Antibody Levels From High-Throughput Variant-Specific SARS-CoV-2 Anti-Spike Immunoglobulin G and Angiotensin-Converting Enzyme 2 Neutralization Assays Correlate With COVID-19 Infection Risk in a Large Population

Abstract

Background

SARS-CoV-2 antibody levels have been proposed as a correlate of protection from infection. Yet, large-scale prospective studies of cost-effective scalable antibody measures as predictors of infection under real-world conditions are limited. We examined whether antibody levels measured by high-throughput variant-specific SARS-CoV-2 anti-spike immunoglobulin G (IgG) and angiotensin-converting enzyme 2 (ACE2) neutralization assays correlate with cell-based neutralizing antibody measurements and whether they can serve as a reasonable correlate of protection from SARS-CoV-2 infection.

Methods

We conducted a large institutional cohort study between January 2022 and March 2023. Participants (N = 2513) provided dried blood spot (DBS) samples for assessment of anti-spike IgG and ACE2 inhibition levels through high-throughput assays. Comparison with authentic cell-based SARS-CoV-2 neutralizing antibody assays was conducted with serum samples (n = 105). Associations between antibody levels and risk of infection were evaluated.

Results

Correlation between serum and DBS sampling and between cell-based neutralizing and high-throughput antibody binding assays was high for anti-spike IgG and ACE2 neutralization, though the degree of correlation varied by variant. Longitudinal evaluation suggested that DBS-based IgG and ACE2 inhibition levels were anticorrelated with infection risk, with higher sensitivity noted for ACE2 inhibition and variant-matched measures. IgG and ACE2 inhibition levels decreased over time, with more durable responses observed in participants whose most recent priming event was infection vs vaccination.

Conclusions

Findings suggest that variant-specific SARS-CoV-2 antibody levels may be a useful correlate of protection for infection, which has important implications for vaccination recommendations and evaluating infection risk. High-throughput assays measured via DBS may have utility in timing of boosters at the population or individual level.

High-throughput variant-specific antibody assays measure SARS-CoV-2 viral neutralization and protection. Antibody cut points can be derived to demarcate infection risk, though cut points differ by assay and variant, with higher sensitivity noted for circulating variant–aligned inhibition assays.

The SARS-CoV-2 pandemic presented an unprecedented global challenge, and meaningful data collected during the height of the pandemic may help inform future responses to emerging SARS-CoV-2 variants and other infectious diseases. Infectious disease experts have highlighted the importance of identifying correlates of protection to inform risk mitigation measures, primarily vaccination strategies [1]. Target binding and neutralizing antibody (NAb) assays have been proposed as potential markers of protection and have been used by regulators to determine vaccine efficacy. Yet, large-scale prospective studies of antibody measures as a predictor of infection risk under real-world conditions are limited. Furthermore, it has been challenging to determine whether levels of antibodies for 1 SARS-CoV-2 variant are predictive of protection against another.

Vaccination and infection result in host antibody responses, although differences in level and duration of protection from each are not entirely clear and may vary by variant [2–6]. Early studies based on statistical models to predict the efficacy of vaccines developed for the ancestral strain against variants of concern suggested that neutralizing antibodies to the ancestral strain were predictive of protection for all variants, though decreased efficacy was noted against symptomatic and severe infections with emerging variants [7, 8]. Similarly, a small study of direct assessment of antibody responses reported that higher levels of anti-spike immunoglobulin G (IgG) and immunoglobulin A to ancestral, Delta, and Omicron variants were associated with protection against likely Omicron infection 4 to 14 months after a third mRNA vaccination dose [9]. Likewise, a larger study found that decreased IgG, immunoglobulin A, and NAb responses to the ancestral strain following 3 BNT162b2 vaccine doses significantly increased risk of the Omicron infection [10], though variant-specific antibody responses were not examined.

