SARS-CoV-2 Preimmune IgM Epitopes CURRENT STATUS: POSTED

While studying the human public IgM igome as represented by a library of 224087 linear mimotopes, three exact matches to peptides in the proteins of SARS-CoV-2 were found: two in the open reading frame 1ab and one in the spike protein. Joining the efforts to fast track SARS-CoV-2 vaccine development, here we describe briefly these potential epitopes in comparison to mimotopes representing peptides of SARS-CoV , HCoV 229E and OC43.


Introduction
The COVID19 pandemic has put to test the capacity of vaccinology to produce as fast as possible relevant vaccines. A number of recent reports predict possible SARS-CoV-2 epitopes for vaccine development but there are no reports on experimentally defined B cell epitopes [1; 2; 3; 4; 5]. The closest to identification of actual epitopes is the finding of pentapeptide sequences from the viral proteome in other known epitopes form IEDB [5]. A library of 224 087 mimotopes corresponding to the human public IgM repertoire as represented in a plasma pool from 10 000 healthy donors was recently designed [6]. The mimotopes were selected from a commercial 7 amino acid random peptide phage display library (Ph.D. 7, New England Biolabs). Conceptually, this mimotope library reflects at a certain level of detail, the repertoire of IgM specificities in the plasma focusing on the recurring ones.
The latter can be just natural antibodies or they may represent the product of fast extrafollicularly expanding of IgM clones that may serve as precursors of highly specific, somatically mutated, classswitched B cells. The preimmune repertoire has to be quasi-complete to provide for rapid expansion of clones reactive with any newly encountered antigen. The same may not be true for our library although, due to the polyspecific binding, most of the available public repertoire may be partially represented in it [6]. Here we report that the IgM mimotope library contains heptapeptides identical to peptides in the proteome of SARS-CoV-2. One of them may serve as a potentially neutralizing epitope on the spike protein.

Methods
The design and the properties of the mimotope database has been published elsewhere [6]. The available sequences of the genomes of SARS-CoV (NC_004718.3), SARS-CoV-2 (ASM985889v3), HCoV229E (NC_002645.1) and HCoVOC43 (AY391777.1) were split into consecutive overlapping 3 heptamers shifted by one residue and the resultant sequence sets were compared to the sequences in the database of natural mimotopes. Only exact matches were considered.
The homologous sequences in the non-redundant databases of the human proteome and Viridae (taxid:10239) were blast searched using the NCBI blastp suite (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE = Proteins).
As part of an ongoing analysis, the natural mimotope database was represented as a graph by connecting the sequences having at least 5 exact matches (i.e. of maximal Hamming distance 2). The The value of p was used subsequently also in a binomial test to calculate the probabilities of finding equal or higher number of adjacent mimotopes ( Table 1).
The structure of the spike of SARS-CoV-2 was recently published (6vsb.pdb [2]). It was used to visualize the molecular context of the spike epitope found. The visualization of the structure and the calculation of the relative solvent exposed surface were done using UCSF Chimera, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311.

