In March 2020, the World Health Organization (WHO) declared the disease known as COVID-19, caused by the Severe Acute Respiratory Syndrome Coronavirus (SARS-CoV-2), as a global pandemic. As of the 2nd of August 2021, there have been over 198.2 million confirmed cases and over 4.3 million deaths reported (WHO, 2020a). COVID-19 patients typically present with symptoms such as fever, dry cough, tiredness, as well as others such as myalgia, sore throat, loss of taste or smell, red eyes, diarrhoea and skin rash (Huang et al., 2020; Meng et al., 2020; WHO, 2020b). The SARS-CoV-2 genome encodes a series of structural and non-structural proteins (Chan et al., 2020; Wu et al., 2020). There are four structural proteins consisting of the surface glycoprotein (spike, SP), nucleocapsid phosphoprotein (nucleoprotein, NP), membrane (M) and envelope (E). There are additionally eight open reading frames (Orf) encoding the non-structural proteins: Orf1ab, Orf3a, Orf3b, Orf6, Orf7a, Orf7b, Orf8 and Orf10. Multiple studies have used serological assays to measure antibody responses to the SARS-CoV-2 virus (Long et al., 2020; Zhao et al., 2020; Post et al., 2021). Antibodies are usually evaluated for reactivity against the spike and the nucleoprotein with the aim of understanding seroconversion, as well as correlating disease severity with neutralizing antibody titres (systematically reviewed in Post et al. (2021)). Additional reports have identified increases in antibodies to other SARS-CoV-2 proteins, including Orf proteins, after infection (Hachim et al., 2020). Since the beginning of the pandemic, variants of SARS-CoV-2 have been emerging and circulating worldwide (van Dorp et al., 2020). Some mutations, such as K417N and E484K in the spike protein, are immune evasion mutations (Altmann et al., 2021). Since the second half of 2020 variants of SARS-CoV-2, containing mutations such as these, have arisen with increased transmissibility, associated with more severe disease and a reduction in neutralizing antibodies. These variants have been called variants of concern (VOC) or variants of interest (VOI). As of July 2021, there are four VOCs and four VOIs classified by the Centers for Disease Control and Prevention (CDC) (SARS-, 2021). The VOCs consist of the strains originating in the United Kingdom (Alpha strain), Brazil (detected in Japan, Gamma strain), South Africa (Beta strain), and India (Delta strain). Whereas VOIs consist of strains originating in Canada (Eta strain), United States (Iota strain), India (Kappa strain) and Peru (Lambda strain).

Viral infections generally can play a role in promoting or exacerbating autoimmune diseases (reviewed in Smatti et al. (2019)). Proposed mechanisms for this association include molecular mimicry and bystander activation (Fujinami et al., 2006; Smatti et al., 2019). Molecular mimicry refers to the phenomenon of immune cross-reactivity between a foreign pathogen and self-protein, where an immune cell recognizes both due to their sequence similarity (Fujinami et al., 2006; Smatti et al., 2019). Bystander activation refers to the activation of self-reacting immune cells driven by the liberation of self-antigens (the targets which stimulate immune responses) from virus-lysed cells which otherwise would not be typically exposed to the immune system (Fujinami et al., 2006; Smatti et al., 2019). Given that a range of extrapulmonary pathologies have been reported in COVID-19 (Cheng et al., 2020; Cheung et al., 2020; Han et al., 2020; Hundt et al., 2020; Nalleballe et al., 2020; Oxley et al., 2020; Poyiadji et al., 2020; Xie et al., 2020; Zheng et al., 2020), molecular mimicry has been hypothesized to be playing a role (Angileri et al., 2020a; Cappello et al., 2020; Marino Gammazza et al., 2020). Additionally, conditions similar to the autoimmune diseases Guillain-Barré Syndrome (Ameer et al., 2020; Korem et al., 2020) and Kawasaki disease (Bordet et al., 2020; Jones et al., 2020) have been reported in COVID-19 patients. Within this Kawasaki-like disease, termed Multisystem Inflammatory Syndrome in children (MIS-C), autoantibodies targeting a range of antigens including, but not limited to, anti-La, anti-Jo-1, anti-MUC15, anti-P2RX4, anti-MAP2K2 and anti-CSNK1A1 have been reported (Consiglio et al., 2020; Gruber et al., 2020). Additionally, autoantibodies targeting other antigens including type I interferons (IFNs) (Bastard et al., 2020) have been identified in COVID-19 positive patients (Zhou et al., 2020a; Bastard et al., 2020; Vlachoyiannopoulos et al., 2020) alongside key autoimmune associated antigens which are often associated with tissue damage, for example, anti-nuclear antigen (ANA) (Zhou et al., 2020a; Vlachoyiannopoulos et al., 2020). It is therefore highly likely autoantibodies that recognize a range of other self-proteins are induced by infection with SARS-CoV-2.

