key: cord-0742113-wehnbuvn authors: Vogl, T.; Klompus, S.; Leviathan, S.; Kalka, I.; Godneva, A.; Shinar, E.; Weinberger, A.; Segal, E. title: Cross-reactive antibody responses against SARS-CoV-2 and seasonal common cold coronaviruses date: 2020-09-03 journal: nan DOI: 10.1101/2020.09.01.20182220 sha: db60561bf223c279cb3a511ec3a003d2c2789fb2 doc_id: 742113 cord_uid: wehnbuvn While cross-reactive T cells epitopes of SARS-CoV-2 and seasonal/common cold human coronaviruses (hCoVs) have been reported in individuals unexposed to SARS-CoV-2, potential antibody-based cross-reactivity is incompletely understood. Here, we have probed for high resolution antibody binding against all hCoVs represented as 1,539 peptides with a phage-displayed antigen library. We detected broad serum antibody responses against peptides of seasonal hCoVs in up to 75% of individuals. Recovered COVID-19 patients exhibited distinct antibody repertoires targeting variable SARS-CoV-2 epitopes, and could be accurately classified from unexposed individuals (AUC=0.96). Up to 50% of recovered patients also mounted antibody responses against unique epitopes of seasonal hCoV-OC43, that were not detectable in unexposed individuals. These results indicate substantial interindividual variability and antibody cross-reactivity between hCoVs from the direction of SARS-CoV-2 infections towards seasonal hCoVs. Our accurate high throughput assay allows profiling preexisting antibody responses against seasonal hCoVs cost-effectively and could inform on their protective nature against SARS-CoV-2. COVID-19 (coronavirus disease 2019), caused by SARS-CoV-2 (severe acute respiratory syndrome 34 coronavirus 2), represents an unparalleled pandemic with millions of cases worldwide. In addition to 35 SARS-CoV-2, six more coronaviruses infect humans (hCoVs) including SARS-CoV-1 responsible for the 36 SARS outbreak in 2003 and MERS-CoV (Middle East respiratory syndrome) (1). Four seasonal endemic 37 hCoVs (OC43, HKU1, NL63, 229E) are widely circulating in the population causing only mild symptoms 38 (common cold)(2). Previous exposures to seasonal hCoVs may elicit immunological memory that could 39 benefit the course of SARS-CoV-2 infections (3), which can range from asymptomatic to life-40 threatening symptoms (4). The exact causes underlying this heterogeneity in COVID-19 severity are 41 incompletely understood and involve factors as age, gender, comorbidities, and preexisting immunity 42 (5, 6). As an important part of the adaptive immune system, it has been demonstrated that up to ca. 43 60% of individuals unexposed to SARS-CoV-2 show CD4 + T cell recognition of its epitopes (7, 8) and 44 cross-reactivity against seasonal hCoVs (9). Preexisting T cell or antibody responses need careful 45 consideration in vaccine development, with recommendations to assess existing immunity in vaccine 46 trial participants to ensure even distributions between testing groups (3). 47 Compared to studying T cell epitope recognition (which involves living cells, antigen presentation by 48 variable MHC alleles, and rather low affinity interactions (10)), antibody-antigen binding is robust, 49 easily detectable, and amenable to high throughput methods (e.g. (11, 12) ). Antibody tests for hCoV 50 cross-reactivity could be valuable tools to assess preexisting immunity at large-scale and stratify 51 vaccine trials cost effectively. Such serological testing could also inform on a possible impact of 52 seasonal hCoVs to herd immunity against SARS-CoV-2 (13). Yet, cross-reactive antibody responses 53 between SARS-CoV-2, seasonal hCoVs, and their protective potential are incompletely understood. It 54 has been demonstrated that antibodies of COVID-19 patients cross-react against full-length spike (S) 55 and nucleocapsid (N) proteins of SARS-CoV-1 and MERS-CoV (14, 15). However, there are differences 56 in the binding of protein segments, with antibodies binding the receptor-binding domain (RBD) of the 57 SARS-CoV-2 S-protein generally failing to bind this region of SARS-CoV-1 or MERS-CoV (14, 16, 17) . 58 Similarly, no cross-reactivity between antibodies against the RBD of SARS-CoV-2 and seasonal hCoVs 59 NL63/229E was detectable (18). Epitope resolved antibody binding data beyond the S-protein/RBD is 60 scarce for SARS-CoV-2, and virtually unavailable for seasonal hCoVs (19) . 