How long COVID-19 and combined infection-vaccine immunity keep neutralizing antibodies in a Canadian group

Samples were collected at The Hospital for Sick Children, Mount Sinai Hospital, and North York General Hospital in Toronto, Ontario, Canada. The samples were divided into three cohorts: (i) symptomatic patients who tested positive for SARS-CoV-2 by PCR or by a serology assay (cohort A; n = 178, 321 samples), (ii) symptomatic patients who tested positive for SARS-CoV-2 by PCR or by a serology assay and had been subsequently vaccinated (cohort B; n = 46, 48 samples), and (iii) healthy naïve controls (cohort C; n = 21, 21 samples) (Table 1).

In our natural infection cohort A (naturally infected SARS-CoV-2 positive cases), the majority of cases had mild symptoms (n = 211 samples, 66%) with a clinical severity score of 1 or 2. The clinical severity score for each patient was assessed and determined according to the WHO criteria (16). Nineteen percent (n = 60 samples) of the cohort was hospitalized with mild disease and had a clinical severity score of 3 or 4; 12% (n = 39) of the cohort was hospitalized with severe disease requiring an intensive care unit (ICU) level of care and had a clinical severity score of 5, 6, or 7; and 3% (n = 11 samples) of the cohort died and had a clinical severity score of 8. In cohort A – natural infection, 49% were female (median age of 42 years) with an age range of 3 months to 100 years (Table 1).

Blood and/or serum samples were collected at multiple time points for some patients, and in cohort A, 51% (n = 113 individuals) of this cohort had >2 visits. Within cohort A, 23% of the samples were collected during the early acute phase of the infection, which was 0–14 days after the onset of symptoms (n = 74); 11% were from the late acute phase between 15 and 27 days (n = 36); 22% of the samples were from the early convalescent phase between 4 and 12 weeks (n = 69); and 44% were collected during the late acute phase, which were samples collected later than 12 weeks PSO (n = 142) (Table 1). The median PSO was 55 days with a range of 2–383 days in cohort A.

In the hybrid immunity cohort B, which included 48 individuals who were naturally infected and subsequently vaccinated with at least one dose of the SARS-CoV-2 vaccines, 62% were female (n = 30), with a median age of 43 years, and the overall cohort ranging from 19 to 76 years old (Table 1). Samples were confirmed positive for SARS-CoV-2 via PCR (83%, n = 45) and/or a serology assay (100%, n = 48). Individuals in cohort B were vaccinated with either AstraZeneca’s Vaxzervria vaccine (4%, n = 2), Moderna’s SpikeVax vaccine (8%, n = 4), or Pfizer/BioNTech’s Comirnaty vaccine (88%, n = 42). Most samples were collected after one dose of the vaccine (83%, n = 48), and 17% were collected after two doses of the vaccine (n = 8). Samples were collected 4 to 91 days post-vaccination with a median of 24 days in cohort B (Table 1).

Cohort C, which was the naïve control sample, was 43% female (n = 9) with a median age of 16 years and an age range of 3 months to 70 years (Table 1).

Serum samples collected from cohort A (natural infection) at different time points were tested to determine the anti-Spike IgG titer ratio using the commercial EUROIMMUN IgG assay. During the early acute phase (0 to 14 days PSO) (n = 74), 57% (n = 42) of the samples had IgG titer ratios that were higher than baseline, meaning these samples were positive for anti-Spike IgGs in the serum but in varying amounts based on the EUROIMMUN titer ratio (Fig. 1A and C). From days 7 to 14, there was a sharp rise in spike-specific IgGs, and IgG concentrations continued to rise until the end of the acute phase of the infection (n = 36) (27 days after the onset of symptoms), where 92% (n = 33) of the samples were above baseline. During the acute phase of the infection, a wide range of EUROIMMUN IgG titer ratios was observed for the different samples, and for one sample, anti-Spike IgGs reached an EUROIMMUN ratio of 12 on day 24. The anti-Spike IgG levels were sustained until the end of the early convalescent phase (3 months PSO). During the early convalescent phase (n = 69), 88% of the samples (n = 61) were above baseline (Fig. 1A and C). The highest level of anti-Spike IgGs was observed in this cohort on day 55 with an EUROIMMUN titer of 12.5. Three months PSO (n = 141 samples), anti-Spike IgG titers began to wane (Fig. 1A and C), where only 62% of the samples (n = 88) were above baseline. Samples in the late convalescent phase seemed to have lower EUROIMMUN IgG titers with a smaller range among the samples. From the subset of cohort A who had multiple visits, the longitudinal kinetics of the anti-Spike IgG response were similar to the kinetics observed when looking at independent values cross-sectionally (Fig. S1A).