Computational biology methods such as immune epitope mapping and Basic Local Alignment Search Tool (BLAST) allow for relatively quick predictive screening analysis. They can help narrow findings and hypotheses, leading to more targeted laboratory-based validation work. In the context of COVID-19, understanding immune epitopes, the sequences that are recognized by antibodies, have been used to predict potential peptide-based vaccine candidates or to increase sensitivity in serological assays (Ahmed et al., 2020; Amrun et al., 2020; Poh et al., 2020). In the initial stages of the COVID-19 outbreak, B-cell epitope mapping was performed in SARS-CoV-2 to check for potential cross-reactivity of immune responses with other coronaviruses (Ahmed et al., 2020; Grifoni et al., 2020). Recently, there have been a small number of studies reporting sequence similarities between SARS-CoV-2 proteins and human proteins, in support of the molecular mimicry hypothesis (Angileri et al., 2020a; Angileri et al., 2020b; Ehrenfeld et al., 2020; Kanduc, 2020; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Lyons-Weiler, 2020; Marino Gammazza et al., 2020). With the exception of one study, which used longer epitopes (Lyons-Weiler, 2020), most of these studies sought to find short regions (5-6aa) of exact identity between the SARS-CoV-2 proteins and human proteins (Angileri et al., 2020a; Angileri et al., 2020b; Kanduc, 2020; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020). Whilst these studies suggested a number of regions, most of the identical sequences identified would be unlikely to sustain a pathogenic autoreactive response (Angileri et al., 2020a; Kanduc, 2020; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020). Specifically, while assessing its potential accessibility for antibody binding, these studies did not discriminate the localization of the epitope within the protein or within the cell (Angileri et al., 2020a; Kanduc, 2020; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Lyons-Weiler, 2020; Marino Gammazza et al., 2020). Moreover, except in the case of Kanduc (2020), who identified identical hexamers between SARS-CoV-2 and human proteins associated with various system disorders, these studies only report small numbers of human proteins with similar (Lyons-Weiler, 2020) or short identical sequences (Angileri et al., 2020a; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020) to SARS-CoV-2 proteins (e.g. association with the immune system, molecular chaperones, brainstem proteins). Therefore, the limited nature of these studies does not show the full extent of similarities between SARS-CoV-2 and human proteins.