61 Here, we have applied a high resolution antibody assay (11, 12) to test for binding against 1,539 62 peptide antigens covering all known proteins of all hCoVs. We have detected a high seroprevalence of 63 seasonal hCoVs in up to 75% of individuals both unexposed to SARS-CoV-2 and recovered from COVID-64 19, variability in antibody repertoires against SARS-CoV-2, and cross-reactivity against seasonal hCoVs 65 upon SARS-CoV-2 infection. 66 Results and discussion 67 Antibody repertoires against hCoVs and cross-reactivities We have generated a PhIP-Seq library covering all open reading frames of hCoVs as 64 amino acid (aa) 77 sections with 20 aa overlaps between adjacent peptides (Fig. 1b) . The library also includes positive 78 controls which confirmed detection of antibody responses against viruses previously reported to elicit 79 population wide immunity (12) and negative controls, that did not show substantial binding (Table S 80 1). 81 We tested IgG antibody binding against this hCoV library with 32 serum samples of individuals 82 unexposed to SARS-CoV-2, that had been collected in 2013/2014 (23) before the COVID-19 outbreak 83 (Fig. 1c) . These antibody repertoires were compared to 32 samples of recovered COVID-19 patients 84 obtained in April and May 2020. In total we have assayed for nearly 100,000 antibody-peptide 85 interactions ( (false discovery rate) correction for being significantly different between the two groups of individuals 102 (Table S 2) . While nearly all COVID-19 patients showed binding against at least one peptide in S-or N-103 proteins and some peptides being bound in up to 72% of recovered patients (Fig. 1e) , no convergence 104 of antibody responses against the same peptide were detected in all individuals. This finding differs 105 from near universal recognition of some viral epitopes previously observed for other human viruses 106 (12) and replicated with controls in this study (S 1), suggesting that the antibody response against 107 SARS-CoV-2 can exhibit substantial inter-individual variability. 108 COVID-19 serum samples showed also common binding against SARS-CoV-1, indicating detection of 109 cross-reactivity of antibodies targeting SARS-CoV-2 ( Fig. 1f) including a peptide of the MERS-CoV S-protein bound in 28% of COVID-19 sera and 3% (1/32) of 117 unexposed individuals (Fig. 1g) . 118 Strikingly, cross-reactive responses from SARS-CoV-2 also extended to the seasonal hCoVs-OC43 (Fig. 119 1h). One hCoV-OC43 spike peptide was bound in 50% of COVID-19 sera and 3% (1/32) of unexposed 120 individuals, passing FDR correction for being differentially enriched between the two groups (Table S 121 2). Another two peptides of hCoV-OC43 spike were not bound in unexposed individuals at all, but 122 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. and 3% (1/32) of unexposed individuals (Fig. 1i ) not passing FDR correction (Table S 2) . We did not 125 detect cross-reactivities against peptides from the alpha coronaviruses (hCoV-NL63 and CoV-229E), 126 with COVID-19 patients and healthy individuals' sera reacting at similar rates (Fig. 1j,k) . Peptides 127 eliciting cross-reactive antibody responses between SARS-CoV-2, SARS-CoV-1, and seasonal hCoVs 128 typically originated from similar regions of S- (Fig. 2a) and N- (Fig. 2b) proteins. This data indicates that 129 epitopes in the SARS-CoV-2 spike S2 region are frequently bound in COVID-19 patients and a target 130 for cross-reactivity. Many recombinant SARS-CoV-2 vaccines in development focus either on the full-131 length spike or the RBD alone (25). The observed S2 reactivity could lead to differences between these 132 designs, with responses against full-length S-protein vaccines potentially benefiting from cross-133 reactivities against seasonal hCoVs, which may not occur for RBD only vaccines. 