Development and Evaluation of a Multiplexed Immunoassay for Simultaneous Detection of Serum IgG Antibodies to Six Human Coronaviruses

Known human coronaviruses (hCoV) usually cause mild to moderate upper-respiratory tract illnesses, except SARS-CoV and MERS-CoV, which, in addition to mild illness can also be associated with severe respiratory diseases and high mortality rates. Well-characterized multiplexed serologic assays are needed to aid in rapid detection and surveillance of hCoVs. The present study describes development and evaluation of a multiplexed magnetic microsphere immunoassay (MMIA) to simultaneously detect immunoglobulin G (IgG) antibodies specific for recombinant nucleocapsid proteins (recN) from hCoVs 229E, NL63, OC43, HKU1, SARS-CoV, and MERS-CoV. We used paired human sera to screen for IgG with reactivity against six hCoVs to determine assay sensitivity, specificity and reproducibility. We found no signal interference between monoplex and multiplex assay formats (R2 range = 0.87–0.97). Screening of paired human sera using MMIA, resulted in 92 of 106 (sensitivity: 86%) as positive and 68 of 80 (specificity: 84%) as negative. This study serves as a proof of concept that it is feasible to develop and use a multiplexed microsphere immunoassay as a next generation screening tool for use in large scale seroprevalence studies of hCoVs.

coupled with molecular detection assays is an effective approach in population based surveillance studies 10 .
We have developed and validated hCoV recombinant nucleocapsid protein (recN) indirect ELISAs that individually detect immunoglobulin G (IgG) antibodies to all six hCoVs. The hCoV nucleocapsid (N) protein is abundantly expressed during infection and is, therefore, one of the ideal candidate antigens for the development of microtiter plate based ELISAs 11 . However, traditional plate ELISAs are limited to detecting IgG antibodies to a single hCoV recN antigen per well. In contrast, a magnetic bead-based multiplexed immunoassay (MMIA) would allow for the simultaneous detection of IgG antibodies to all six hCoV recN in a single test well, reducing sample consumption. In addition, multiplexing would allow higher sample throughput at lower reagent, labor, and material costs, without losing sensitivity and specificity of traditional ELISAs. The present work describes the development and validation of a MMIA to detect human serum IgG against six hCoV recN as a next generation screening tool in population-based seroprevalence studies.

Results
titration of Antigen Concentration. A monoplex assay for each hCoV recN conjugated bead set was developed and standardized to determine optimum antigen quantity per bead coupling reaction, optimum serum dilution and assay cut-off mean fluorescent intensity (MFI) values. Each bead set was conjugated with 1.0 µg, 2.5 µg, or 5.0 µg of antigen per 100 µl of conjugation reaction (1.25 × 10 6 beads). A checkerboard titration assay was carried out for each antigen against a series of 2-fold (1:100-1:800) positive and negative control sera. The optimum working antigen amounts were determined to be 1.0 µg for SARS-CoV recN, His-BoNT, NL63 recN; 2.5 µg for 229E recN, MERS-CoV recN, OC43 recN3; pET E. coli, and 5.0 µg for HKU1 recN per conjugation reaction (1.25 × 10 6 beads) (Fig. 1). Based on positive serum control titration results, optimum dilutions for each antigen were determined to be 1:400 for OC43, NL63, HKU1 and SARS-CoV; and 1:800 for 229E, and MERS-CoV (Fig. 1). The optimum antigen quantities and positive control serum dilutions were selected based highest signal to background ratio, most reproducible results and economical usage of antigen stocks.
Correlation between monoplex and multiplex assay. We compared the MFI values generated by monoplex assays with multiplex assay for each antigen to investigate possible loss of sensitivity due to drop in MFI values. Positive control serum pools for each hCoV were serially diluted in 2-fold dilutions (1:100-1:51200) and screened using monoplex and multiplex microsphere immunoassays. In each assay format, 200 beads (100 beads in duplicate) were counted for each recN conjugated bead set. Linear regression analysis was performed to determine correlation coefficient for each recN antigen. For all recN bead sets, statistically significant correlations were observed between the monoplex and multiplex formats with a correlation coefficient (R 2 ) ranging from 0.88 to 0.97 (Fig. 2).

