Homologies between SARS-CoV-2 and allergen proteins may direct T cell-mediated heterologous immune responses

The outbreak of the new Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2) is a public health emergency. Asthma does not represent a risk factor for COVID-19 in several published cohorts. We hypothesized that the SARS-CoV-2 proteome contains T cell epitopes, which are potentially cross-reactive to allergen epitopes. We aimed at identifying homologous peptide sequences by means of two distinct complementary bioinformatics approaches. Pipeline 1 included prediction of MHC Class I and Class II epitopes contained in the SARS-CoV-2 proteome and allergens along with alignment and elaborate ranking approaches. Pipeline 2 involved alignment of SARS-CoV-2 overlapping peptides with known allergen-derived T cell epitopes. Our results indicate a large number of MHC Class I epitope pairs including known as well as de novo predicted allergen T cell epitopes with high probability for cross-reactivity. Allergen sources, such as Aspergillus fumigatus, Phleum pratense and Dermatophagoides species are of particular interest due to their association with multiple cross-reactive candidate peptides, independently of the applied bioinformatic approach. In contrast, peptides derived from food allergens, as well as MHC class II epitopes did not achieve high in silico ranking and were therefore not further investigated. Our findings warrant further experimental confirmation along with examination of the functional importance of such cross-reactive responses.

We have previously reported on heterologous immune responses induced by 113 influenza, another respiratory RNA virus, against allergens, which mediated protection 114 from experimental allergic asthma. 4 Indeed, virus-induced T cell mediated 115 heterologous immunity has been widely described in a variety of settings, which can 116 confer protection or drive immunopathology against other antigens. 5,6 Given that the 117 host immune response to SARS-CoV-2 and associated disease course can be so 118 varied from patient to patient, this spectrum of presentations raises the question of 119 what drives the differential host immune response. There is still little known about 120 asthma phenotypes and severity of COVID-19. In general, asthma has not been shown 121 to be a risk factor for COVID-19 in several published cohorts. 7,8 However, recent 122 studies from the UK and the USA indicated higher numbers of asthmatics in COVID-Interestingly, the UK Biobank recently reported that non-allergic patients had a higher 125 risk of severe COVID-19, compared to patients with allergic asthma 10 . Moreover, 126 increased numbers of activated T cells were found among asthmatic COVID-19 127 patients who showed a less severe disease, suggesting that activated T cells have a 128 positive impact on severity of SARS-CoV-2 infection 11 . These preliminary clinical 129 observations along with our prior experimental evidence involving RNA viruses led us 130 hypothesize that SARS-CoV-2 may share a degree of protein sequence homology to 131 allergens, which may lead to the generation of cross-reactive T cell epitopes. Pre-132 existing T cells specific for such cross-reactive allergen-derived epitopes may have an 133 impact on COVID-19 outcome via aberrant cytokine responses to the virus peptides. 134 Indeed, these cytokines could prevent an overshooting T1 inflammatory reaction, both 135 locally (as in the case of preexisting pulmonary CD4 + T cells specific to inhalant 136 allergens) and/or systemically. Therefore, we sought to predict potentially cross-137 reactive allergen-and SARS-CoV-2-derived MHC Class I and Class II T cell epitopes, 138 which can be presented by the most prevalent HLA alleles. 139 140

Methods and Results 141
In order to examine our working hypothesis, we applied two distinct independent, 142 complementary and systematic bioinformatics approaches (Figure 1  further processed for T cell epitope prediction using netMHC 23 and netMHCpan 24 for 158 MHC Class I, and netMHCII and netMHCIIpan 25 for MHC Class II prediction (affinity 159 score threshold for strong binders: 0.500; for weak binders: 2.000). Viral and allergen 160 epitopes were pairwise aligned with Biopython module pairwise 2 26 and for pairs with 161 a score > 8, a final pair combined score (pcs) was calculated (Supplementary 162 Methods). Duplicates among the resulting candidate epitope pairs were removed 163 before further processing. Therefore, possible sequence repetition due to isoforms and 164 Table S1) do not influence further analyses. In total, we 165 obtained more than 5000 candidate pairs for each, MHC Class I and Class II. The top 166 30 candidate epitope pairs, as per pair combined score, are listed for aero-and food 167

