key: cord-0995001-ur8xde95
authors: Pagano, S.; Yerly, S.; Meyer, B.; Juillard, C.; Suh, N.; LeTerrier, C.; Daguer, J.-P.; Farrera Soler, L.; Barluenga, S.; Piumatti, G.; Hartley, O.; Lemaitre, B.; Eberhardt, C. S.; Siegrist, C.-A.; Eckerle, I.; Stringhini, S.; Guessous, I.; Kaiser, L.; Pugin, J.; Winssinger, N.; Vuilleumier, N.
title: SARS-CoV2- infection as a trigger of humoral response against apolipoprotein A-1
date: 2021-02-16
journal: nan
DOI: 10.1101/2021.02.12.21251298
sha: 65388f55fbb2b857bc88cedad20b824dc3649a3e
doc_id: 995001
cord_uid: ur8xde95

Aims: Unravelling autoimmune targets triggered by SARS-CoV-2 infection may provide crucial insights in the physiopathology of the disease and foster the development of potential therapeutic candidate targets and prognostic tools. We want to determine whether SARS-CoV-2 exposure could trigger a humoral response against apolipoprotein A-1 (anti-apoA-1 IgG) through molecular mimicry and assess its relationship to patient prognosis. Methods and Results: Anti-Spike domain 1 (SD1) IgGs, anti-apoA-1 IgGs and against mimic peptides, as well as cytokines were assessed by immunoassays on a case-control (n=101), an intensive care unit (ICU; n=126) with a 28-days follow-up, and a general population cohort (n=663) with available samples in the pre and post-pandemic period. Linear sequence homologies and antibodies cross-reactivity between apoA-1, TLR2, and Spike epitopes were identified. Overall, anti-apoA-1 IgG levels were higher in COVID-19 patients or anti-SARS-CoV-2 seropositive individuals than in healthy donors or anti-SARS-CoV-2 seronegative individuals (p<0.0001). Significant and similar associations were noted between anti-apoA-1, anti-SARS-CoV-2 IgG, cytokines, and lipid profile. In ICU patients, anti-SARS-CoV-2 and anti-apoA-1 seroconversion rates displayed similar 7-days kinetics, reaching 82% for anti-apoA-1 seropositivity. C-statistics (CS) indicated that anti-Spike/TLR2 mimic-peptide IgGs displayed a significant prognostic accuracy for overall mortality at 28 days (CS: 0.64; p=0.02). In the general population, SARS-CoV-2 exposure increased baseline anti-apoA-1 IgG levels. Conclusions: COVID-19 induces a marked humoral response against the major protein of high-density lipoproteins. As a correlate of poorer prognosis in other clinical settings, such autoimmunity signatures may relate to long-term COVID-19 prognosis assessment and warrant further scrutiny in the current COVID-19 pandemic.

The validation cohort consisted of 126 consecutive patients admitted to the intensive care unit (ICU) of HUG, between the 1 st of March and the 30 th of April 2020 for which serum sample leftovers were available from within 24 hours of admission to the ICU of patients who also completed a 28-day follow-up. This study was approved by the Geneva Ethic Council for Research (CCER) (Protocole Nr: CCER 2020-00917), Rue Adrien-Lachenal 8, 1207 Geneva, Switzerland. The written informed consent was obtained by patients prior enrollment. Informed consent was obtained either from the patient or from the next-of-kin, or it was deemed to have been given by default, unless a formal opposition was formulated. The inclusion criteria consisted of admission to the ICU with acute respiratory failure due to SARS-CoV-2 infection requiring mechanical ventilation or a fraction of inspired oxygen (FiO2) >80% between 1 st of March and the 30 th of April 2020. Exclusion criteria consisted of documented refusal to participate to the study, a negative SARS-CoV-2 RT-PCR, serum sample unavailability upon ICU admission, or incomplete 28-day follow-up. All the relevant medical information, including days post symptoms onset (DPSO) and ICU admission, as well as the 28-day followup, were retrieved in the electronic patient medical files by ICU ward physicians, which were blinded to the biochemical analyses. During the inclusion period, 138 patients were admitted to the ICU. Among them, one patient was admitted for a non-COVID-19 pneumonia, 5 patients did not complete the 28-day follow-up due repatriation to their country of origin, and six additional patients did not have available serum leftovers, leaving 126 ICU COVID-19 patients available for the analysis.

The primary biological objective of the study was to explore the associations between anti-SARS-CoV-2 and anti-apoA-1 IgG responses, as well as between the IgG responses against two mimic peptides, including Spike-apoA-1 mimic, and Spike-TLR2 mimic. The secondary biology objectives consisted of exploring the associations between the aforementioned serologies with inflammation biomarkers and the lipid profiles. The third biological objective All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 was to determine the short-term serological kinetics in a subgroup of 51 patients for which additional samples at day 3 and day 7 of ICU admission were available. The only primary clinical study endpoint was overall mortality at 28 days.