While the association between decreased antibody levels and increased infection risk is consistent in prior studies, they did not explore whether cut points could be identified to demarcate infection risk or whether use of variant-specific measures improves prediction, thus limiting their ability to inform vaccination strategies. Additionally, while prior research has highlighted the potential for antibody levels to be used as a correlate of protection, traditional cell-based methods of NAb assessment are sample, resource, and time intensive, restricting their potential utility in a population-based manner. Two proposed methods for scale-up of antibody assessment are (1) the use of dried blood spot (DBS) samples, which are easier to collect than serum samples, and (2) high-throughput assays, which facilitate cheaper and more feasible testing. While small proof-of-concept studies suggest acceptable precision and reliability of DBS as compared with serum on immunoassay-based quantification [11], large-scale testing and comparison with cell-based neutralization are necessary to support widespread use.

We examined whether variant-specific antibody levels measured with high-throughput anti-spike IgG binding and angiotensin-converting enzyme 2 (ACE2) assays on DBS samples correlate with cell-based NAb measurements and are associated with subsequent SARS-CoV-2 infection in a large intensively surveilled population.

METHODS

The present study uses data from the Neutralizing Antibody Project for COVID-19, conducted at the University of California San Diego (UCSD) between January and December 2022. The overarching goal of the study was to estimate levels of SARS-CoV-2 antibodies in the university population and determine whether they were associated with risk of contracting SARS-CoV-2. Students and employees aged ≥18 years were eligible to participate. Participants were recruited via flyers, newsletters, and email messaging in a unique workflow integrated in the electronic health record (EHR) [12]. Following initial enrollment and sample collection, participants were invited to provide follow-up samples for longitudinal assessment. The present analysis uses results of the first sample only to address the primary study aims. The study was approved by the Institutional Review Board at UCSD (No. 801672); all participants provided electronic informed consent prior to participation.

Assessment of SARS-CoV-2 Antibody Levels

To evaluate antibody levels, study participants provided a capillary blood sample via finger-stick blood collection on a Whatman card for DBS. Specific details of sample processing have been provided in the supplementary material. Samples were analyzed with the MSD V-PLEX SARS-CoV-2 Panel 25 Kit (IgG K15583 U or ACE2 neutralization K15586 U; Meso Scale Diagnostics) and processed according to the manufacturer's instructions. The panel includes 10 SARS-CoV-2 spike variant targets: ancestral (Wuhan), B.1.1.7, BA.1, B.1.351, AY.4, B.1.640.2, BA.1 + L452R, BA.1 + R346 K, BA.2, and BA.3. The present analysis primarily focused on 2 SARS-CoV-2 variants: ancestral WA1/2020 (WA), the target of the original mRNA vaccines, and Omicron (BA.1), the primary circulating variant during the study period. Anti-spike IgG levels are reported as arbitrary units (AU) per milliliter. NAb levels were calculated as the percentage inhibition of baseline ACE2 assay signal: [1 – (signal in assay well with participant sample / background signal in assay well with buffer)] × 100.

A subset of participants (n = 105) who had consented under a separate protocol (UCSD Institutional Review Board, No. 200477) that allowed for collection of additional biospecimen types was asked to provide serum samples on the same day as DBS collection for comparison. Serum samples were analyzed by the MSD IgG and ACE2 plates in a similar manner to the DBS samples.

Authentic SARS-CoV-2 Neutralizing Antibody Assay

Neutralization titers against SARS-CoV-2 WA and BA.1 were determined by a focus reduction neutralization test (FRNT) with serum samples. Duplicate aliquots of each sample were heat inactivated at 56 °C for 30 minutes and then frozen at −80 °C. Fourfold serial dilutions of the sera were made in Dulbecco’s Modified Eagle Medium + 1% fetal bovine serum + 1× penicillin/streptomycin by a Hamilton Bravo liquid handler. Dilutions were incubated with 100 to 250 focus-forming units of SARS-CoV-2 in Dulbecco’s Modified Eagle Medium for 1 hour at 37 °C, 5% CO₂. The virus + serum mixture was applied to confluent monolayers of VeroE6-TMPRSS2 cells in 96-well plates for 1 hour with gentle rocking. Inputs were removed, and cells were overlaid with 1% methylcellulose in Minimum Essential Medium + 2% fetal bovine serum + 1× penicillin/streptomycin and incubated 24 hours at 37 °C, 5% CO₂. Plates were then fixed in 4% formaldehyde and stained with a primary antibody against SARS-CoV-2 nucleocapsid protein (gtx135357; GeneTex) and AlexaFluor 594 secondary antibody (Thermo Fisher Scientific) with Sytox Green nuclear counterstain to visualize foci of infection. Whole-well images were acquired on an Incucyte S3 or SX5 imager. Foci were counted with the Incucyte software, and percentage neutralization was calculated relative to infected no-serum control wells on each plate.

