Beads- and oil-free single molecule assay with immuno-rolling circle amplification for detection of SARS-CoV-2 from saliva

Digital enzyme linked immunosorbent assays (ELISA) can be used to detect various antigens such as spike (S) or nucleocapsid (N) proteins of SARS-CoV-2, with much higher sensitivity compared to that achievable using conventional antigen tests. However, the use of microbeads and oil for compartmentalization in these assays limits their user-friendliness and causes loss of assay information due to the loss of beads during the process. To improve the sensitivity of antigen test, here, we developed an oil- and bead-free single molecule counting assay, with rolling circle amplification (RCA) on a substrate. With RCA, the signal is localized at the captured region of an antigen, and the signal from a single antigen molecule can be visualized using the same immune-reaction procedures as in the conventional ELISA. Substrate-based single molecule assay was theoretically evaluated for kd value, and the concentration of capture and detection antibodies. As a feasibility test, biotin-conjugated primer and mouse IgG conjugates were detected even at femto-molar concentrations with this digital immuno-RCA. Using this method, we detected the N protein of SARS-CoV-2 with a limit of detection less than 1 pg/mL more than 100-fold improvement compared to the detection using conventional ELISA. Furthermore, testing of saliva samples from COVID-19 patients and healthy controls (n = 50) indicated the applicability of the proposed method for detection of SARS-CoV-2 with 99.5% specificity and 90.9% sensitivity.


Introduction
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) have been pandemic since 2020, necessitating the urgent development of vaccines and antiviral drugs as well as methods for early detection of SARS-CoV-2 to prevent the spread of disease (Kevadiya et al., 2021;Krammer, 2020;Wu et al., 2020b). To date, reverse transcription polymerase chain reaction (RT-PCR)-based molecular diagnosis has been the gold standard for the detection of SARS-CoV-2 infection (Bwire et al., 2021;Falzone et al., 2020). Despite of its high accuracy and sensitivity, RT-PCR-based methods suffer from complicated sample preparation procedures to extract and purify viral RNA from human nasopharyngeal swab. Detection of spike (S) or nucleocapsid (N) protein of SARS-CoV-2 could provide an alternative method for rapid screening of SARS-CoV-2 patients. Owing to its simplicity, such antigen test can be widely applicable in point-of-care testing systems, such as lateral flow assays. These tests are more suitable for early diagnosis of SARS-CoV-2 before symptom onset from a variety of specimens, such as saliva, nasonasopharyngeal and, anal swab, blood, urine, and tears (Huang et al., 2021;Kaya et al., 2020;Peng et al., 2020). The simplicity of lateral flow assay based antigen tests makes it widely used for the screening of COVID-19 patients, but RT-PCR is still used as gold standard for the accurate diagnosis of COVID-19; however, because the sensitivity is much poorer than RT-PCR, it can lead to a high ratio of false-negative results (Cerutti et al., 2020;Peto et al., 2021). For increased sensitivity, various types of antigen tests, such as enzyme linked immunosorbent assay (ELISA) (Hingrat et al., 2021;Liu et al., 2022), electrical (Kim et al., 2022;Zhao et al., 2022), electrochemical (Samper et al., 2022;Zeng et al., 2022), plasmonic (Ahmadivand et al., 2021;Behrouzi and Lin, 2022), and proximity assay and CRISPR based (Liu et al., , 2021 sensors, have been used; however, the performance of these tests still lags behind that of RT-PCR considerably.
Digital ELISA is an improvisation of conventional immunoassay, with better sensitivity, in which the number of positive signals, indicating a single target molecule and its immune-complex, are counted with high signal to noise ratio (Chang et al., 2012;Rissin et al., 2010). The accuracy of the immunoassay can be improved to as low as single-molecule level. Capture antibody conjugated beads sequentially capture the antigen and detection antibody. Each bead is loaded into a micro-well and compartmentalized with oil. Because signal from a single molecule is continuously generated upon enzyme reaction and is diffused out, compartmentalization is necessary to localize the positive signal. However, during compartmentalization, a large number of beads is not captured in microwells and only <10% of beads used in the assay are actually analyzed. The loss of beads during compartmentalization can lower the sensitivity of assay because single molecule-conjugated beads can also be lost. To mitigate this loss and thereby increase the number of beads that is analyzed, droplet microfluidics has been incorporated to compartmentalize all the beads used in the assay through water-in-oil interfaces (Cohen et al., 2020;Liu et al., 2018;Yelleswarapu et al., 2019). Increasing the number of analyzed beads enhances the accuracy of the assay; however, because of the Poisson's noise, 10 times more droplets than the number of beads should be generated and analyzed, which makes the analytical procedure longer and tedious. The need for complicated compartmentalization procedure can be eliminated if the signal is localized near the captured region of an antigen and does not diffuse out. For this purpose, rolling circle amplification (RCA) and tyramide signal amplification (TSA) have been applied wherein fluorescence signal is localized on the surface of beads (Akama et al., 2016;Maley et al., 2020;Wu et al., 2020aWu et al., , 2022. Using RCA or TSA, the performance vis-à-vis dELISA, can also be improved because there is no loss of beads during compartmentalization and more beads can be analyzed. Without the need for compartmentalization, a single molecule can be successfully analyzed with improved performance. However, the requirement of beads in these methods necessitating the use of bulky instruments, such as flow cytometer, or a hard fixation procedure on substrate, for fluorescence analysis or imaging, makes them non user-friendly. In this study, we developed an oil-and bead-free single molecule counting assay with digital immuno-RCA (Fig. 1). By using capture antibodies coated glass slide instead of beads, the RCA signal can be localized on the glass slide where immuno-complexes of the target antigen are formed. For digital immuno-RCA, the detection antibody was modified with a primer and circular DNA and the immuno-reaction was performed following the same steps as in a conventional ELISA. After the immuno-and RCA reaction, the localized signal from single molecules can be imaged and the number of signals can be counted. Single molecule assay was also studied theoretically based on the antigen-antibody kinetics model. The capture efficiency of molecules as well as the ideal limit of detection was calculated depending on the concentration and k d value of the capture antibody and the concentration of the detection antibody. For characterization of digital immuno-RCA before moving to the sandwich format assay, biotin-primer as well as mouse IgG conjugates were detected at various concentrations. Finally, the N protein of SARS-CoV-2 was detected with digital immuno-RCA and the results were compared to that obtained with the conventional ELISA. Saliva samples from COVID-19 patients were tested to verify the developed assay platform.