Sensitivity, Specificity and cut-off MFI values.
To determine the sensitivity, specificity and cut-off MFI values for each antigen, receiver operating characteristic (ROC) analysis was carried out based on screening of human sera collected from individuals whose respiratory specimens were tested by reverse real time polymerase chain reaction (rRT-PCR) for presence or absence of hCoV infections. Acute and convalescent phase sera from individuals tested positive for hCoVs 229E, NL63, OC43, and HKU1 infections as well as convalescent phase sera, from confirmed positive infections of MERS-CoV and SARS-CoV were screened using MMIA. Additionally, paired sera, from individuals tested negative for hCoV infection were also screened using MMIA. Percentage sensitivity (%Se) and specificity (%Sp) for each recN was estimated over a wide range of cut-off MFI values and plotted as Two Graph -Receiver Operating Characteristic curve (TG-ROC) using SPSS software (Data not shown). The MFI value at which the paired values of (%Se) and (%Sp) were optimal (intersection of %Se and %Sp graph lines) were selected as the cut-off MFI for respective recN antigen (data not shown). The cut-off MFI values of 1333, 513, 228, 1171, 1333 and 520 were selected for 229E, NL63, OC43, HKU1, SARS-CoV and MERS-CoV recN monoplex assays, respectively. The corresponding sensitivity and specificity values are described in Table 1.

Reproducibility of Multiplex Microsphere Immunoassay.
To assess the reproducibility of the MMIA, a panel of positive control serum samples for each hCoV was run in duplicate at optimum serum dilutions on two different plates on the same day and was repeated with a new batch of conjugated beads on a different day. Coefficients of variation (CV) for each replicate for each antigen was calculated and averaged ( Table 2). All assays except NL63 recN gave CV values below 10% indicating MMIA precision.
Cross-reactivity between recN antigens. To examine cross reactivities between all hCoV recN in human sera, we screened each positive control of hCoV in a multiplex format at dilutions ranging 1:100-1:800. The results show overall minimum cross reactivity amongst hCoV recN (Fig. 4). However, positive controls of the hCoVs of the same antigenic groups showed cross reactivity with each other. Positive controls of group 1 (alpha) human coronaviruses, 229E and NL63, showed reactivity with each other but not with group 2 hCoVs. However, SARS-CoV (group 2b) positive control serum reacted with group 1 CoVs, 229E and NL63. Positive control serum for OC43, a group 2a (beta) coronavirus, reacted with HKU1 recN, another member of group 2a whereas positive control for MERS-CoV (group 2c) reacted with HKU1 recN.

Discussion
Use of bead-based multiplexed immunoassay technology has been gaining recognition as a next generation screening tool in seroprevalence studies. Several research groups have reported the development of bead-based multiplexed immunoassays to detect antibodies against a variety of human viral pathogens. A duplex assay was developed to detect IgG against recombinant nucleocapsid proteins of human respiratory syncytial virus (hRSV) and human metapneumovirus (hMPV) 10 . A quadruplex assay was developed to detect IgGs against measles, mumps, rubella and varicella-zoster virus 12 . A thirteen-plexed immunoassay has been developed to detect IgM and IgGs for serodetection of arboviruses 13 . The results of the current study shows that it is feasible to use a multiplexed bead immunoassay to screen human sera for IgGs against the six human coronavirus recombinant nucleocapsid proteins. The use of ROC analysis in our study not only facilitated selection of optimal MFI cut-off values, but also helped determine the accuracy and usefulness of our test method by calculating the area under the ROC curves 14 .
A perfect test has an AUC of 1 and a useless test has an area of 0.5, the equivalent of a coin toss. Based on arbitrary criteria, the following guidelines have been suggested for interpretation of AUC values in context of test accuracy: low (0.5 < AUC ≤ 0.7), moderate (0.7 < AUC ≤ 0.9), and high (0.9 < AUC ≤ 1) 15 . Based on our assay test results, the calculated AUC ranged from 0.87 to 0.94, suggesting a high to high-moderate levels of accuracy and overall assay performance. The cross reactivity experiment results showed positive control of SARS-CoV reacting with group 1 hCoVs 229E and NL63 recN. A major limitation of these assays is that we are unable to distinguish the difference between cross reactivity and evidence of previous infections. SARS-CoV recN has been shown to react with polyclonal antisera of group 1 coronaviruses in a western blot analysis 16 . However, this cross reactivity was one-way from SARS-CoV to other hCoV recN in this study which has also been previously confirmed by Che et al., in a separate study 17 . The multiple sequence identity matrix of all hCoV nucleocapsid protein sequences show amino acid homology ranging from 20% to 63% (data not shown). In contrast, group 1 hCoVs showed only 20% to 22% sequence homology with group 2 hCoVs which explains the lack of detectable cross-reactivity between  these two antigenic groups. However, sequence homology was high between nucleocapsid sequences of OC43 and HKU1 (%) and that of HKU1 and MERS-CoV (%). Higher percentage of nucleocapsid protein sequence homology may indicate the shared epitopes in the conserved and/or overlapping regions between hCoVs of the same antigenic groups. Since the previous hCoV infection history of patients enrolled in the study was unavailable, dual infection with OC43 and HKU1 could also have contributed to the cross-reactivity, though that is unlikely for HKU1 and MERS-CoV. Our hCoV MMIA offers high throughput testing at reduced cost; the assay can be completed in less time with less labor using less clinical specimen and assay reagents. Moreover, MMIA allows the flexibility of mixing and matching the bead sets depending upon the analysis requirements. Overall, the results of our study showed that development of a bead-based multiplex immunoassay for screening human sera for IgG antibodies against hCoV recN proteins is feasible and may serve as an effective tool for large scale seroprevalence studies such as examining prevalence of hCoVs in young children of different ages.

