Development and Validation of an Enzyme Immunoassay for Detection and Quantification of SARS-CoV-2 Salivary IgA and IgG

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of the coronavirus disease 2019 (COVID-19) pandemic, is a betacoronavirus related to SARS and Middle East respiratory syndrome coronavirus (MERS-CoV) (1–3). As of July 30, 2021, SARS-CoV-2 infections had caused more than 200 million cases worldwide and an estimated 4.2 million deaths. The clinical spectrum of SARS-CoV-2 infection ranges from asymptomatic infection to symptomatic disease (4). The high proportion of asymptomatic individuals not only results in a high transmission rate but also suggests differences in the host immune response compared with other coronaviruses (5). Because the duration of immunity to SARS-CoV-2 dictates the overall course of the pandemic as well as postpandemic strategies, a comprehensive understanding of the relationship between systemic and mucosal Ab responses becomes important. Because the oral and nasal cavities are considered the main sites for SARS-CoV-2 entry and replication, locally produced mucosal Abs may protect against infection. Therefore, saliva samples can be used as a noninvasive tool for virus detection as well as for measuring the immune response (mucosal and systemic) (6).

Salivary Ab levels can be 100- to 1000-fold lower than serum levels (7). Salivary IgG is mainly derived from serum by leakage across capillaries and enters saliva through gingival crevices. At mucosal membranes, IgA is the main Ig class and is found most often in the secretory form. Within 2–3 wk after onset of disease, SARS-CoV-2–specific IgG Abs can be detected in saliva, persist for at least 9 mo, and show high correlation with serum Ab levels in most patients with COVID-19 (8–11). Salivary IgA Abs, on the contrary, rapidly increase 1 wk after onset of disease, become undetectable 4–5 wk later, and show a moderate correlation with serum levels (8, 9). Saliva provides a noninvasive collection method that is easy to implement in remote areas and community settings without a need for extensive training. These features, while additionally evaluating both mucosal and systemic immune responses, make salivary Ab testing an ideal approach to evaluate population immunity, transmission, asymptomatic infections, and vaccine performance.

We previously demonstrated the value of saliva-based Ab assays to evaluate immune responses mounted against norovirus (12). In this article, we describe the development and validation of an enzyme immunoassay (EIA) to quantitatively evaluate the presence of SARS-CoV-2–specific IgA and IgG Abs in saliva and to describe the salivary immune response to SARS-CoV-2 mounted in different cohorts of infected patients.

A total of 333 saliva samples were collected from 187 participants who had positive test results for SARS-CoV-2 by real-time RT-PCR (rRT-PCR) (13) or Ag test in two cohorts (Fig. 1). In cohort I, 235 samples were collected from (1) 113 participants at a single time point and (2) 32 participants on a weekly basis for 4–5 wk (n = 122) after diagnosis. In cohort II, 98 saliva samples were collected from 42 participants with either asymptomatic (n = 8), mild (n = 29), or severe disease (n = 5) at different times after the onset of disease (range, 0–203 d). Disease severity was defined according to the World Health Organization criteria (4), and clinical data were obtained using a standardized questionnaire. In addition, 373 prepandemic archived samples collected between 2009 and 2010 were included as negative controls (12).

Saliva was collected at least 30 min after consumption of food or liquids. Prepandemic archived saliva samples were collected using the Oracol saliva collection device, processed, and stored at −80°C according to the manufacturer’s instructions (Malvern Medical Developments, Worcester, UK). Cohort I samples were collected using the Oracol S14 collection device by gently rubbing the swab along the gumline around the entire mouth for ∼1 min. This collection device specifically harvests gingival crevicular fluid, which resembles serum composition (14). Saliva samples collected by the Oracol swabs were separated by centrifugation (10 min at 1500 × g), transferred to the attached microtube (10–200 μl), and stored at −80°C until analysis. For cohort II, participants were asked to cough deeply and spit into a collection cup containing virus isolation media (PBS plus 2% FBS, gentamicin, amphotericin B). Saliva samples were clarified by centrifugation (10 min at 3000 × g), aliquoted, and stored at −80°C until analysis. All samples collected during the pandemic were inactivated by γ-irradiation (2 × 10⁶ rad) before testing (15). Samples were initially tested for SARS-CoV-2–specific salivary IgA. If enough sample volume (≥100 μl) was available, samples were tested for SARS-CoV-2–specific salivary IgG.