Assessment of Vaccination and Infection Status

Self-reported demographic information was collected at enrollment. Information about prior SARS-CoV-2 infection (positive polymerase chain reaction or antigen test result), SARS-CoV-2 vaccination, and SARS-CoV-2 infections during follow-up was queried from EHRs at UCSD. UCSD provides all students and employees with no-cost, easy-to-access SARS-CoV-2 vaccination and testing. Services provided by UCSD are recorded in a shared EHR for student and employee health [13, 14]. Students and employees who access outside services are encouraged to enter them into the UCSD EHR. To decline vaccination (primary series plus booster, when eligible), students and employees must formally request an exemption. Prior to 2022, unvaccinated students and employees were required to test twice per week, while those vaccinated were encouraged to test weekly or if feeling symptomatic. These policies resulted in high rates of vaccination (98.7% completed the primary series or more) and testing (median number of follow-up tests, 8) in the study population.

Statistical Analysis

To assess authentic SARS-CoV-2 neutralizing antibodies, graphs were constructed to show averages of 2 independent experiments done with 2 replicates. SARS-CoV-2 neutralization titers were defined as the sample dilution at which a 50% reduction in foci (FRNT₅₀) was observed relative to the average of infected control wells. Best-fit curves determining FRNT₅₀ were generated in Prism version 10 (GraphPad). Correlations between FRNT₅₀ and serum and DBS anti-spike IgG and ACE2 assays were compared by Spearman coefficients.

The primary outcome for the present study was a positive SARS-CoV-2 polymerase chain reaction or antigen test result within 90 days of DBS collection. EHRs were queried through 31 March 2023 to allow for at least 90 days of follow-up for all participants (median follow-up, 419 days; range, 105–447). Differences in baseline antibody levels between those infected and not infected during follow-up were assessed by generalized linear regression. Associations between antibody levels and risk of infection within 90 days were estimated with log-binomial regression. Logistic regression models were used to construct receiver operating characteristic (ROC) curves for antibody levels. Youden index was used to determine the point on the curve that maximized sensitivity and specificity for infection, which was used to group participants. Kaplan-Meier survival curves and Cox proportional hazards models were used to examine risk of infection over the entire follow-up period by antibody level. Follow-up was censored at the time of EHR extraction, based on time since DBS. Participants who received a booster vaccination during follow-up and had not been infected were censored at the time of vaccination. Antibody levels over time since the most recent priming event (vaccination or infection) prior to DBS were assessed in exploratory analyses. All analyses were conducted for WA and BA.1 variants separately for comparison of variant-specific findings. IgG was analyzed on a log scale and back transformed for reporting. Analyses were completed with SAS 9.4. P < .05 was considered significant.

RESULTS

Between January and December 2022, 2513 participants provided a DBS sample (Table 1). Almost all participants had been vaccinated for SARS-CoV-2 (n = 2481, 98.7%), and 516 (20.5%) had a documented prior positive SARS-CoV-2 test result. During follow-up, 120 (4.8%) had a positive SARS-CoV-2 test result within 90 days of DBS collection, and 613 (24.4%) had a positive test result at any point during follow-up (mean time to infection, 151.6 days). No hospitalizations or COVID-19–related pneumonias were documented during follow-up.

Table 1.

