To assess the antibody response against SARS-CoV-2 Wuhan, Gamma, Mu, and Delta variants, we assembled two ambispective cohorts in Bogotá, Colombia, with adult donors from the Universidad de Los Andes, Hospital Universitario San Ignacio (HUSI) recruited by Pontificia Universidad Javeriana, and from Hospital Universitario Mayor-Méderi and Clínica del Occidente, these last two being recruited by Universidad del Rosario. Cohort 1 is composed of naturally infected individuals who underwent follow-up and antibody measurements at approximately 1- and 6 – 8 months post-symptoms onset (PSO) to determine the seropermanence of antibodies induced by natural infection. Cohort 2 is composed of vaccinated individuals, in whom measurements of antibodies were performed during the pre-vaccination, and at the early (15 – 30 days after the second dose), and late post-vaccination periods (6 – 8 months and 12 months after the second dose), to determine the seropermanence of antibodies induced by vaccination. A fraction of the individuals also received a third vaccine boost and a sample was also obtained 1-3 months after this dose. The CD4⁺ T cell response was evaluated in a subgroup of individuals belonging to cohort 2.

For serological assays, whole blood was collected in clot activator and gel separation tubes, and serum was collected and stored at -80°C. To evaluate the CD4⁺ T cell response, whole blood was collected in EDTA tubes, and the cellular fraction was separated from plasma by centrifugation at 10,000 rpm for 10 minutes. Next, peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation using Ficoll-Paque™ (Cytiva). Isolated PBMCs were cryopreserved in cell recovery media containing 90% fetal bovine serum (FBS) and 10% DMSO (ATCC) and stored in liquid nitrogen until cellular assays were performed.

To evaluate if an individual was naturally infected by SARS-CoV-2, an IgG test that recognizes the viral nucleoprotein was performed (09203095190 – ROCHE), using a Roche Cobas e analyzer, following the manufacturer’s protocol. The positive and negative controls were performed at the beginning of each lot of antibody testing for quality control. Samples with a cutoff index (COI) ≥ 1 were considered positive. This assay was performed in a commercial clinical laboratory.

To assess the presence of RBD-specific antibodies induced by natural infection or vaccination against SARS-CoV-2 Wuhan, Gamma, Mu, and Delta variants, a hemagglutination Test (HAT) that measures total antibodies (IgG, IgM, IgA) was employed (18, 19). O-negative red blood cells from a healthy volunteer were isolated from whole blood by density gradient centrifugation according to the manufacturer’s instructions. Red blood cells were preserved in Alsever’s solution (Sigma Aldrich) at 2°C for up to 30 days. After standardization, the HAT was performed using red blood cells from the same volunteer. For the HAT, codon-optimized IH4-RBD sequences of VOC and VOIs containing amino acid changes in the RBDs Gamma (K417T, E484K, N501Y), Mu (R346K E484K N501Y), and Delta (L452R, T478K) were expressed in Expi293F cells and purified by their c-terminal 6xHis tag using Ni-NTA chromatography (18, 19). The point HAT was performed in a V-bottomed 96-well plate. Serum samples were double diluted in duplicate from 1:40 in 50 µl of phosphate-buffered saline (PBS), up to 1:1,280. Samples with titers >1:1,280 were re-tittered until negative (1:10,240). Equal volumes of human O-negative red blood cells and 1 µg/mL of IH4-RBD of Wuhan, Gamma, Mu, and Delta variants were premixed, adding 100 µL per well, as described in our previous publication (20). Plates were incubated for 1 hour at room temperature to allow red blood cells to settle; then, the plate was tilted for 30 seconds, read, and photographed. Positive agglutination HAT titer (here reported as the reciprocal titer) was defined as the last dilution in which the teardrop did not form. Partial teardrops were considered negative (18). A representative example of HAT is shown in Supplementary Figure 1 . To validate the assay, we tested reference serum samples with different levels of antibodies against SARS-CoV-2, the serum 20/130 provided by WHO, as well as the monoclonal antibodies CR3022 and EY-6A (kindly provided by Dr. Alain Townsend), with expected results (20).