Results And Discussion
A simple comparison for exact matches to peptides from the SARS-CoV-2 proteome yielded 3 heptapeptides -two in the open reading frame 1ab ( 3518 AQTGIAV 3524 and 5198 TKGPHEF 5204 ) and one in the spike protein ( 108 TTLDSKT 114 ). The E value of such a search result is very high and the mere number of sequences is not statistically significant. Yet, this does not refute the fact that 3 heptapeptides which are operationally defined mimotopes of human preimmune antibodies, are part of the viral proteome and, thus, represent (parts of) possible epitopes. On the other hand, the mimotopes in the database sometimes form non-random clusters of homologous sequences much like the mimotopes selected by a single monoclonal antibody. Each one among 224087 randomly selected heptamers should have on the average 1.33 homologous sequences in that same database that differ from it by up to 2 mismatches. As seen from Table 1, all SARS sequences but not those from trivial HCoV were members of clusters significantly greater than random (Binomial test, p<0.05, false discovery rate adjusted). This is an indication that the presence of these sequences is non-random and they represent clusters of mimotopes representing individual (poly)specificities.
An important prerequisite for the functionality of these epitopes is their degree of exposure to the solvent. The recently published structure of the spike (S protein) of SARS-CoV-2 [2] shows that 108 TTLDSKT 114 forms a loop exposed to the solvent (Fig 1A.). The relative solvent exposed surface greatly exceeds the threshold of 5% for participating in contacts (Fig.1B). This loop is adjacent to the loop representing the epitope of the neutralizing antibody LCA60 on the SARS-CoV spike [7; 8].
Presumably, it is similarly exposed further in the open conformation of the spike domains. The adjacent N-glycosylation sites are N165 and N234. Dependent on the size of the carbohydrate sidechains, they may partially occlude the epitope.
The closest sequences in the human proteome are 540 tlTLDSKT 547 of the prostate-specific transglutaminase (TGM4) and 462 TTLDSKi 468 of mucin-16 (also known as ovarian tumor marker CA125, Q8WXI7.3, [9]). Both are on tumor associated antigens [9; 10]. While TGM4 is an intracellular 5 antigen, mucin-16 is highly accessible on cell surfaces and in a soluble form. The mucin sequence 462 TTLDSKI 468 is T/S biased, represents part of the highly O-glycosylated N-terminal part of mucin-16 and is predicted to be O-glycosylated itself. Normally, such mucin protein core epitopes are occluded by glycosylation and thus, cryptic with respect to immune tolerance. Yet, monoclonals to similar epitopes turned out to bind specifically to tumour expressed mucins [11; 12; 13; 14] which are aberrantly/hypo glycosylated.
The sequences 108 TTLDSKT 114 has several exact matches in viruses outside the family Coronaviridae in hypothetical proteins of various phages. At least one of them infects L. plantarum which is a common species in the gut microbiome.
It is not surprising that the public IgM repertoire has clones potentially capable of binding to nonconserved regions of novel viruses. Similarly, the IgM igome contained sequences found also in SARS-CoV, although the epidemic was too restricted to be reflected in the antibody repertoires of the donors ( Table 1). As expected, no signs of persistent antibody titers after exposure were observed.
The representation of clones reactive with the trivial human coronaviruses 229E and OC43 was rather narrower than that of the unknown strains. Some of the epitopes were conserved between SARS-CoV and SARS-CoV-2 (AQTGIAV and TKGPHEF) but they were found in non-structural proteins and are hardly targets for neutralizing antibodies (Table 1). On the other hand, all potential epitopes found could play a role in targeting the viral proteins to specific B1 cells which produce the bulk of natural IgM. The latter are known to be excellent antigen presenting cells able to prime CD4 + T cells, and initiate immune responses [15] in antigen specific manner much like activated specific B2 cells [16]. It has been shown that B1 cells secreted IgM is a non-redundant and essential arm of the humoral responses to influenza in mice [17]. This implies that natural antibody epitopes might be essential components of subunit vaccines even though they may not represent typical dominant epitopes. The role of overlapping T and B cell epitopes is not clear except when the B cell receptor has a high enough affinity for the epitope to protect it during processing [18], but it is interesting that one of the SARS-CoV natural epitopes ( 922 TTSTALG 928 ) is also part of a CD4 T cell epitope in the context of HLA-6 DR 0401 [19].
The current thinking separates the repertoire of natural and induced antibodies [20]. The preimmune IgM epitopes we describe could be epitopes of naïve B clones which are poised to initiate an immune response or epitopes of natural antibodies. As to the capacity of these epitopes to induce fully mature antibody response, it is interesting to note that the two preimmune IgM epitopes found for the spike of SARS-CoV (922TTSTALG928 and 389VKGDDVR395) are proven antibody targets in approximately one fourth of the SARS patients [21]. Furthermore, a recent report indicates the importance of IgM antibodies in the control of the diseases in mild case of COVID19 [22]. Thus, it is quite possible that the SARS-CoV-2 spike epitope TTLDSKT is bound by B cells that will contribute to the induced immune response. The correlation between the actual reactivities in SARS-CoV patients' sera and the Bepipred score ( Fig. 1C) confirms the low power of linear B cell epitope predicting algorithms, and underlies the necessity to base the proposals of new epitopes as much as possible on actual binding data.
These considerations make the novel SARS-CoV-2 epitopes valid targets in the search for a vaccine for COVID-19. The whole paradigm followed here focuses exclusively on the relatively rare linear epitopes. A lot more information about conformational epitopes may be hidden in the natural mimotope database but the approaches for sorting out clusters of mimotopes defining a conformational epitope are still being developed. The proposed actual preimmune IgM epitopes of SARS-CoV-2 can be instrumental both as parts of subunit vaccines or in the design of nanoparticlebased vaccines but also in the development of therapeutic monoclonal antibodies.  (based on [21]). The two predicted natural epitopes are overlaid in red. There is antibody reactivity in patients sera to these epitopes although one of them has Bepipred score far bellow the threshold of 0.5.