An analytical retrospective observational cross-sectional study of South African patients exposed to SARS-CoV-2 (n=186) was conducted. Patients were grouped into two categories: 1. previously SARS-CoV-2 infected with positive serology and no symptoms; 2. acutely SARS-CoV-2 infected, hospitalized for COVID-19 and severe symptoms. The demographic and clinical characteristics of all patients are presented in Table 1 .

Group 1 consisted of a total of 94 non-hospitalized HIV-infected adults with positive serology for SARS-CoV-2 presenting for routine HIV treatment at the Gugulethu Community Health Centre Antiretroviral clinic (Desmond Tutu Centre) between October 2020 and June 2021. The patient cohort has been described previously (22). Group 2 was comprised of 92 hospitalized HIV-positive and negative adult patients with RT-qPCR-proven SARS-CoV-2 infection recruited from Groote Schuur Hospital between June and August 2020 as described before (23). Patients’ recruitment coincided with the first wave (group 2) and second wave (group 1) of COVID-19 disease in Cape Town, South Africa. None of the patients were vaccinated against SARS-CoV-2.

The study was conducted according to the declaration of Helsinki, conformed to South African Good Clinical Practice guidelines. The recruitment of both patient groups was approved by the University of Cape Town’s Health Sciences Research Ethical Committee (HREC 134/2020 and HREC 207/2020). Written informed consent was obtained from all participants. For patients in group 2, relatives provided proxy consent for participants without capacity to consent for themselves.

The sample size was estimated with the calculation used for classical analytical cross-sectional studies in epidemiology: n = [DEFF * Np (1 – p)]/[(d ²/Z ² _{1- α /2}*(N-1) + p * (1 – p)], with DEFF (design effect), N (population size), p (anticipated frequency), d (confidence limits as % of 100) and Z (level of confidence according to the standard normal distribution). For this estimation the OpenEpi software (Open-Source Epidemiologic Statistics for Public Health) v. 3.01 from the US Centers for Disease Control and Prevention (CDC) (http://www.openepi.com/SampleSize/SSPropor.htm) was used. Since serological data from South Africa were lacking, the expected proportion (p) of patients previously infected with a common coronavirus was established as 3.7% according to the epidemiological study of acute respiratory virus infections detected among hospitalized patients in South Africa (18) who found the following acute HCoV prevalences: HCoV-229E (3.7%) and other HCoV (< 2%). Since 3.7% was the highest prevalence among the four human coronaviruses, this value was taken as reference for the sample size estimation. The study population (N) corresponded to 322444 people (i.e. confirmed COVID-19 cases in the Western Cape, South Africa as of June 30, 2021, https://sacoronavirus.co.za/2021/06/30/update-on-covid-19-30-june-2021/). For this study, the confidence interval (CI) of 95% was used and a total of 186 serums samples was analyzed.

Clinical data of the asymptomatic patients (group 1) was collected at enrollment, including any self-reported symptoms. Routine clinical analyses of peripheral blood samples taken at enrolment were performed by the National Health Laboratory Services (NHLS). Absolute CD4 count was determined using the Aquios PLG panel (CD45-FITC/CD4 PE monoclonal antibodies), while HIV-1 Viral Load (VL) was measured by the ALINITY mHIV-1 assay (Abbott Molecular Inc., Des Plaines, IL, USA). Full blood count and differential cell count were also performed. Previous exposure to SARS-CoV-2 was determined by in-house ELISA against SARS-CoV-2 IgG antibodies using RBD and S1 proteins as previously described (22). Only SARS-CoV-2 seropositive patient samples (n=94) were used in the present study.

Patients hospitalized with COVID-19 (group 2) were tested for SARS-CoV-2 by diagnostic RT-qPCR (Seegene, Roche or Gene Xpert) on the day of enrollment using nasopharyngeal or oropharyngeal aspirates. Full blood count, differential cell count and HIV-1 VL was also determined upon enrolment. All tests were performed by NHLS. Absolute CD4 count (for HIV-infected patients) and white cell counts (WCC) were obtained from patients’ medical files. Detailed clinical data not relevant to this study have been reported before (23). All data are reported using NHLS defined thresholds and ranges as previously determined to be applicable for the general population of South Africa. Clinical, demographic and experimental data were recorded and stored on an electronic REDCap database (24), hosted by the University of Cape Town.