In the present study we have pioneered a different immunoinformatics approach to overcome some of these limitations and further explore the potential for autoimmune cross-reactivity driven by SARS-CoV-2 molecular mimicry. B-cell epitopes within the structural proteins and several Orf proteins (shown to generate antibody responses in COVID-19 positive patients (Hachim et al., 2020)), were first predicted using the IEDB B Cell epitope tool. Epitopes predicted in the structural proteins were compared to the current VOCs (Alpha, Beta, Gamma and Delta) and VOIs (Eta, Iota, Kappa and Lambda). Eight of the fifteen epitopes contained at least one mutation in one variant. The identified 9-53aa long sequences, in both structural and Orf proteins, were then compared to sequences in human proteins using the NCBI protein BLAST tool (Altschul et al., 1990). Assessment of potential for cross-reactivity between SARS-CoV-2 and self-proteins with capacity to perpetuate autoimmune pathology was based on a combination of immunologically relevant sequence similarity (not just identity) (Angileri et al., 2020a; Kanduc, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020) and the localization of the protein itself, with a focus on extracellular targets. 11 human proteins, containing amino acid sequences similar to nine predicted SARS-CoV-2 B cell epitopes, were identified based on these selection criteria. These findings indicate that antibodies induced by SARS-CoV-2 could directly interfere with cell function, including that of immune cells, and could help explain some of the additional pathologies identified in COVID-19 patients (Cheng et al., 2020; Cheung et al., 2020; Han et al., 2020; Hundt et al., 2020; Nalleballe et al., 2020; Oxley et al., 2020; Poyiadji et al., 2020; Xie et al., 2020; Zheng et al., 2020). Finally, comparing the sequences of both predicted spike epitopes and full length spike protein to various human proteins implicated in immune thrombocytopenia purpura (ITP) and thrombocytopenia syndrome (TTS), our results indicate that molecular mimicry is unlikely to be the cause of TTS, or vaccine induced prothrombotic immune thrombocytopenia (VIPIT) following vaccination with the COVID-19 adenovirus vector vaccines. To our knowledge, this is the first study to compare immune epitopes across the circulating VOCs and VOIs; highlight multiple similarities between the selected Orf proteins and human proteins; identify proteins with reported associations to autoimmunity as sharing sequences with SARS-CoV-2 epitopes; and to highlight novel extracellular human proteins which may have antibody cross-reactivity with SARS-CoV-2 immunogenic regions.

Using in silico immunoinformatic tools, potential B cell immunogenic epitopes in the SARS-CoV-2 proteome were predicted and further used to compare to global variants as well as explore the similarity to human proteins. In doing so, we identified eight structural epitopes containing mutations in at least one strain within these immunogenic regions. When comparing the epitopes to the human proteome, a variety of human proteins were identified to share sequences similar to SARS-CoV-2 proteins. Many of the identified human proteins were found to be associated with diseases, some of have which been reported to be related to COVID-19 symptoms and complications. Additionally, we show associations of these proteins to autoimmune diseases, such as SLE and MS. We further identified sequence similarities between SARS-CoV-2 immunogenic regions and human proteins which are localized in the extracellular region. These similarities and potential ease of access to circulating antibodies suggests the potential damaging cross-reactivity that can perpetuate a pathological condition. Finally, we analyzed and found that molecular mimicry may not be the mechanism for the thrombosis and thrombocytopenia syndrome occurring following vaccination with the AstraZeneca and Johnson & Johnson COVID-19 vaccines.