Demographic, Clinical, and Infection Characteristics (N = 2513)

	No.	%
Demographics
Age, y
Mean (SD)	38.5	12.7
Median (range)	37	18–80
Sex
Male	800	31.8
Female	1706	67.9
Unknown	7	0.3
Ethnicity
Hispanic	431	17.2
Non-Hispanic	1928	76.7
Unknown	154	6.1
Race
White	1434	57.1
Black	66	2.6
Asian	674	26.8
Native American/Alaska Native	14	0.6
Native Hawaiian/Pacific Islander	30	1.2
Other/mixed	178	7.1
Unknown	117	4.7
Prior exposure status
Vaccinated	2481	98.7
Days from most recent vaccination
Mean (SD)	128.3	92.6
Median (range)	101	2–671
Documented prior positive SARS-CoV-2 test result	516	20.5
Days from most recent infection
Mean (SD)	165.6	175.5
Median (range)	100	0–916
Follow-up infection
Infection within 90 d	120	4.8
Any follow-up infection	613	24.4
Time to infection, d
Mean (SD)	151.6	86.9
Median (range)	141	2–414
Events during follow-up
First follow-up event
Infection	492	19.6
Vaccination	1129	44.9
None	892	35.5
Time to first event infection, d
Mean (SD)	130.9	71.5
Median (range)	129	2–384

Data are presented as No. (%) unless noted otherwise.

Correlations Between FRNT Results and High-Throughput IgG and ACE2 Inhibition Results From Serum and DBS

Neutralizing titers in serum samples by FRNT (FRNT₅₀) varied over a range of 3 to 4 logs and were on average higher against WA than BA.1 (Figure 1A). Matched titers against WA and BA.1 showed a median difference of approximately 6-fold (Figure 1B and 1C), and the relationship of matched log-transformed FRNT₅₀ values against WA and BA.1 was approximately linear (Figure 1D).

Figure 1.

Neutralization of SARS-CoV-2 WA and BA.1 by participant serum fold dilution achieving half-maximal reduction in infection (FRNT₅₀): A, comparison of distributions; B, aligned values for each participant; C, fold difference between WA and BA.1; D, XY relationship with Spearman correlation of log-transformed values of FRNT₅₀ against WA and BA.1 variants for each participant. Each point is 1 participant. Points show average FRNT₅₀. A and C, Error bars show geometric mean ± SD of the population. ****P < .0001 in Wilcoxon test of log₁₀-transformed values. B and D, Error bars indicate mean ± SD of each participant's FRNT₅₀ values from independent neutralization assay experiments. For 72 participants, experiment was performed twice in duplicate for both variants. For 33 participants, experiment was performed twice in duplicate for WA and once in duplicate for BA.1. BA.1, Omicron; WA, ancestral WA1/2020.

Neutralization of WA and BA.1 was highly correlated to variant-specific anti-spike IgG levels measured from serum and DBS (all ρ > .70 or similar; Figure 2A and 2B). FRNT₅₀ correlated closely with ACE2 neutralization assay results for pre-Omicron variants (all ρ > .80) but less so for Omicron variants (Figure 2C and 2D), potentially due to the higher proportion of responses with little to no signal in the ACE2 inhibition assays for Omicron variants. Specific details for all variants are provided in Supplementary Figures 1 to 4. Correlations were lower in the ACE2 inhibition assay for the Omicron variants, likely due to a narrower dynamic range for the ACE2 inhibition assay than for the FRNT assay (Supplementary Figures 3b and 4b) and/or suboptimal titration of serum samples (Supplementary Figures 3a and 4a).

Figure 2.

Correlation between serum-derived FRNT₅₀ values and V-PLEX IgG and ACE2 neutralization assay results from serum (SER; white bars) and dried blood spot (DBS; black bars) samples from 104 participants: A, WA FRNT₅₀ vs variant-specific IgG; B, BA.1 FRNT₅₀ vs variant-specific IgG; C, WA FRNT₅₀ vs ACE2 percentage inhibition; D, BA.1 FRNT₅₀ vs ACE2 percentage inhibition. X-axes show spike proteins used in V-PLEX assays. Spearman r values are shown above bars. Error bars indicate 95% CI. ACE2, angiotensin-converting enzyme 2; BA.1, Omicron; FRNT₅₀, 50% reduction in foci; IgG, immunoglobulin G; WA, ancestral WA1/2020.