134 When looking at the principal components of an analysis (PCA, Fig. 3a Overall, this data indicates that substantial cross-reactivity between hCoVs observed for T cells (7, 8) 140 also extends to antibody responses against seasonal hCoVs. We show that infection with SARS-CoV-2 141 mounts cross reactive antibodies against hCoV-OC43 antigens. The reverse direction of preexisting 142 antibody responses targeting seasonal hCoVs recognizing SARS-CoV-2 is more difficult to assess. We 143 observed binding of two SARS-CoV-2 epitopes in a few unexposed individuals (peptides from NSP2 in 144 19% and N protein in 6% (2/32), Table S 2). The abundance of responses against NSP2 did not change 145 in recovered COVID-19 patients, but binding of the N-protein antigen increased to 66% of recovered 146 patients. We also detected additional SARS-CoV-2 peptides bound in single healthy individuals 147 (including a spike peptide, Fig. 2a ) similar to a population-wide abundance of 3% (1/32) in unexposed 148 individuals reported by a recent study comparing four antibody tests for the SARS-CoV-2 S/N proteins 149 (26). hCoV peptides bound at higher abundance in unexposed individuals than recovered COVID-19 150 patients, would suggest a protective nature. As we did not detect any such peptides our data does not 151 support such a simple protective mechanism. 152 Assessing the protective nature of these population wide preexisting responses would require 153 comparing samples of the same individuals before and after contracting COVID-19 and information on 154 the course (severity) of the disease. The COVID-19 cohort of this study consisted of patients who had 155 experienced mild symptoms, and had not required hospitalization (testing positive in PCR and 156 serological tests). Employing our antigen library to compare antibody profiles against seasonal hCoVs 157 between mild and severe COVID-19 cases could also inform on protective effects of cross-reactivity, 158 as demonstrated for other aspects of the anti-SARS-CoV-2 immune response (27) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. . https://doi.org/10.1101/2020.09.01.20182220 doi: medRxiv preprint 5/15 resolved antibody binding data reported by our assay extends beyond conventional serological tests, 169 as binding against various epitopes of all hCoV needs to be weighed. We used machine learning to 170 build a predictor that highly accurately separated COVID-19 patients from healthy controls based on 171 antibody signatures (AUC=0.96, Fig. 3b ). Depending on the intended application and cutoffs employed 172 (Fig. 3c) , this assay can display 100% specificity at 72% sensitivity (reporting virtually no false-positives) 173 or 94% specificity at 91% sensitivity. Hence, in addition to informing on cross-reactivity between 174 hCoVs, our antigen library could also represent a tool for SARS- provides a unique layer of information unobtainable from working with full length antigens or isolated 188 domains: Given the high resolution of the peptide approach, we pinpoint the exact bound regions 189 revealing crucial targets of cross-reactivities. Our hCoV antigen library can be leveraged to study 190 extended cohorts of patients with mild/severe disease courses or samples collected pre/post 19 infection, and could thereby inform on the potential protective nature of preexisting antibody 192 responses against seasonal hCoVs. In addition, our approach can be extended to other antibody 193 isotypes such as IgA, the primary mucosal antibody. Given the low cost of processing phage displayed 194 libraries in parallel (31), high accuracy (Fig. 3b,c) , and its excellent amenability for robot automation 195 (12, 31), serological testing based on this hCoV library could be a broadly applicable tool to assess 196 preexisting immunity at population-scale (with implications towards protection and herd immunity of 197 SARS-CoV-2) as well as stratifying vaccine trial costs effectively. 198 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. (#0658-12-TLV). Our cohorts of unexposed individuals and COVID-19 patients showed a different sex 330 distribution and minor age differences (Fig. 1c) . While age/sex may influence COVID-19 serology of 331 severe cases (5, 6), we do not expect these parameters to affect key conclusions of our study in mildly 332 affected patients (with both cohorts also showing similar antibody responses against viral controls 333 Table S ORFs reported in the literature (1) (but not annotated in RefSeq NC_045512.2) were added. 347 The final list of proteins were cut to peptides of 64 amino acids (aa) with 20 aa overlaps (to cover all 348 possible epitopes of the maximal length of linear epitope (37)) between adjacent peptides. The 349 peptide aa sequences were reverse translated to DNA using the Escherichia coli codon usage (of highly 350 expressed proteins), aiming to preserve the original codon usage frequencies, excluding restriction 351 sites for cloning (EcoRI and HindIII) within the coding sequence (CDS). The coding was re-performed, 352 if needed, so that a barcode was formed in the CDS, by the 44 nt at the 3' end of each oligo. Every 353 such barcode is a unique sequence at Hamming distance three from all prior sequences in the library, 354 which allows for correcting of a single read error in sequencing the barcode. For similar peptide 355 sequences, alternative codons were used following E. coli codon usage, to achieve discrimination. 