the exchange and improving reproducibility.Following the virus measurements, we observed the G-FET array by scanning electron microscopy (SEM) to confirm whether the signal was responsible for the virus binding.Our custom-made portable microfluidic measurement system enables one to measure the virus outside of the laboratory including at home.

G-FET array fabrication
The G-FET arrays were fabricated as follows.Source and drain electrodes were formed by depositing 10 nm/90 nm of Ti/Au on a SiO 2 /Si substrate.The graphene film was transferred using poly methyl methacrylate (PMMA) onto the prepared SiO 2 /Si substrate, and patterned via oxygen plasma etching, resulting in the integration of 32 G-FETs on the substrate.The PMMA film was dissolved in ethanol.A photo of a G-FET array on the substrate is shown in Fig. 1(a).G-FETs were arranged in an array of two rows and 16 columns, and numbered from #0 to #31, as depicted in Fig. 1(b).An optical image of a typical G-FET is shown in Fig. 1(c).The channel length and width are 10 μm and 100 μm, respectively.

Antibody modification to the G-FET array
The prepared G-FET array was modified with the antibody using the following procedure.To develop the antibodymodified G-FETs and the non-modified G-FETs individually, silicone rubber was used to separate the G-FET array into two regions: #0-#14 and #19-#31.The G-FETs of #15-#18 located under the rubber were not functionalized and not used in the analysis.PBASE was dissolved in 2-methoxyethanol to a concentration of 5 mM.PBASE solution was subsequently dropped and incubated on the G-FETs (#0-#14) at room temperature for 1 h.On the other hand, the G-FETs (#19-#31) were incubated with 2-methoxyethanol alone, without PBASE.Following the PBASE modification, the antibody was immobilized as follows.The antibody of a concentration of 40 μg ml −1 in 0.2 × PBS was incubated on the PBASE-modified G-FETs (#0-#14) overnight at room temperature.The antibody was immobilized on the graphene surface through the process.On the other hand, the G-FETs (#19-#31) were incubated with 0.2 × PBS alone, resulting in the non-modified G-FETs.After the antibody modification, the rubber with two chambers was removed and then the microfluidic channel was set, as shown in Fig. 1(d).Finally, the G-FET array was treated with 0.01% tween 20 in 0.01 × PBS at room temperature for 1 h to block nonspecific adsorption.

Preparation of inactivated SARS-CoV-2
For safe experiments, SARS-CoV-2 (JPN/TY/KW-521) was inactivated by using β-propiolactone, which is widely used for the inactivation of viruses. 24)The inactivated viruses were then filtrated and purified by ultracentrifugation.The resultant virus concentration was 1 × 10 9 virus ml −1 .Prior to the FET experiments, the affinity between the inactivated virus and the spike antibody was confirmed by using bio-layer interferometry.The virus solution was diluted with 0.01% tween 20 in 1 × PBS and applied to the FET measurement.

Portable microfluidic measurement system
A microfluidic channel was set on the G-FET array, as shown in Fig. 1(d).Figure 1(e) provides an overview of the portable microfluidic measurement system.The microfluidic channel comprises four through holes, where the side two holes function as an inlet and outlet for 0.01 × PBS.0.01 × PBS flowed through the channel at a constant rate of 7 μl s −1 by using micropumps.The central hole is for inserting a Ag/AgCl electrode.In this study, the Ag/AgCl electrode was used as a reference electrode without KCl salt bridge.In the present study, only 0.01 × PBS was used during the measurement for the viral detection.Therefore, the potential disturbance does not occur by the solution exchange.The hole beside the inlet is for injecting the virus solution or 1 × PBS.The cavity volume is approximately 100 μl.

Electrical characterization
All electrical measurements were performed in either 0.01% tween 20 in 0.01 × PBS or 0.01% tween 20 in 1 × PBS.A bias voltage V DS was set to 100 mV.A top gate voltage V GS was applied through the solution via a Ag/AgCl electrode.While sweeping V GS from 0 to 500 mV, I DS was measured for each G-FET with the custom-built measurement equipment.The V GS at the minimum I DS is known as the charge neutrality point or Dirac point (DP).The value of DP for each G-FET was calculated by polynomial fitting, and used to evaluate the G-FET response.I DS of DP changes when gm and/or the resistance change.Therefore, the measurement at a constant I DS could not be used.During the experiments, a few G-FETs were disconnected and those were not used for the analysis.

SEM observation
After the electrical measurements, the viruses on the device were observed by using SEM (S-4800, Hitachi High-Tech).Multiple SEM images were taken to observe the entire channel for each G-FET.From the images, the number of adsorbed viruses was manually counted.