This cohort consisted of individuals initially enrolled in the BUS-Santé study in the pre-COVID-19 era (n=2531; recruited in years 2016-2018) which were then subsequently enrolled in the SEROCoV-POP study (n=2076) , which aimed at assessing the seroprevalence of anti-SARS-CoV-2 antibodies in the Geneva cantonal population during the first pandemic wave (31) . Briefly, for these 2076 individuals, pre-pandemic serum samples stored at -80°C after being collected between February 2016 and January 2018 were available from 663 individuals to assess the anti-apoA-1 IgG response. This study was approved by the Geneva Ethic Council for Research (Protocole Nr: CCER16-363), and written informed consent was obtained by patients prior enrollment. Further details of the initial SEROCoV-POP have been published previously (31) . The two predefined study objectives were to assess at the population level: i) the impact of SARS-CoV-2 infection on anti-apoA-1 IgG levels, and ii) the impact of prepandemic anti-apoA-1 serological status on the subsequent anti-SARS-CoV-2 humoral response.

The study flow chart summary of the three cohort is presented in figures 3 and 4.

Whole blood samples were obtained: at the time of donation for healthy donors; at the time of hospital and ICU admission, respectively for cases of the derivation and validation cohorts; and at the time of outpatient recruitment for the general population cohort. Plasma and serum were generated from whole blood via centrifugation. Aliquots were frozen at −80 ˚C until analysis. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Nasopharyngeal swabs for cases of the derivation and the validation cohort, were obtained at the time of hospital and ICU admission.

As previously reported (30, 31) , SARS-CoV-2 RT-PCR was performed according to the manufacturers' instructions on various platforms, including initially in house method using the BD SARS-CoV-2 reagent kit for BD Max system (Becton, Dickinson and Co, US) and Cobas 6800 SARS-CoV-2 RT-PCR (Roche, Switzerland).

As previously reported (30, 31) , we used the Euroimmun IgG enzyme-linked immunosorbent assays (ELISA) (Euroimmun AG, Lübeck, Germany # EI 2606-9601 G; CE-marked) to assess SARS-CoV-2 IgG serology against the S1-domain of the Spike protein (anti-S1 IgG). EDTAplasma (derivation cohort) and serum (validation cohort) were diluted at 1:101 according to the manufacturer's instructions. Results of patient sample's immunoreactivity are expressed as the ratio of the optical density at 450 nm (OD450) divided by the calibrator's OD450. The ratio is interpreted as follows: OD450 ratio: <0.8 = negative; ≥0.8 and < 1.1 = indeterminate; ≥1.1 = positive. For the purpose of this study, indeterminate results were considered as negative (30) . Inter-assay variation was 15.6% for IgG at a ratio of 2.09 (n=17).

Total antibodies against the N antigen of SARS-CoV-2 were measured on a Cobas e801 analyzer (Roche Diagnostics, Rotkreuz, Switzerland) according to the manufacturer's instructions. Results are reported as numeric values in form of a cut-off index (signal sample/cutoff or signal calibrator ratio) and are considered as positive when equal to or above 1. Inter-assay variation was 14.3% at a ratio of 2.97 (n=17). All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Anti-apoA-1 IgG assessment Anti-apoA-1 IgGs were measured as previously described (21-27), using frozen EDTA plasma (derivation cohort) and serum (validation and general population cohort). Maxisorp plates (Nunc TM , Roskilde, Denmark) were coated with purified, human-derived delipidated and unmodified apoA-1 (20 µg/mL; 50 µL/well) for 1 h at 37 C. After being washed, all wells were blocked for 1 h with 2% bovine serum albumin (BSA) in a phosphate buffer solution (PBS) at 37 C. Participants' samples were also added to a non-coated well to assess individual nonspecific binding. After six washing cycles, a 50 µL/well of signal antibody (alkaline phosphatase-conjugated anti-human IgG; Sigma-Aldrich, St Louis, MO, USA, ref: A-3150), diluted 1:1000 in a PBS/BSA 2% solution, was added and incubated for 1 h at 37 C. After washing six more times, phosphatase substrate p-nitrophenyl-phosphate-disodium (Sigma-Aldrich, St Louis, MO, USA) dissolved in a diethanolamine buffer (pH 9.8) was added and incubated for 30 min at 37 C (Molecular Devices TM Filter Max F3, Molecular Devices, San Jose, CA, USA). OD450 was determined at 450 nm, and each sample was tested in duplicate.

Corresponding non-specific binding was subtracted from the mean OD450 for each sample. The specificity of detection against lipid-free and unmodified apoA-1 has been previously determined by conventional saturation tests, Western blot, and LC-MS analyses (23). At an intermediate ratio of 0.6 OD450, the interassay coefficient of variation was 9% (n = 5), and the intra-assay CV was 5% (n = 5). For serum, the anti-apoA-1 IgG seropositivity cut-off was previously specified and validated, and was set at an OD450 ratio >0.64 for the 97.5 th percentile of anti-apoA-1 IgG of healthy blood donors (21-27).

