A high-throughput liquid bead array assay confirms strong correlation between SARS-CoV-2 antibody level and COVID-19 severity.

A detailed understanding of the adaptive host response to SARS-CoV-2 infection in humans is urgently needed. We developed a sensitive, high-throughput, and efficient assay using liquid bead array technology. We observed advantages over traditional ELISA for the detection and quantification of binding IgG against the receptor binding domain (RBD) of SARS-CoV-2. To determine whether COVID-19 symptom severity correlates with SARS-CoV-2 IgG, we measured anti-RBD IgG levels from 67 subjects recovered from PCR-confirmed COVID-19. We found that COVID-19 symptom severity strongly correlated with RBD IgG level (p < 0.001). These findings have substantial implications for public policy surrounding assessments of antibody responses and possible immunity, as not all cases of COVID-19 can be assumed to generate a protective antibody response, and mild disease in particular is capable of generating very low-level anti-RBD IgG levels. These findings also have important implications for the selection of donors for convalescent plasma to be used therapeutically.

Introduction

Much is unknown regarding the adaptive immune response to the coronavirus SARS-CoV-2 (Coronaviridae Study Group of the International Committee on Taxonomy of, 2020). Due to the novel nature of this virus, the population has little or no pre-existing immunity to SARS-CoV-2, which has major implications for optimal management and prevention of this globally important pathogen. As with other coronaviruses, the spike protein (S) of SARS-CoV-2 plays a crucial role in pathogenesis and serves as the immunodominant antigen in the host response to infection. The spike protein is a large, membrane-bound glycoprotein that is divided into the S1 and S2 domains: the S1 domain, or receptor binding domain (RBD), binds to the human angiotensin-converting enzyme 2 (ACE2) receptor and facilitates viral attachment to human cells (Jiang et al., 2020; Wrapp et al., 2020; Zhou et al., 2020; Letko et al., 2020; Yan et al., 2020).

During the SARS outbreak in 2003, it was reported that anti-SARS-CoV antibody levels correlated with disease severity, with clinically sicker patients mounting a higher-titer response (Lee et al., 2006). A number of neutralizing monoclonal antibodies (mAbs) were also reported to bind specifically to the RBD of SARS-CoV (Pak et al., 2009; Hwang et al., 2006). Emerging data strongly suggest that RBD is a highly specific, immunodominant target of the host response following COVID-19 (Premkumar et al., 2020). Antibody titers to RBD correlate with reduction of plaque formation by SARS-CoV-2 in ACE2-producing Vero cells (Okba et al., 2020), and sera with high-titer anti-RBD antibodies neutralize SARS-CoV-2 more potently (Hansen et al., 2020; Yuan et al., 2020). It remains unclear, however, whether disease severity is clearly correlated with antibody production. This correlation has major implications as a correlate of immunity is sought, as mild disease may not generate optimal protection against future SARS-CoV-2 exposures.

To investigate this, we developed SARS-CoV-2 SRBD ELISA and liquid bead array assays to quantify binding antibody levels in human sera. The bead-based assay is intended to offer a superior degree of discrimination and accuracy, particularly at low-level antibody concentrations, along with a much wider dynamic range, compared with traditional indirect ELISA methodology. We then obtained human sera from subjects with proven, symptomatic SARS-CoV-2 infection at disease convalescence (~6 weeks post symptom onset). We used these samples to determine whether clinical symptom severity is correlated with SARS-CoV-2 RBD antibody titer following COVID-19.

Results

Assay performance characteristics

The liquid bead array assay was designed to detect the presence of anti-SARS-CoV-2 SRBD -specific antibodies from human sera. Recombinant RBD was conjugated to Luminex MagPlex microspheres and incubated with serially diluted serum samples and a cross-reactive SARS-CoV monoclonal standard. Over 20 replicate runs were conducted to define the accuracy, precision, and range of the bead array assay. Accuracy, defined as percent recovery of the known value of the mAb reference, ranged between 94.9% and 107.5%. The lower limit of detection (LOD), defined as the mean fluorescent intensity (MFI) of IgG-depleted serum plus 3 times the standard deviation of the mean, was <3.07 × 10−7 ng/mL. Precision was assessed by repeated measurements of a high control, negative control, and three samples known to span the range of the assay. An inter- and intra-assay coefficient of variation (CV) of under 25% for all samples was obtained. Last, the assay was found to be highly efficient, requiring only 0.8 μL of human sera per run.