We next looked at the neutralizing capacity of serum from cohort A using a pseudovirus (PsV) neutralization assay against the wild-type SARS-CoV-2 virus. Serum samples from naturally infected patients displayed a broad range of ID₅₀ values ranging from >0.02 (non-neutralizing) to 5.016 × 10⁻⁵ µg/mL against the wild-type PsV. The neutralization status of all samples was divided into one of three groups: non-neutralizers (with an ID₅₀ value of >0.02 µg/mL), weak neutralizers (with an ID₅₀ value of 0.01 µg/mL), and neutralizers (with ID₅₀ values determined by the assay; see Materials and Methods for criteria used for neutralization status classification). A subset of the samples (n = 26, 35%) strongly neutralized the wild-type PsV during the early acute phase where ID₅₀ values spanned three log units from 0.00436 to 9.27 × 10⁻⁵ µg/mL, while 17% (n = 13) were weak neutralizers (Fig. 1B and D). During the late acute phase (n = 36), 61% (n = 22) were neutralizers with ID₅₀ values ranging from 0.00426 to 5.02 × 10⁻⁵ µg/mL, and 22% (n = 8) were weak neutralizers. The top neutralizer of cohort A was observed during the late acute phase of the infection on day 26 with an ID₅₀ value of 5.02 × 10⁻⁵ µg/mL. During the early convalescent (n = 69) and the late convalescent (n = 141) phases of the infection, samples above baseline decreased where only 30% (n = 21) and 7% (n = 10) were neutralizers, respectively, and 20% (n = 14) and 16% (n = 23) were weak neutralizers, respectively. Looking longitudinally at the 46 individuals who had multiple visits during the study (n = 136 samples), 24 individuals developed neutralizing antibodies at some point during the infection, 10 individuals developed a weak neutralizing antibody response, while 12 individuals did not develop a neutralizing antibody response during any of their tested visits (Fig. S1B). Neutralizing antibody kinetics of the longitudinal subset of cohort A were similar to the kinetics observed at the population level, confirmed a progressive decrease, and eventually sustained levels or no presence of neutralizing antibodies during time points sampled after the acute phase and the early convalescent phase of the infection (Fig. 1B and D; Fig. S1B).

For the top neutralizing samples in the natural infection cohort A (n = 29), antibodies present in the serum bound to both the entire wild-type spike protein (Fig. 1E) and the RBD (Fig. 1F), showing that antibodies were mounted against the whole spike protein and the RBD of the virus during natural infection. In this cohort, neutralization was positively correlated with the IgG titers with an R² value of 0.52 (Fig. S2A).

Several factors including disease severity can influence the strength and breadth of the humoral response (17–21). Many studies have suggested that the humoral response differs in patients with distinct clinical disease progression and infection kinetics (18, 22–28). In our study, we considered different disease severities of COVID-19 as a co-variate for S1 subunit-specific IgG titers (Fig. 2A) or neutralizing antibody titers (Fig. 2B).

In weeks 1 and 2 post-symptom onset, a wide range of EUROIMMUN IgG titer ratios was observed in hospitalized patients with mild and severe disease (Fig. 2A). Samples for patients with mild symptoms or who were deceased revealed low or no anti-S1 IgGs in the serum (Fig. 2A). Notably, a significant difference was seen in anti-S1 IgGs between the hospitalized (mild disease) samples and the deceased patient samples (P = 0.0388). Additionally, a significant difference in anti-S1 IgGs was observed between the hospitalized (severe disease) patient samples and the deceased patient samples (P = 0.0029) (Fig. 2A). Looking at the ability of the mounted antibodies to neutralize the virus in weeks 1 and 2 PSO, samples from the different disease severity groups exhibited variable neutralizing activity. Higher neutralizing titers were seen in patients who were hospitalized with either mild or severe disease with a median of 0.01 µg/mL (Fig. 2B). In comparison, the majority of mild-symptom patients or deceased patient samples showed no neutralizing antibody activity in weeks 1 and 2 post-symptom onset (Fig. 2B). A significant difference in neutralizing antibody titers was observed between the hospitalized (severe disease) samples and the deceased samples (P = 0.0443).