Serum samples were serially diluted in DMEM containing 5% FBS and combined with ~100 focus forming units (FFU) of SARS-CoV-2 variants Mu (BEI#NR-56225) and Gamma (BEI#NR-54982) and allowed to complex at 37°C and 5% CO2 for 1 hour in a 96-well round bottom plate. The antibody-virus complex was then added to each well of a 96-well flat bottom plate containing a monolayer of Vero-hACE2-TMPRSS2 cells. Following 1 hour of incubation at 37°C and 5% CO2, the cells were overlaid with 2% methylcellulose and returned to the incubator. After 24 hours of infection, the cells were fixed with 5% electron microscopy grade paraformaldehyde in PBS for 15 minutes at room temperature. The cells adherent to the plate were then rinsed with PBS and permeabilized with 0.05% Triton-X in PBS (FFA Wash Buffer). Foci of infected Vero-hACE2-TMPRSS2 cells were stained with anti-SARS-CoV-2 human mAb 2165 (Leinco) overnight at 4°C and washed 3 times with FFA Wash Buffer. Cells were then stained with horseradish peroxidase-conjugated goat anti-human IgG for 2 hours at room temperature. Cells were washed again with 0.05% FFA Wash Buffer prior to the addition of TrueBlue KPL peroxidase substrate, which allows the visualization of infected foci as blue spots. The foci were visualized and counted using an ImmunoSpot CTL Elispot plate reader. Percent infection was calculated by the number of foci per well/number of foci in negative control wells*100.

Spike mega pool (CD4S) sets consisting of 253 overlapping peptides of 15-mer with 10 amino acid overlap were synthesized to span the entire S protein of the Wuhan, Gamma, and Mu variants sequences, as previously reported (4). In addition, for the rest of the Wuhan variant proteome (CD4R), a mega pool of 221 dominant HLA class II predicted epitopes from the Immune Epitope Data Base (IEDB) was used to assess the complete response against the SARS-CoV-2 proteome, excluding the S protein.

The Activation-Induced Markers (AIM) assay has been previously described in detail (4). The PBMCs that were preserved in liquid nitrogen were thawed by diluting the cells in 10 mL of complete RPMI 1640 (Invitrogen) with 5% human AB serum (Gemini Bioproducts) in the presence of benzonase-nuclease (Merck Millipore). The PBMCs were cultured in the presence of SARS-CoV-2 Wuhan, Gamma, and Mu variants peptide mega pools (at 1 µg/mL) in 96 well U bottom plates, at 1x10⁶ PBMCs/200 µL/well. DMSO was used as a negative control, while Staphylococcal Enterotoxin B (SEB, at 4 µg/mL) was used as a positive control. Cells were stimulated for 24 hours at 37°C, and 5% CO₂. Upon incubation, cells were washed with PBS, and stained for 20 minutes at 4°C with the viability marker Live/Dead Aqua (Invitrogen, at 1/40 dilution). Next, the cells were washed with PBS, and incubated for 30 minutes at 4°C with fluorochrome-labeled antibodies against human CD19, CD14, CD3, CD4, CD8, CD134 (OX40), CD137, CD45RA, and CCR7 ( Supplementary Table 1 ). Next, the cells were washed with FACS buffer (PBS and 3% FBS) and fixed with paraformaldehyde 1%. Cells were acquired on an Aurora spectral flow cytometer (Cytek) and analyzed in SpectroFlo software (v.3.0). Gates for AIM⁺ cells (OX40⁺ CD137⁺) were set according to the negative and positive controls for each donor. A representative example of the gating strategy is shown in Supplementary Figure 2 . For the resulting AIM data, the background of AIM⁺ cells in the DMSO condition of each sample was removed, with a minimal DMSO level set to 0.005%. A positive response was defined as a stimulation index (SI) >2, and above 2 times the standard deviation of DMSO values (6). In SEB-stimulated cells from all the samples included the geometric mean (95% CI) was 3.89% (3.38-4.47%).

Data were analyzed and plotted in GraphPad Prism software, version 9.3.1. In all the cases, the geometric mean and 95% confidence intervals (CI) are shown. To determine statistical differences between the groups studied, non-parametric tests were performed. The Mann-Whitney and Wilcoxon tests were used for the comparison of two unpaired and paired groups, respectively. The Kruskal-Wallis and Friedman tests were used for the comparison of three or more unpaired or paired groups, respectively. The Dunn test was applied to correct for multiple comparisons. In all the cases, a p-value <0.05 was considered as significant.