356 Including the sequencing barcode as part of the CDS, rather than a separate barcode, allowed to use 357 the entire oligo for encoding a peptide (and as opposed to completely omitting a barcode, it did not 358 require sequencing the complete CDS). After finalizing the peptide sequence, the EcoRI and HindIII 359 restriction sites, stop codon, and annealing sequences for library amplification were added and 360 obtained from Agilent Technologies as 230mer pool (library amplification primers, fwd: 361 GGACCGCGACTGGAATTCT, rev: CCCGGGCATGAAGCTTTCA) and cloned into T7 phages following the 362 manufacturers recommendations (Merck, T7Select®10-3 Cloning Kit, product number 70550-3). 363 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. was computed for all peptides which passed the threshold p-value, all other peptides were given a 404 log-fold-change value of 0. 405 All oligo creation code, and analysis code was written in Python, using the libraries scikit-learn (39), 406 scipy, statsmodels, pandas, numpy and matplotlib. 407 Alignments shown in Fig. 2 were created with CLC Main Workbench 6 (default settings). 408 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. indicates that peptides appearing at low signal strengths in single individuals may arise from 419 nonspecific binding. To eliminate such peptides, the following analyses of viral antigens (with 420 population wide seroprevalence previously reported (12)) was performed using a cut-off of at least 421 two peptides appearing per virus per individual (or peptides occurring in at least four individuals) to 422 count as positive. Following analyses of hCoV binding and predictions were performed with these or 423 even more stringent cutoffs. 424 Serum prevalence for non-hCoV viral antigens were similar between unexposed individuals and 425 recovered COVID-19 patients (b). The slight differences of seroprevalence in unexposed individuals 426 and recovered COVID-19 patients for Human adenovirus C and Influenza A virus antigens may be due 427 the cohort sizes or age/gender differences. 428 429 . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted September 3, 2020. . https://doi.org/10.1101/2020.09.01.20182220 doi: medRxiv preprint Genome Composition and Divergence of the Novel Coronavirus (2019-nCoV) 244 Originating in China Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein. 246 Cell Pre-existing immunity to SARS-CoV-2: the knowns and unknowns Immunology of COVID-19: current state of the science Considering how biological 252 sex impacts immune responses and COVID-19 outcomes The lethal sex gap: COVID-255 19 Targets of T Cell Responses to SARS-CoV-2 Coronavirus in Humans with 257 COVID-19 Disease and Unexposed Individuals SARS-CoV-2-specific T cell immunity in cases of COVID-19 and SARS, and 259 uninfected controls Selective and cross-reactive SARS-CoV-2 T cell epitopes in unexposed 261 humans. Science (80-. ) Sequence analysis of T-cell repertoires in health 263 and disease PhIP-Seq characterization of serum antibodies using oligonucleotide-265 encoded peptidomes Comprehensive serological profiling of human populations 267 using a synthetic human virome Projecting the transmission 269 dynamics of SARS-CoV-2 through the postpandemic period Human neutralizing antibodies elicited by SARS-CoV-2 infection Broad neutralization of SARS-related viruses by human monoclonal 274 antibodies. Science (80-. ) The receptor binding domain of the viral spike protein is an 276 immunodominant and highly specific target of antibodies in SARS-CoV-2 patients Proteolytic processing, deubiquitinase and interferon antagonist activities of 441 Middle East respiratory syndrome coronavirus papain-like protease The Nonstructural Proteins Directing Coronavirus RNA 444 Synthesis and Processing Dissecting antibodies with regards to linear and conformational epitopes Autoantigen discovery with a synthetic human peptidome API design for machine learning software: experiences from the scikit-learn