Sex Differences in the Evolution of Neutralizing Antibodies to Severe Acute Respiratory Syndrome Coronavirus 2

Abstract

We measured anti-spike (S), nucleoprotein (N), and neutralizing antibodies in sera from 308 healthcare workers with a positive reverse-transcription quantitative polymerase chain reaction result for severe acute respiratory syndrome coronavirus 2 and with mild disease, collected at 2 timepoints up to 6 months after symptom onset. At month 1, anti-S and -N antibody levels were higher in male participants aged >50 years and participants with a body mass index (BMI) >25 kg/m². At months 3–6, anti-S and anti-N antibodies were detected in 99% and 59% of individuals, respectively. Anti-S antibodies and neutralizing antibodies declined faster in men than in women, independent of age and BMI, suggesting an association of sex with evolution of the humoral response.

SARS-CoV-2, antibodies, sex-related differences, COVID-19

The duration of humoral immune responses to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is debated. Patients with severe coronavirus disease 2019 (COVID-19) produce more antibodies than asymptomatic or mildly symptomatic individuals [1–3]. Some studies showed a rapid decrease in convalescents regardless of disease severity, and others reported stable antibody titers within the first 3 months [1–7]. Anti-spike (S) antibody amounts correlate with neutralization capacity, since S is the main, if not unique, target for neutralizing antibodies. Neutralizing antibody titers also vary depending on the time post–onset of symptoms (POS) and disease severity [1–3, 8]. Little is known about the influence of sex, age, or body mass index (BMI) on the longevity of anti–SARS-CoV-2 antibodies, particularly in mildly symptomatic individuals, who represent the majority of COVID-19 cases.

MATERIALS AND METHODS

Study Design and Participants

We assessed the persistence of anti–SARS-CoV-2 antibodies in sera from mild COVID-19 healthcare workers in Strasbourg University Hospital. Three hundred eight donors with reverse-transcription quantitative polymerase chain reaction (RT-qPCR)–confirmed SARS-CoV-2 infection were enrolled in the study, with samples longitudinally collected at month 1 (M1) (median, 31 days [range, 11–58 days]) and months 3 to 6 (M3–6) (median, 107 days [range, 78–172 days]) POS (Supplementary Table 1).

SARS-CoV-2 RT-PCR testing on nasopharyngeal swabs was performed at least 10 days before inclusion. Participants completed a questionnaire covering sociodemographic characteristics, virological findings, and clinical data including myalgia, difficulty of breathing, fever, asthenia, rhinitis/pharyngitis, cough, headache, anosmia/dysgeusia, and diarrhea.

Ethics Committee Approval

Results are part of an ongoing prospective, interventional, monocentric, longitudinal, cohort study enrolling staff from the Strasbourg University Hospitals (ClinicalTrials.gov identifier NCT04441684). The protocol was approved by the institutional review board of Strasbourg University Hospitals.

Serological Assays

Commercial Assays.

The Biosynex (COVID-19 BSS immunoglobulin G [IgG]/immunoglobulin M [IgM]) lateral flow assay (LFA) detects IgM and IgG against the receptor-binding domain with 99% specificity and 96% sensitivity after 22 days POS [9]. The EDI Novel Coronavirus COVID-19 IgG enzyme-linked immunosorbent assay (ELISA) (Epitope Diagnostics) detects IgG against the nucleocapsid protein (N), with 96% specificity and 81% sensitivity after 28 days POS [9].

S-Flow Assay.

The S-Flow assay [10] was adapted by using 293T cells stably expressing the S protein (293T spike cells) and 293T empty cells as control. The positivity of a sample was defined as a specific binding above 40%, with 100% specificity (95% confidence interval [CI], 98.5%–100%) and 99.2% sensitivity (95% CI, 97.69%–99.78%) [10, 11]. Binding units (BU) were calculated to standardize the results. A standard curve with serial dilutions of a human anti-S monoclonal antibody (mAb48) was acquired in each assay. The logarithm of the median of fluorescence of each sample was reported on the curve to obtain an equivalent value (in nanograms per milliliter) of mAb48 concentration in logarithm.

Cells.

293T cells (American Type Culture Collection [ATCC] CRL-3216) were transduced to express SARS-CoV-2 S (GenBank accession number QHD43416.1) or with an empty lentivector and selected with puromycin [10, 12]. 293T cells stably expressing angiotensin-converting enzyme 2 and inducible transmembrane protease, serine 2 (293T-ACE2-iTMPRSS2) were produced by lentiviral transduction and selection with puromycin and blasticidin. Cells were regularly tested for absence of mycoplasma.