The correlation between serum and DBS measurements was >0.9 for variant-specific anti-spike IgG across all variants in the V-PLEX assay and for ACE2 inhibition against pre-Omicron variants (Wuhan, B.1.640.2, B.1.1.7, B.1.351, and AY.4; Figure 3). Variant-specific antibody levels measured by high-throughput assays on DBS samples correlated well with the same assays performed on matched serum samples (Supplementary Figures 5 and 6), with possible overdilution of serum samples noted for some of the ACE2 inhibition assays (Supplementary Figure 6).

Figure 3.

Correlation between dried blood spot and serum analyte in V-PLEX assays: variant-specific (A) anti-spike immunoglobulin G and (B) angiotensin-converting enzyme 2 percentage inhibition for the indicated variants (n = 103). Spearman r values are shown above bars. Error bars indicate 95% CI. BA.1, Omicron; WA, ancestral WA1/2020.

Variant-Specific SARS-CoV-2 Antibody Levels in DBS Samples and Risk of Infection

DBS-based anti-spike IgG and ACE2-competing antibody levels for WA and BA.1 variants were significantly lower at baseline in participants who became infected with SARS-CoV-2 (P < .0001; Table 2). Despite lower measured values, BA.1-specific antibody levels were more predictive of follow-up infection than WA-specific antibody levels, and ACE2 measures performed slightly better than IgG. A 1% decrease in risk of SARS-CoV-2 infection within 90 days was noted for a 100-unit increase in WA anti-spike IgG (relative risk [RR], 0.99; 95% CI, .98–.99), whereas a 5% decrease was noted for the same unit increase in BA.1 IgG levels (RR, 0.95; 95% CI, .93–.98; Table 3). Moreover, a 21% decrease in risk of infection within 90 days was noted for every 10% increase in WA-specific ACE2 inhibition (RR, 0.79; 95% CI, .73–.86), while a 16% decrease was seen for the corresponding BA.1-specific assay (RR, 0.84; 95% CI, .76–.93). Greater decreases in risk were seen with increasing antibody quartiles for all measures except ACE2 BA.1 inhibition, where risk appeared to decrease in the highest quartile (Supplementary Table 1).

Table 2.

Baseline Antibody Levels by Follow-up Infection Status

	Infection Within 90 d, Mean (95% CI)			Any Infection, Mean (95% CI)
	No	Yes	P Value	No	Yes	P Value
IgG, AU/mLa
WA	3415 (3259–3578)	1917 (1557–2361)	<.0001	3508 (3329–3698)	2803 (2556–3075)	<.0001
BA.1	881 (842–921)	493 (404–603)	<.0001	911 (866–958)	709 (649–775)	<.0001
ACE2, %
WA	78.3 (77.5–79.1)	67.9 (64.4–71.4)	<.0001	78.8 (77.9–79.6)	74.8 (73.3–76.4)	<.0001
BA.1	20.9 (19.9–21.8)	13.3 (9.1–17.6)	.0008	21.9 (20.9–23.0)	16.1 (14.2–18.0)	<.0001

Effect estimates from generalized linear regression models. Bold indicates P < .05.

Abbreviations: ACE2, angiotensin-converting enzyme 2; AU, arbitrary unit; BA.1, Omicron; IgG, immunoglobulin G; WA, ancestral WA1/2020.

^aEvaluated on a log scale.

Table 3.