In this study, 549 participants were enrolled and classified into naturally infected ( Table 1 ) and vaccinated cohorts ( Table 2 ), according to medical records and study surveys. Naturally infected donors (n=113) included adults with a history of mild (Javeriana and Andes participants from cohort 1) and severe hospitalized (Rosario participants from cohort 1) SARS-CoV-2 infection confirmed by PCR and/or serologic tests. Samples from these individuals were collected at 1- and 6-month PSO ( Table 1 ). The vaccinated cohort has donors with and without a history of natural SARS-CoV-2 infection. The part of the cohort recruited by Javeriana (HUSI from cohort 2) had individuals vaccinated with two doses of the BNT162b2 vaccine, in whom samples were collected at pre-vaccination, at 15-30 days, and 6-8 months after the second vaccine dose ( Table 2 ). The part of cohort 2 recruited by El Rosario- consisted of individuals who, in addition to two doses of the BNT162b2 vaccine, also received a third dose of mRNA-1273, ChAdOx1 nCoV-19, or Coronavac vaccines (approximately 10 months after the second vaccine dose). Samples from these donors were collected 15-30 days after the second vaccine dose, and 1-3 months after the third dose ( Table 2 ). A fraction of these individuals also had pre-vaccination samples ( Table 2 ). Nucleoprotein-specific IgG antibodies were assessed to determine asymptomatic and BTI in the vaccination cohort, as previously reported (17). The characteristics of the subgroup of individuals belonging to cohort 2, to whom the CD4⁺ T cell response was evaluated, are summarized in Supplementary Table 2 .

We first evaluated the durability of RBD-specific antibodies against the Wuhan, Gamma, Mu, and Delta variants in naturally infected individuals, according to disease severity. One-month PSO in individuals with mild ( Figures 1A, B ) or severe disease ( Figure 1C ), the antibody response against the Wuhan variant was predominant. However, most of the participants also showed RBD recognition for the viral variants, indicative of a cross-reactive antibody response. Interestingly, while RBD-specific antibody titers decreased (3.3 to 6.3-fold) at 6 months PSO in donors who developed mild disease ( Figures 1A, B ), the individuals who suffered severe disease maintained this response over time ( Figure 1C ), and indeed, antibody titers for the Wuhan variant tended to increase at 6 months PSO in them ( Figure 1C ). Our results showed that mild COVID-19 subjects reduced the RBD-specific antibody levels at 6 months PSO compared to 1-month PSO, but subjects who presented severe COVID-19 maintained antibody levels, suggesting that the durability of anti-RBD antibodies upon natural infection depends on disease severity.

In Colombia, in 2020, the circulation of the Gamma, Mu, and Delta variants was absent or low, and then, during 2021 these variants circulated, with Mu predominating during the third wave of the pandemic (21). Thus, we aimed to further evaluate the anti-RBD antibody cross-recognition in naturally-infected individuals during these two periods using the HAT. Despite the low or absent circulation of these viral variants (Gamma, Mu, and Delta) in the first period, there was antibody cross-recognition of all variants, with Delta exhibiting the highest recognition. Notably, Mu exhibited a significantly higher antibody cross-recognition than the Gamma variant ( Figure 2A ). During the period of active circulation of Gamma and Mu, the pattern seemed similar, although differences between Gamma and Mu did not differ statistically ( Figure 2B ). We were also able to evaluate the RBD antibody binding in serum from an individual with a confirmed infection by the Gamma variant, obtained 1-month PSO. Expectedly, we obtained high antibody titers against Gamma; although there was some level of cross-recognition for the Wuhan, Mu, and Delta variants, the antibody titers for these variants were lower than for Gamma ( Supplementary Figure 3 ). These results further confirm the reproducibility and specificity of the HAT, as well as the cross-reactive nature of anti-RBD antibodies following natural infection with SARS-CoV-2. Moreover, these results suggest that natural infection with heterologous viruses elicits binding antibodies that are more cross-reactive for the RBD of the Mu than for the RBD of the Gamma variant.

Importantly, while we observed a higher antibody binding to Mu than to Gamma ( Figures 2A, B ), previous studies have shown that the Mu variant has higher antibody evasion than Gamma when evaluated by neutralizing assays (12). Thus, to extend the findings obtained by the HAT, we evaluated the neutralizing capacity of antibodies against the Gamma and Mu variants in serum samples obtained at 1-month PSO from SARS-CoV-2 naturally-infected individuals ( Figure 2C ). Interestingly, despite the higher antibody binding for the Mu variant ( Figures 2A, B ), the antibodies against Mu had a lower neutralizing capacity relative to Gamma ( Figure 2C ). Moreover, we observed a weak positive correlation between the neutralizing and binding antibodies for both variants (rho=0.30, p-value=0.02 for Gamma; and rho=0.30, p-value=0.03 for Mu). Thus, there is a weak correlation and a different pattern of response between the binding and neutralizing antibodies against the Mu and Gamma variants in Colombian individuals with natural SARS-CoV-2 infection.