allergens, MHC Class I and Class II presentation background in Supplementary 168
Tables S4-7, respectively. The top 30 MHC Class II restricted predicted virus-allergen 169 pairs achieved relatively low pcs (24-657) as compared to Class I epitope pairs (1036-170 10816). Among our top 30 MHC Class I potentially cross-reactive allergen derived 171 epitopes, we identified more than 20 distinct protein families (Allfam database). In 172 addition to MHC binding affinity and homology between peptide sequences, also other 173 factors (e.g. conservation, association with clinical reactions) are important for the to capture this information level in our ranking, all allergen peptides and associated 176 sources listed among the top 30 candidate epitope pairs were evaluated further with a 177 scoring system (Supplementary Fig. S1 and Supplementary Methods). We found 178 that the top 5 Class I aeroallergens were on average associated with higher pcs as 179 compared to the top 5 potentially cross-reactive food allergens ( Table 1 for MHC Class 180 I and Table 2 for MHC Class II peptide pairs). 181

Pipeline 2 182
We obtained all known allergen-derived linear T cell epitope peptides from the IEDB, 183 containing peptides known to bind MHC molecules with at least one published 184 experimental evidence (e.g. based on the results of a T cell assay) (Supplementary 185 Table S8). A total of 8,207 antigenic peptides from 142 antigens were selected for 186 evaluation, among which, peptides with ambiguous amino acids (e.g. with unknown 187 amino acid 'X' or any special character) were removed from the subsequent analysis. 188 Therefore, all included peptides could be defined in full. Next, SARS-CoV-2 protein 189 sequences were analyzed for the potential antigenic regions by splitting each of the 190 sequence into sequential k-mers (length=15), and homology with allergen antigenic 191 peptides was then profiled. Within a given threshold range, we found 43 unique SARS-192 CoV-2 peptides that belong to replicase poly protein and spike glycoprotein 193 Table S9). These peptides demonstrate homology with antigenic 194 peptides of 6 different allergens, all of which are known to be respiratory allergens (e.g. 195 aeroallergens; Figure 1). However, despite the homology, it is likely that some of the 196 peptides may not have strong MHC Class I binding affinity, and thus be less likely to 197 be presented as antigens by HLA molecules. Therefore, we assessed the binding 198 affinity of these peptides with human MHC Class I molecules, across a broad range of 199 alleles that are known to bind viral proteins (52 most common HLA-A and HLA-B 200 alleles). We observed that some of these peptides (n=79) were predicted to have MHC Class I binding epitope regions associated with at least one of the Class I HLA alleles 202 with IC50 < 500nm (Supplementary Table S10 Table  212 S11) with 14 high confidence HLA Class I binding peptides with IC50 < 50nm (Table  213 3). Quite importantly, the SARS-CoV-2 Nsp6141-149, which was identified among our top 225 potentially cross-reactive epitope pairs, has been recently described by an 226 independent group. 29 To our knowledge, this is the first report on in silico predicted T cell epitope cross-reactivity between SARS-CoV-2 and allergens. While a limitation of 228 our study is the in silico nature of the work, the sequence homology between SARS-229

CoV-2 and clinically relevant respiratory allergens is along the lines of previously 230
reported cross-reactivity between RNA virus-and allergen-derived peptides at the level 231 of T memory cells. 4 Moreover, our current findings generate further hypotheses in how 232 the adaptive immune system responds differentially with respect to the atopy status of 233 the host. Our present study warrants an immediate investigation of these predicted T 234 atopy/asthma and COVID-19. Indeed, the role of SARS-CoV-2-specific T cells in 258 exposed and non-exposed individuals, thereby underlining the importance of 259 heterologous immunity, has been very recently described 33 Table 3. HLA-I binding high confidence (IC50 < 50nm) SARS-CoV-2 antigenic peptides (pipeline 2)