IgG against Spike/apoA-1 and Spike/TLR2 mimic peptides assessment All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 The same protocol used for anti-apoA-1 IgG assessment was applied for the assessment of IgG against these two mimic peptides, mutatis mutandis, where they were used as coating antigen in our ELISA plates followed by our conventional protocol (21-27).

To assess the degree of cross reactivity between anti-SARS-CoV-2, anti-apoA-1 IgG (and control IgG) with their respective antigens, delipidated apoA-1 (purified from plasma of healthy blood donors), Spike protein was kindly provided by the institute of technology EPFL (Lausanne, Switzerland), and the two mimic peptides were coated on Maxisorp plates (NuncTM, Roskilde, Denmark) according to the anti-apoA-1 IgG protocol (17, (21) (22) (23) (24) (25) (26) . Competition experiments were done using our standard anti-apoA-1 IgG protocol (21-27), which involved coating the plate with apo-A-1 or Spike protein, then increasing concentrations of polyclonal anti-Spike antibodies or polyclonal anti-apoA-1 IgG (5 to 40 µg/ml) were added to the plate for 30 min, and then pooled sera from seven anti-apoA-1 IgG were subsequently incubated onto the plate for one hour at 37°C. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Cytokines and anti-pneumococcal IgG (P14 serotype) assessment As a negative control for anti-SARS-CoV-2 and anti-apoA-1-related serologies, IgG against pneumococcus belonging to the P14 serotype, the most prevalent serotype in Switzerland (32) , was assessed for the derivation cohort only, using a SQ120 instrument (MSD platform) according to the manufacturers' instructions. The seropositivity cut-off was 0.3 mg/l and the intra-run CV was 3.8% and inter-run CV was 12%.

Results were reported as proportion, median, range and interquartile range (IQR), unless stated otherwise. Differences between groups were calculated using Fisher's exact bilateral test (or Yates' Chi-Squared test when appropriate) or the Mann-Whitney test for categorical and continuous variables, respectively. Spearman correlation was used to assess correlations between the variables. Receiving operating characteristics (ROC) analyses were used to determine the association with patients' prognosis on the validation cohort according to the area All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 under the curve (AUC). Cox regression analyses were used to determine the associations between pre-pandemic and post pandemic anti-apoA-1 IgG status in unadjusted mode, and after adjustment for age, gender and smoking. The same univariate analyses were used to determine whether pre-pandemic serology status could predict subsequent anti-SARS-CoV-2 serology status. Results were expressed as hazard ratios (HR) with 95% confidence intervals (95%CI).

Due to the exploratory nature of this work and the pre-specified analysis planned, adjustments for multiple tests were not performed. All analyses were performed using Statistica software (version 13.5.0.17, TIBCO Software Inc., Palo Alto, CA, USA). Statistical significance was defined as P < 0.05.

Capitalizing on prior findings indicating that: i) anti-apoA-1 IgG has to bind to TLR2 due to molecular mimicry in order to generate a pro-inflammatory response by inducing the formation of a TLR2/TLR4/CD14 heterotrimer (18, 19) , and ii) that anti-apoA-1 IgG are preferentially oriented against the c-ter of apoA-1 in humans (28, 29); we searched for linear sequence similarities between the Spike protein epitopes (13, 14, 15 , 16) , apoA-1 and the extracellular part of TLR2.

As shown in Figure 1 , these analyses revealed that the amino acid (aa) sequence 1139-1162 of the Spike protein shares sequence homology with the c-ter part of apoA-1 (amino acids 216-243; Figure 1d ), and that the aa region spanning 579-587 of Spike protein had a good alignment with those of human TLR2 (aa 456-464; 7 out of 9 amino acids; Figure 1d ). Closer inspection All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 of the position of the sequence showed that the secondary structure of the peptide was comparable for both sequence matches ( Figure 1d ). Specifically, the aa sequence 1139-1162 in Spike is part of an alpha helical bundle with a portion of the peptide sequence unresolved whereas the corresponding apoA-1 sequence is also part of an alpha helical bundle with an unstructured portion. The Spike aa 579-587 peptide is part of a beta turn and continues into a beta pleated sheet. There is remarkable structural homology with the structure of the corresponding peptide on TLR2, namely a beta turn followed by a beta pleated sheet segment ( Figure 1d ). The same analysis was performed to search for homology between TLR2 or apoA-1 and the N-protein of SARS-CoV-2, using epitopes experimentally detected (15, 16) . A good alignment between apoA-1 (aa 131-143) and N (aa 400-412) was identified with 7 identical and 2 similar residues out of 11, however, the lack of structural data for the segment of the N protein precluded further analysis. Likewise, a good match was identified between TLR2 (aa [549] [550] [551] [552] [553] [554] [555] [556] [557] and N (aa 217-225) with 5 identical and 2 similar residues in a 8 amino acid stretch, but its structural homology could not be validated. In light of the lack of structural data for these N homology regions, further experimental validation of cross-reactivity was not pursued. Peptide sequences and structures are presented in supplemental figures 1 and 2.

