To determine whether humoral cross-reactivity against SARS-CoV-2 and HCoV could be observed during COVID-19, we sequentially analysed the sera of eight severe COVID-19 patients ( Supplementary Table 1 ) for their IgG reactivity against SARS-CoV-2 and HCoVs ( Figure 1 and Supplementary Figure 1 ). We found that IgG reactivity against the S2 domain of SARS-CoV-2 Spike protein preceded that against S1 and/or Receptor Binding Domain (RBD). This was also followed by a parallel increase in IgG antibody titres directed towards other SARS-CoV-2 antigens and beta-coronaviruses HCoV-OC-43 and HCoV-HK-U1. However, we did not detect an increase in responses to alpha-HCoV NL-63 or 229-E. The rapid IgG responses to common beta-HCoVs identified were most likely due to cross-reactivity and not as a result of ongoing infections with other HCoVs because their nasopharyngeal RT-PCR was specifically positive for SARS-CoV-2 and negative for all other HCoVs ( Supplementary Tables 1 and 2 ). Finally, early sera from all patients showed reactivity against HCoV-OC-43 and HCoV-HK-U1 as they were already present before the appearance of anti-SARS-CoV-2 antibodies ( Figure 1 and Supplementary Figure 1 ).

We postulated that the appearance of anti-SARS-CoV-2 IgG before IgM could be due to their cross-reactivity against beta-HCoVs. To test this, we retrospectively analysed the titres of anti-SARS-CoV-2 IgM, IgG and IgA, as well as anti-HCoV IgG in the earliest available sera (mean day from symptoms onset: 10.6 days) from 167 patients with COVID-19. Amongst them, 41 had mild COVID-19 that did neither required hospitalisation nor oxygen therapy by nasal cannula. 62 had severe COVID-19 requiring hospitalisation with ward-based oxygen therapy only whereas the remaining 64 patients required admission to an intensive care unit ( Supplementary Table 3 ). As demonstrated by the heatmap in Figure 2A , the collected sera from all patients confirmed a pattern of high titres of anti-SARS-CoV-2 IgG antibodies, either recognizing the Full Spike, S1, S2 and RBD spike domains or NC in severe and critical COVID-19 patients (10). A strong correlation was also observed between the anti-HCoV-OC-43 and anti-HCoV-HK-U1 IgG responses and the serum levels of anti-SARS-CoV-2 IgG antibodies, in particular those directed against the S2 domain of the SARS-CoV-2 spike protein (r > 0.7, p <0.0001). In contrast, we did not identify any correlation between the IgG responses to SARS-CoV-2 antigens, to HCoV-NL-63 (r=0.05, p-value= 0.051) or HCoV-229-E (r=0.19, p-value = 0.01).

Unlike the IgG responses, there was a poor correlation between IgM and IgA responses to SARS-CoV-2 and those to common alpha- or beta-HCoVs (r<0.27). In comparison, antibody responses to beta-HCoV OC-43 and HK-U1 strongly correlated regardless of the isotype ( Figure 2B and Supplementary Figure 2 ). Based on these data, we concluded that the IgG responses to SARS-CoV-2 antigens strongly correlated with those to beta-HCoV -OC-43 and -HKU-1, but not common cold alpha-HCoV -NL-63 and -229-E.

It was widely expected that during the COVID-19 outbreak, all patients would develop a primary type antibody response, characterized by the production of anti-SARS-CoV-2 IgM antibodies prior to anti-SARS-CoV-2 IgG and/or IgA antibody seroconversion. However, antibody responses to SARS-CoV-2 in COVID-19 patients were in fact heterogeneous. We observed patients harbouring an early IgG response in the absence of detectable IgM response. In comparison, some had early IgM, but no IgG responses, or even both ( Figure 2A ).

To analyse the role of such pre-existing immunity to HCoVs in the heterogeneous humoral response to SARS-CoV-2 antigens, we stratified all patients based on the levels of circulating anti-SARS-CoV-2 IgM/IgG (but not IgA) in relation to the timing of blood testing after the clinical onset of disease ( Figure 3A and Supplementary Figure 3 ). A first subset of patients could be defined as the IgM⁺/IgG^- group, characterised by the presence of circulating anti-SARS-CoV-2 IgM, but not IgG antibodies ( Figure 3A ). The titres of anti-HCoV-OC-43 and anti-HCoV-HKU-1 IgG in this group were significantly lower, compared to those in the IgM^-/IgG⁺⁺ and IgM⁺/IgG⁺⁺ groups ( Figure 3C and Supplementary Figure 4A ). Put together, this suggested that patients with a predominant IgM response to SARS-CoV-2 antigens would not have encountered HCoV-OC-43 or HcoV-HKU-1 prior to the outbreak of SARS-CoV-2.

