Rapid, high-sensitivity detection of biomolecules using dual-comb biosensing

Rapid, sensitive detection of biomolecules is important for biosensing of infectious pathogens as well as biomarkers and pollutants. For example, biosensing of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is still strongly required for the fight against coronavirus disease 2019 (COVID-19) pandemic. Here, we aim to achieve the rapid and sensitive detection of SARS-CoV-2 nucleocapsid protein antigen by enhancing the performance of optical biosensing based on optical frequency combs (OFC). The virus-concentration-dependent optical spectrum shift produced by antigen–antibody interactions is transformed into a photonic radio-frequency (RF) shift by a frequency conversion between the optical and RF regions in the OFC, facilitating rapid and sensitive detection with well-established electrical frequency measurements. Furthermore, active-dummy temperature-drift compensation with a dual-comb configuration enables the very small change in the virus-concentration-dependent signal to be extracted from the large, variable background signal caused by temperature disturbance. The achieved performance of dual-comb biosensing will greatly enhance the applicability of biosensors to viruses, biomarkers, environmental hormones, and so on.

Biosensors are biomolecular sensors that utilize the skilful molecular identification function of living organisms; they are applied to a wide range of fields, such as medical care, food industry, and environmental monitoring.However, further enhancement of the biosensing performance is still required for improved testing methods for infectious pathogens as well as biomarkers and RNA.One timely and urgent application that benefits from improved performance is testing for coronavirus disease 2019 .COVID-19, caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has rapidly spread and is still occurring all over the world.
One reason for the failure to suppress the rapid spread of COVID-19 is the timeconsuming testing process for SARS-CoV-2 because it hinders sooner quarantine and thus presumably prevention of the spread of COVID-19.The current standard for COVID-19 testing is reverse-transcription polymerase chain reaction (RT-PCR) [1][2][3], which sensitively detects SARS-CoV-2 RNA within a limit of detection (LOD) range from sub-aM to several aM.However, RT-PCR requires multiple time-consuming steps (required time = 4~5 hours).Although a qualitative antigen test enables rapid, simple testing (required time = 15~30 min), it requires a certain amount of virus due to the limited sensitivity.To facilitate earlier quarantine, there is a considerable need for rapid, high-sensitivity detection of SARS-CoV-2.
Among the potential methods that may improve the SARS-CoV-2 detection time and sensitivity is the use of optical biosensors due to both their rapidity and high sensitivity [4,5].For example, optical biosensors based on surface plasmon resonance (SPR) [6,7] have been widely used for analysing viruses [8][9][10][11][12][13][14], proteins [15,16], DNA [17,18], and whole cells [19,20]; the sample-concentration-dependent optical spectral shift of the SPR trough in the wavelength or angular spectrum is measured through the combined effect of SPR and the molecular identification function on the sensor surface.SPR enables real-time, label-free analysis of intermolecular interactions.The LOD for SARS-CoV-2 has reached 85 fM [14]; however, it is not yet at the RT-PCR level.A reason for the limited sensitivity is the optical instrumentation resolution as well as the relatively broad spectrum of the SPR trough compared to its slight spectrum shift.
One promising method to overcome the limitation of optical instrumentation resolution is to transform the sample-concentration-dependent optical spectral shift into its equivalent photonic radio-frequency (RF) spectral shift because such photonic RF biosensing would benefit from the high precision and real-time nature provided by mature electrical frequency measurements with frequency standards.Recently, optical frequency combs (OFCs) [21][22][23][24] have attracted attention for use as photonic RF sensors based on a frequency conversion function between the optical and RF regions [25][26][27][28].An OFC is composed of a series of optical frequency modes (freq.= nm) with a constant mode spacing of frep in the RF band.The relation between nm and frep is given by (1) where fceo is the carrier-envelope-offset frequency and m is the mode number.Thus, the OFC acts as an accurate frequency converter between nm and frep.For example, a refractive-index (RI)-dependent optical spectrum shift was converted into a change in frep by placing a multimode-interference (MMI) fibre sensor [27,28] inside a fibre OFC cavity to realize an RI-dependent variable-optical-bandpass filter [29][30][31].Then, the frep signal was rapidly and precisely measured by an RF frequency counter.The photonic-to-RF conversion reduces the spectral linewidth to below 1 Hz [29].
Furthermore, the intracavity fibre sensor enables multiple interactions between the sample and the light, reducing the resolution to 4.9×10 -6 refractive index units (RIU), which is two orders of magnitude better than that in a previous study [27].Such RIsensing OFCs would have the potential to be further extended to optical biosensing through surface modification of the MMI fibre sensor with a molecular identification layer in terms of biomolecule interactions, similar to SPR.However, there are no attempts to apply such RI-sensing OFCs for optical biosensing because the residual temperature drift of frep (typically, a few hundreds of Hz/hour) is larger than the sampleconcentration-dependent frep shift (typically, a few to a few tens of Hz), hindering their extension to biosensing OFCs.Largely reducing the temperature drift of frep is essential to achieve both a sensitivity close to that of RT-PCR and a measurement time considerably shorter than that of RT-PCR.Thus, in this article, we first developed a dual-comb configuration [32] with an active sensing OFC and a dummy OFC to suppress the temperature drift of frep, namely, dual-comb biosensing; this function is similar to the active-dummy temperature compensation of strain sensors.Then, for a preliminary test, we applied the active-dummy dual-sensing OFCs to RI sensing of a glycerol solution.Finally, as a proof of concept, we demonstrated rapid detection of the SARS-CoV-2 nucleocapsid protein (N protein) antigen by combining dual-comb biosensing and surface modification with the SARS-CoV-2 N protein antibody.