Circularization of DNA template
A single-stranded DNA (ssDNA) template was circularized using a ssDNA circularization kit. Briefly, 50 μL of 1 pmol/μL ssDNA template was mixed with 10 μL of circligase 10X reaction buffer, 5 μL of 1 mM ATP, 5 μL of 50 mM MnCl 2 , 5 μL of ssDNA ligase, and 25 μL of distilled water. The mixture was incubated at 60 • C for 1 h and subsequently at 80 • C for 10 min to inactivate the ligase. Thereafter, 1 μL of exonuclease I was added to remove the non-circularized ssDNA template. The mixture was incubated at 37 • C for 1.5 h and deactivated at 85 • C for 15 min. Finally, 500 nM of the circular DNA template was obtained.

Conjugation of DNA primer to antibodies
DNA-antibody conjugates were obtained using DBCO-Azide copperfree click chemistry (Fig. 1B). To 48 μL of 1 mg/mL detection antibody solution, 2.6 μL of 4 mM DBCO-NHS ester in DMSO was added, and the mixture was incubated on a shaker for 30 min at room temperature.
Thereafter, 1.5 μL of Tris-HCl was added to the mixture to remove the free end of NHS-ester. The excess amount of DBCO-NHS ester was removed using a 7k Zeba spin column kit. For this process, the spin column kit was washed three times with PBS (1500×g, 1 min) and the mixture was cleaned to get purified DBCO-modified antibodies (1500×g, 2 min). Next, 4.8 μL of 100 μM azide-modified DNA primer was added to the purified solution and the mixture was incubated overnight at 4 • C. About 0.7 mg/mL of primer-conjugated antibodies was obtained.