Convalescent sera from three SARS-CoV-2 cases with IgA and IgG Abs against SARS-CoV-2 spike protein were used for the initial assay development. The prefusion stabilized ectodomain of SARS-CoV-2 spike protein (Wuhan-Hu-1 strain; GenBank MN908947.3) that was used in the assays was expressed in suspension-adapted HEK-293 cells as described previously (16). Ag concentrations ranging from 0.125 to 1.00 μg/ml in PBS and 1:1,000 to 1:20,000 diluted HRP-conjugated goat anti-human IgA or IgG were initially tested. Positive and negative controls (SARS-CoV-2 convalescent serum and prepandemic saliva samples, respectively), as well as blank controls (only blocking buffer), were also included in each run. All volumes were 100 μl per well, except where indicated. All washes were performed three times with 250 μl of PBS with 0.05% Tween 20 using a BioTek 405 plate washer. All dilutions were prepared in blocking buffer (5% w/v powdered milk/PBS with 0.05% Tween 20). All incubations were carried out for 1 h at 37°C except where indicated. All concentrations and incubation times were optimized to maximize the OD difference between prepandemic negative samples and SARS-CoV-2 convalescent sera.

Immunolon 2 HB flat-bottomed 96-well plates (Fisher Scientific) were coated with 100 µl of SARS-CoV-2 spike protein (0.5 µg/ml in PBS, rows A–D, positive coated wells) or PBS (rows E–H, negative coated wells) and incubated overnight at 4°C in a humidified chamber. Plates were washed, then blocked with 200 μl/well blocking buffer for 2 h at 37°C. After blocking, plates were washed three times, and fourfold serial dilutions (1:10–1:160) of each saliva sample were added to both Ag- and PBS-coated wells. Plates were incubated and washed, and bound Abs were detected using HRP-conjugated goat anti-human IgA (SeraCare Life Sciences, Milford, MA) diluted 1:4,000 in blocking buffer or anti-IgG (SeraCare Life Sciences) diluted 1:16,000. After incubation, plates were washed again before 100 μl of 3,3′,5,5′-tetramethylbenzidine substrate (SeraCare Life Sciences) was added. The colorimetric reaction was stopped 5 min later by adding 100 µl of Stop solution (SeraCare Life Sciences). The plates were read at 450 nm and 630 nm using an Epoch2 instrument (BioTek). After background correction (OD _450nm − OD_630nm), the adjusted OD value (OD value positive coated well minus OD value negative coated well) was determined.

For the detection of total IgA, Immunolon 2 HB flat-bottomed 96-well plates (Fisher Scientific) were coated with goat anti-human IgA (α-chain) (0.5 µg/ml in PBS, rows A–F, positive coated wells) or PBS 1× (rows G and H, negative coated wells) and incubated overnight at 4°C in a humidified chamber. Plates were washed and blocked with blocking buffer (200 μl/well) for 2 h at 37°C. Three dilutions (1:1,280, 1:5,120, and 1:20,480) of each saliva sample were prepared. A standard curve for IgA was prepared by serial dilution of purified human IgA from colostrum (Sigma-Aldrich, St. Louis, MO). After incubation, plates were washed, and saliva samples were added to rows A–D. Purified human IgA dilutions were added to both positive coated (anti-IgA, rows E and F) and negative coated (PBS, rows G and H) wells. Plates were incubated and washed, and bound Abs were detected using 1:4000 diluted HRP-conjugated goat anti-human IgA (SeraCare Life Sciences). After incubation, plates were washed again before 3,3′,5,5′-tetramethylbenzidine substrate (SeraCare Life Sciences) was added. The colorimetric reaction was stopped 5 min later by adding Stop solution (SeraCare Life Sciences). After background correction, the adjusted OD values were calculated as described above.