To the best of our knowledge, this is the first study to predict B cell epitopes and compare to the highlighted VOCs and VOIs known to escape the immune response and be more infectious. However, a recent study aligned 10,664 SARS-CoV-2 genomes, to identify conserved regions and predicted both B and T cell epitopes specifically within these regions (Ghosh et al., 2021). Of their highlighted B cell epitopes, only our single predicted membrane epitope crossed over. This may be due to different epitope prediction algorithms used or single mutations in these genomes eliminating regions of interest due to not being conserved. While this method, may help in the design of epitope-based synthetic vaccines, due to targeting conserved regions, it does not indicate the immunogenic regions of the full proteins and how mutations may impact immune responses. Typically, the mutations reported in the VOCs and VOIs are those that lie in the spike protein due to its key role in infection and pathogenesis (Shang et al., 2020). D614G was one of the earliest mutations in variants that emerged as more infectious than the initial SARS-CoV-2 variant and became globally dominant (Korber et al., 2020). Each of the VOCs contain this mutation along with various others. The Alpha strain contains a N501Y mutation in the ACE2 receptor binding domain (RBD) (Garcia-Beltran et al., 2021). Examples of other mutations characteristic to the different strains used in the present study include K417N/T and E484K (Garcia-Beltran et al., 2021), and P681H, L425R, P681R, E484Q, among others (COVID-19 Weekly Epidemiol, 2021). Of these, one mutation of interest in the spike protein, P681H/R, is located next to the furin cleavage site which is important for invading host cells (Papa et al., 2021; Peacock et al., 2021). Specifically, P681R is of interest as it is found within the Delta variant. Following the wave of devastation in India, with its increased transmissibility, the Delta variant has become the dominant variant in multiple countries (Bolze et al., 2021; Mishra et al., 2021; Sheikh et al., 2021). An early study suggests mutation P681R is associated with the increased viral pathogenicity (Saito et al., 2021). This mutation site can be identified within the predicted epitope SP675-687. This epitope is the only spike predicted epitope that contained at least one mutation in multiple VOCs (Alpha, Beta and Delta) and VOIs (Eta and Kappa) but is also the only predicted spike epitope not overlapping a validated epitope. Despite this, there may still be antibody responses to this epitope in the population, and therefore the mutations within the SP675-687, such as P681R, may be impacting antibody binding due to structure and charge changes that could be impacting a functional role of these antibodies. Among the other predicted spike epitopes only SP249-261 contained at least one mutation in VOIs, Iota and Lambda. With no mutations identified in the other three spike epitopes, this suggests that although found overlapping validated regions, these antibodies may not play a role in viral neutralization. We additionally found mutations in at least one variant in six of the nine predicted nucleoprotein epitopes. NP164-216 was found to contain multiple mutations among all VOCs and VOIs examined, all of which were around the same region (position 199-205 and 214). It has recently been reported that high anti-NP responses may be associated with poorer outcomes in COVID-19 (Batra et al., 2021). With more severe disease associated with these variants, it raises questions of whether the mutations in these epitope regions increase antibody binding, rather than decrease binding, or whether the different strains have unique epitope regions.

COVID-19 is associated with a series of multi-organ complications (Huang et al., 2020; Zaim et al., 2020). Many of the human proteins identified in this study, that share amino acid sequence similarities with the SARS-CoV-2 virus, play key roles in cellular functions, which if interrupted may result in altered cell function and therefore pathology. We found that clusters of proteins could be grouped based on their relationship to similar diseases and overlap to multiple body systems, some of which have been implicated in COVID-19 pathology, including respiratory, cardiovascular, gastrointestinal tract and nervous systems. Some of the broad examples of such diseases include epilepsy, cardiomyopathy and cerebellar ataxia, all of which have been reported in COVID-19 patients (Siripanthong et al., 2020; Mithani et al., 2021; Povlow and Auerbach, 2021; Werner et al., 2021). However, some of the diseases associated with the similar proteins may not result in a complication but instead confer a higher risk. Alzheimer’s Disease, macular degeneration and cardiovascular diseases were all diseases identified with proteins that shared sequence similarities to SARS-CoV-2 capable of making them the targets of autoantibodies. Pre-existing diagnosis for each of these have been found to predict higher risk of infection and greater severity and risks in COVID-19 (Chen et al., 2020; Huang et al., 2020; Ramlall et al., 2020; Yu et al., 2021). Many of the proteins identified to be associated with disease are intracellular and are therefore less likely to be immune targets. However, as the SARS-CoV-2 virus is an intracellular pathogen, the sequence similarities could alternatively have an impact on cellular functions which may result in the observed pathologies, independently of having the potential to be recognized by antibodies.