Digital immuno-RCA procedures
The capture antibody-coated glass slide was incubated with the antigen solution for 2 h and then washed three times with PBST. The slide was further incubated with 0.01 μg/mL detection antibody conjugates in dilution buffer for 1 h and washed three times with 0.1% TBST. The detection antibody conjugates were prepared by incubating 2 μg/mL of detection antibody-primer conjugates and 20 nM circularized DNA template in TET buffer with 1 M NaCl for 1 h. After washing out of the detection antibody conjugates, the slide was incubated with the RCA mixture (0.25 mM dNTPs, 200 μg/mL BSA, and 400 U/mL phi29 polymerase in 1x RCA reaction buffer) for 2 h at 37 • C. After washing off the RCA mixture three times with TBST, the slides were incubated with 100 nM of biotin-conjugated primers 1 and 2 in TBST for 1 h at 37 • C. After further washing three times with TBST, the slide was incubated with 0.01 mg/mL SA-PE in TBST for 30 min at room temperature. The glass slide was ready for imaging after washing three times each with TBST and distilled water. For the patient sample test, collected saliva samples were mixed with PBS at 1:1 ratio. The mixture was then centrifuged at 10,000×g for 5 min, and the supernatant was mixed with lysis buffer (100 mM Tris-HCL, 800 mM NaCl, 1% BSA pH 9.0, 1% Triton X-100) at 1:1 ratio for the digital immuno-RCA. The primers and probes used for RCA are listed in Table S1.

Theoretical analysis of the single molecule assay
The single molecule counting assay was analyzed theoretically based on the equilibrium aspects between the capture and binding molecules. The capture molecules on the glass substrate (Ab) and binding molecules (Ag) in solution form complexes (AbAg) and their kinetics can be predicted with the following equation (Chang et al., 2012). Based on the above equation, the efficiency of the capture of binding molecules by capture molecules and the limit of detection can be calculated depending on the k d value and concentration of the capture and binding molecules. The assay was performed in a PDMS well attached over the glass slide, and 10 μL solution was used for each process. The diameter of each well was 5 mm and the concentration of total capture antibodies ([Ab total ]) was calculated to be 14.4 nM, assuming the capture antibodies to be densely loaded onto the well occupying a square of 15 nm each. Additionally, we initially performed theoretical analysis assuming the reaction efficiency between the capture and binding molecules to be 100%. The reaction efficiency can be speculated in comparison with the experimental results. First, we analyzed the efficiency of capture of the antigen by the capture antibodies, depending on the k d value of the capture antibody with respect to increasing concentration of binding molecules ( Fig. 2A). We analyzed the k d value from 10 pM to 10 nM-the typical range of antibodies-and 1 fM for the streptavidin-biotin interaction. Almost 100% of the capture efficiency was calculated at low value of k d , but the capture efficiency decreased as the k d value increased. The capture efficiency did not vary with the concentration, except at a high concentration of antigen similar to that of the capture antibody. The limit of linearity (LoL) was also analyzed, which was determined at the concentration of antigen at which the capture efficiency started decreasing (indicated by arrows). A lower k d value showed higher LoL, which is beneficial for quantification over a wide range of antigen concentration. The limit of detection (LoD) was analyzed theoretically by calculating the number of captured binding molecules per imaging frame (Fig. 2B). The number of captured binding molecules should be greater than 1 to be imaged; the concentration of LoD was approximately 100 atto-molar. As expected, the LoD decreased as the k d value of capture antibodies decreased (Fig. 2C). The same analysis was performed for different concentrations of the capture molecules (14.4 nM to 14.4 pM) in the assay area ( Fig. 2D-F). The capture efficiency decreased dramatically as the concentration of the antigen, rather than the k d value, decreased, especially at higher concentrations of the binding molecules. The LoL also decreased as the concentration of capture antibodies decreased. The LoD was calculated to be 10-100 atto-molar (reaction efficiency = 100%), and it decreased with the increase in the concentration of capture molecules. Finally, the labeling efficiency and LoD were analyzed at different concentrations of detection antibodies by assuming the situation as in a sandwich immunoassay ( Fig. 2G-I). For a sandwich assay, immuno-complexes of capture antibodies and antigens should be labeled with detection antibodies to be imaged. This analysis can be done by applying the same principle as in the capture and binding molecule kinetics that immuno-complex of capture antibody and antigen would be the capture molecules and detection antibodies would be the binding molecules. The most drastic change was observed in the labeling efficiency as the concentration of detection antibodies decreased rather than in the above two cases, but LoL did not change as much as the labeling efficiency. The LoD changed with respect to the decreasing concentration of detection antibodies even though the same number of antigens were captured by the capture antibodies. At least 1 nM of detection antibodies were required to label about 100% of the captured antigens.