Neutralization of Pseudotyped Lentiviral Particles

The assay was performed as described previously [10]. 293T-ACE2-TMPRSS2 (2 × 10⁴) was plated in 96-well plates. Sera were diluted at 1:100 and incubated with spike-pseudotyped lentiviral particles (provided by Theravectys-Pasteur laboratory) for 15 minutes at room temperature before addition to cells. After 48 hours, the luciferase signal was measured with EnSpire Plate Reader (PerkinElmer). The percentage of neutralization was calculated as: $100\times (1-\frac{mean\left(luciferasesignalinsampleduplicate\right)}{mean\left(luciferasesignalinvirusalone\right)})$⁠. A titration of a mAb48 was performed on each plate as control.

Statistical Analysis

Baseline characteristics between men and women were compared using a χ ² test for categorical variables and Student t test for continuous variables. Correlations between antibody measures at M1 and characteristics of participants were estimated using linear regression models for factors associated with BU and neutralization levels, and logarithmic regression models for factors associated with IgM and IgG positivity. The difference in BU and neutralization levels between M1 and M3–6 was then estimated and standardized by the time interval between the 2 timepoints. Factors associated with these standardized differences were investigated using linear regression models. Factors that were associated with the outcome with a P value <.15 in univariate analysis were introduced in the multivariate model. P < .05 was considered statistically significant. Subjects were divided into “sustainer” or “decayer” categories, for anti-S, anti-N IgG, and neutralizing antibodies, as reported previously [7]. We calculated the fraction of antibody value at M3–6 divided by value at M1. “Sustainers” were defined by a fraction ≥1, and “decayers” by a fraction <1. The half-life of decayers, extrapolated from the equation of the segment formed by the 2 timepoints, corresponds to the week for which antibodies reach half of M1 level.

Analyses were performed using Stata (StataCorp, College Station, Texas), Excel 365 (Microsoft), RStudio Desktop 1.3.1093 (R Studio, PBC), or Prism 8 (GraphPad) software.

RESULTS

We analyzed the longitudinal antibody response in a monocentric cohort of 308 SARS-CoV-2 RT-qPCR–confirmed staff from Strasbourg University Hospitals (Supplementary Figure 1). The cohort included 75% women, with a median age of 39 years (Supplementary Table 1). The participants were nurses, doctors, caregivers, and administrative staff. Contact with a COVID-19 patient, within or outside of the hospital, was reported by 37% of individuals and 94% had mild symptoms consistent with COVID-19 (Supplementary Table 1). Sixteen participants were hospitalized for moderate disease. None progressed to severe illness. The median time from onset of symptoms to RT-qPCR testing was 3 days. All individuals were sampled twice, first at M1 with a median of 31 days POS (range, 11–58) and second at M3–6 at a median of 107 days POS (range, 78–172).

Seropositivity rates were estimated with 4 assays. The dynamics of the immune response were assessed by comparing antibody levels at different times POS (Figure 1). All participants had anti-S IgG by S-Flow at M1 and 3 participants (1%) became negative at M3–6. Quantitative measurement with standardized BU demonstrated a slight but significant decrease of anti-S IgG amounts between M1 and M3–6 (Figure 1B). The LFA (Biosynex) detecting anti-S IgG and IgM was less sensitive than S-Flow, with 85% of individuals IgG seropositive at the 2 timepoints. IgM was detected in 93% of participants at M1 and only 79% at M3–6, likely reflecting the contraction of the IgM response. Measurement of anti-N IgG with an ELISA (EDI) gave similar results to LFA at M1, but only 59% of individuals remained positive at M3–6 (Figure 1A). The neutralization activity, measured with pseudotyped lentiviral particles, also declined over time. With a positive neutralization threshold set at 20%, 95% and 84% of the sera were positive at M1 and M3–6, respectively (Figure 1A). Applying more stringent thresholds (50% or 80%) confirmed this decline (Supplementary Figure 2). We observed a correlation between neutralization activity and anti-S or anti-N IgG in the sera (not shown). Plotting the median values of anti-S, neutralizing, and anti-N antibodies at different time intervals confirmed a slow decline over time, with large interindividual variations (Supplementary Figure 3A).

Figure 1.

Temporal evolution of anti–severe acute respiratory syndrome coronavirus 2 antibodies. A, Number of individuals displaying anti-S (S-Flow or Biosynex tests), anti-N (EDI N enzyme-linked immunosorbent assay [ELISA]), or neutralizing antibodies were plotted at month 1 (M1) and months 3–6 (M3–6) post–onset of symptoms (POS). The percentages of positive cases are indicated in the bars. Neutralization positivity was defined as a neutralizing activity against lentiviral pseudotypes >20%, at a 1:100 serum dilution (defined as inhibitory dose). Differences between timepoints were analyzed with χ ² test. ****P < .0001. B, Levels of antibodies defined as binding units for anti-S immunoglobulin G (IgG) (S-Flow assay), percentage of neutralization, and optical density for anti-N IgG (ELISA) were plotted against the days POS. Pink and purple points stand for M1 and M3–6 timepoints, respectively. Each gray line connects the timepoints from a same donor. The black line represents the median of all samples for each timepoint. Paired Wilcoxon test was performed between M0 and M3–6. ****P < .001. Abbreviations: Abs, antibodies; BU, binding units; ID₂₀, inhibitory dose; IgG, immunoglobulin G; IgM, immunoglobulin M; M1, month 1; M3–6, months 3 to 6; Nb, number of; ns, not significant; OD, optical density; POS, post–onset of symptoms.