General principle of operation
In this study, we sought to design a biosensor that combines photonic-to-RF conversion and antigen-antibody interactions in an OFC.The biosensing OFC operates through three steps: (1) antigen-antibody interactions on the antibodymodified sensor surface, (2) RI-dependent optical spectrum shift of the OFC provided by the intracavity MMI fibre sensor [27,28,33], and (3) photonic-to-RF conversion via the wavelength dispersion of the fibre cavity [29][30][31], as depicted in Fig. 1a

Temperature drift in the single-comb configuration
We first evaluated the cavity temperature dependence of frep in a single sensing OFC because temperature disturbances cause frep to fluctuate via thermal changes of nL.To this end, we measured the temporal fluctuations in the frep of the single sensing OFC under an uncontrolled cavity temperature, as shown in Fig. 2a.
Pure water was used as a standard sample with a stable RI and placed in a glass sample cell together with the MMI fibre sensor without surface modification for RI sensing.The output light from the OFC was detected by a photodetector (PD), and frep was measured by an RF frequency counter synchronized to a rubidium frequency standard working in the RF band. Figure 2b shows the frep shift (dfrep) when the cavity temperature changed over a range of 1 °C.dfrep represents the frequency deviation from the initial measurement value.The temporal behaviour of dfrep in synchronization with the cavity temperature indicated a temperature sensitivity of approximately -400 Hz/°C.This cavity-temperature-dependent frep drift is considerably larger than the sample-concentration-dependent frep shift in biosensing (typically, a few to a few tens of Hz).Although the cavity temperature was actively controlled within a range of 25.0±0.1 °C in the following experiments, this was still insufficient to suppress the cavity-temperature-dependent frep drift to below the sample-concentration-dependent frep shift.Thus, to further reduce the temperature drift, we applied a dual-comb configuration for active-dummy temperature compensation together with active control of the cavity temperature, as described in the following subsection.