During weeks 3 and 4, hospitalized patients with severe disease and hospitalized patients with mild disease mounted robust S1-IgG-specific antibodies, while those with mild symptoms demonstrated a lower abundance of S1-IgG-specific antibodies (Fig. 2A). Deceased patient samples at weeks 3 and 4 PSO showed higher levels of S1-IgG-specific antibodies compared to samples from deceased patients tested at weeks 1 and 2 PSO and all samples during this timepoint (Fig. 2A). Samples from deceased patients seemed to have only mounted anti-S1 IgGs 3–4 weeks PSO, which was later compared to other groups where anti-S1 IgGs were mounted earlier during weeks 1–2 PSO. A significant difference was seen between mild-symptom samples and hospitalized (severe disease) patients with anti-S1 IgGs (P = 0.0280). Assessing antibody neutralization titers during weeks 3–4 PSO, there were higher neutralizing antibody titers in each group during this time-point compared to each group assessed during weeks 1–2. Here again, the trend follows for neutralizing antibody titers where the hospitalized (mild and severe disease) and deceased patient samples had higher neutralizing antibody titers than the mild-symptom patients with a clinical severity score of 1 or 2 (Fig. 2B).

Greater than 4 weeks PSO, a significant difference in EUROIMMUN anti-S1 IgG titers was seen between the mild-symptom samples and hospitalized (mild disease) samples (P = 0.0193). A wide range of anti-S1 IgG titers was observed for mild-symptom samples, while hospitalized (mild disease) samples had EUROIMMUN titer ratios between 5 and 10. Similarly, when assessing the neutralizing antibody response in patients greater than 4 weeks PSO, neutralizing antibody titers were higher in all groups other than the mild-symptom patients, and this was consistently true at all the other time points tested. Additionally, overall neutralizing antibody titers in the greater than 4 weeks PSO had lower median values among the different clinical severity categories compared to the other time points. A significant difference was observed between mild-symptom patients and hospitalized patients with severe symptoms greater than 4 weeks post-symptom onset (P = 0.0193). These results substantiate the observation that patients experiencing more severe disease are more likely to develop a highly neutralizing antibody response.

With the development and approval of SARS-CoV-2 vaccines, it was pertinent to understand the antibody immune response elicited by vaccination, particularly in the context of previous infection. Therefore, we evaluated the anti-S1 IgG response and neutralizing antibody titers in cohort B – hybrid immunity (naturally infected and then subsequently vaccinated). All hybrid patient samples were positive for anti-S1 IgGs in the sera up to 127 days post-vaccination, and a wide range of anti-S1 IgG titers was observed (Fig. 3A). In the same cohort, individuals who had two doses of the vaccine showed stable levels of anti-S1 IgGs (Fig. 3A). Hybrid-immunity samples strongly neutralized the wild-type (D614G) spike protein, except for two cases that showed no neutralization and another two cases where they were classified as a weak neutralizer (Fig. 3B). Hybrid samples that mounted no or weak neutralizing antibodies may be, in part, due to the early timepoint at which the samples were tested post-immunization; indeed, the two weak neutralizers were taken 4 to 12 days post-vaccination, while the two non-neutralizing cases were from patient samples taken 28 days post-vaccination. The second dose of the vaccine did not seem to increase neutralizing antibody titers but instead helped sustain the neutralizing antibody titers at similar levels as the first dose of the vaccine. Testing the top neutralizing samples in this cohort (n = 14), antibodies present in the serum bound to both the entire wild-type spike protein (Fig. 3C) and the RBD of the virus (Fig. 3D). Additionally, in this cohort, the titers of neutralizing antibodies were positively correlated with the IgG titers with an R² value of 0.62 (Fig. S2A).