We then determined whether these variations may be attributed to biological or clinical characteristics of the participants. We analyzed the associations between antibody levels (anti-S IgG, neutralizing activity, or anti-N IgG) and sex, age, BMI, and type of symptoms, at M1 and M3–6. We calculated the slope of the curves between the 2 timepoints, to assess the impact of the participants’ characteristics on the evolution of the response. Levels of anti-S and neutralizing antibodies were higher in men than in women at M1 but not at M3–6 (Figure 2A). Accordingly, the slope of antibody decline was significantly steeper in men (Figure 2B). A multivariate analysis showed that anti-S and neutralizing antibodies were higher at the first timepoint and declined faster in men, independent of other factors (Supplementary Table 3). There was no significant difference between men and women regarding the decline of anti-N IgG.

Figure 2.

Sex differences in anti– severe acute respiratory syndrome coronavirus 2 antibody levels at the 2 samplings and their temporal evolution. A, Anti-S immunoglobulin G (IgG) (in binding units), percentages of neutralization, and anti-N IgG (optical density) were compared between men (green dots) and women (orange dots) at month 1 (M1) or months 3 to 6 (M3–6). The black line represents the median of all samples for each timepoint. Samples from women and men were compared with a Mann–Whitney test. *P < .05. P value or nonsignificance is indicated on the segments. B, Weekly evolution of antibody levels between M1 and M3–6 was calculated as (level at M3–6 – levels at M1) / (number of weeks post–onset of symptoms [POS] M3–6 – number of weeks POS M1). Color coding and graphical parameters are as in A. The dotted line represents a stable antibody level (evolution of 0). Statistical analysis by Mann–Whitney test. **P < .01. C, Each subject was defined as “sustainer” (purple) if the fraction antibody at M3–6 / antibody at M1 was ≥1 or “decayer” (yellow) if the fraction was <1. The proportion of sustainers and decayers is compared between women and men. Differences were assessed with χ ² test. *P < .05; ***P = .0001. Abbreviations: Abs, antibodies; BU, binding units; IgG, immunoglobulin G; M1, month 1; M3–6, months 3 to 6; Nb, number of; ns, not significant; OD, optical density.

The majority of individuals showed a decline whereas others displayed stable antibody amounts. We categorized the subjects, based on the stability of the humoral response, into “sustainers” and “decayers” [7] (Supplementary Figure 3B). The proportion of decayers varied between 71% and 83% for the 3 assays. Among decayers, the median half-life of antibody levels was 41.0 (interquartile range [IQR], 24.3–71.8) weeks for anti-S IgG, 19.9 (IQR, 14.4–36.0) weeks for neutralizing antibodies, and 18.4 (IQR, 15.2–25.7) weeks for anti-N IgG (Supplementary Figure 3C). We also noted that female subjects were in higher proportion sustainers than decayers compared to male subjects (Figure 2C), in line with our observation that antibodies persist for longer periods of time in women (Figure 2B).

Categorization of the participants by age (≤30, 30–50, and >50 years old) and BMI (17–25 kg/m², ≥25 kg/m²) further showed that older participants and those with a high BMI had higher antibody titers at M1, as seen with anti-S, neutralization, and anti-N IgG (Supplementary Figure 4). However, the decline of antibody levels occurred at the same rate, regardless of age or BMI (Supplementary Table 3 and Supplementary Figure 4).

There was no association of reported clinical signs, except anosmia/ageusia or cough, with the amount of antibodies at M1 nor with their evolution (Supplementary Tables 2 and 3). This likely reflects the homogeneity of symptoms, as all participants suffered from a mild-to-moderate disease. As reported [10], the antibody levels at M1 were higher in hospitalized individuals but decreased at the same rate as nonhospitalized patients (Supplementary Table 1 and Supplementary Figure 5). Multivariate analyses indicated that high antibody levels at M1 were associated with a more rapid decline, independent of any other parameters (Supplementary Table 3).