Following the bioinformatics identification of common epitopes between Spike and apoA-1, we experimentally assessed the degree of cross-reactivity between polyclonal anti-SARS-CoV-2 and polyclonal anti-apoA-1 IgG with their respective antigens in our ELISA format. As shown in Figure 2 , anti-Spike IgGs minimally cross-reacted with apoA-1 and both mimic peptides.

Such cross-reactivity was weak and similar to the one observed for polyclonal control IgGs. On the other hand, polyclonal anti-apoA-1 IgG displayed a substantial cross-reactivity with Spike protein and both mimic peptides ( Figure 2a ). This experimental evidence was furthered by the fact that anti-Spike IgG pre-incubation onto the plate was able to outcompete the anti-apoA-1 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 IgG signal from anti-apoA-1 IgG seropositive individuals (p=0.01) (Figure 2b ), while anti-apoA-1 IgG pre-incubation of patients sera positive for anti-SARS-CoV-2 IgG did not affect the anti-SARS-CoV-2 IgG signal (P=0.4) (Figure 2c ). Taken together, these results indicate that, due to the presence of common linear epitopes on Spike, apoA-1 and TLR2, molecular mimicry-driven cross-reactivity between anti-apoA-1 IgG and S1 antigen seemed to be minimal, and preferentially concerns anti-apoA-1 IgG cross-reacting with S1 antigen rather than the opposite.

To further investigate the possible clinical impact of such cross-reactivity of IgG against S1 and apoA-1 on anti-S1 IgG results provided in routine diagnosis, 11 samples identified since the beginning of the COVID-19 pandemic that were considered as anti-S1 IgG false-positives, based upon negative recombinant immunofluorescence and negative anti-N serology (28-29), were tested for anti-apoA-1 IgG. As shown in supplemental Table 1 , none of these eleven samples were found to be positive for anti-apoA-1 IgG.

Taken together, these results indicate that despite molecular mimicry and a certain degree of cross-reactivity observed in vitro, our data do not indicate that anti-apoA-1 IgG could generate analytical interference with anti-S1 IgG in humans.

In order to validate and extend our previous findings to humans, we explored the associations between anti-apoA-1 and anti-SARS-CoV-2 serologies on three different cohorts, including a derivation and a validation cohort, as well as a general population cohort described in figure 3 and 4. In the derivation and validation cohort, we aimed at replicating the previously reported correlations between anti-apoA-1 IgG with cytokine and lipid profile (20, 33) and explored their possible extension to Spike/apoA-1 and Spike/TLR mimic peptides as an additional orthogonal assessment of closely related serologies. The general population cohort was All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 instrumental to generalize the results observed in acute settings at the population level and to determine whether the pre-pandemic anti-apoA-1 serological status could modulate the anti-SARS-CoV-2 response.

The derivation cohort consisted in 50 COVID-19 RT-PCR confirmed cases, and 51 unmatched healthy donors recruited over the same period (figure 3 panel a). The baseline demographic characteristics of participants are summarized in table 1. Briefly, RT-PCR confirmed COVID-19 patients were older with an over representation of male gender, and displayed a higher systemic pro-inflammatory state when compared to the healthy blood donors (Table 1 ). Among COVID-19 patients, the median delay between a positive SARS-CoV-2 RT-PCR diagnostic test and current biomarker assessment was 10 days (IQR 5-15 days). The proportion of patients within each days post-diagnosis subgroup (delta between their molecular testing and serological testing), was 43.5% for 0-6 days (n=20), 30.4% for 7-14 days (n=14) and 26.1% for >14 days (n=12). Table 1 , COVID-19 patients had higher median levels of all the pro-inflammatory cytokines and serologies tested, with the exception of P14 pneumococcal IgG used as an unrelated serological control. The distribution of serological values between cases and controls is available in supplemental figure 3. No difference between cases and controls was observed for circulating INF- levels. Furthermore, when the cohort was split according to anti-S1 IgG seropositivity status, identical differences were observed (Table 1, 

These results were further corroborated by the significant and substantial correlations observed between anti-apoA-1 IgG and both (anti-S1 and anti-N) anti-SARS-CoV-2 serologies, as well as by similar correlations with both anti-Spike/TLR2 and anti-Spike/apoA-1 IgGs (supplemental table 2 ). On the other hand, none of the aforementioned antibodies correlated All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 with anti-pneumococcal IgG (supplemental table 2 ). Furthermore both anti-SARS-CoV-2 serologies, anti-apoA-1 IgG, anti-Spike/apoA-1 mimic IgG, anti-Spike/TLR2 mimic IgGs displayed similar strength of associations with most of the cytokines measured. Finally, antipneumococcal IgG was not correlated to any of the cytokines tested (supplemental table 2) .