In the subset of patients who enrolled in our study after natural infection and followed up with a subsequent visit after vaccination, 79% of individuals had detectable levels of anti-S1 IgGs before vaccination, and the remainder (7/34 or 21%) tested negative for anti-S1 IgG antibodies before vaccination; the latter ranged from early acute to late convalescent phase (Fig. 4A). Correspondingly, only 24% of these samples were positive for neutralizing antibodies against the wild-type D614G strain, and 18% mounted weak neutralizing antibodies(Fig. 4B).

The antibody responses, both anti-S1 IgG titers and neutralizing antibody titers, significantly increased after the first dose of the vaccine. Indeed, a significant difference was observed in S1-binding IgGs among the late acute (P = 0.0002), early convalescent (P = 0.0002), and late convalescent phases of the infection (P = 0.0003), compared to samples derived after the first dose of a SARS-CoV-2 vaccine (Fig. 4A). No additional increase or a significant difference in antibody levels was detected after a second dose of the vaccine (Fig. 4A). The same trend was observed when assessing the neutralizing antibody response where individuals had a robust neutralizing antibody response after primary immunization, with no further increase in neutralization titers against the wild-type D614G strain after the second dose of the vaccine (Fig. 4B). All patient samples in this subset were positive for anti-S1 IgGs after vaccination, although among these, one person mounted weak neutralizing antibodies and two individuals did not mount neutralizing antibodies against the D614G pseudotyped virus after one dose of the vaccine (Fig. 4B), as described above. Furthermore, after two doses of the vaccine, all but one sample was able to induce neutralizing antibodies against the D614G pseudotyped virus. Like the anti-S1 IgG-binding data, a significant difference was observed in neutralizing antibodies among the late acute (P = 0.0176), early convalescent (P = 0.0005), and late convalescent phases of the infection (P < 0.0001), compared to the first dose of the SARS-CoV-2 vaccine (Fig. 4B). Additionally, there was no significant difference observed between the neutralizing antibodies mounted after the first and second doses of the SARS-CoV-2 vaccine.

The top 43 neutralizing patient serum samples against the SARS-CoV-2 ancestral wild-type (D614G) variant were tested against the Delta and Omicron BA.1 VOCs pseudotyped viruses. These two variants of concern were chosen to test our samples because these SARS-CoV-2 strains quickly dominated all other circulating strains worldwide during the time of this study. In this smaller subset of our cohort, n = 29 patient samples were derived from cohort A – natural infection only (7–360 days post-infection, with a median of 18 days), and the other n = 14 patient samples were derived from cohort B – patients with hybrid immunity (8–52 days post-vaccination, with a median of 17 days) (Table 1). All samples from cohort A neutralized the wild-type D614G PsV strain (Fig. 5A). Against the Delta PsV strain, there was a wider range of ID₅₀ values: 62% of the cohort was positive for neutralizing antibodies, 28% of the cohort was weak neutralizers, and 10% of the cohort was non-neutralizers (Fig. 5A). Of note, only 10% (n = 2) of cohort A was positive for Omicron BA.1-neutralizing antibodies, while the rest of the patient samples were unable to neutralize the Omicron BA.1 PsV strain (Fig. 5A). In contrast, in the hybrid cohort (cohort B), all samples neutralized the wild-type (D641G) and the Delta PsV strains, albeit with a wider range of ID₅₀ values against the Delta PsV compared to the wild-type PsV (Fig. 5B). Strikingly, all samples but one neutralized the Omicron BA.1 PsV strain (Fig. 5B). These data highlight how antibodies elicited by hybrid immunity are more resilient against different VOCs compared to neutralizing antibodies mounted from a single natural infection event.

Persons with a diagnosis of COVID-19 were prospectively enrolled and followed with a plan for serial sampling during the acute and convalescent phases of COVID-19. They were eligible for enrollment if COVID-19 was diagnosed by PCR testing or rapid antigen testing on nasopharyngeal swabs in the presence of respiratory or other symptoms consistent with COVID-19. Controls were collected from the same participant pool and were deemed naïve controls if they were symptom free, tested negative in either a PCR or a rapid antigen nasopharyngeal swab test, or were negative through serological testing. Samples were collected at The Hospital for Sick Children, Mount Sinai Hospital, and North York General Hospital in Toronto, Ontario, Canada.