Active-dummy compensation of the temperature drift with the dual-comb configuration
A dual-comb configuration [32] with an active sensing OFC with a frequency spacing of frep1 and a dummy sensing OFC with a frequency spacing of frep2 was adopted to compensate for the temperature drift.Figure 3a shows a schematic drawing of the dual-comb configuration, in which a pair of fibre OFC cavities were arranged in a temperature-controlled box so that they were affected by similar temperature fluctuations.In this configuration, although frep1 and frep2 fluctuate depending on the residual fluctuation of the cavity temperature, their fluctuations are similar because they experience the same thermal disturbances.Therefore, the frequency difference ∆frep between frep1 and frep2 remains constant regardless of the temperature drift of frep1 and frep2.Thus, when the active sensing OFC evaluates a sample solution in a certain temperature environment and the dummy sensing OFC evaluates a reference material in the same temperature environment, ∆frep reflects the sample concentration without influence from temperature drift.Figures 3b, 3c, and 3d show the MMI and sample cell for dual-comb RI sensing of pure water, dual-comb RI sensing of glycerol solution, and dual-comb biosensing of the SARS-CoV-2 N protein antigen, respectively.Table 1 summarizes frep1, frep2, ∆frep, the MMI, and the sample cell used in the following three dual-comb sensing experiments; these frequency values were selected for stable operation and better temperature compensation.A pair of output light beams from the active and dummy sensing OFCs was detected by a pair of PDs; then, frep1, frep2, and ∆frep were measured by the RF frequency counter (see the Methods section).
The blue and green lines in Fig. 4 show the temporal shifts in frep1 and frep2, namely, dfrep1 and dfrep2, respectively, when pure water was used as a sample for both the active and dummy sensing OFCs without surface modification (see Table 1 and Fig. 3b).dfrep1 and dfrep2 suffered from a temperature drift of over -38 Hz due to an increase in the cavity temperature; however, they behaved almost the same in terms of drift.The resulting ∆frep shift (d∆frep) was stable, with a variation of up to 1.18 Hz, as shown by the red line in Fig. 4.This level of ∆frep stability is sufficient for precise measurement of the sample-concentration-dependent ∆frep shift.
We next tested active-dummy temperature compensation for RI sensing of a liquid sample different from the reference sample.For RI sensing, the active and dummy sensing OFCs have no surface modification (see Table 1 and Fig. 3c).We used glycerol solutions consisting of glycerine and pure water at different ratios, corresponding to different RIs, as target samples in the active sensing OFC.We exchanged the target sample and the reference sample (pure water) by using a pair of peristaltic pumps.The blue and green lines in Fig. 5a represent dfrep1 and dfrep2 as the glycerol solution concentration increased from 0 vol% to 5 vol%.dfrep2, in the dummy sensing OFC, exhibited a slow drift with some rapid fluctuations.Since these rapid fluctuations were synchronized with the operation of the peristaltic pump, they were caused by disturbances from the water flow when the samples were exchanged.
In contrast, dfrep1, in the active sensing OFC, exhibited a combination of a step-like change with the sample RI and the slow drift shown by dfrep2.This combined behaviour of dfrep1 is detrimental to the RI sensing performance in the single sensing OFC configuration.From the mean of d∆frep at each concentration, we calculated a relationship between the sample RI and d∆frep, indicated by the red line in Fig. 5c.The good fitting result indicated that the dual-comb effect minimizes the effect of temperature fluctuations.