To extend and validate these findings in severe COVID-19 disease, we used a cohort of 126 consecutive patients admitted to the ICU for severe COVID-19 disease who completed a follow-up at 28 days, figure 3 panel b. The baseline demographic and biological characteristics of ICU COVID-19 patients are summarized in Table 2 . Anti-apoA-1 IgG seropositivity upon ICU admission was found in 26.9% of the ICU patients (34/126), while anti-S1 IgG and anti-N seropositivity was 36.5% (46/126) and 42% (53/126), respectively. When split according to anti-apoA-1 IgG seropositivity status upon patient admission at the ICU, seropositive patients tended to have a more severe Simplified Acute Physiology Score II score, a higher number of DPSO at ICU admission, displayed higher median D-dimers levels, anti-S1 IgG levels anti-Spike/apoA1 and anti-Spike/TLR2 IgG levels, but lower total cholesterol, LDL, and triglycerides levels when compared to anti-apoA-1 seronegative individuals ( Table 2 ). The proportion of anti-S1 and anti-N seroconversions were two-fold higher in anti-apoA-1 IgG seropositive individuals compared to those tested negative for these autoantibodies (55.8% vs 28.2%, P=0.006; and 64.7 vs 33.7%; P=0.01, respectively). No other significant differences for the remaining parameters were identified between anti-apoA-1 IgG seropositive and seronegative individuals.

When split according to anti-S1 serological status, anti-S1 IgG seropositive patients were less likely to be known to have chronic kidney disease, they tended to have a shorter length of ICU stay, and they were less likely to require mechanical ventilation (supplemental table 3). On the other hand, anti-S1 IgG seropositive patients displayed higher number of DPSO before ICU All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 admission; higher medians levels of anti-apoA-1 IgG, anti-Spike/apoA-1, anti-Spike/TLR2 mimic IgG, and D-dimers; but lower median levels of INF- and INF-α2a. No significant differences were noted regarding the lipid profile and other biological parameters tested (supplemental table 3 ). The proportion of anti-apoA-1 IgG seropositivity was increased by 2fold in anti-S1 seropositive individuals compared to anti-S1 IgG seronegative ones ( 

In the subgroup of 54 ICU patients for which additional serum samples were available at day 3 and day 7 of ICU admission, significant increases in median values of anti-SARS-CoV-2 and all anti-apoA-1 IgG-related serologies were observed ( Figure 5 ). Accordingly and as expected, anti-SARS-CoV-2 seropositivity rates for both anti-S1 and anti-N serologies were high and reached values above 92% at Day 7, while anti-apoA-1 IgG seropositivity displayed a very similar trend, reaching 82.4% at Day 7, indicating that anti-apoA-1 IgG serology kinetic closely follows the occurrence of anti-SARS-CoV-2 antibodies over 7 days of severe COVID-19 disease (supplemental table 5) . No such kinetics were observed for anti-P14 pneumococcus IgG.

As anti-apoA-1 IgG has been associated with worse overall and cardiovascular prognoses in different studies (22-25) and because the humoral response against SARS-CoV-2 has been All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 equivocally related to COVID-19 disease severity (34, 35) , we investigated their possible prognostic accuracy for overall-mortality at 28 days in ICU patients. ROC curve analyses using continuous biomarker values indicated that the only significant predictor was IgG against the Spike/TLR2 mimic peptide, but only modestly so, with an area under the curve of 0.64. The remaining antibodies were not associated with ICU patients' prognosis, although a trend was observed for anti-S1 IgG (supplemental table 6). ROC analyses indicated that the optimal cutoff for anti-spike/TLR2 cut-off would correspond to an arbitrary unit value of 0.11 OD450 ratio displaying a sensitivity of 100% (95%CI: 81-100), a negative predictive value of 100% (95%CI: 78-100) a specificity of 17% (95%CI: 13-28) and positive predictive value of 19% (95%CI: [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] . Given the fact that this cut-off provided no false negatives results, logistic regression analyses could not be performed.

Because anti-apoA-1 IgG seropositivity has been shown to be partially genetically driven, concerning about one fifth of the general population, and to be associated with a poorer prognosis over 5 years (23, 36), our results prompted us to investigate whether the SARS-CoV-2-induced anti-apoA-1 IgG response could be replicated in the general population. With this aim, we identified participants who had been recruited in the "Bus Santé" study between 2016 and 2018 and were subsequently included in the SEROCoV-POP study (31) The median age of this cohort was 50 years-old (range 24-78), 297 (44.7%) participants were male, while the baseline (pre-pandemic) anti-apoA-1 IgG seropositivity rate was 25.0 % (166/663) and was not associated with any factors commonly ascribed to auto-antibodies such as age, gender or smoking (data not shown). The median time between the first anti-apoA-1 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 IgG assessment in the pre-COVID-19 period and the anti-S1 IgG plus the second anti-apoA-1