Following the prospective enrollment of patients with confirmed COVID-19, medical records were reviewed to obtain pertinent clinical and related data. Collected data included date of birth, age at enrollment, sex, COVID-19 PCR status, date of symptom onset or date of positive PCR if asymptomatic and tested, clinical severity score, visit date, date of vaccines dose(s), and type of vaccine. After centrifugation for collection of serum, samples were stored at −80°C.

Inactivation of potential infectious viruses in serum was performed by heating the previously frozen samples to 60°C for 30 minutes. Heat-inactivated serum samples were then stored at −80°C until further use in downstream immunological assays.

FreeStyle 293-F Cells (FreeStyle 293-F cells, Thermo Fisher Scientific; HEK 293S, GnT I^-/- cells, American Type Culture Collection [ATCC]) were cultured in suspension in FreeStyle 293 Expression Medium (Gibco) and were maintained at 37°C in a Multitron Pro Shaker (Infors HT) with 8% CO₂ and 70% humidity, on a shaker platform rotating at 130 rpm.

For pseudovirus neutralization assays, human embryonic kidney (HEK) 293 cells containing the SV40 T-antigen (HEK293T) (ATCC) and HEK293T-ACE2 cells (BEI NR2511) were cultured in Dulbecco’s Modified Eagle’s Medium High Glucose (DMEM; Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Gibco), 2.5% HEPES (Gibco), and 0.5% gentamicin (Thermo Fisher) in a 37°C, 5% CO₂ incubator. All cells were negative for mycoplasma.

FreeStyle 293-F cells were split to a density of 0.8 × 10⁶ cells/mL at least 1 hour before transfection. Cells were transfected using FectoPRO Reagent (Polyplus) following the manufacturer’s instructions at a 1:1 DNA to FectoPRO ratio. Plasmid DNA (50 µg) encoding the SARS-CoV-2 Spike (S) protein or RBD was used per 200 mL of cell culture. Recombinant proteins were purified by affinity chromatography via a HisTrap Ni-NTA column (Cytiva) and were eluted using 20 mM Tris pH 8.0, 500 mM imidazole buffer. Subsequent size exclusion chromatography was performed in 20 mM Tris, pH 8.0, 150 mM NaCl on a Superdex 200 Increase column (Cytiva). All purified proteins were validated for integrity and purity via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and were stored at −80°C until use.

The ELISA procedure in this study was modified from a protocol published by Burn Aschner et al. (64). Briefly, Nickel-coated 96-well plates (Pierce, Thermo Fisher Scientific) were coated overnight with 50 µL per well of His-tagged SARS-CoV-2 RBD wild-type (D614G) protein or His-tagged SARS-CoV-2 wild-type (D614G) spike (S) protein in Tris-buffered saline (TBS) (20 mM Tris, pH 8.0, 150 mM NaCl) at 4°C. Plates were washed one time with TBS + 0.05% Tween 20 (TBS-T) and were subsequently blocked with TBS + 5% bovine serum albumin for 1 hour at room temperature. Plates were washed one time with TBS-T, incubated with serum diluted 1:50, and subsequently serially diluted 1 in 4 for 1 hour at room temperature. Plates were washed two times with TBS-T, and bound antibodies were detected using a horseradish peroxidase-conjugated anti-human Fab secondary antibody (Abcam, 87422) for 1 hour at room temperature. Plates were washed three times with TBS-T and were developed using 3,3′,5,5′-Tetramethylbenzidine (TMB) substrate reagent set (BD Biosciences) following the manufacturer’s instructions. Reactions were stopped by the addition of 2N HCl. Data were collected using a Synergy Neo2 Multi-Mode Assay Microplate Reader (Biotek Instruments).

The SARS-CoV-2 PsV production and neutralization assay procedure in this study was modified from a protocol published by Burn Aschner et al. (64). PsVs were produced using a lentiviral vector backbone, as previously described by Corman et al. (65). Briefly, SARS-CoV-2 PsVs were generated by transient co-transfection of HEK293T cells with a lentiviral backbone encoding luciferase reporter gene (BEI NR52516), plasmids encoding Gag-Pol (BEI NR52517), Tat (BEI NR52518), Rev (BEI NR52519), and a plasmid expressing SARS-CoV-2 spike protein (BEI NR52310) using the BioT transfection reagent (Bioland Scientific) according to the manufacturer’s directions. Transfected cells were incubated at 37°C for 24 hours, followed by the addition of 5 mM of sodium butyrate. Transfected cells were further incubated for an additional of 24 to 30 hours at 30°C. SARS-CoV-2 Spike mutant D614G was kindly provided by D. R. Burton (Scripps Research) and was used to generate the SARS-CoV-2 wild-type PsV. Delta (B.1.617.2) SARS-CoV-2 PsV variant was generated by substituting the wild-type spike plasmid with the respective VOC spike using plasmids kindly provided by David Ho (Columbia). Omicron (BA.5) was generated using a spike plasmid synthesized at GeneArt (Invitrogen). PsVs were harvested, filtered through 0.45 µm sterile filters, and concentrated using 100 K Amicon filters (Millipore Sigma).