We also evaluated antibody levels before and after vaccination. A first analysis including 57 individuals from cohort 2 recruited by the Javeriana, indicated that vaccination induced an increase in antibody titers at 15 – 30 days after receiving the second dose of the BNT162b2 vaccine, but they decreased at 6 months post-vaccination, showing at this time point similar titers to the pre-vaccination period ( Supplementary Figure 4A ). On the other hand, 38 individuals from cohort 2 recruited by El Rosario who received a third dose of the vaccine exhibited high antibody titers 1-3 months after the boost, which were comparable to those observed 15-30 days after the second vaccine dose ( Supplementary Figure 4B ). These results are consistent with a transitory boosting effect on the humoral response by sequential immunizations against SARS-CoV-2.

Next, we evaluated the antibody response pattern according to the antecedent of SARS-CoV-2 natural infection before receiving the first vaccine dose ( Figure 3 ). As expected, individuals with a preceding infection had detectable antibody titers that cross-recognized all SARS-CoV-2 variants before vaccination, while most individuals without a preceding infection had undetectable anti-RBD antibodies ( Figure 3A ). Of note, in a few individuals that did not have a history of infection before vaccination and had a negative IgG nucleoprotein test, we could detect low levels of HAT antibodies against the viral variants RBD ( Figure 3A ), which may be indicative of a previous infection that was not detected by the nucleoprotein IgG assay, or of potential cross-reactivity with other HCoVs.

One month after the second vaccine dose, individuals with a history of infection generated higher levels of anti-RBD antibodies relative to those without a history of infection ( Figure 3B ). At 6 months after the second dose, antibody titers in individuals without a history of infection returned to almost undetectable levels, but those with previous infection were still able to maintain antibody titers comparable to those observed 1 month after the second dose ( Figure 3C ). There were no differences in antibody titers after a third vaccine dose when comparing those with and without pre-vaccination infection ( Figure 3D ). These results indicate that the durability of the anti-RBD response to vaccination depends on the history of pre-vaccination SARS-CoV-2 infection and a vaccine boost. Finally, as seen after natural infection, a higher cross-recognition of the Mu versus the Gamma variant was observed after vaccination ( Figures 3B–D ).

To determine whether vaccine-elicited antibody levels measured 15-30 days after the second vaccine dose are associated with protection against SARS-CoV-2 infection, we classified individuals according to the occurrence or not of a BTI up to 6-8 months after vaccination. BTI that occurred after vaccination were identified based on clinical records of mild infections and seroconversion in SARS-CoV-2 nucleoprotein-specific IgG between samples taken 15-30 days and 6-8 months after the second vaccine dose. As shown in Figure 4 , individuals with BTIs had lower antibody titers against the Gamma, Mu, and Delta variants in comparison with individuals without BTI. No statistical differences were observed for antibodies against the Wuhan variant ( Figure 4 ). These results indicate that the levels of RBD-specific antibodies post-immunization are associated with protection against SARS-CoV-2 BTI.

We next evaluated the CD4⁺ T cell response in individuals before and after receiving the BNT162b2 vaccine. Using the AIM assay, we measured the frequency of SARS-CoV-2-specific CD4⁺ T cells (defined as OX40⁺ CD137⁺ cells) for S peptides of the Wuhan, Gamma, and Mu variants, in 20 donors pre-vaccination, and 16 of them 1 - 2 months after the second vaccine dose ( Figure 5 ). In the pre-vaccination group, we included four samples of individuals (represented with uncolored squares) obtained before the onset of the pandemic ( Figure 5 ). Of note, in the cohort of individuals in whom we performed cellular analyses, 13 out of 20 donors exhibited an antibody response against the Wuhan, Gamma, or Mu variants, indicative of previous coronavirus exposure.