Rapid detection of the SARS-CoV-2 N protein antigen
Antibody modification of the intracavity MMI fibre sensor creates a photonic RF biosensor for the detection of target antigens through antibody-antigen reactions (see Fig. 1a).The enhanced RI precision above covers the RI change expected for antigen-antibody interactions, enabling us to apply these dual-sensing OFCs to rapid, high-sensitivity detection of viruses/pathogens and biological molecules.
The concept of antigen-antibody interactions (in this case, a viral protein) was applied for detection of SARS-CoV-2 protein.Among several proteins in SARS-CoV-2, the N protein, which functions to package the viral RNA genome within the viral envelope into a ribonucleoprotein complex, is a promising candidate for antigenantibody interactions because of its abundance, low probability of mutation, and relatively low molecular weight.Thus, we used the combination of a commercialized N protein monoclonal antibody and a commercialized N protein recombinant antigen, which exhibited high affinity in an enzyme-linked immunosorbent assay, for the intracavity MMI biosensor (see Fig. 1b).The MMI fibre sensors with and without surface modification by the immobilized antibody were placed together in the same sample cell as the active and dummy sensing OFCs, respectively (see Table 1 and Fig. 3d).Solution samples of the N protein antigen in phosphate-buffered saline (PBS) at different molar concentrations were consecutively introduced into the sample cell with a peristaltic pump.The N protein antigen-antibody interaction could only occur on the sensor surface of the active sensing OFC because the surface of the dummy sensing OFC did not include immobilized antibodies.Thus, the dummy sensing OFC was used as a negative control.
Figure 6a shows the sensorgram of dfrep1 and dfrep2 as the molar concentration of the antigen/PBS solution increased from 1 aM ~ 1 nM after starting with pure PBS; this range of molar concentrations was selected for the sample considering the LOD for SARS-CoV-2 of RT-PCR.The time period for data acquisition (see colour-highlighted zones in Fig. 6a) was set to 10 min.Each grey zone (time period = 8 min) includes the sample change by the pump, the waiting time for completion of the antigen-antibody interaction, and rinsing of the sensor surface with PBS.Since the step-like change in dfrep1 with the antigen concentration was completely overshadowed by the background temperature drift, we calculated the frequency difference (d∆frep) between dfrep1 and dfrep2 to eliminate the influence of temperature drift as described in the previous subsection.Figure 6b shows the sensorgram of d∆frep.
Focusing on the zones highlighted in colours other than grey, a slightly dull stepped change in d∆frep dependent on the molar concentration was observed, although a small drift in d∆frep within the range of a few Hz remained at each molar concentration.To evaluate the validity of this behaviour, we calculated the relationship between the molar concentration and d∆frep, as shown by the red circles plotted in Fig. 6c.The negative slope was consistent with the RI dependence of d∆frep (see the red line plotted in Fig. 5c) because the progression of the antigen-antibody reaction increases the effective RI near the MMI fibre sensor and hence decreases frep [34].In this way, we demonstrated the potential for rapid detection of the SARS-CoV-2 N protein antigen within this range of molar concentrations.

Discussion
To achieve rapid, high-sensitivity biosensing, we developed dual-comb biosensing and applied it to the detection of the SARS-CoV-2 N protein antigen with a molar concentration from 1 aM to 1 µM in a measurement time of 10 min.We performed the quantitative analysis of the result shown in Fig. 6c.The antigenantibody reaction is represented by a sigmoidal curve, often used when discussing the biosenser performance [35]; thus, a sigmoidal curve was applied to the experimental data to evaluate the ability of dual-comb biosensing to sense the SARS-CoV-2 N protein.The sigmoidal function of the Hill plot is given by (2) where d∆frep_max and d∆frep_min are the maximum and minimum of d∆frep, n is the Hill coefficient, C is the concentration of the antigen, and CKa is the dissociation constant.
The purple line in Fig. 6c represents the sigmoidal fit.At the end of the grey zones in Fig. 6b, the d∆frep signal reflects the amount of antigen adsorbed on the sensor surface after desorption.From the curve fitting analysis, we determined CKa to be 8.4 fM, indicating the molar concentration at the middle between d∆frep_max and d∆frep_min (see pink dashed lines in Fig. 6c).When the linear range (LR) was defined as the molar concentration at 10~90 % of the dynamic signal range of d∆frep_max to d∆frep_min (see blue dashed lines in Fig. 6c) [35], it was determined to be 34 aM ~ 2.1 pM.Additionally, the LOD was calculated to be 38 aM from the crossover point between d∆frep_max and the linear approximation given by the Hill coefficient n (= 0.40, see the green lines in Fig. 6c).
We next compare the dual-comb biosensing with other SARS-CoV-2 testing methods, as shown in Table 2. RT-PCR testing shows an LOD of ~ 100 copies/ml [36], corresponding to 0.17 aM; however, it requires a long analysis time.SPR [37] and its enhancement [14,38] could achieve an LOD of sub-pM ~ pM in rapid analysis.The LOD of the colorimetric assay remains at approximately a few tens of pM [39].
Importantly, only dual-comb biosensing achieves an LOD close to that of RT-PCR with rapid measurement considerably shorter than that of RT-PCR.
We also discuss the potential for dual-comb biosensing to be used for the detection of other viruses and biomolecules of interest.The considerably low CKa implies the potential for a wide variety of biosensing applications, for example, early detection of cancer cells from a droplet of blood or other body fluids by detecting a sugar chain specifically expressed on the cell surface.Such blood biopsy or liquid biopsy will be a powerful tool for detection of important biomarkers, such as proteins or RNA, in addition to cancer cells.As another interesting application, biosensing of exosomes via miRNA is expected to make a great contribution to the diagnosis (marker) and treatment (drug delivery) of diseases such as cancer and Alzheimer's disease because exosomes play an important role as an intercellular communication tool.