The same plate design and steps were used for IgG. Plates were coated with goat anti-human IgG (γ-chain) at 0.5 µg/ml in PBS. A standard curve for IgG was prepared by serial dilutions of purified human IgG (Sigma-Aldrich), and bound Abs were detected using 1:16,000 diluted HRP-conjugated goat anti-human IgG (SeraCare Life Sciences). After background correction, the adjusted OD values were calculated as described above.

The sensitivity (ability to identify samples with Abs to SARS-CoV-2) and the specificity (ability to identify samples without Abs to SARS-CoV-2) were defined as the values for which there is 95% probability that the estimated value can be obtained (17). For IgA EIA validation, 373 SARS-CoV-2 rRT-PCR negative saliva samples (prepandemic) collected from healthy adults (2009–2010) and 44 saliva samples from SARS-CoV-2 rRT-PCR-confirmed cases were used. Similarly, for validation of the IgG assay, 373 prepandemic saliva samples and 68 saliva samples from confirmed cases were used. Saliva samples were collected from 0 to 63 d after diagnosis.

The limit of detection (LOD) was defined as the lowest predicted value for which there is 95% probability that an estimated value can be obtained. To determine the LOD for the IgA (or IgG) EIA, five IgA (or IgG) positive SARS-CoV-2 saliva samples were fourfold serially diluted (1:10–1:10,240). The adjusted OD values were extrapolated from the linear portion of the IgA (or IgG) standard curve. The LOD was determined from the lowest concentration of Ab above the cutoff value (mean of the adjusted OD values for the prepandemic samples + 3 SD).

To evaluate potential cross-reactivity between Abs against the spike protein of SARS-CoV-2 and other human coronaviruses, serum samples (n = 12) were selected from a SARS-CoV-2 household transmission study performed during 2020 (18). Participants’ ages ranged from 1 to 67 y, and reactivity to each endemic coronavirus spike protein (HCoV 229E, HCoV NL-63, HCoV OC43, and HCoV HUK1) was evaluated using the V-PLEX COVID-19 Coronavirus Panel 2 Kit (Meso Scale Diagnostics, Rockville, MD).

Descriptive statistics for continuous variables are presented as median with interquartile range (IQR). For categorical variables, number (percent) was used for descriptive statistics, and differences were evaluated by χ² test or Fisher exact test. Nonparametric data from more than two groups were compared by Kruskal-Wallis test. Receiver operating characteristic (ROC) curve analysis was used to evaluate the sensitivity and specificity of each assay when establishing a cutoff for positivity. Each saliva sample was tested at two different dilutions. To quantify total and SARS-CoV-2–specific Abs, adjusted OD values from serially diluted purified human IgA (or IgG) were plotted against concentration and fit to a sigmoidal four-parameter logistic model. Total and SARS-CoV-2–specific salivary IgA (or IgG) titers were extrapolated from the linear portion of the IgA (or IgG) standard curve. To account for participant differences (e.g., severity of illness, immunocompetency, collection time, Ab secretion levels), SARS-CoV-2–specific IgA or IgG were normalized to 100 µg of total IgA or IgG, respectively. When virus-specific Abs could not be detected but total Abs were detected, SARS-CoV-2–specific IgA (or IgG) per 100 µg of total IgA (or IgG) was arbitrarily assigned as half the lower LOD, based on the standard curve of purified human IgA (19). Statistical analysis was performed using GraphPad Prism version 8.0 (GraphPad Software, La Jolla, CA), and p values <0.05 were considered significant.

This activity was reviewed by the Centers for Disease Control and Prevention (CDC) and was conducted consistent with applicable federal law and CDC policy [see, e.g., 45 C.F.R. part 46.102(1), (2); 21 C.F.R. part 56; 42 U.S.C. §241(d); 5 U.S.C. §552a; 44 U.S.C. §3501 et seq.]. The study was approved by the institutional review boards of the Oregon Health & Science University (IRB 21230) (cohort II). Participants who were cognitively or decisionally impaired were excluded. Written informed consent was obtained from each participant. The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the CDC.