A further group of patients with no detectable serum IgG and IgM antibodies was defined as the IgM^-/IgG^- group. Most of these patients were tested before day 12 after symptoms onset ( Figures 3A, B ). It can be assumed that these patients, due to the lack of an early recall-type IgG response, would eventually develop a primary IgM response. This “non-recall-type response” group also had low titres of anti-HCoV-OC-43 and HCoV-OKU-1 IgG antibodies ( Figure 3C and Supplementary Figure 4A ).

Additional groups of patients with weak or strong IgM responses as well as concomitant elevated IgG responses could be defined as IgM⁺/IgG⁺⁺ and IgM⁺⁺/IgG⁺⁺ respectively. It is already known that simultaneous production of IgM and IgG antibodies can be observed during primary responses on one hand and recall antibody responses on the other hand. Therefore, these patient groups could be further subdivided into; (1) early (within 12 days after clinical onset) IgM and IgG response subgroup that corresponds to early recall IgG responses with emerging recall IgM responses and, (2) a late (later than 12 days after clinical onset) IgM and IgG response subgroup that comprises both patients with primary IgM responses seroconverting to IgG responses and those with late recall IgG responses with emerging IgM responses ( Figures 3A, B and Supplementary Figure 3 ). Moreover, the early IgG and IgM response subgroups had higher titres of anti-beta-HCoVs IgG antibodies than the IgM primary response groups defined above (IgM+/IgG-, IgM-/IgG- “not recall-type response” groups). This further indicated that pre-existing immunity to beta-HCoV was present in COVID-19 patients with IgG recall-type responses ( Figure 3D and Supplementary Figure 4B ).

To further examine the role of pre-existing beta-HCoV immunity in the determination of primary or recall-type antibody responses to SARS-CoV-2, we first merged patients with IgM only responses (IgM⁺/IgG^- group) or absence of both IgM and IgG (IgM^-/IgG^-) within 12 days after clinical onset into a primary response group. Patients with early IgG responses (IgM⁺⁺/IgG⁺, IgM⁺/IgG⁺, IgM^-/IgG⁺, IgM^+-/IgG⁺⁺, IgM⁺/IgG⁺⁺, IgM⁺⁺/IgG⁺⁺, groups) within 12 days after clinical onset were merged into a recall-type response group (see Supplementary Figure 3 ). We found that anti-HCoV-OC-43 and -HK-U1 IgG antibodies were mostly absent in the primary response group but abundant in the recall-type response group (p<0.0001 for both beta-HCoV) ( Figure 3E and Supplementary Figure 4C ). Importantly, the distribution of mild, severe, and critical cases differed significantly (p<0.001) between the two groups with a higher proportion of critical cases in the early recall-type IgG response group. This further indicated that pre-existing immunity to beta-HCoV was not protective against COVID-19 ( Figure 3F ).

Taken together, these results indicate that early anti-SARS-CoV-2 IgG production in COVID-19, reminiscent of a recall-type IgG response was more likely to be found in patients with pre-existing anti-common beta-HCoV IgG. In comparison, primary SARS-CoV-2 infection with dominant IgM early response was more likely to be observed in patients without pre-existing anti-common beta-HCoV IgG antibodies. This pre-existing immunity to common beta-HCoVs, although leading to cross-reactive IgG responses to SARS-CoV-2, failed to prevent the onset of COVID-19 ( Supplementary Table 3 ).

To determine whether the presence of pre-existing immunity to HCoV could also lead to cross-reactivity against SARS-CoV-2 antigens in healthy individuals, we analysed a cohort of 76 healthy French donors (48 males; 28 females; median age of 39 years; age range 19-65, Supplementary Table 4 ) established in 2015. Although we did not detect anti-RBD reactivity in the sera of these individuals, six serological samples (7.9%) were found to be reactive against one or several of the other SARS-COV2 antigens that is, the S2 domain, full-length Spike, and/or Nucleocapsid (NC). These sera also recognized all HCoVs, indicating that pre-COVID-19 cross-reactivity to SARS-CoV-2 antigens was neither specific for a unique family, nor for a particular type of coronavirus ( Figure 4A ).