Bead array assay versus standard indirect ELISA

To evaluate the performance of the liquid bead array assay, a subset of sera were tested by traditional indirect ELISA. The subset was chosen to include high-, medium-, and low-titer samples based on the bead array data. Standard curves were determined for both assays using the mAb CR0322 (Figure 1); although the linear portion of the curves largely overlap for both assays, the bead array assay shows a higher degree of sensitivity for all samples tested. For equivalent curves, standard ELISA required a higher concentration of 2 μg/mL compared with the bead assay initial concentration of 0.33 μg/mL (see Table S1 for specific sample data from dilutions across the linear portion of each sample curve). The performance of the bead assay at the lower end of the curve was also more sensitive than ELISA, with the bead assay achieving a LOD of less than 3.07 × 10−7 ng/mL when compared with the LOD by ELISA of 1.86 × 10−6 ng/mL. Thus, indirect ELISA confirmed the results of the bead array assay, while highlighting the potential advantages of a bead array platform.

Figure 1

Liquid bead array assay performs favorably in comparison with standard indirect ELISA

Standard curves were generated by 4-fold serial dilutions of a cross-reactive SARS-CoV-1 monoclonal antibody known to bind SARS-CoV-2 RBD. Both the bead array and standard ELISA generate valid binding curves, although the bead array exhibits higher accuracy (defined as percent recovery against known antibody concentrations) and greater sensitivity with a lower limit of detection.

Correlation of disease severity with SARS-CoV-2 IgG

Convalescent sera were obtained from 67 subjects following PCR-confirmed COVID-19. The median age in the study population was 36.0 years (interquartile range [IQR] 24.0–52.5), and 33% of the subjects were male (see Table S2 for full demographic information and RBD IgG levels per subject). Symptom scores ranged from very mild (scores of 1 or 2; n = 2 and 7, respectively) to severe symptoms that warranted hospitalization (scores of 7 or 8; n = 8 and 4, respectively). The most frequent symptom scores were 4, 5, and 6 (n = 14, 10, and 13, respectively).

Raw MFI from the bead array assay was converted to EU/mL by standardization against a monoclonal antibody. Samples from subjects who had recovered from COVID-19 had a markedly wide range of IgG levels against SARS-CoV-2 RBD, ranging from 9.6 to 731,768 EU/mL. Increasing COVID-19 symptom severity was strongly and significantly correlated with quantity of SARS-CoV-2 RBD binding IgG in both unadjusted and covariate-adjusted analyses (p < 0.001, Figure 2; Table 1). For example, the mean increase in IgG levels for a patient who reports a 2 versus a 7 on the COVID-19 severity scale was 40,906 EU/mL (95% confidence interval: 16,785, 69,951). Furthermore, this correlation was independent of time to sample collection. All samples were collected approximately 6 weeks post symptom onset (median days from symptom onset: 45; IQR 41–55]. Four subjects were hospitalized for COVID-19; these four subjects exhibited four of the five highest SARS-CoV-2 IgG levels in the cohort.

Figure 2

SARS-CoV-2 receptor binding domain IgG levels are strongly correlated with COVID-19 symptom severity

Partial effect plot of log antibody measurement (EU/mL, calculated via liquid bead array normalized to a monoclonal standard) of binding IgG against SARS-CoV-2 RBD in samples obtained ~6 weeks post-COVID-19. Increasing symptom severity is strongly associated with increased anti-SARS-CoV-2 RBD IgG (p < 0.001, see Table 1). Measurements were performed in duplicate, with a triplicate value obtained if the duplicate values differed by >15%, and the repeated measures were accounted for in the model by compound symmetric correlation. Gray region indicates point wise 95% confidence interval limits. Red points indicate IgG values from subjects hospitalized for COVID-19.