Characterization of digital count of RCA products
To characterize digital immuno-RCA, we first checked the digital count of DNA molecules without any antibodies. We performed the digital count of biotin-conjugated DNA primer using the streptavidincoated glass slide (Fig. 3A). After incubation with various amounts of biotin-conjugated DNA primer, a constant amount of circularized DNA template (50 nM) was loaded for RCA reaction (2 h). Thereafter, RCA products were labeled with biotin-conjugated detector probes and SA-PE sequentially, and imaged. The number of counted points was increased with the increasing amount of DNA primer (Fig. 3B and Fig. S1). Although many non-specific signals were also observed, the signal from 1 fM of DNA primer could be distinguished from the negative signal. In comparison with the theoretical LoD (tens to hundreds of atto-molar), the efficiency of reaction between the capture and binding molecules was supposed to be between 1 and 10%. The RCA reaction was then performed in conjugation with antibodies. The DNA primer was conjugated to mouse IgG and its concentration was detected using anti-mouse IgG antibodies immobilized on the glass slide (Fig. 3C). Mouse IgG-DNA primer conjugates and circle template were sequentially loaded and the RCA reaction was performed. However, contrary to the above results, considerably larger non-specific signal was observed and the signal from 600 fM mouse IgG was distinguishable from the negative signal (Fig. S2). This could be attributable to the nonspecific binding of the circle template to the anti-mouse IgG antibody-coated glass slide. To address this problem, we preincubated the circle template with mouse IgG-primer conjugates with at ~1:1 concentration ratio and then assessed different amounts of mouse IgG conjugates with RCA (Fig. 3D). The number of nonspecific signals was dramatically reduced and the signal of 60 fM mouse IgG conjugates was distinguishable from the negative signals. Compared with the detection of biotin-conjugated DNA primer that relies on streptavidin-biotin interaction having a lower k d value, interaction with antibodies, which has a higher k d value, may be attributable to the decrease in the number of positive signals.