We also assessed the reactivity of intravenous immunoglobulins (IVIG) manufactured prior to the COVID-19 outbreak. This was done to indirectly identify coronavirus reactivity in a large cohort as therapeutic IVIG consists of IgG isolated from about 10 000 pooled plasma samples of healthy donors. The IgG antibodies of the three different batches of IVIG demonstrated strong reactivity against all HCoVs, as well as detectable reactivity against the SARS-CoV-2 S2 domain and the full-length spike antigen ( Figure 4B ). Collectively, these results indicated that pre-COVID-19 cross-reactivity against SARS-CoV-2 antigens was present in the general population.

In order to experimentally confirm that pre-existing cross-reactivity did not lead to cross-protection against SARS-CoV-2 infection, we assessed the neutralising capacities of the six pre-COVID-19 sera reactive against SARS-CoV-2 antigens and the three IVIG batches through an in vitro SARS-CoV-2 neutralisation assay. Whilst full neutralisation of SARS-CoV-2 was observed with sera from COVID-19 patients containing anti-RBD antibodies, pre-COVID-19 serum devoid of detectable anti-RBD antibodies was ineffective. This effect is best illustrated in Supplementary Figure 5 with sera of a patient being able to neutralise SARS-CoV-2 replication when drawn 17 days after symptom onset but not at day 7. In addition, neither the pre-COVID-19 sera, nor the IVIG batches, were able to neutralise SARS-CoV-2 in vitro.

We also assessed the neutralising capacities of IgG isolated from this patient diluted with increasing volumes of IVIG to exclude the interference of a possible inhibitor, inadvertently introduced during the manufacturing process. As shown in Figure 5 , purified IgG from this patient was able to potently neutralise SARS-CoV-2 in vitro even when diluted with IVIG. The half-maximal inhibitory concentrations IC₅₀ of the patient’s serum alone (11.67 µg/mL) or diluted with IVIG (IVIG1: 9.94 µg/mL, IVIG2: 9.14 µg/mL, IVIG3: 10.15 µg/mL) were similar. Collectively, these results confirmed that IVIG products manufactured before the COVID-19 outbreak had no neutralising capability.

As shown in Figure 5 , the addition of neutralising IgG from COVID-19 patients to IVIG preparations could confer this neutralising capability. We had already identified that ~10 µg/mL of COVID-19 patient’s IgG was sufficient to neutralise 50% of the viral cytopathic effect and 64 µg/mL to neutralise all of the viral cytopathic effect in the presence of IVIG ( Figure 5 ). Due to our IVIG neutralisation tests being done at 12 mg/mL, we estimated that the presence in serum of only a single SARS-CoV-2-immunized individual out of 1200 donors (0.08%) would be sufficient to confer minimal, but detectable, anti-SARS-CoV-2 activity, while its presence among sera from 187 IVIG donors (0.5%) would confer potent neutralising capacities to IVIG.

For the healthy donor blood samples, we used previously cryopreserved cells obtained from the French Institute of Blood Donation (EFS, Etablissement Français du Sang, Paris, France). All samples were collected from patients referred to the Pitié-Salpêtrière Hospital. All patient demographic and clinical characteristics are detailed in Supplementary Table 2 . The provision of samples complied with the guidance from our research ethics committees at the Pitié-Salpêtrière Hospital and Sorbonne Université (CPP - Ile de France-VI and n°2020-CER2020-21). All patients or their relatives gave written informed consent. The three batches of IVIG pharmaceutical products, manufactured before 2019 in France, were obtained from Tegeline-LFB, Clairyg-LFB and CSL Behring. Each batch contained IgG at a concentration of 12 mg/ml.

The presence of serum antibodies specific for the viral antigen was determined using the Maverick SARS-CoV-2 Multi-Antigen Serology Panel (Genalyte Inc. USA). This technology uses an antigen-bound chip to detect the following antibodies for SARS-CoV-2; nucleocapsid, spike S1 RBD, full length spike S1S2, spike S2, and spike S1, as well as the those specific for the common coronavirus HCoV-NL-63; nucleocapsid, HCoV-OC-43, HCoV-229-E and HCoV-HK-U1 spike proteins (27, 28).. It detects and measures changes in resonance when antibodies bind to their respective antigens. All threshold values for positivity were set by the manufacturer. The raw data are shown in Supplementary Tables 1 – 4 .