Conclusion
We have demonstrated dual-comb biosensing for rapid, high-sensitivity detection of biomolecules.To the best of our knowledge, this is the first application of an OFC as a biosensor itself.The integration of photonic-to-RF conversion, an intracavity sensor, and active-dummy dual-comb compensation in the OFC enables detection of the SARS-CoV-2 N protein antigen with an LOD of 38 aM, close to that of RT-PCR, in a measurement time of 10 min, considerably shorter than that of RT-PCR.
The current COVID-19 pandemic may diminish in the future; however, we are always at risk of facing another emerging and re-emerging infectious disease again.As dual-comb biosensing can be used for other viruses through selection of the antigenantibody reaction or other molecular identifications, it will be important as a proactive measure against unknown infectious diseases.Simultaneous achievement of high sensitivity and rapid measurement by dual-comb biosensing will greatly enhance the applicability of biosensors to viruses, biomarkers, environmental hormones, and so on.1a) [29].Furthermore, if the surface of the MMI fibre sensor is modified with a virus antibody, then the RI-dependent frep shift is converted into a virus-antigen-concentration-dependent frep shift through antibody-antigen reactions.In other words, the intracavity MMI fibre sensor with antibody surface modification enables a photonic RF biosensor for viruses.

Dual-comb configuration of active and dummy sensing OFCs
We used a pair of linear-cavity sensing OFCs (frequency spacing = frep1 and frep2, frequency difference between them = ∆frep = frep1 -frep2) for the active sensing OFC and the dummy sensing OFC in the dual-comb configuration (see Fig. 3a).The configuration of each linear cavity was similar to that in Fig. 2a.The output light of the LD pump source was split into two beams and used for these two OFCs, which eliminates the influence of power fluctuations in the LD pump source through commonmode rejection.These OFC fibre cavities were enclosed in an aluminium box with a lid, and its temperature was controlled to 25.0±0.1 °C by a combination of a Peltier heater (Kaito Denshi Inc., Hasuda, Saitama, Japan, TEC1-12708, power = 76 W), a thermistor (YHTC Co., Ltd., Machida, Tokyo, Japan, PB7-42H-K1), and a temperature controller (Thorlabs Inc., Newton, NJ, USA, TED200C, PID control) (not shown in Fig. 3a).We set the frequency spacings of the active and dummy sensing OFCs to approximately 31.7~32.5MHz in RI sensing and 29.6~29.7 MHz in biosensing by adjusting the cavity length; the resulting ∆frep was -851.9 kHz in RI sensing and -88.6 kHz in biosensing.We set the cavity fibres of the dual OFCs in the same arrangement so that they would be equally affected by environmental temperature disturbances [32].
The light output of the dual OFCs was detected by a pair of PDs, and the resulting frequency signals of frep1, frep2 and ∆frep were measured through a combination of an RF frequency counter and a rubidium frequency standard.