To determine serology against the four common coronaviruses HCoV-OC43, HCoV-NL63, HCoV-HKU1 and HCoV-229E, in-house indirect ELISA assays were developed, which were based on previously published methodologies (25, 26). Briefly, 3 μg/mL of recombinant N proteins (recombinant HCoV-OC43 N His-tag Protein (Bio-techne, USA), recombinant HCoV-NL63 N protein (ab270843) (Abcam, UK), recombinant HCoV-HKU1 N protein (MyBioSource, USA), and HCoV-229E N protein (MyBioSource, USA)) in a total volume of 100 μL per well was used for coating high binding 96-well plates (Thermo Fisher, USA), followed by incubation overnight at 4 °C. The following day, plates were washed three times using 1X Phosphate Buffered Saline (PBS) (Lonza BioWhittaker^®, Switzerland) with 0.05% Tween^® 20 (Sigma-Aldrich, USA), followed by one wash with 1X PBS. Plates were blocked with 100 μL of blocking solution per well containing 1X PBS, 0.05% Tween^® 20 and 10% Fetal Bovine Serum (FBS) (Thermo Fisher, USA) at room temperature for 1 h. Thereafter, the plates were washed three times (as before). Serum aliquots (previously stored at -20 °C) were thawed and diluted in 1X PBS containing 0.05% Tween^® 20 and 2.5% FBS. Based on optimization experiments, 1:200 serum dilutions were used for HCoV-229E ELISA assays, while for the other HCoV tests, 1:500 dilutions were used. 50 μL of serum dilution per well were added to the plates and incubated at room temperature for 1 h. After the serum incubation, the plates were washed five times (as before) and incubated at room temperature for 1 h with a goat anti-human IgG antibody HRP conjugate (Sigma-Aldrich, USA) at a dilution of 1:3000 in a solution containing 1X PBS, 0.05% Tween^® 20 and 2.5% FBS. Thereafter, the plates were washed five times (as before). To develop the plates, 50 μL of TMB Substrate Reagent Set (BD OptEIA™, USA) was added per well and incubated at room temperature in the dark for 30 min. The reaction was stopped with 50 μL of 2 N H₂SO₄ per well (Sigma-Aldrich, USA). Finally, the plates were read at 450 nm using a Multiskan™ FC Microplate Photometer (Thermo Fisher, USA). Six serum-free negative controls were included per HCoV assay. A cut-off for positivity for each HCoV was set at 2 SD above the mean OD of 12 pre-pandemic samples (27). Adjusted OD values were then normalized to cut-off which were set as 1.8 (HCoV-OC43), 1.85 (HCoV-NL63), 2.03 (HCoV-HKU1) and 1.89 (HCoV-229E).

All patient samples from both groups were selected based on previous positive SARS-CoV-2 tests. Selected group 1 patients (n=94) were reactive to SARS-CoV-2 S protein using an in-house ELISA as previously reported (22). Group 2 patients (n=92) were determined SARS-CoV-2 positive by diagnostic RT-qPCR (23). In order to confirm seropositivity against SARS-CoV-2 N protein in both groups, an in-house ELISA was applied (27). Briefly, high binding 96-well plates were coated with 50 μL of N protein (Cape Bio Pharms, Cape Town, South Africa) at a concentration of 2 μg/mL at 4 °C overnight. The following day, plates were washed five times using 1X PBS containing 0.1% Tween^® 20, and then incubated in blocking buffer (1% casein, Hammarsten bovine (Sigma-Aldrich, USA), 1% PBS-Tween) at room temperature for 1 h. Thereafter, the blocking buffer was discarded and 100 μL of 1:50 diluted serum samples (in 0.5% casein and 1% Tween^® 20 in PBS) were added to the plates and incubated at room temperature for 2 h. Plates were washed five times (as before) and incubated at room temperature for 1 h with a goat anti-human IgG (Fc specific) antibody (1:5000; IgG-HRP; Sigma-Aldrich, USA). To develop the plates, 100 μL of O-phenylenediamine dihydrochloride OPD (Sigma-Aldrich, USA) per well was added at room temperature for 12 min. The reaction was stopped with 50 μL 3 M hydrochloric acid (HCl). Plates were read at 490 nm using a GloMax^® microplate reader (Promega, USA). The cut-off for positivity was set at 2 SD above the mean OD of 12 pre-pandemic samples. Adjusted OD values were then normalized to cut-off which was set as 0.067.

The seroprevalence (P) of the individual HCoV-229E, HCoV-NL63, HCoV-OC43, HCoV-HKU1 as well as total HCoV, respectively, was expressed as percentage of the entire patient cohort (n = 186), using the following formula, with a 95% confidence interval (CI).