The neutralising activity of the sera samples and IVIG was tested with a whole virus replication assay for which a SARS-CoV-2 strain isolated from a SARS-CoV-2 positive patient was used. The virus was isolated by inoculating Vero E6 cells with the patient’s sputum sample in a Biosafety Level-3 (BSL-3) facility. The serum samples were heat-inactivated at 56°C for 30 minutes and following two-fold serial dilutions (from 1:5 to 1:2560), pre-incubated on a 96-well plate with 50 µl of diluted virus (2x10³ Fifty percent Tissue Culture Infective Dose 50/mL at 37°C for 60 minutes. Next, 100 µL of the Vero E6 cells suspension (3x10⁵ cells/mL) was added to the mixture and incubated at 37°C with 5% CO₂ until a microscopic examination was performed on day 4 to measure (or determine) the cytopathic effect (CPE). All neutralising antibody titres were expressed as the highest serum dilution that showed 100% inhibition of CPE. An identical positive serum was added to each experiment as an internal control to assess the reproducibility of the test.

IgG were isolated from serum samples diluted in 1X-PBS as previously described (29). Briefly, serum samples were loaded onto Protein G/Agarose column (Invivogen) after column equilibration. Chromatography steps were then performed at a flow rate of 0.5 ml/min. Next, 20 column volumes of 1X-PBS were used to wash the column. IgG were then eluted with 5ml of 0.1M glycine (pH 2-3, Sigma-Aldrich) and pH was immediately adjusted to 7.5 with 1M Tris. 1X-PBS buffer exchange was achieved using Amicon^® Ultra centrifugal filters (Merck Millipore) through a 100-kD membrane according to the manufacturer’s instructions. Quantification of purified IgG was determined using the NanoVue Plus microvolume spectrophotometers.

Pseudotyped vectors were produced by triple transfection of HEK 293T cells as previously described (30). Briefly, cells were co-transfected with plasmids encoding for lentiviral proteins, a luciferase Firefly reporter, and a plasmid expressing a codon-optimized SARS-CoV-2 spike gene. Pseudotyped vectors were then harvested on day 2 post-transfection. Functional titre (TU) was determined by qPCR after the transduction of a stable HEK 293T-hACE2 cell line. To generate this cell line, HEK 293T cells were transduced at a multiplicity of infection (MOI) 20 with an integrative lentiviral vector expressing the human ACE2 gene under the control of the UBC promoter. Clones were generated by limiting dilution and selected on their permissiveness to SARS-CoV-2 S pseudotyped lentiviral vector transduction.

Firstly, serum dilutions were mixed and co-incubated with 300 TU of the pseudotyped vector at room temperature for 30 minutes. The serum and vector were then diluted in culture medium (DMEM-Glutamax (Gibco), supplemented with 10% foetal calf serum and penicillin/streptomycin (both from Gibco) or with IVIG batches at a 12 mg/mL concentration of the IgG. The samples were then transferred to a tissue culture-treated black 96-well plate (Costar) containing 20x10³ HEK 293T-hACE2 cells in suspension. To prepare the suspension, the cell flask was washed with DPBS twice (Gibco) and the cells were individualised with DPBS and supplemented with 0.1% EDTA (Promega) to preserve the hACE2 protein. After 48 hours, the media was removed and bioluminescence was measured using a Luciferase Assay System (Promega) on an EnSpire plate reader (PerkinElmer). The half maximal inhibitory concentrations IC₅₀ were determined using Graphpad Prism (version 5).

All analyses were performed using the R programming language (version 4.2). The heatmaps and correlation plots were generated using “pheatmap” (version 1.0.12). The figures and plots were generated using ggplot (version 3.3.0). For statistical comparisons, we implemented the Wilcoxon test using the “stat_compare_means” function from the ggpubr package. We categorised the patients using reference cut-off values for the NC and RBD, IgG, IgM and IgA titres. The patients were stratified using only the IgM and IgG values given that there was a statistically significant correlation between the IgA and IgG values. The intensity of antibody responses was defined as follows; (-) negative, (+) 1-2 fold above the threshold value and (++) more than 2 fold above the threshold value. When considering the global antibody responses to both RBD and NC ( Figure 4 ), the global anti-SARS-CoV-2 antibody response was defined as follows; (-) if both anti-RBD and anti-NC antibody responses were negative, (+) if either anti-RBD or anti-NC or both were positive (but not strongly) and, (++) if at least one antibody response was strongly positive. This translated into the following setup:

For each Ig subtype, three categories were set:

if NC value AND RBD value < threshold for NC/RBD; then the category is 0;

else the category is 1.