Neutralization assays were performed using 293T-ACE2 cells (BEI NR52511) as previously described with a few modifications (66, 67). Briefly, 96-well plates were coated with poly-L-Lysine (Sigma-Aldrich), and cells were seeded at a density of 10,000 cells/well in 100 µL. The following day, serial dilutions of the heat-inactivated human sera or control IgG antibodies in duplicates were incubated with the PsV for 1 hour at 37°C. Cell culture medium was then replaced with the PsV-sera or PsV-IgG mixture supplemented with 10 µg/mL polybrene (Sigma-Aldrich). Cells were incubated for 48 hours at 37°C, before the addition of 50 µL of Britelite plus reagent (PerkinElmer) for 2 minutes. Supernatants were transferred to 96-well white plates (Sigma-Aldrich), and luminescence in relative light units (RLUs) was measured using a Synergy Neo2 Multi-Mode Assay Microplate Reader (Biotek Instruments). Absorbance data were converted to percent inhibition using the following formula:

Percent inhibition = (1 – [OD sample – OD min] / [OD max – OD min]) × 100%.

Percent inhibition was plotted, and data were fitted using the non-linear 4PL regression, constrained at top = 100% and bottom = 0% in GraphPad Prism version 10.02 to determine the ID₅₀ values of each sample. ID₅₀ values were assigned to samples based on three criteria:

For neutralizing samples, all three criteria are met, which means an ID₅₀ value could be confidently calculated for the sample. Samples are classified as weakly neutralizing if criteria 3 is met; however, criteria 1 or 2 is not met. An ID₅₀ value cannot be confidently assigned to this sample because the sample is still neutralizing at the lowest dilution used, thus we assigned an ID₅₀ value of 0.01 to these samples. Lastly, non-neutralizers are samples where all three criteria are not met; we assigned these samples an ID₅₀ value of >0.02 as the ID₅₀ is below the limit of detection for this assay. Plotted ID₅₀ values are the average of at least two to three biological replicates with two technical replicates each when performed.

Statistical analyses were performed using GraphPad Prism version 10.02 and R (Foundation for Statistical Computing, Vienna, version 4.2.2). We reported participant characteristics using median (with interquartile range [IQR]) for continuous variables, and counts (with percentage) for categorical variables. Wherever applicable, the normality of the data was determined by the Shapiro-Wilk test to select the most appropriate statistical test. For comparison of two groups, the Student’s t-test (normal distribution) or Mann-Whitney (non-normal distribution) was performed. For comparison of more than two groups, a one-way analysis of variance (ANOVA) (normal distribution) or a Kruskal-Wallis followed by Dunn’s multiple comparisons test (non-normal distribution) was performed. Locally estimated scatterplot smoothing (LOESS) non-parametric function was used to visualize trends in EUROIMMUN anti-S1 IgG levels and neutralizing antibody levels over days post-symptom onset. All LOESS analyses were completed using R (Version 4.2.2) within Rstudio (Version 2023.03.01+446) and were visualized using ggplot2. Statistical significances are indicated in the figures by asterisks as follows: *P < 0.05, **P < 0.01, ***P < 0.001, or ****P < 0.0001. Across all LOESS analyses, a span of 0.5 or 0.6 with a 95% confidence interval was applied. Correlations were assessed using the Pearson correlation test to assess the linear relationship between variables, specifically to look at the correlation between the EUROIMMUN anti-S1 IgG titer ratios and the neutralizing antibody titers (ID₅₀ values). The goodness of fit for linear regression analyses was reported as the coefficient of determination, R².