Table 1

Multivariate model of association with SARS-CoV-2 RBD IgG level with a priori selected variables

Predictor	Chi-square statistic	p value
Severity of COVID-19 symptoms	50.730	<0.001
Agea	0.509	0.775
Sexb	0.735	0.391
Time from symptom onset to sample collectionc	1.127	0.288
Total	51.584	<0.001

Median age in the study population was 36.0 years [IQR 24.0–52.5].

67% of subjects were female and 33% male.

Median time from symptom onset: 45 days [IQR 41–55 days].

Discussion

As physicians and public health officials across the globe work to rapidly identify best practices for COVID-19 pandemic response and containment, a detailed understanding of the natural adaptive host response to SARS-CoV-2 infection is crucial. We identified a highly significant correlation between clinical severity of COVID-19 disease and the amount of SARS-CoV-2 binding antibody against the receptor binding domain. This finding has significant implications for defining immunity following infection. Although RBD titer has been shown to correlate with viral neutralization in vitro (Premkumar et al., 2020; Okba et al., 2020), a true “correlate of protection” remains unknown. It appears increasingly likely, however, that total antibody amount (particularly anti-RBD IgG) will serve as a surrogate for functional protection following infection. Our data suggest that more severe COVID-19 symptoms may be associated with increased protection from subsequent SARS-CoV-2 infection, although these responses are known to wane over time, and true correlates of durable protection are urgently needed.

The liquid bead array assay described in the report offers advantages over traditional ELISA, including improved discrimination at low-level antibody titers. We found the reported assay methodology to be highly accurate and reproducible, and the assay allowed for discrimination of anti-RBD IgG at very low levels. Furthermore, the wide dynamic range of the assay (with read-outs in this study ranging from 10 to 410,700 EU/mL) allows for a highly granular visualization and analysis of data, which can be advantageous for research questions such as the one reported in this article. While the bead array is unlikely to be deployable in clinical settings due to a relatively higher cost and required expertise compared with standard techniques, this assay is ideally suited to a research environment for situations in which higher dynamic range and immunologic granularity are desirable.

Clinical aspects of COVID-19 are highly variable between individuals (Huang et al., 2020; Holshue et al., 2020), and asymptomatic disease is also common. Recent reports have suggested that asymptomatic disease generates a fundamentally different host response compared with critical illness, but it was previously unclear if this distinction persisted across milder, but symptomatic, patients with COVID-19 (Long et al., 2020). Our study benefits from access to nearly 70 samples from patients with confirmed COVID-19 and associated symptom severity scoring. Our clinical correlation data are somewhat limited because the majority of patients were not hospitalized, symptom scores were subjective, and judgments of severity may vary across study subjects. However, it is notable that of the 67 patients, 4 required hospitalization (a more objective marker of increased severity), and each of the hospitalized subjects mounted a very high-titer IgG response to SARS-CoV-2 RBD.

In conclusion, we report an efficient and reproducible platform for the measurement of SARS-CoV-2 antibodies in human serum and have found, consistent with data from SARS-CoV-1, that the severity of COVID-19 correlates with the amount of antibody produced against the viral receptor binding domain. As pandemic responses begin to investigate the use of serology and consideration of “immunity passports,” it is important to note that not all cases of COVID-19 can be assumed to generate a protective antibody response, and mild disease in particular may generate virtually no detectable anti-RBD IgG. This has widespread implications, including the use of COVID-19 symptom severity as a simple screen for which patients might represent ideal donors for convalescent plasma, as hospitals move toward use of plasma as a potential therapeutic. Further work to define a true correlate of protection and predictors of protective immunity is urgently needed.

Limitations of the study

This study uses convalescent plasma from infected patients to quantify anti-SARS-CoV-2 antibody titer. Presence of antibody alone does not confer an assumption of protection (i.e., virus neutralization), and ongoing work will investigate correlation between RBD-IgG-mediated neutralization and protection from SARS-CoV-2 and to determine whether the reported liquid bead assay can be bridged to a correlate of protection.

Resource availability

Lead contact

Further information and request for resources can be directed to Isaac Thomsen, MD, MSCI, Vanderbilt University Medical Center (isaac.thomsen@vumc.org).

Materials availability

The study did not generate new or unique reagents or materials.

Data and code availability

This study did not involve the use of any custom code, algorithms, or software.

Methods

All methods can be found in the accompanying Transparent methods supplemental file.