Virus measurement with the G-FET array
Using the fabricated devices, we monitored responses to SARS-CoV-2.To enhance the detection efficiency, buffer exchange was implemented during the measurements.It is known that viruses can bind effectively to receptors when the ionic strength approaches a physiological condition, i.e. 1 × PBS (150 mM) 22) .However, in 1 × PBS, Debye length, which is an electrolytic screening length, is estimated to be 0.7 nm, 27) which is considerably shorter than the size of IgG antibodies. 28,29)Therefore, the viral charge is screened by the electrolyte and not effectively transduced as the G-FET response.In contrast, Debye length in 0.01 × PBS (1.5 mM) is estimated to be approximately 7 nm, 27) which is 10 times as long as that in 1 × PBS, thus improving the transducing efficiency.However, the binding efficiency is reduced in a lower ionic strength environment.Consequently, there is a trade-off between the binding and transducing efficiency.To mitigate the trade-off, we used 1 × PBS for the virus binding and 0.01 × PBS for the electrical measurements.The buffer exchange was conducted by using the microfluidic system.Figure 3(a) displays the time series of DP for the virus measurement while changing the ionic strengths of the buffers.The measurement includes seven phases denoted as (i)-(vii).During the measurements, the blank sample, i.e. 1 × PBS with no virus, was introduced twice at the (ii) and (iv) phases to evaluate how much the buffer exchange process affected the DP values.Prior to the (i) phase measurement, the device was treated with 0.01% tween 20 for blocking, which resulted in a gap in the initial DP values between the antibody-modified and non-modified G-FETs.In the (i) phase, the G-FETs were measured while keeping 0.01 × PBS flowing.Following the (i) phase, 0.01 × PBS was completely drained and 1 × PBS was subsequently injected and measured for 1 h in the (ii) phase.In the (iii) phase, the G-FET array was washed with 0.01 × PBS flowing and measured in 0.01 × PBS, as in the (i) phase.The (iv) and (v) phases followed the same protocol as in the (ii) and (iii) phases, respectively.Followingly, in the (vi) phase, SARS-CoV-2 in 1 × PBS with a concentration of 1 × 10 8 virus ml −1 was injected and incubated for 1 h.Finally, in the (vii) phase, the virus solution was exchanged to 0.01 × PBS, in which unbound viruses were washed away, and then the device was measured for a while.It is noted that in the (ii), (iv), and (vi) phases, where 1 × PBS was used, the DP values were not calculated because those were shifted downwards and out of the V GS sweep range, although the electrical measurements were performed during these phases.
To analyze the response of each G-FET against the sample, we used DP values at the endpoints of the measurement phases ((i), (iii), (v) and (vii)) and defined ΔV DP , which is a change against the samples, i.e. the difference between before and after the sample injection phases ((ii), (iv) and (vi))., to cancel out noises due to baseline drifts and physisorption of substances.S for the first and the second blank samples were determined as 10.1 mV and 7.0 mV, respectively.Consequently, we supposed the mean S for the blank sample (S 0 ) was equal to 8.6 mV, which can be considered as a perturbation of the buffer exchange.For the results of the virus sample in Fig. 3(d), ΔV DP w and ΔV DP w/o were calculated as 19.3 mV and 2.8 mV, respectively.Both the ΔV DP w and ΔV DP w/o were considerably upshifted by the virus introduction, in contrast to the blank samples.The upshifts may be attributed to the negative charge of the virus. 30)In addition, the positive value of ΔV DP w/o implies the presence of physisorbed viruses.Based on the obtained ΔV DP w and ΔV DP w/o , S was calculated as 16.5 mV, in which the response due to the physisorption of the virus was canceled out by subtraction.In this study, the important parameter in the preset study is not the fluctuation of the baseline, but the amount of S. S of the first and the second blank samples is the low value of 10.1 mV and 7.0 mV, respectively.In contrast, when the SARS-CoV-2 was introduced to the microfluidic channel, S is increased as high as 16.5 mV.Thus, it is said that S was increased by the binding of SARS-CoV-2 to antibody selectively.
It is noted that the S of 16.5 mV should include S 0 , which is the perturbation of the solution exchange process.Consequently, the net S against the virus injection was estimated to be S − S 0 = 7.9 mV.It is important to note that the size of the viruses approximately 100 nm 31) was much larger compared to the Debye length of about 7 nm in 0.01 × PBS.Therefore, the charges of the underside of the virus may be accountable for the net S.
The detection of SARS-CoV-2 in saliva by G-FET is the future target.SARS-CoV-2 in saliva have been tried to detect by using G-FET modified with SARS-CoV-2 spike RBD antibody and the reference G-FET modified with influenza antibody.G-FET modified with influenza antibody is the more accurate reference than non-modified G-FET used in the present study.It was confirmed that saliva does not affect the baseline because DP did not shift before and after the introduction of the saliva.