We first performed an analysis with all the donors included. As expected, the frequencies of SARS-CoV-2-specific CD4⁺ T cells were low during the pre-vaccination period, while the frequencies of S-specific, but not R-specific CD4⁺ T cells increased upon vaccination ( Figure 5A ). Importantly, the CD4⁺ T cell response cross-recognized peptides from the Gamma and Mu variants ( Figure 5A ). We next assessed the impact of natural infection before vaccination on the dynamics of the SARS-CoV-2-specific CD4⁺ T cell response. We classified the individuals as those with or without a history of natural infection, based on the detection of nucleoprotein-specific IgG antibodies in samples before and after vaccination. Of note, we observed detectable SARS-CoV-2-specific CD4⁺ T cells for the Wuhan S and R proteins in the pre-vaccination samples of some individuals without a history of natural infection and in some pre-pandemic samples ( Figure 5B ). This result might be explained by undetected asymptomatic SARS-CoV-2 infections and/or cross-reactive responses to previous infections with HCoVs. Interestingly, while unexposed individuals exhibited an increase in the frequencies of Wuhan, Gamma, and Mu S-specific CD4⁺ T cells upon vaccination ( Figure 5B ), this was not the case for individuals with a history of natural infection, since they exhibited high frequencies of Wuhan, Gamma, and Mu S-specific CD4⁺ T cells during the pre-vaccination period, and only had a modest increase in the frequency of Wuhan S-specific CD4⁺ T cells upon vaccination ( Figure 5C ). These results indicate that the dynamic of the SARS-CoV-2-specific CD4⁺ T cell response is dependent on the history of SARS-CoV-2 exposure.

To assess the quality of the SARS-CoV-2-specific CD4⁺ T cell response, we analyzed the memory phenotype (CD45RA and CCR7) of AIM⁺ (OX40⁺ CD137⁺) CD4⁺ T cells in response to peptide stimulation. The analysis of all the individuals included showed that peptide-responding cells at the pre-vaccination assessment had predominantly a central memory phenotype (T_CM; CD45RA^- CCR7⁺), followed by an effector memory phenotype (T_EM; CD45RA^- CCR7^-), and these profiles were higher than those found in bulk CD4⁺ T cells ( Supplementary Figure 5A ). This pattern was similar for cells specific for S-derived peptides from Wuhan, Gamma, and Mu variants, as well as for Wuhan R-derived peptides ( Supplementary Figure 5A ). Upon vaccination, SARS-CoV-2-specific CD4⁺ T cells were enriched in a T_EM and T_CM phenotype relative to bulk CD4⁺ T cells, but the T_CM profile (suggestive of long-lived responses) predominated along CD4⁺ T cells specific for the Wuhan, Gamma, and Mu variants ( Supplementary Figure 5B ). Low frequencies of naïve/T stem cell memory (CD45RA⁺ CCR7⁺) and T_EMRA (CD45RA⁺ CCR7^-) SARS-CoV-2-specific CD4⁺ T cells were observed at both time points ( Supplementary Figure 5 ).

We observed that during the pre-vaccination period, some individuals without a history of natural infection had detectable SARS-CoV-2-specific CD4⁺ T cells ( Figure 5B ). Thus, we aimed to evaluate the phenotype of these potentially SARS-CoV-2 cross-reactive CD4⁺ T cells. Interestingly, SARS-CoV-2-specific CD4⁺ T cells, including those specific for R-derived peptides, in individuals without a history of natural infection were enriched in a T_EM phenotype, when compared with bulk CD4⁺ T cells, while a high proportion also exhibited a T_CM phenotype ( Figure 6A ). These data support that these cross-reactive cells are part of the memory compartment. On the other hand, SARS-CoV-2-specific CD4⁺ T cells at the pre-vaccination time point from donors with a history of natural infection were enriched in a T_CM phenotype, further supporting the notion that they are long-lived memory cells ( Figure 6B ). Finally, we evaluated the changes in the memory phenotype of SARS-CoV-2-specific CD4⁺ T cells upon vaccination. Interestingly, we noted that in individuals without a history of SARS-CoV-2 infection, the phenotype of SARS-CoV-2-specific CD4⁺ T cells tended to augment in a T_CM profile after vaccination, relative to pre-vaccination values ( Figures 6A, C ). On the other hand, in individuals with a history of infection, the proportion of T_EM among SARS-CoV-2-specific CD4⁺ T cells tended to increase upon vaccination ( Figures 6B, D ), and this increment was statistically significant for S Wuhan-specific cells (p-value=0.03). These data indicate that the number of exposures to the viral antigens induces changes in the memory profile of SARS-CoV-2-specific CD4⁺ T cells.