We developed and validated EIAs for the quantitative detection of SARS-CoV-2–specific IgA and IgG Abs in saliva. The final assay conditions are summarized in Table I. Prepandemic saliva (negative for SARS-CoV-2 Abs) and convalescent serum from COVID-19 patients were used for the initial standardization of both EIAs. All reagent concentrations (SARS-CoV-2 spike protein, anti-human IgA [α-chain] or IgG [γ-chain], HRP-labeled secondary Ab, purified human IgA, and purified human IgG concentration for standard curves) and incubation times were optimized to reduce background, obtain maximum anti-SARS-CoV-2–specific signal, and maximize the OD difference between prepandemic negative samples and SARS-CoV-2 convalescent sera.

A dilution of 1:10 of true negative (prepandemic) and true positive (SARS-CoV-2 rRT-PCR–confirmed cases) (Fig. 1) saliva samples resulted in OD values with no background signal in the negative coated wells (PBS coat). Adjusted OD values for both SARS-CoV-2–specific IgA and IgG Abs from true-positive samples were significantly higher in saliva from rRT-PCR confirmed cases than in that from prepandemic samples (Fig. 2A; p < 0.0001). ROC curve analysis was applied to determine the cutoff value that maximizes sensitivity and specificity (Fig. 2B and 2C). For the IgA EIA, the assay specificity improved from 91.5% (95% confidence interval [CI], 88.2–93.9%) to 97.3% (95% CI, 95.0–98.5%) when the cutoff value increased from 1 to 3 SD above the mean (⁠$x¯$⁠) of the prepandemic samples, whereas the sensitivity remained at 95.5% (95% CI, 84.9–99.2%). For the IgG EIA, a significant increase of the specificity from 97.9% (95% CI, 96.6–99.3%) to 98.9% (95% CI, 97.3–99.6%) (p < 0.05) was achieved when the cutoff value was modified from $x¯+SD$ to $x¯+3SD$⁠. Although the sensitivity slightly decreased from 91.2% (95% CI, 82.1–95.9%) to 89.7% (95% CI, 80.2–94.9%), this difference was not significant (p = 0.0625). On the basis of these data, we used a cutoff value of $x¯+3SD$ of the adjusted OD values from prepandemic saliva samples for both the IgA and IgG EIAs. The overall diagnostic accuracy was 98.3% (95% CI, 95.6–100%) and 99.1% (95% CI, 98.0–100%) for the IgA and IgG assays, respectively (Table I).

We also determined if there was any impact on the sensitivity and overall diagnostic accuracy of each assay based on the collection method. Although a slight increase in both parameters was observed for the IgA assay when using passive drools as a collection device and for the IgG assay when collecting samples with Oracol swabs, none of these differences (swabs versus passive drools or swabs versus passive drools versus both together) were significant (data not shown).