Autoantibody (AAb) targeted proteins found within a range of autoimmune diseases (SLE, Myasthenia Gravis, T1D etc.) were found to share similar sequences to some of the new predicted SARS-CoV-2 epitopes, as well as known SARS-CoV-2 epitopes (Shrock et al., 2020; Wang et al., 2020). Studies have shown that within COVID-19 patients, known AAbs associated with autoimmune diseases, including but not limited to, anti-cardiolipin, anti-SSA/Ro and anti-nuclear antibody (Zhou et al., 2020a; Vlachoyiannopoulos et al., 2020) are increased, indicating a breaking of immune tolerance (the mechanisms which regulate responses and ensure immune cells do not attack self). As identified in the present study, histone H3 shares an identical 6aa sequence with the SARS-CoV-2 Orf8 protein, which additionally sits in a region identified as an epitope in COVID-19 patients (Wang et al., 2020), indicating the possibility of immune cross-reactivity. Gravin, the myasthenia gravis autoantigen, was also found to have sequence similarity with the Orf3b viral protein. While, to our knowledge, anti-gravin AAbs have not been reported in COVID-19 patients, there has been a case report of post-COVID-19 infection onset of myasthenia gravis (Huber et al., 2020). With several viral infections being associated with autoimmune diseases, these data support the possibility that the COVID-19 pandemic will lead to an increase in autoimmune diseases. Anti-La and anti-Jo-1, typically associated with SLE/Sjogren’s Disease and inflammatory myopathies respectively, have been identified in children with COVID-19 Multisystem Inflammatory Syndrome (MIS-C) (Consiglio et al., 2020; Gruber et al., 2020). However, it is not only AAbs with known associations to autoimmune diseases found in COVID-19 patients (Consiglio et al., 2020; Gruber et al., 2020). AAbs to a range of tissue specific and immune related mediators have been identified in MIS-C patients. Additionally, increased levels of anti-interferon (IFN) antibodies have been reported to be associated with more severe COVID-19 disease (Bastard et al., 2020). This shows that it is not only important to understand the targets associated with autoimmune diseases, but that any similarity or recognition of self may contribute to additional COVID-19 pathology and/or greater disease severity.

The majority of proteins found to fit within our distinct criteria for the potential of cross-reactivity were intracellular, with only 11 found to be extracellular. By applying these criteria, two proteins associated as antigens in autoimmune diseases, Tubby-related protein 3 and Alpha-internexin, were also identified as potential cross-reactive proteins. We additionally found that tubby-related protein 3, as well as two other potential cross-reactive targets, bone morphogenetic protein 1 and Mucin-12, have alignments similar to SARS-CoV-2 epitopes which overlap with in vivo validated epitopes (Li et al., 2020; Shrock et al., 2020; Wang et al., 2020), potentially indicating an opportunity for cross-reactivity, especially as these targets are extracellular. Additionally, as seen in autoimmune diseases, many autoantibodies to various intracellular targets are identified (Suurmond and Diamond, 2015). During COVID-19 infection, the release of intracellular proteins may be playing a role in breaking immune tolerance, allowing for the potential cross-reactivity of SARS-CoV-2 antibodies to self, or the increase of AAbs. This may be perpetuated through the recognition of the regions that are extracellular and therefore more likely to be visible to circulating antibodies. Interestingly, the SARS-CoV-2 spike and nucleoprotein alignments with human protein matches conforming to our criteria, are ones that do not contain mutations across different global variants, suggesting the potential for cross-reactivity to these proteins irrespective of the virus variant. Furthermore, according to the IEDB, there are several confirmed discontinuous B cell epitopes within the spike protein, which may also further expand the number of cross-reactive epitope targets to human proteins to be explored in future studies.