Antibody modification of the MMI fibre sensor
A schematic diagram of the intracavity MMI fibre sensor with antibody surface modification is shown in Fig. 1b.First, a UV ozone cleaner (Sun Energy Corp., Minoo, Osaka, Japan, SKB1101N-01) was applied to the MMI fibre sensor for 30 min to remove any organic compounds on the surface of the clad-less MMF and modify the resulting surface with hydroxy groups.Second, the surface of the clad-less MMF was modified with amino-terminated groups through a silane coupling reaction using 1 % (v/v) 3-aminopropyltriethoxysilane (APTS) in ethanol for 1 hour, followed by washing with ultrapure water and drying at 110 °C for 10 min.Third, the monoclonal antibody specific for the N protein antigen was immobilized on the amino-group-coated MMF by a dehydration-condensation reaction using 10 mM 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium chloride (DMTMM) in PBS buffer (pH 7.4).

Data analysis
The frequency spacings of the OFCs (frep, frep1, and frep2) were continuously acquired by an RF frequency counter with a gate time of 100 ms and a sampling interval of 2.8 s.The frequency difference ∆frep between frep1 and frep2 was calculated from acquired frep1 and frep2.Finally, dfrep, dfrep1, dfrep2, and d∆frep were calculated as the frequency deviations from the initial values of frep, frep1, frep2, and ∆frep, respectively.In the dual-comb biosensing of SARS-CoV-2 N protein antigen, we calculated the 99.9 % confidence interval for the first 100 data of the d ∆ frep sensorgram measured in the PBS, and then used it as a criterion of rejection test to judge whether d∆frep value acquired at each molar concentration is considered as a measurement error.was used as a sample for the active and dummy sensing OFCs (see Table 1 and Fig. 3b).dfrep1, dfrep2, and d∆frep were calculated as the frequency deviations from the initial values of frep1, frep2, and ∆frep, respectively.Glycerol solutions consisting of glycerine and pure water at different ratios were used as the target samples in the active sensing OFC; pure water was used as the reference sample in the dummy sensing OFC (see Table 1 and . The functions of steps (1) and (2) are implemented by an intracavity MMI fibre sensor with antibody surface modification, as shown in Fig. 1b.Finally, the change in the antigen concentration can be read out as the frep shift via frep = c/nL, where c is the velocity of light in vacuum and nL is the optical cavity length.The frep linewidth is much smaller than the frep shift expected due to antigen concentration changes.

Figure
Figure 5b shows a sensorgram of d∆frep calculated by subtracting the green

Figure 1b shows a
Figure1bshows a schematic diagram of the intracavity MMI fibre sensor with

Figure 1 .
Figure 1.Principle of operation for the biosensing OFC.a, Block diagram of the

Figure 2 .
Figure 2. Basic performance of single-comb RI sensing of pure water with

Figure 3 .
Figure 3. Experimental setup of dual-comb RI sensing and biosensing.a,

Figure 4 .
Figure 4. Basic performance of dual-comb RI sensing of pure water with

Figure 5 .
Figure 5. Temperature-drift-free dual-comb RI sensing of glycerol solution.a, Fig. 3c).Grey zones indicate the time period for sample exchange by peristaltic pumps.dfrep1 and dfrep2 were calculated as the frequency deviations from the initial values of frep1 and frep2, respectively.b, Sensorgram of d∆frep with mixtures of glycerine and pure water at different ratios.d∆frep was calculated as the frequency deviation from the initial value of ∆frep.The mean and the standard deviation of d∆frep were 0.76±0.19Hz at 0 vol%, -7.58±0.24Hz at 1 vol%, -16.48±0.52Hz at 2 vol%, -25.64±0.53Hz at 3 vol%, -34.59±0.31Hz at 4 vol%, and -43.12±0.34Hz at 5 vol%.c, Relationship between the sample RI and d∆frep.A linear