IgG measurement during the post-COVID-19 period (after the first pandemic wave) was 3.2 years (IQR: 2.8-3.6; Range: 2.3-4.3). In the post-COVID-19 period, 7.5% (50/663) were seropositive against SARS-CoV-2 according to anti-S1 IgG levels; anti-apoA-1 IgG seropositivity, as well as median anti-apoA-1 IgG levels (data not shown), were significantly lower than in the pre-COVID-19 period ( Table 3 ). As shown in Table 3 , in the post-COVID-19 samples, a modest but significant correlation was observed between anti-apoA-1 IgG and anti S1 IgG levels. Furthermore and as shown in Table 3 , in the post-COVID-19 period, anti-S1 seropositive individuals displayed higher median anti-apoA-1 IgG levels (0.35 vs 0.57 OD;

P=0.0002) and higher median anti-apoA-1 seropositivity rates than anti-S1 seronegative individuals (34%.0 vs 16.8%, P=0.004). Moreover, the strength of correlation between anti-apoA-1 IgG and anti-S1 IgG in anti-S1 seropositive individuals was of 0.31 (P=0.03), whereas no association between these two serologies was found in anti-S1 seronegative individuals (Table 3) 

The major findings of this study can be summarized by the fact that SARS-CoV-2 infection triggers humoral responses against native and unmodified apoA-1, the major HDL lipoprotein, in up to 80% of severe COVID-19 patients and in up to one third of the infected population.

Our study indicates that such phenomenon can be at least partly ascribed to molecular mimicry between immunodominant linear epitopes of SARS-CoV-2 and those of apoA-1 and TLR2. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 The c-ter sequence identified in Spike, aa 1139-1162 (15) , is known to be conserved in HCoV-OC43 but not amongst other coronaviruses, and was found to share linear homology with aa 216-243 of apoA-1. This c-ter region of apoA-1 structurally corresponds to an alpha helix bundle playing a key role in the cellular cholesterol efflux regulation by ATP-binding cassette transporter A1 (ABCA1) and in HDL maturation (37), and was shown to be preferentially targeted by the polyclonal anti-apoA-1 IgG response (28, 29). Such regional targeting could explain the inverse relationships between anti-apoA-1 IgG and HDL cholesterol reported previously (21, 22, 33) , and lately implicated in the disruptive effects of anti-apoA-1 IgG on cellular cholesterol homeostasis (33) . The second linear sequence homology identified concerned the aa 579-587 of Spike, located in the S1 domain and close to the receptor binding domain, which has very homology to the aa 456-464 sequence of TLR2. This sequence is part of the leucine reach repeats (LRR) ectodomain of TLR2, known to be key for proper pathogenassociated or damage-associated molecular pattern recognition and TLR2's function (38). The engagement of this region by anti-apoA-1 IgGs due to sequence homology with apoA-1 (20) is believed to mediate their pro-atherogenic response (18-20, 33), affecting myocardial necrosis and mice survival through specific signaling pathways (18-20, 39). Taken together, these results indicate that these two immunodominant SARS-CoV-2 epitopes previously identified as potential T cell and B cell epitopes (13, 14, 15 , 40) , share linear sequence homologies with host antigens playing key roles in inflammation and lipid homeostasis regulation, and that the COVID-19-induced humoral response may elicit unwanted pathogenic antibodies potentially cross-reacting with numerous autoantigens due to pathogenic B cell activation (41). While the linear epitope homology between apoA-1, TLR2 and the N antigen could not be validated because of a lack of structural information, significant and substantial associations were still observed between the anti-N and anti-apoA-1-related serologies. These results could also be explained by the possible existence of common conformational epitope(s) that our bio-All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ;  https://doi.org/10.1101/2021.02.12.21251298 doi: medRxiv preprint informatics and epitope mapping systems could not detect, and/or by the likely existence of additional mechanisms, such as intermolecular epitope spreading, allowing the initial targeted humoral response to quickly broaden to antigens other than the inducing epitope (42) .