For both IgA and IgG Ab assays, a sigmoidal four-parameter logistic curve was fitted to the resulting adjusted OD values to yield a standard curve of Ab concentration versus OD (Fig. 2D). The lower LODs for the assays were 1.98 ng/ml for SARS-CoV-2 IgA Abs and 0.3 ng/ml for SARS-CoV-2 IgG Abs. At the upper limits of quantification, above which the detectors on the plate reader for IgA and IgG Abs are saturated, samples were tested at higher dilutions (>1:1280) for total Abs to fit within the linear portion of the sigmoidal curve and without compromising the detection of Abs against SARS-CoV-2. We tested three dilutions for SARS-CoV-2 and total Abs (Table I). At 1:10 dilution, 42 saliva samples had positive test results for SARS-CoV-2 IgA, whereas at 1:40 and 1:160 dilutions, 16 of 42 (38%) and 25 of 42 (59%) false-negative test results were detected. For total IgA Abs, 10 of 42 (24%) saliva samples were above the upper limit of quantification at 1:1,280 dilution, whereas at 1:5,024 dilution, 42 of 42 (100%) were detected, and at 1:20,480 dilution, 2 of 42 (5%) false-negative test results were detected. At 1:10 dilution, 60 saliva samples had positive test results for SARS-CoV-2 IgG, whereas at 1:40 and 1:160 dilutions, 33 of 60 (55%) and 40 of 60 (67%) false-negative test results were detected. For total IgG Abs, 12 of 60 (20%) saliva samples were above the upper limit of quantification at 1:1,280 dilution, whereas at 1:5,024 dilution, 60 of 60 (100%) were detected, and at 1:20,480 dilution, 7 of 60 (12%) false-negative test results were detected (Fig. 2E). On the basis of these data, we chose to test each saliva sample at 1:10 and 1:40 dilution for SARS-CoV-2 Abs and 1:1280 and 1:5120 dilution for total Abs.

To evaluate the potential cross-reactivity with Abs against endemic coronaviruses (229E, HKU1, OC43, NL63), serum samples with low, medium, and high reactivity to each endemic coronavirus spike were tested for the presence of SARS-CoV-2 IgA and IgG (Fig. 2F and 2G). At the lowest serum dilution (1:100), both SARS-CoV-2 IgA and IgG adjusted OD values were significantly lower in endemic coronavirus (0.039 ± 0.023 and 0.044 ± 0.01) serum than in SARS-CoV-2 serum (0.742 ± 0.133 and 0.841 ± 0.203) (p < 0.0001), suggesting that our assays show no cross-reactivity with Abs against seasonal coronaviruses.

Cohort I yielded 205 saliva samples from 118 SARS-CoV-2 confirmed cases available to be tested for SARS-CoV-2–specific salivary IgA and 166 saliva samples for which enough saliva was available were also tested for IgG (Fig. 1). The mean time between SARS-CoV-2 diagnosis and saliva collection was 22.9 d (median, 20 d; IQR, 14–31.75; range, 2–65 d). Overall, 86.4% (102 of 118) of participants had measurable salivary Abs for one (IgA, 19.5%, n = 23 of 118; IgG, 20.3%, n = 24 of 118) or both isotypes (46.6%; 55 of 118). Conversely, 13.6% (16 of 118) participants had negative test results for both of SARS-CoV-2 salivary IgA and IgG. During the first week after diagnosis, the positivity rate for both IgA and IgG was 28.6% (2 of 7 samples). For IgA, the positivity rate rapidly increased to 62.1% at week 2, peaked at week 4 (68.6%), and decreased to 0% at week 9 and week 10. At week 3, the positivity rate for IgG was 80%, peaked at 100% at week 8, and remained positive to the end of the study (week 10) (Fig. 3A). The median amount of SARS-CoV-2–specific IgA was 0.1 ng (IQR, 0.1–49.44)/100 µg of total IgA at week 1 and increased to 7.95 ng (IQR, 0.1–25.74)/100 µg of total IgA at week 3 (IQR, 0.1–25.74). After fluctuating at week 4 and week 6, the IgA levels became undetectable at week 9. The salivary IgG titers increased between week 1 (median, 0.1 ng/100 µg of total IgG; IQR, 0.1–14.32) and week 5 (median, 123.4 ng/100 µg of total IgG; IQR, 39.32–269.5), peaked to 162.9 ng (IQR, 31.93–322.9)/100 µg of total IgG at week 6 and then remained stable throughout the follow-up period. Compared with week 1, significant differences between IgG levels were observed at week 5 (p < 0.05) and week 6 (p < 0.01) (Fig. 3B).

In cohort II, 85 saliva samples from 42 rRT-PCR–positive participants who were either asymptomatic (n = 8) or had mild (n = 29) or severe (n = 5) clinical COVID-19 disease symptoms were tested for IgA and IgG Abs (Fig. 1; Table II). In the group of patients with mild symptoms, the mean time between onset of disease and sample collection was 146.8 d (median, 178; IQR, 81.25–200.8; range, 1–208 d) and 25.6 d (median, 18; IQR, 14.5–40.5; range, 12–60 d) in the group with severe symptoms.