Detection of nucleocapsid protein of SARS-CoV-2 using digital immuno-RCA
Based on the results of the characterization of digital count of RCA products, we performed a sandwich assay to detect the SARS-CoV-2 N protein. The N protein was chosen as the target analyte because it is much less mutagenic compared to other SARS-CoV-2 antigens (Thakur et al., 2022). First, 0.1 μg/mL (~1 nM) of detection antibody conjugates were used because this concentration offers almost 100% capture efficiency of detection antibodies by immuno-complexes based on the theoretical analysis. However, because of the nonspecific signals, we used 0.01 μg/mL of detection antibody conjugates. The signal from immuno-RCA was counted at increasing concentrations of the N protein (Fig. 4A,B). The number of bright points increased with the increasing antigen concentration. The signal was distinguishable in the images at concentrations as low as 1 pg/mL and the LoD was less than 1 pg/mL. The blank + 3σ value was close to zero; therefore, the 4 PL fitting of the standard curve could not cover the value as low as that of blank + 3σ. Compared with the LoD of ELISA (114 pg/mL), the LOD was improved more than 100-fold. The performance of the assay was comparable when the lysis buffer was spiked with the antigen (Fig. S3). The LoD of the assay when the antigen was spiked in lysis buffer was 0.16 pg/mL. The signal to noise ratio of digital immuno-RCA is higher than ELISA that attribute to higher sensitivity of the assay (Fig. 4C). The reason for higher signal to noise ratio of digital immuno-RCA is that the localized signal from each immunocomplex is visualized whereas the signal from immunocomplexes diffuse out to total assay solution in ELISA. In addition to ELISA, many SARS-CoV-2 N protein detection methods (Liu et al., , 2021Samper et al., 2022;Zeng et al., 2022) have been reported by amplifying the signal from immunocomplexes, but they analyze the average signal from the total assay unit so that it is hard to get high signal to noise ratio at low antigen concentration. In this manner, single molecule assay based on digital immuno-RCA is beneficial for highly sensitive SARS-CoV-2 N protein detection. The sensitivity of assay can be improved as much as that shown in our theoretical study when we could reduce background signal around zero that arise from nonspecific binding of detection antibody conjugates. The performance of digital immuno-RCA was also confirmed with SARS-CoV-2 S protein that showed 172 times LoD improvement compared to the conventional ELISA (Fig. S4). Additionally, the specificity of the developed assay was validated with 1 ng/mL of SARS-CoV-2 S protein, Amyloid beta, p-Tau, and TnI (Fig. 4D). Furthermore, the product of immuno-RCA on a glass slide was analyzed using a scanning electron microscope (SEM), and around 0.2 μm of DNA bundle was visualized (Fig. S5).

COVID-19 patient sample test
To validate the developed digital immuno-RCA assay for the detection of the SARS-CoV-2 N protein, we tested 30 positive and 20 negative saliva samples. The positive samples showed a higher count of the RCA product whereas the negative samples showed a lower count (Fig. 5A,  Table 1, Fig. S6). Some positive samples showed a very high number of RCA product that saturated the image frame. Based on the standard curve for the N protein spiked in the lysis buffer, the number of RCA product counts per frame was converted to the concentration (Fig. 5B). The samples that showed saturated images were treated to the upper limit of the assay in the calibration curve (1,000 pg/mL), and samples with the RCA product below the LoD were treated to be the LoD of the calibration curve (0.16 pg/mL). The sensitivity and specificity were 90.9% and 99.5%, respectively, when the cut-off was set to 8.2 pg/mL. The measured concentration of the N protein was also compared to the Ct value of RT-PCR, and correlated with the Pearson's r value of − 0.73 (Fig. 5C). At current stage, there has been various types of mutation in SARS-CoV-2 antigens and it is progressing. So, we used a detection antibody that can bind multiple variants (D3L, P13L, P80R, S197L, S202N, R203K, G204R, T205I, L230F, S235F, Q384H etc.) with similar affinity as compared to the wild type N protein (information provided by Acrobiosystems). The use of universal antibodies targeting highly conserved regions of the N protein can improve the sensitivity and specificity of the assay.

Conclusions
In summary, an oil-and beads-free single molecule assay with digital immuno-RCA was developed, and it was applied for assaying the N protein of SARS-CoV-2. The single molecule assay was also theoretically analyzed; digital immuno-RCA was characterized with biotinconjugated primer as well as mouse IgG conjugates and could detect the molecules even at a few femto-molar concentration. The RCA signal was localized on a glass slide and could be easily imaged and analyzed for highly sensitive detection of antigens using a procedure that is almost the same as in the conventional ELISA; the assay did not require compartmentalization procedures using micropatterned surface, microbeads, or oil. The N protein of SARS-CoV-2 was detected using the digital immuno-RCA, with a limit of the detection less than 1 pg/mL. Notably, the LoD was improved more than 100-fold compared with that achieved using the conventional ELISA although the assay procedures are almost the same. The results of clinical studies highlight the potential of the proposed assay as a highly sensitive antigen test that can be used to screen COVID-19 patients in whom the Ct value is lower than 25. The LoD should be further improved to achieve a performance comparable to that of RT-PCR. Nonetheless, we expect that the proposed method should be widely applicable for highly sensitive detection of biomolecules.