Proposed Antibody Cut Points and Overall Risk of Infection Within 90 Days of Testing

	Cut Point–Based Metrics for Infection Within 90 d of Testing, No. (%)					Increasing Antibody Level
Cut-offa	No	Yes	P Valueb	Sensitivity, %	Specificity, %	RR (95% CI)c
Anti-spike IgG, AU/mL						100-unit increase
WA			<.0001	57.5	64.7	0.99 (.98–.99)
<2550	845 (92.4)	69 (7.6)
≥2550	1548 (96.8)	51 (3.2)
BA.1			<.0001	59.2	63.8	0.95 (.93–.98)
<650	866 (92.4)	71 (7.6)
≥650	1527 (96.9)	49 (3.1)
ACE2, %						10% increase
WA			<.0001	73.3	48.8	0.79 (.73–.86)
<85	1225 (93.3)	88 (6.7)
≥85	1168 (97.3)	32 (2.7)
BA.1			<.0001	89.2	26.2	0.84 (.76–.93)
<25	1767 (94.3)	107 (5.7)
≥25	626 (98.0)	13 (2.0)

Abbreviations: ACE2, angiotensin-converting enzyme 2; AU, arbitrary unit; BA.1, Omicron; IgG, immunoglobulin G; RR, relative risk; WA, ancestral WA1/2020.

^aCut points derived from receiver operating characteristic curves maximizing sensitivity and specificity.

^b P values based on χ². Bold indicates P < .05.

^cRelative risk of infection with increasing antibody levels estimated from log-binomial regression.

Development of Variant-Specific Cut Points for Infection Risk

For infection within 90 days, ROC curves suggested an optimal IgG cut point of 2541 AU/mL for WA (area under the ROC curve [AUROC], 0.64; sensitivity, 59.2%; specificity, 64.5%) and 642 AU/mL for BA.1 (AUROC, 0.65; sensitivity, 59.2%; specificity, 64.3%). By rounding to 2550 and 650 AU/mL for analysis (Table 3), infection rates were >2 times higher in participants with levels below these cut points (WA: 7.6% vs 3.2%, P < .0001; BA.1: 7.6% vs 3.1%, P < .0001), though sensitivity and specificity were relatively low (WA: 57.5% and 64.7%; BA.1: 59.2% and 63.8%, respectively). For ACE2 inhibition levels, ROC curves suggested a WA cut point of 88% (AUROC, 0.65; sensitivity, 79.2%; specificity, 43.8%) and a BA.1 cut point of 24% (AUROC, 0.55; sensitivity, 89.2%; specificity, 27.3%). With 85% as the cut point for WA and 25% for BA.1, participants with WA-specific ACE2 inhibition levels <85% and BA.1-specific levels <25% were significantly more likely to have a positive SARS-CoV-2 test result within 90 days of testing (WA: 6.7% vs 2.7%, P < .0001; BA.1: 5.7% vs 2.0%, P < .0001). While sensitivity was high for both variants (WA, 73.3%; BA.1, 89.2%), specificity remained low (WA, 48.8%; BA.1, 26.2%).

Longitudinal Assessment of Infection Risk Based on Derived Cut Points

Based on the aforementioned cut points, Kaplan-Meier survival curves for infection at any time during follow-up (n = 492 prior to receiving a booster vaccination) suggested an increased likelihood of infection among participants in the low antibody groups (all P < .0001; Figure 4). The highest risk of infection was seen for participants with BA.1 ACE2 inhibition levels <25%, for whom a 73% increased risk of infection during follow-up was noted (hazard ratio [HR], 1.73; 95% CI, 1.38–2.18; Figure 4D). Infection risk for other measures was relatively similar, ranging from 41% to 46% (WA IgG: HR, 1.45 [95% CI, 1.22–1.74]; BA.1 IgG: HR, 1.46 [95% CI, 1.23–1.75]; WA ACE2 inhibition: HR, 1.41 [95% CI, 1.18–1.69]; Figure 4A–C, respectively).

Figure 4.

Kaplan-Meier survival curves of SARS-CoV-2 free follow-up and numbers at risk by antibody risk category: A, anti-spike IgG WA; B, anti-spike IgG BA.1; C, ACE2 WA; and D, ACE2 BA.1. Vaccination during follow-up was censored. ACE2, angiotensin-converting enzyme 2; AU, arbitrary unit; BA.1, Omicron; HR, hazard ratio; IgG, immunoglobulin G; WA, ancestral WA1/2020.