Overall, for cohort II, the positivity rate for IgA showed a clear pattern with increasing values during weeks 1–3 and a peak at week 4, whereas the positivity rate for IgG was less defined (Fig. 3C). For patients with mild disease symptoms, the positivity rate for IgA and IgG was 33.3% at week 1 after onset of symptoms. The IgA titer rapidly increased and peaked at week 4 (100%), sharply decreased to 0% at week 9, and remained negative until week 30. The positivity rate for IgG peaked at 50% at week 3, fluctuated for several weeks, and returned to 0% 30 wk after onset of disease. In patients with severe clinical symptoms, the positivity rates for IgA and IgG were consistently higher than in mild cases at each time point. The positivity rate for IgA peaked at week 3 (100%), then slowly decreased to baseline at week 10, whereas IgG peaked (80%) at week 5 and remained elevated until the end of the study (week 10).

In cohort II, salivary IgA levels in patients with severe clinical symptoms peaked at week 3 (median, 346 ng SARS-CoV-2–specific IgA/100 µg of total IgA; IQR, 128.2–855.9) and remained positive for the entire study (10 wk), whereas salivary IgG titers peaked at week 4 (median, 810.5 ng SARS-CoV-2–specific IgG/100 µg of total IgG; IQR, 598.2–1778) and remained positive until week 10. In patients with mild disease, salivary IgA titers peaked later, at week 5 (median, 143.6 ng SARS-CoV-2–specific IgA/100 µg of total IgA; IQR, 9.94–277.3) and became undetectable at week 7, whereas salivary IgG Abs peaked earlier at week 3 (median, 507.9 ng SARS-CoV-2–specific IgG/100 µg of total IgG; IQR, 58–957), although with fluctuating values, salivary IgG titers remained detectable until week 36 after onset of disease. In asymptomatic participants (n = 8), salivary IgA titers were detected 1 wk earlier than IgG. Overall, during the first 6 wk after onset of symptoms, salivary IgA and IgG titers were higher in patients with severe symptoms than in patients with mild symptoms (Fig. 3D).

We report the development and validation of in-house EIAs to quantitatively assess the presence of SARS-CoV-2–specific IgA and IgG in saliva and show its value to describe the salivary immune response after natural SARS-CoV-2 infection. Both the IgA and IgG assays were highly accurate, sensitive, and specific. The sensitivity of both assays was high (IgA, 95.5%; IgG, 89.7%) without compromising specificity (IgA, 99%; IgG, 97%). Other published studies have shown high sensitivity for IgG (88–98.4%) but low sensitivity (17–59%) for IgA, with a high specificity for both isotypes (96–100%) (8–11). The difference in sensitivity between our and other IgA assays could be explained by the assay platform, type of sample, or collection time after infection. The LODs of 1.98 ng/ml for SARS-CoV-2 IgA and 0.3 ng/ml for SARS-CoV-2–specific IgG suggest that both assays are suitable for the detection of salivary Abs in samples collected early after infection up to several weeks after recovery.

Overall, we detected SARS-CoV-2–specific IgA and IgG responses as early as 1 wk after onset of disease or diagnosis when disease data were not available. The IgA positivity rate decreased to zero after 10 wk, whereas the IgG positivity rate remained high for at least up to 30 wk. Similarly, data from previous cross-sectional studies showed detectable SARS-CoV-2–specific IgA and IgG levels in saliva 2–4 wk after the onset of symptoms, with only IgG response Abs persisting beyond 60 d (8, 9). A different study reporting results from single saliva samples collected <3 to 9 mo after onset of disease showed a consistently high IgG positivity but a significant decrease of IgA (10). In a longitudinal study including 95 participants, the mean time from disease onset to IgG detection in saliva was 9–11 d, and IgG Abs remained detectable until day 90 (20).