Materials and Methods
Human serum samples. All methods were carried out in accordance with CDC's institutional review board (IRB) guidelines and regulations. All protocols were approved by CDC's IRB. All patients were hospitalized with pneumonia symptoms, gave informed consent, and their respiratory specimens were tested by rRT-PCR to determine etiology. Paired acute and convalescent phase human sera collected from individuals whose respiratory specimens were tested positive for hCoV-229E (n = 4), hCoV-NL63 (N = 9), hCoV-OC43 (n = 21), and hCoV-HKU1 (n = 14), by rRT-PCR, were included in a positive control panel for the study. Human sera from patients with confirmed SARS-CoV (n = 5) and MERS-CoV infections (n = 7), tested during outbreak investigation by our laboratory, were also included to the positive control panel. Acute sera were collected upon hospital admission, and convalescent phase sera were obtained 3-7 weeks later. From this panel of positive sera, convalescent phase sera exhibiting ≥ 4 fold increase in IgG titers, were pooled together to create a positive control for each hCoV recN. Similarly, acute and convalescent phase human sera, collected from individuals with acute respiratory illness whose respiratory specimens tested negative for all six hCoVs by real-time RT-PCR and exhibited no IgG reactivity when screened by in-house hCoV recN enzyme immunoassays, were included to create a negative control panel.

Expression and purification of recombinant nucleocapsid protein antigens. For this study, we chose
to use recombinant coronavirus nucleocapsid (N) proteins. N antigens The full length SARS-CoV recN and a negative control antigen, the nontoxic 50-kDa C-terminal fragment of the botulinum neurotoxin serotype A (BoNt/HcA), were expressed and purified as described previously 18 . Expression and purification of full length hCoV-229E recN, hCoV-NL63 recN, hCoV-HKU1 recN, MERS-CoV recN and purified protein from E. coli containing pET-28 vector without the N gene, as a negative antigen control, were carried out as described previously for hCoV-OC43 recN3 19 .  control wells (n) were subtracted from and divided by the OD values of antigen-coated wells (p) and the average OD values of the antigen-coated wells were calculated as the difference (p-n) and ratio (p/n).
All washing steps were performed using a magnetic separator (Luminex ® ). The beads were vortexed vigorously recN Multiplex Immunoassay. All recN conjugated beads were brought to room temperature prior to performing the multiplex immunoassay. Beads were vortexed for 30 sec and then sonicated in a water bath sonicator (Branson ultrasonic cleaner, VWR International, West Chester, PA) for 20 sec. To prepare a working dilution, each bead set was diluted in PBS (pH 7.4) to a final concentration of 50 beads/region/µl of PBS (pH7.4). All diluted bead sets were then mixed into a single 15 ml polypropylene centrifuge tube in equal quantities. The 15 ml centrifuge tube was vortexed and sonicated as described above to create homogeneous mixture of all conjugated bead sets. All the test sera were diluted to an optimal dilution and tested in duplicate. To perform multiplexed bead immunoassay, 50 µl of homogeneous mixture of conjugated beads, 50 µl of diluted test serum sample and 50 µl of PBS (pH 7.4), were added to each of the well of the plate. The rest of the assay was carried out using the same steps as described above in the monoplex assay, except the results were expressed as MFI of 100 beads/region/well. Assay Sensitivity and Specificity. A panel of positive paired sera and negative paired sera from individuals with confirmed hCoV infection, was tested using multiplexed microsphere immunoassay (MMIA). Using rRT-PCR as the reference test, %sensitivity (%Se) and %specificity (%Sp) of the MMIA were calculated and plotted as Two Graph Receiver Operating Characteristic (TG-ROC) analysis plots. The intersections, of %Se and %Sp plots, were selected as the cut-off MFI for each recN. The results were also plotted as ROC curve and the Area under the Curve (AUC) was calculated for each recN to determine overall assay performance.
Assay Reproducibility and correlation coefficient. A series of 2-fold dilutions (1:100-1:12800), of positive and negative controls of all six hCoVs, was prepared in SurModics ® Assay Diluent (cat#SM01-1000; SurModics, MN) and tested using monoplex and multiplex immunoassay formats to determine correlation between two formats. The same panel was used in multiplex format for determining the cross-reactivity between hCoV recN beads as well as within and between assay variations. Each sample of the panel was run in triplicate on two different plates. The trial was repeated after seven days using another set of conjugated beads.

Data Availability
Raw data will be made available upon request. Authors also agree to share reagents used in this manuscript.