Change in Variant-Specific DBS Antibody Levels Following Infection and Vaccination

A steady decline in anti-spike IgG levels was noted following vaccination and infection through approximately 120 days, at which point levels stabilized for participants postinfection and continued to decline for participants postvaccination, dropping below the identified “risk” cut points approximately 130 days after vaccination (Supplementary Figure 7). WA ACE2 inhibition levels dropped below the 85% cut point at 60 days for the infection and vaccination groups. Conversely, BA.1-specific inhibition levels dropped below the 25% cut point approximately 150 days postinfection, as compared with 30 days following vaccination.

DISCUSSION

Findings from this study demonstrate the feasibility and utility of high-throughput anti-spike IgG and ACE2 neutralization assays performed on DBS in determining SARS-CoV-2 antibody levels and risk of subsequent infection. Comparison of sampling and measurement methodologies indicated a high degree of correlation between serum and DBS sampling and between gold standard authentic SARS-CoV-2 neutralizing and high-throughput antibody assays for anti-spike IgG and ACE2 inhibition, though degree of correlation varied across variants. Longitudinal evaluation suggested that IgG and ACE2 inhibiting antibody levels were associated with infection risk. Higher sensitivity was noted for ACE2 inhibition as compared with IgG levels and for variant-matched measures, while higher but still moderate specificity was noted for IgG and ancestral variant measures. As expected, anti-spike IgG and ACE2 inhibition levels decreased over time, with more durable responses observed in participants whose most recent priming event was infection. These findings support and build on prior studies suggesting that higher SARS-CoV-2 antibody levels are associated with decreased risk of infection and thus may be a useful correlate of protection for infection [7–10], though utility depends on identifying measures that demarcate increased risk.

It is important to recognize that this study was launched in January 2022, meaning that all of the vaccinations were with the original mRNA vaccines designed against the ancestral variant and that Omicron subvariants were likely responsible for infections during follow-up. Thus, results indicate that the original mRNA vaccines continue to provide some protection against Omicron infection, though calculated risk cut points were relatively high for WA anti-spike IgG levels (≥2550 AU/mL) and ACE2 inhibition levels (≥85%). Variant-specific cut points for BA.1 anti-spike IgG and ACE2 inhibition were much lower (≥650 AU/mL and ≥25%, respectively), with only slightly improved sensitivity and HRs, suggesting that either WA or BA.1 assays can be used to predict protection from infection by using different assay-specific cutoffs. Yet, while assays for the ancestral strain may still provide valuable information regarding risk, results suggest that, when available, a variant-specific assay or an assay for a more closely related variant may be preferable as a screening test given the increased sensitivity provided, especially as future variants emerge that may have even lower correlations with the ancestral variant. This finding is supported by prior research suggesting that vaccines against the ancestral variant remain protective, albeit to a lesser extent, against emerging variants [4, 15–18], including Omicron [3, 9, 19, 20], though titers and protection are consistently lower for Omicron. This may help explain why high levels, as measured by assays specific to the ancestral strain, were needed to mitigate Omicron infection risk as compared with the more moderate levels based on BA.1-specific assays.

Furthermore, while the correlations between gold standard cell-based neutralizing assays and IgG levels and the WA strain were higher than for ACE2 inhibition levels and the BA.1 variant, respectively, BA.1-specific ACE2 inhibition levels appear to better differentiate risk of infection with Omicron variants. We note that this finding may be due to the greater spread of values in the tested population in BA.1-specific ACE2 inhibition levels as compared with WA-specific IgG or ACE2 inhibition levels or BA.1-specific IgG levels, which clustered around the higher range of the assays. Additionally, while reporting in binding antibody units per the World Health Organization/National Institute for Biological Standards and Control would permit comparison of anti-spike IgG levels across studies, the assays used here allowed for binding antibody unit measurement to the ancestral strain only, thus limiting its utility in comparing across variants. Reporting ACE2 inhibition levels as “percentage neutralization” may allow for broader comparison, given that this measure can be more readily standardized and it allows for a more limited range of values that may better translate to risk cut points across populations. For these assays to be validated for clinical use, comprehensive optimization of parameters would need to be performed, including a more comprehensive dilution series to mitigate ceiling effects.