Graphical representations were performed in Prism (v 9.4.1; GraphPad Software Inc., San Diego, CA, USA; https://www.graphpad.com/scientific-software/prism/) and SPSS software version 29 (IBM Corp, New York, NY, USA; https://www.ibm.com/products/spss-statistics). Descriptive statistics was used to manage data associated with the clinical status due to COVID-19 in the two groups, as well as their seropositivity to HCoV. To determine whether seropositivity to HCoV-OC43, HCoV-NL63, HCoV-HKU1 and HCoV-229E was related to the severity or absence of clinical signs present in the two patient groups, the prevalence odds ratio (OR) was used as a risk association measure. OR was calculated through the binomial logistic regression model, with a 95% CI using the SPSS software. Continuous variables were transformed, where appropriate, to approximate normal distributions. The two groups (non-hospitalized versus hospitalized COVID-19 patients) were assessed for association with positive HCoV serology using Chi-square, Fisher’s exact, or Mann–Whitney U test, as appropriate. p-values are two-tailed and were considered significant if <0.05.

The clinical characteristics of the two patient groups according to COVID-19 severity are listed in Table 1 (group 1: asymptomatic with positive SARS-CoV-2 serology; group 2: hospitalized COVID-19 patients). Briefly, 55% of patients were women and 45% men, with a median age of 45 years (range: 20 - 85). Most of the patients (47%) were in the range between 41 and 60 years. Regarding HIV-1 status, 65% of patients were HIV-1 positive (100% in group 1 and 28% in group 2), 95% of whom were on antiretroviral therapy (ART). The median HIV-1 VL was 49 copies/mL (range 1 – 1648565). The median CD4 count was 144 cells/mL (range: 3 – 903). Neither HIV VL nor CD4 count differed between the HIV-infected patients among the two groups. However, immune cell parameters such as white cell count, lymphocyte and monocyte counts significantly differed between both groups indicating severe inflammatory disease in group 2 patients ( Table 1 ).

The overall anti-HCoV IgG seroprevalence in the entire patient cohort (n = 186) was 60.8% (n = 113; [95% CI: 53.7 – 67.8]), with 37.1% HCoV-NL63 (n = 69; [95% CI: 30.0 – 44.0]), 30.6% HCoV-229E (n = 57; [95% CI: 24 – 37.3]), 22.6% HCoV-HKU1 (n = 42, [95% CI: 16.6 – 28.6]), and 21.0% HCoV-OC43 (n = 39, [95% CI: 15.1 – 26.8]). 73 patients (39.2%) were found to be HCoV seronegative. The proportion of patients exposed to only one HCoV was 29.0% (n = 54), exposed to two HCoV: 15.6% (n = 29), exposed to three HCoV: 13.4% (n = 25) and exposed to all four HCoV: 2.7% (n = 5). In total, 59 coinfection events (31.7%) were observed. The majority of the HCoV seropositive patients (60.2%; 68/113) as well as the majority of coinfections (69.5%; 41/59) were found in group 1 (asymptomatic, non-hospitalized COVID-19 patients), Table 1 and Figure 1 . Interestingly, group 1 patients had a significantly higher prevalence of HCoV-NL63 serology compared to group 2 patients (n=63 in group 1 versus n=6 in group 2; p<0.0001, Table 1 ) of whom 40% occurred as single infection in group 1 and only 17% in group 2. While HCoV-229E seroprevalence was comparable in both groups, 0% occurred as single infection in group 1 but 52% were found as single-infection in group 2 ( Figure 1 ). Significantly more group 1 patients were infected with HCoV-HKU1 compared to group 2 patients (n=30 in group 1 versus n=12 in group 2, p=0.0027, Table 1 ); however, the coinfection prevalence of HCoV-HKU1 was comparable in both groups, as was the total and coinfection prevalence of HCoV-OC43 ( Figure 1 ).

All patients were previously determined to be SARS-CoV-2 positive by in-house ELISA against the S protein [group 1 (22)] or by diagnostic RT-qPCR [group 2 (23)], respectively. In order to directly compare reactivity against the N protein of all HCoV and SARS-CoV-2, all patient samples were re-tested for seropositivity against SARS-CoV-2 N protein by in-house ELISA. 94.1% (n= 175) were confirmed to be SARS-CoV-2 N protein seropositive ( Figure 2A ).

The optical density as a measure for relative quantification of N-specific IgG antibodies against HCoV and SARS-CoV-2 is represented in Figure 2 . Hospitalized COVID-19 patients (group 2) expectedly showed a significantly higher SARS-CoV-2 antibody titer compared to asymptomatic previously infected SARS-CoV-2 patients (group 1) ( Figure 2A ); at the same time, group 2 patients presented with a significantly lower titer for HCoV-NL63 ( Figure 2B ; Table 1 ), while all other tested HCoV showed no differences in the two patient groups ( Figures 2C–E ). The median OD for HCoV-NL63 in asymptomatic patients (group 1) was 1.90 (95% CI: 0.62 – 2.45) and 1.32 (95% CI: 0.30 – 2.01) in hospitalized patients (group 2) (p <0.0001). Since the cut-off for determining seropositivity was 1.85, a total of n=63 (67.0%) of group 1 patients was considered positive for HCoV-NL63, while only n=6 (6.6%) of group 2 patients were seropositive (p <0.0001), Figure 2B and Table 1 . When stratified by age group, there was a trend to higher HCoV-NL63 seroprevalence in the 20-40 age group compared to the other age groups, but this did not reach statistical significance ( Supplementary Table S1 ). Likewise, no significant differences in HCoV-NL63 titers were found between the different age groups when assessing the entire patient cohort or the individual groups ( Supplementary Figure S1 , Supplementary Table S1 ). As mentioned before, most cross-reactive responses in patients have been described to target SARS-CoV-2 N (13); we therefore cannot exclude that there might also be some cross-reactivity in our serological assay. However, we assume this to be minimal as the clear difference seen in SARS-CoV-2 serology between the two patient groups ( Figure 2A ) was not reflected in the individual HCoV assays ( Figures 2B–E ).