Other studies have reported sequence similarities between SARS-CoV-2 and human proteins (Angileri et al., 2020a; Angileri et al., 2020b; Ehrenfeld et al., 2020; Kanduc, 2020; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Lyons-Weiler, 2020; Marino Gammazza et al., 2020), the majority through identifying identical 5 or 6 amino acid segments found within immunogenic regions (Angileri et al., 2020a; Angileri et al., 2020b; Ehrenfeld et al., 2020; Kanduc, 2020; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020). Apart from the study of Kanduc (2020), who formed overlapping hexamers (offset by one residue) from SARS-CoV-2 sequences that are identical in validated SARS-CoV-1 epitopes, these studies began by aligning overlapping pentamers/hexamers to human proteins to identify identical sequences (Angileri et al., 2020a; Angileri et al., 2020b; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020). These identical alignments were then determined if they were found in immunogenic regions. In contrast, Lyons-Weiler (2020) predicted immunogenic SARS-CoV-2 epitopes which were then compared to human proteins. In the present study, we used a similar approach to Lyons-Weiler (2020) by identifying sequences in key SARS-CoV-2 proteins with higher predicted immunogenic regions. However, the comparison between these SARS-CoV-2 sequences and human proteins was further expanded, in contrast to the other reports (Angileri et al., 2020a; Angileri et al., 2020b; Ehrenfeld et al., 2020; Kanduc, 2020; Kanduc and Shoenfeld, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020), as we did not only rely on short identical sequences. Instead, to explore potential immune cross-reactivity, we applied a novel combination of criteria, which considered the ability of antibodies to potentially recognize amino acid differences, as well as the localization of the alignments for antibody accessibility. Although our criteria were unique compared to the previous studies, we’d expect some crossover with those who report identical hexamers (Angileri et al., 2020b; Ehrenfeld et al., 2020; Kanduc, 2020; Lucchese and Flöel, 2020; Marino Gammazza et al., 2020), and had a similar start methodology (Lyons-Weiler, 2020). However, to the best of our knowledge only five previously reported proteins were also found in our potential targets: Hermansky-Pudlak Syndrome I protein, Unconventional myosin XVI, Transmembrane Protein KIAA1109 (Ehrenfeld et al., 2020), Ankyrin repeat and sterile alpha motif domain containing 1A (Lyons-Weiler, 2020) and CLOCK (Kanduc, 2020). Each of these proteins reported the same alignment to the same SARS-CoV-2 protein, except for CLOCK which was reported with a different hexamer alignment suggesting some proteins may have multiple shared sites with SARS-CoV-2. Other similar findings may not have been identified as these reports focus only on human proteins of a certain function or location (e.g. molecular chaperones or adaptive immune system proteins), some of which were found to match SARS-CoV-2 proteins not studied in the present study (e.g. orf1ab). Differences in findings may also be due to different in silico tools used for the comparison of SARS-CoV-2 and human proteins, where we used the NCBI blastp tool while others used the Pir Peptide Match program (Kanduc, 2020; Kanduc and Shoenfeld, 2020; Marino Gammazza et al., 2020). It was found that the Pir Peptide Match program preferred short sequences (∼6aa length) to identify the similar human proteins but could not obtain the same results when using the full-length epitopes (>10aa long). Whereas the NCBI blastp tool allowed us to apply criteria that not only relied on identical sequences but considered how similar charge and structure between amino acid variations may not impact antibody binding and therefore explore sequence similarity between the full length of the predicted epitopes and human proteins. Additionally, not all pentamers or hexamers from the SARS-CoV-2 proteins, reported to align with human proteins, can be found within our predicted epitopes, further explaining differences in the human proteins identified. In some cases, this may be due to different prediction algorithms used, or the sequences were part of prediction epitopes that did not fit our specific epitope criteria. One sequence, SRSSSR, found within the nucleoprotein, has been highlighted in multiple reports as having a shared alignment with human proteins (Angileri et al., 2020b; Kanduc, 2020). This hexamer can be found within our predicted epitope NP164-216, however this was one of the epitopes that obtained no significant blast results. This example demonstrates how there can be major differences in outputs between different bioinformatic methods.