DISCUSSION

Assessing the long-term humoral response is critical to evaluate immune protection at the population level. Commercially available assays have been validated with sera collected from acutely or recently infected individuals. Differences in the sensitivity of ELISA tests, including those detecting anti-N antibodies, may be dependent on the days POS [13]. We performed here a longitudinal analysis of the humoral response in a cohort of 308 RT-qPCR–confirmed SARS-CoV-2–infected patients with mild disease. Antibodies declined over the 172 days of analysis. We observed a sharp decrease of anti-N seroprevalence between M1 and M3–6 that may reflect a lower abundance of anti-N antibodies in mild disease, a different kinetic of the anti-N response, or a lower sensitivity of the test. The poor performance of some serological assays for long-term analyses may explain discrepant results regarding the stability or waning of antibody titers in convalescent patients.

Neutralizing antibody levels were assessed using pseudotyped lentiviral particles. One limitation of our study is the use of a single dilution of the sera (1:100). However, we and others previously reported a correlation between the percentage of neutralization at this nonsaturating dilution and titers obtained with pseudovirus or infectious virus and serial dilutions of the sera [10, 13]. Neutralizing antibody levels decreased twice as fast as anti-S IgG, with half-lives of 19.9 and 41 weeks, respectively. We further report sex differences in the longevity of the immune response. Males displayed higher antibody levels shortly after infection, but a steeper decrease, so that the difference was no longer visible at M3–6. In line with our results, previous reports showed a stronger induction of the immune response against SARS-CoV-2 in male patients [14]. Multiple studies have demonstrated that women develop more robust responses to infections and vaccination and are more sensitive to autoimmune diseases than men [15]. This may be linked to sex hormones, X chromosomal, and environmental factors. SARS-CoV-2–infected women mount more robust T-cell activation than male patients [14], which will impact the duration of the response. We limited our analyses to the first 6 months of convalescence after infection. Future work will help determining whether the sex differences reported here are amplified over time and may be linked to differences in antigen persistence [8]. It will also be of interest extending our analysis on antibody longevity to other categories of persons, including asymptomatic individuals who represent most of SARS-CoV-2 cases, patients who recovered from severe forms of COVID-19, and vaccine recipients. Whether vaccines provide a longer protection in women than in men and whether this different evolution will impact the sensitivity to viral variants remain outstanding questions.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. Conceptualization and methodology: A. F., O. S., S. F. K. Cohort management and sample collection: A. V., F. G., C. S.-M., M. J. W., D. R., N. M., Y. H., J. D. S., A. F., M. G., S. F. K. Lateral flow assay and EDI assay: A. V., F. G., M. J. W., L. Gl. S-Flow and seroneutralization: T. B., L. Gr., I. S., O. S. Data assembly and manuscript writing: L. Gr., A. V., F. G., T. B., Y. M., A. F., S. F. K., O. S. Funding acquisition: A. F., O. S., S. F. K. Supervision: O. S., S. F. K.

Acknowledgments. We thank the patients for participation in the study; Maaran Michael Rajah for critical reading of the manuscript; the Direction de la Recherche Clinique et de l’Innovation; the Centre d’Investigation clinique and Médecine du travail teams for cohort management; Sophie Bayer from the Unité de Coordination de la Biologie des Essais Cliniques team; Anne Moncolin, Veronique Sohn and Axelle Grub for management and distribution of samples; and Pierre Charneau for the gift of pseudoviruses.

Disclaimer. The funders had no role in study design, data collection, interpretation, or the decision to submit the work for publication.

Financial support. S. F. K.’s laboratory is funded by Strasbourg University Hospitals (Hôpitaux Universitaires de Strasbourg; PRI 7782), Agence Nationale de la Recherche (ANR-18-CE17-0028), Laboratoire d’Excellence TRANSPLANTEX (ANR-11-LABX-0070_TRANSPLANTEX), and Institut National de la Santé et de la Recherche Médicale (UMR_S 1109). O. S.’s laboratory is funded by Institut Pasteur, Urgence COVID-19 Fundraising Campaign of Institut Pasteur, Agence Nationale de recherches sur le sida, Sidaction, the Vaccine Research Institute (ANR-10-LABX-77), Labex IBEID (ANR-10-LABX-62-IBEID), “TIMTAMDEN” Agence Nationale de la Recherche ANR-14-CE14-0029, “CHIKV-Viro-Immuno” ANR-14-CE14-0015-01, the Gilead HIV cure program, and ANR/Fondation Pour la Recherche Médicale Flash COVID. L. Gr., is supported by the French Ministry of Higher Education, Research and Innovation.

Potential conflicts of interest. L. Gr., I. S., T. B., and O. S. are holder of a provisional patent on the S-Flow assay. All other authors report no potential conflicts of interest.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Author notes

L. G., A. V., and Y. M. contributed equally to this work as co–first authors.

O. S. and S. F.-K. contributed equally to this work as co–last authors.