In order to assess whether HIV infection in group 2 patients might have had a confounding effect on the observed significantly lower HCoV-NL63 seroprevalence compared to group 1 patients ( Table 1 ), we excluded all HIV negative patients from the comparison ( Figure 3A ). Similar to our previous observations ( Figure 2B ), we found a significantly lower HCoV-NL63 seroprevalence in HIV-infected group 2 patients compared to all HIV-infected group 1 patients ( Figure 3A ), p <0.0001. Moreover, when all group 2 patients were stratified according to their HIV status, no statistically significant differences in HCoV-NL63 seroprevalence was detected ( Figure 3B ). While the majority of samples derived from group 2 patients displayed HCoV-NL63 antibody levels below cut-off, there was a tendency to higher titers in HIV positive individuals ( Figure 3B ) which was also reflected in the entire group 1 ( Figure 3A ). These results were also observed when the ART-negative patients from group 2 (n=6) were excluded from the analysis ( Supplementary Figure S2 ).

Given the relatively small sample size we cannot generally exclude that HIV infection affects HCoV-NL63 antibody levels; however, in the context of this study we found that HIV infection does not impact on the association of pre-exposure to HCoV-NL63 and COVID-19 severity.

Assessment of parameters that differed between group 1 (asymptomatic COVID-19 patients) and group 2 (hospitalized COVID-19 patients) indicated that male sex, age range between 40 - 61 years, age range between 61 - 85 years, HCoV status, HCoV coinfection frequency and white cell counts were statistically different between both groups ( Table 1 ); these parameters were therefore considered for logistic regression analysis. In patients exposed to SARS-CoV-2, previous HCoV-NL63 exposure was associated with reduced severity [ Table 2 , Model A, p <0.001, adjusted OR = 0.0230 (95% CI: 0.0046 – 0.1146)]. On the other hand, previous HCoV-229E exposure was associated with increased severity [ Table 2 , Model A, p = 0.0052, adjusted OR = 14.4211 (95% CI: 2.2218 – 93.6044)]. Additionally, age range between 40 - 61 years [ Table 2 , Model A, p <0.001, adjusted OR = 8.7738 (95% CI: 2.9782 – 25.8480)], and age range between 61 - 85 years [ Table 2 , Model A, p = 0.0015, adjusted OR = 13.6463 (95% CI: 2.7197 – 68.4704)], were as well associated with increased COVID-19 severity.

To avoid overfitting the model, variables that were not significant in model A were removed and a stripped-down logistic regression was run confirming that previous HCoV-NL63 exposure was associated with reduced COVID-19 severity (protection factor) [ Table 2 , Model B, p <0.001, adjusted OR = 0.0176 (95% CI: 0.0039 – 0.0786)], while previous HCoV-229E exposure [ Table 2 , Model B, p = 0.0051, adjusted OR = 7.3239 (95% CI: 1.8195–29.4800)], male sex [ Table 2 , Model B, p = 0.0467, adjusted OR = 2.5242 (95% CI: 1.0136 – 6.2862)], age range between 40 - 61 years [ Table 2 , Model B, p <0.001, adjusted OR = 7.2237 (95% CI: 2.7511 – 18.9680)] and age range between 61 - 85 years [ Table 2 , Model B, p = 0.0012, adjusted OR = 12.5092 (95% CI: 2.7022 – 57.9072)] were associated with increased COVID-19 severity (risk factors). The protective association of previous exposure with HCoV-NL63 also remained significant when only adjusted for age [ Table 2 , Model C, p <0.001, adjusted OR = 0.034 (95% CI: 0.012 – 0.095)]. Interestingly, when this logistic regression was performed for death outcome in group 2 patients (n=26, 28.3%), no significant associations were identified ( Supplementary Table S2 ). Figure 4 shows a summary of the association between previous HCoV exposure and other relevant demographic parameters with COVID-19 severity (Adjusted OR and 95% CI).