Computational methods, such as epitope mapping and blasting are useful techniques. They can allow narrowing of questions and potential proteins of interest in hypotheses before doing experimental studies. However, while bioinformatics studies can narrow these lists, the current study is limited to predictions alone and ultimately need to be confirmed experimentally. Since the initial epitope predictions were performed, new literature emerged to identify new SARS-CoV-2 B cell epitopes in vitro that were also identified by our approach which have not been reported previously (Amrun et al., 2020; Poh et al., 2020; Shrock et al., 2020; Wang et al., 2020; Yi et al., 2020; Yoshida et al., 2021). Several of our predicted epitopes, including: Spike 249-261, 597-606, 805-816, 1256-1265; NP 164-216, 232-269, 361-390; Orf3a 172-197, 216-225, Orf8 23-45, 48-56 overlap with the new epitopes highlighted in these reports. Regions within spike 553-579 and 806-835 have been identified across several studies as immunodominant regions (Amrun et al., 2020; Poh et al., 2020; Yi et al., 2020), whereas other epitopes show differing ranges of reactivity (Shrock et al., 2020). This indicates that, although not immunodominant or reported in the literature, these predicted epitopes may still be real, especially as individual immune responses are polymorphic. This is important to consider when mapping potential cross-reactivities between SARS-CoV-2 and self-antigens, as there may still be protein cross-reactivities to identify due to host or virus polymorphisms. Additionally, to allow for direct sequence similarities between virus epitopes and human proteins, only linear B cell epitopes were explored in the present study. However, B cell epitopes can be composed of multiple discontinuous segments (Rubinstein et al., 2008). As more protein structures covering the full-length SARS-CoV-2 proteins are identified, performing discontinuous epitope analysis may further uncover other potential human proteins that may become targets of antibodies through cross-reactivity. Furthermore, T cells may also cross-react with self-protein T cell epitopes through mechanisms such as bystander activation (Smatti et al., 2019). A future study looking specifically at T cell epitopes and cross-reactivity to self may provide further evidence of the immune system’s role in pathology.

For the first time, to our knowledge, we further explored and identified that some of the alignments in self-proteins predicted to be cross-reactive with SARS-CoV-2 alignments are potential epitopes within their own complete sequence. Additionally, where possible, we applied structural analysis to further explore the potential for cross-reactivity to the human proteins. In doing so, we identified the motif alignments were towards the surface of the protein, or in unstructured regions, and therefore would provide ready access for antibody binding. As crystal structures become available for the human proteins of interest, both full coverage of those mapped, as well as the remaining proteins, it would be interesting to further apply this method. However, the Bepipred Linear Prediction Algorithm 2.0 derives its epitope predictions from crystal structures (Jespersen et al., 2017). It is therefore likely that the Orf proteins and the human cross-reactive alignments predicted to be epitopes, are more likely to be found on the surface where antibodies have easier access.

The mechanism behind the thrombotic events occurring in some people following vaccination with the AstraZeneca and Johnson & Johnson adenovirus vectored vaccines is not known (Scully et al., 2021). Our findings suggest that molecular mimicry between the SARS-CoV-2 spike protein and proteins implicated in TTS and ITP, including PF4 and ADAMTS13, is unlikely to be the cause of these events. The vaccine induced thrombosis thrombocytopenia syndrome has been reported to be similar to autoimmune heparin-induced thrombocytopenia (aHIT) (Schultz et al., 2021), and therefore similar mechanisms may be involved. In aHIT, it has been suggested that structural changes in PF4 may be involved (Greinacher et al., 2017). Anti-PF4 antibodies are part of the diagnostic criteria for TTS (Greinacher et al., 2021). It has been hypothesized that free DNA in the vaccines may be a possible trigger of the anti-PF4 antibodies (Greinacher et al., 2021). However, further research into the formation of these anti-PF4 autoantibodies and the causes behind TTS following COVID-19 vaccination is required to identify potential interventions to prevent these events.

In the present study, linear B cell epitopes in SARS-CoV-2 proteins were predicted and compared to human proteins, identifying a number of new targets for potentially cross-reactive autoantibodies. Since many of the predicted epitopes reported in this study overlap with ones validated in COVID-19 patients, future studies may synthesize peptide sequences from the SARS-CoV-2 epitopes and the equivalent human sequences to perform in vitro analysis for validation.