In cohort II, when no saliva sample was collected during the first week after onset of disease, we assumed that participants would have had negative results for both of SARS-CoV-2–specific IgA and IgG. Salivary IgA levels peaked 3 wk after onset of disease and remained elevated until at least week 10 in participants with severe disease symptoms. Conversely, in patients with mild disease, salivary IgA levels were transient, peaked late at week 5, and returned to baseline levels at week 7. Participants with severe disease showed a later (week 4) peak for IgG than those with mild disease, and both remained positive until the end of the study. These initial increases in salivary IgA and IgG were similar in asymptomatic individuals, although the number of samples available after 1 wk was limited. Our data agree with the study by Pisanic et al., who found that salivary IgA and IgG Abs reached a peak at 3 wk after infection, but only IgG remained above baseline levels for at least 60 d after onset of symptoms (9). Another group found similar IgG and IgA kinetics in sera from mild and severe cases (21). Several studies reported higher correlation for IgG than for IgA when testing paired serum and saliva samples (8, 9), leading them to question the advantage of testing IgA in saliva. Early SARS-CoV-2 humoral immune responses are dominated by IgA Abs, which remained detectable in saliva for a longer time than in serum (days 49–73 after symptoms). IgA Abs also contribute to virus neutralization to a greater extent than IgG, although they circulate for a shorter time than IgG (22). Given the less invasive collection method and the transient presence of IgA in contrast to long-lasting IgG, testing for IgA in saliva might allow a better understanding of the timing of infection.

Our study has several limitations. First, our assays were designed and validated to detect Abs against the Wuhan-Hu-1 strain, which may become a limitation, considering the continues evolution of SARS-CoV-2. Although the assays can be adapted and validated for new strains, the applicability of the assays described here to currently circulating strains is unknown. Second, data on the onset and severity of disease were not always available. Second, collection of saliva samples did not always start on day 0 after onset of disease, not all participants provided samples at the same time points, and a limited number of longitudinal samples from asymptomatic individuals were available. Third, because of low sample volume, IgG testing could not be performed on all saliva samples. Fourth, saliva samples were not screened for the presence of Abs against SARS-CoV-1 or MERS-CoV. Others have shown some cross-reactivity with SARS-CoV-1 and MERS-CoV, which would limit the applicability of these assays in a population with high SARS-CoV-1 and/or MERS-CoV incidence. Cross-reactivity with endemic coronaviruses has also been reported particularly in younger populations (23, 24). We tested a limited number of serum samples (ages 1–67) with low, medium, and high reactivity to each endemic coronavirus spike protein. Although we showed no cross-reactivity in our assay in samples collected from children, and considering the fact that Ab levels are lower in saliva than in serum (25), more pediatric samples need to be tested to confirm if these assays can be applied confidently in samples from this age group. Finally, paired serum samples from individuals who provided saliva samples were not available; therefore, a correlation between saliva and serum SARS-CoV-2 Abs could not be evaluated.

In summary, we developed and validated EIAs that are sensitive and specific to detect SARS-CoV-2–specific IgA and IgG Abs in saliva. Significant fluctuations of salivary IgA and IgG Ab levels were observed after infection. Detection of salivary Abs may serve as an easy-to-employ screening method for population and transmission studies as well as for evaluation of vaccine response, especially when collection of blood is challenging or not feasible.

We are grateful to the participants for their willingness to participate during the challenging time of the pandemic. We thank Jennifer Harcourt and Mohammed Rasheed at the Respiratory Immunology Team and Owen Herzegh and David Petway at the Division of Scientific Resources for their support with the γ-irradiation process. We thank John Jernigan, Sujan Reddy, Farrell Tobolowsky, Kelly Hatfield, and Signature HealthCARE for their assistance with recruiting participants. We are also grateful to Peter Sullivan, Matthew Strnad, Felicity Coulter, and Sarah Siegel for logistical and administrative support for patient consent and sample processing at Oregon Health & Science University.

The authors have no financial conflicts of interest.