Given the timing of our study, it is likely that Omicron subvariants were responsible for most of the recent prior infections. It is therefore not entirely unexpected that the difference in durability of neutralizing antibody response between the vaccination and infection groups was greater for the BA.1-specific assays than WA. We noted a nonlinear decline in antibody levels, which stabilized approximately 3 to 6 months following the most recent event, consistent with other studies [5, 21]. Antibody levels dropped below our designated cutoff more rapidly following vaccination as compared with infection. Prior studies report that those with a previous infection demonstrate higher antibody levels and slower rates of decline against emerging variants even after vaccination [6, 10, 22], though decreased protection against Omicron has been noted [2, 23, 24], suggesting that postvaccination infection may serve as a “priming event” against newer variants. These findings suggest that vaccination boosters may be needed sooner in patients without evidence of infection to maintain protection, though results are limited by use of a single time-point measurement to infer decline.

A major strength of this study is the large number of participants from a well-characterized, stable population with high vaccination and testing rates. All COVID-19–related data were available from a single EHR, with vaccination and testing provided at no cost by a centralized organization. Nonetheless, there are limitations, including the potential for missed infections if participants were tested elsewhere or were asymptomatic and did not test. Results may also be influenced by unmeasured participant-level bias, such as more frequent testing or risk-mitigating behaviors in certain populations. Our analysis is limited by use of the same population to define and analyze cut points and should thus be considered proof of concept rather than formally defined cut points. Additionally, though the study population is from a single, highly vaccinated institution, which may limit generalizability, the high vaccination rates and mix between vaccination with and without prior infection are likely reflective of the condition of the general population, making findings more applicable to the current SARS-CoV-2 situation.

This study has important implications for vaccination recommendations and understanding infection risk. Our findings suggest that variant-specific SARS-CoV-2 anti-spike IgG and ACE2 neutralization assays measured via DBS may have value in timing of boosters, either at the population level or on an individual basis. However, this is complicated by the emergence of new, more transmissible variants, as indicated by the difference in variant-specific cut points and variable protection based on more recent infections. Moving forward, it will be important to continue to evaluate correlations among vaccination, variant-specific antibody response, and susceptibility to infection, particularly given the continuing waves of new SARS-CoV-2 variants and updated vaccines that target these infections.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The following reagent was deposited by the Centers for Disease Control and Prevention and obtained through BEI Resources, National Institute of Allergy and Infectious Diseases, National Institutes of Health: SARS-related coronavirus 2, isolate USA-WA1/2020, NR-52281.

Author contributions. M. B. J., A. E. C., P. D. H., and L. C. L. wrote the manuscript, and M. B. J. and A. E. C. were responsible for primary data analysis. M. B. J. participated in data collection, study conceptualization, methodology, and study oversight. N. H. G. and H. D. V. facilitated participant enrollment, data collection, and study logistics. A. E. C., P. D. H., A. F. C., and L. C. L. optimized and provided oversight of experiments. T. B., J. H., A. R., T. N., B. M., A. F. G., and W. B. assisted with optimization and conduct of experiments and generation of assay data. R. K., C. A. L., A. F. C., and L. C. L. contributed to the conceptualization of the study, methodology, and study oversight. All authors reviewed and approved the final version of the manuscript.

Disclaimer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Financial support. This work was supported by University of California San Diego Chancellor Dr Pradeep Khosla; UC San Diego Health CEO Patricia Maysent; and former UC San Diego Health Sciences Vice Chancellor Dr David Brenner. The project described was partially supported by the National Institutes of Health (grant UL1TR001442), and A. F. C. is partially supported by a Career Award for Medical Scientists from the Burroughs Welcome Fund.