They are multiple potential clinical implications of the present findings. Firstly because, anti-SARS-CoV-2 IgG responses against the regions containing these two epitopes were previously shown to correlate with COVID-19 severity (15) and because the presence of anti-apoA-1 IgG has been predictive of poorer outcomes (21-25, 36, 39) , these findings have implications for COVID-19 patients' prognosis and risk stratification. Indeed, corroborating these previous observations, our results indicate that SARS-CoV-2-induced IgGs against aa 456-464 of TLR2 (and cross-reacting with anti-apoA-1 IgG (20)) could significantly predict 28-day mortality in ICU patients despite a limited sample size, with a NPV reaching 100% using a post-hoc defined cut-off. Although requiring further prospective validations, these preliminary results indicate that assessing such autoantibodies on top of existing stratification tools could pave the way for enhanced risk stratification and resource allocation optimization, a major focus point in severe COVID-19 disease. Furthermore, because the anti-apoA-1 IgG response in settings other than COVID-19 was shown to be associated with long-term outcomes varying between 1 to 7 years of follow-up (21-25, 36, 39), knowing whether the presence of these autoantibodies could predict outcomes on period longer than 28 days and in less severe forms of disease deserves further investigations. Thirdly, from an analytical standpoint, our results indicate that despite the molecular mimicry-driven cross-reactivity between anti-apoA-1 IgG towards S1 antigen, our results do not indicate that anti-apoA-1 IgG could potentially yield false-positive results in anti-SARS-CoV-2 S1 serology. Along the same line, although SARS-CoV-2 infection increased anti-apoA-1 IgG levels in the general population too, our observations did not indicate that pre-COVID-19 anti-apoA-1 IgG status could affect the subsequent anti-S1 IgG response, which was important to rule-out. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in (33) . Because membrane free-cholesterol is a key regulator of membrane ACE2R trafficking into dedicated lipid rafts for optimal SARS-CoV-1 endocytosis (48), such All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 results suggest that by modulating the membrane free-cholesterol content anti-apoA-1 IgG may influence the course of COVID-19. Although devoid of any experimental demonstration, such a hypothesis remains to be further investigated.

We acknowledge several limitations of the present work. Firstly, our analysis used linear epitope mapping and doesn't consider potential conformational epitopes which may further contribute to the association between the anti-apoA-1 and anti-N serology. Nevertheless, as those two linear immunodominant epitopes are exposed to the immune system regardless of the conformation of spike (monomeric or trimeric) (13, 14, 15 ) , potential conformational issues are unlikely to have blunted the present conclusions. Secondly, the relatively limited number of patients enrolled in this study impeded our ability to complete the survival analyses with the desired level of details regarding the prognostic value of anti-Spike/TLR2, which needs replication in larger prospective cohorts. Thirdly, due to the fact that our institution became a fully COVID-19 dedicated infrastructure, we could not identify non-COVID-19 matched controls during the studies inclusion period. Fourthly, because the aa579-587 sequence of Spike shared sequence homology with the aa 456-464 of TLR2 corresponding to the PAMP/DAMP sensing of TLRs, knowing whether direct anti-S1 IgG-mediated TLR2 stimulation could per se elicit a pro-inflammatory response remains to be explored. Fourth, as anti-apoA-1 IgG are not to be predominantly ascribed to CV prognosis, knowing whether the SARS-CoV-2-induced anti-apoA-1 IgG seroconversion could modulate the long-term CV prognosis of COVID-19 patients remains elusive. Furthermore, being primarily driven by hour bio-informatics modelling, we did not investigate the possible homology between Spike and other auto-antigens and therefore other autoantibodies in the present work, which deserves further attention. Lastly, if our results indicate that an acute exposure to SARS-CoV-2 rapidly increases the anti-apoA-1 IgG response, they do not allow inferring any conclusions about the possible longer term persistence of anti-apoA-1 IgG levels after COVID-19 disease. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 In conclusion, this report shows for the first time that in a substantial proportion of SARS-CoV-2 exposed individuals, a marked humoral autoimmune response against the major lipoprotein of what is often referred to the "good cholesterol" occurs partly due to molecular mimicry between immunodominant epitopes of Spike and apoA-1. Knowing whether the pre or coexistence of anti-apoA-1 IgG may modulate the course of COVID-19 disease remains uncertain. However, as correlates of poorer prognosis in different settings, such autoimmunity signatures may relate to COVID-19 risk stratification and warrant further scrutiny in the current perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 Zhang Z. Antibody responses to SARS-CoV-2 in patients of novel coronavirus disease 2019.

Clinical infectious diseases. 2020.

Antiochos perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 

NV received restricted research grants unrelated to this study from Roche.

All the other co-authors have no conflicts of interest to declare.

All rights reserved. No reuse allowed without permission.

perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 highlighted in green (a and b) and conserved sequences between Spike and ApoA-1 are represented in red (a and c). Panel d) Sequence alignment of SARS-CoV-2 Spike protein (QIV65088.1) with human TLR2 (H33756AA.1) using Clustal W. Conserved residues are indicated in green and the semiconserved one in yellow. BlastP sequence alignment of SARS-CoV-2 Spike sequence with human apoA-1 (P02647.1). Conserved residues are shown in red and semiconserved (functional equivalent) ones in yellow.