SEM observation to validate the G-FET response
To confirm whether the signal was responsible for the virus binding, the G-FET array after the electrical measurements was observed by using SEM.Figures 4(a 4(b), the size of the virus was approximately 100 nm, which aligns with previously reported values. 31)For comparison, a typical SEM image of a G-FET without the antibody modification is presented in Fig. 4(c), showing fewer adsorbed viruses.The numbers of the viruses in each G-FET channel are summarized in Fig. 4(d).In the antibody-modified G-FETs, the number of the viruses ranged  from 21 to 56, with an average of 37.In contrast, the number of viruses ranged from 1 to 9, with an average of 4, in the non-modified G-FETs.The results indicate that the adsorption of viruses on the antibody-modified graphene surface was roughly 10 times more than on the non-modified surface, as shown in Fig. 4(d).Although the SEM analysis is unable to distinguish between specifically adsorbed and physisorbed viruses, the SEM results support that the viruses were preferentially adsorbed onto the antibody.Figures 4(a)-4(c) is the SEM image after the washing process of the viral measurement.Even after the washing process, SARS-CoV-2 selectively binding to antibody can be observed.Figure 4(e) plots ΔV DP in Fig. 3(d) as a function of the number of viruses.The plot confirms that larger ΔV DP w was attributed to increased virus adsorption.Additionally, Fig. 4(e) also suggests the presence of physisorbed viruses on the non-modified graphene surface, which may be responsible for the positive values of ΔV DP w/o against the virus injection.Based on these results, it can be concluded that our G-FET devices responded to the virus specifically adsorbed onto the antibody.

Conclusions
We have developed G-FET arrays with the portable microfluidic measurement system for SARS-CoV-2 detection.Two types of G-FETs, i.e. with and without the antibody modification were integrated on a single substrate.Based on the configuration, we aimed to minimize noise due to baseline drifts and physisorption of substances, by computing the difference between the average DP values of the antibody-modified and the non-modified G-FETs.Upon introducing the virus at a concentration of 1 × 10 8 virus ml −1 to the G-FET array, the virus response was obtained as the net DP shift of 7.9 mV.The response was verified through SEM observation.The SEM analysis demonstrated that the number of viruses on the antibody-modified G-FETs was approximately 10 times greater than that on the non-modified G-FETs, indicating the selective binding of the virus to the antibody.Therefore, the SEM analysis supported the FETresponse against the virus.Furthermore, our measurement system offers several advantages, such as a simple protocol and short measurement time, and portability, when compared to conventional PCR tests.Owing to these advantages, the presented G-FET biosensor system has the potential to contribute to the development of reliable point-of-care rapid test technology for pandemic prevention.

Fig. 1 .
Fig. 1.(a) Photo of a G-FET array.(b) Schematic of the G-FET arrangement on the electrode pattern.The G-FETs were numbered from #0 to #31.(c) Optical image of a typical G-FET before the modification.(d) Photo of the microfluidic channel over the G-FET array.(e) Photo of the portable microfluidic measurement system during electrical measurements.
Figures 3(b)-3(d) represents the ΔV DP for the first and second blank samples, and for the virus, respectively.According to Figs. 3(b), 3(c), the means of ΔV DP of the antibody-modified G-FETs (ΔV DP w ) for the first and second
), 4(b) is typical SEM images of a part of the antibody-modified G-FET channel.The arrows in Fig. 4(a) indicate the positions of the viruses.As shown in Fig.

Fig. 3 .
Fig. 3. (a) The time series of DP values for the virus measurement including (i)-(vii) phases.In the (ii) and (iv) phases, 1 × PBS was injected as a blank sample, and in the (vi) phase, the virus with a concentration of 10 8 virus ml −1 was injected.The mean DP values of the G-FETs with the antibody and those of the G-FETs without the antibody were plotted as red and blue lines, respectively.The shades represent their standard deviation.The gray bars represent the period when the electrical measurements were not performed properly due to the airflow through the microchannel.(b), (c), (d) ΔV DP for the first and second blank samples, and for the virus, respectively.

Fig. 4 . 5 ©
Fig. 4. (a) SEM image of the G-FET with the antibody (#5).(b) The enlarged SEM image of (a).(c) SEM image of the G-FET without the antibody (#19).The arrows in (a) and (c) indicate the position of the viruses.(d) The number of adsorbed viruses for each G-FET channel.(e) ΔV DP as a function of the number of adsorbed viruses.Red and blue colors represent the G-FETs with the antibody and those without the antibody.