Panel a). ApoA-1 protein is recognized in a statistical significant way and in a concentration dependent manner by polyclonal antibodies anti-Spike (**P=0.002) or by control rabbit IgG (*P=0.01). Spike protein is also recognized in a statistical significant way and in a concentration dependent manner by polyclonal anti-apoA-1 IgG (***P = 0.0004) and by its control goat IgG (***P = 0.0002). Anti-apoA-1 IgG and goat IgG also bind the Spike/apoA-1 mimic (**P = 0.001 and **P = 0.003 respectively) and the Spike/TLR2 mimic (***P = 0.0006 and **P=0.003 respectively) in a concentration dependent manner. The Kruskal-Wallis test was used to compare the three groups. Grey bars indicate the difference in the binding toward the specific antigen between the antibody and its control at the highest concentration of 10 mg/mL, all the differences are significant and the P-values are as follow:

P=0.02, P=0.01, P=0.02 and P=0.03, the Mann-Whitney test was used to compare the two groups. The results from at least three independent experiments (n = 3-6) are presented as median with range.

Panel b). Polyclonal anti-Spike antibody and not the control IgG slightly but significantly compete for apoA-1 binding sites with anti-apoA-1 autoantibodies present in the pool patient All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 serum (*P=0.01, Kruskal-Wallis test) while anti-apoA-1 IgG didn't compete for Spike protein as shown in panel c). The results from three independent experiments (n=3) are presented as median with range. Results are expressed as median with interquartile range and the Kruskal-Wallis test was used to compare the three groups.

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 All continuous variables are expressed as median (interquartile range; and range); *P value derived from the comparison between anti-apoA-1 IgG seropositive verse seronegative individuals.

DPSO: days post symptom onset; IC: ischemic cardiopathy; HF: heart failure; APACHE II: Acute Physiology And Chronic Health Evaluation II; SOFA: Sequential Organ Failure Assessment; SAPS: Simplified Acute Physiology Score. Table 3 . Associations between anti-apoA-1 IgG and anti-S1 IgG status in the general population before and after the first COVID pandemic The general population cohort (identical individuals from the pre and post-COVID era)

Pre-COVID era individuals (n=663)

Post-COVID era individuals split according to S1 serology status (n=663)

Anti-S1 seropositive individuals (n=50) Anti-S1 seronegative individuals (n=613) LDL 

Hs-cTnT, ng/L 16.0 (9.7-34.9;3.31-971) 13 .0 (8.1-40.5; 3.7-665) 17-9 (9.7-33.5;3.3-971) 0.53 NT-proBNP, pg/ml 308 (95. 6-1015; 15.1-18772) 363 (110-1437;22-5928) 278. 5 (93.3-909;15.1-18772) 0.42

Anti-S1 IgG, ratio 0. 

Creatinine; µmol/l 81.0 (66.5-105; 38-769) 81. 5 (67.5-110.5; 47-173) 80.5 (65-98, 38-769) 0.36 All rights reserved. No reuse allowed without permission.

perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted February 16, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 

Identification of 663 the same individuals of the «Bus santé study» subsequently enrolled in the SEROCoV-POP study with serum samples available in both studies SEROCoV-POP study (n=2076)

Pre-covid era samples (n=663)

Anti-apoA-1 IgG assessment Post-covid era samples (n=663)

Anti-apoA-1 IgG assessment Anti-S1D IgG ssessment

The general population cohort (n=663) 

A role for autoantibodies in atherogenesis

Autoimmunity and HIV

Frequency and clinical correlates of antiphospholipid antibodies arising in patients with SARS-CoV-2 infection: findings from a multicentre study on 122 cases

Immune-mediated neurological syndromes in SARS-CoV-2-infected patients

Kanduc D. From Anti-SARS-CoV-2 Immune Responses to COVID-19 via Molecular Mimicry

A highly conserved cryptic epitope in the receptor binding domains of SARS-CoV-2 and SARS-CoV

Identification of immunodominant linear epitopes from SARS-CoV-2 patient plasma

Two linear epitopes on the SARS-CoV-2 spike protein that elicit neutralising antibodies in COVID-19 patients

MGH COVID-19 Collection & Processing Team

SARS-CoV-2 proteome microarray for mapping COVID-19 antibody interactions at amino acid resolution. ACS Central Science. 2020. All rights reserved. No reuse allowed without permission

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted

Diagnostic accuracy of

Augurix COVID-19 IgG serology rapid test

Seroprevalence of anti-SARS-CoV-2 IgG antibodies in Geneva, Switzerland (SEROCoV-POP): a population-based study

Serotype epidemiology of invasive pneumococcal disease in Swiss adults: A nationwide population-based study

Anti-ApoA-1 IgGs in Familial Hypercholesterolemia Display Paradoxical Associations with Lipid Profile and Promote Foam Cell Formation

Extrafollicular B cell responses correlate with neutralizing antibodies and morbidity in COVID-19

Viruses and Autoimmunity: A Review on the Potential Interaction and Molecular Mechanisms

Lipoprotein Receptors Redundantly Participate in Entry of Hepatitis C Virus

HDL-scavenger receptor B type 1 facilitates SARS-CoV-2 entry

Lipoprotein concentrations over time in the intensive care unit COVID-19 patients: Results from the ApoCOVID study

Length of stay at ICU, days 16