The entire proteomes of 7 bacteria species and 8 respiratory viruses were obtained from NCBI after the entries indicated in Table 1 and assembled into individual files in FASTA format.

To identify shared peptides between virus and bacteria proteomes, the entire viral proteomes were first fragmented into overlapping 17-mer peptides with a 10-residue overlap. Subsequently, the peptides were used as queries in sequence similarity searches using BLASTP (35) against the target bacteria proteomes, previously formatted as BLAST databases. BLAST searches were performed with default parameters and the e-value set to 10,000. BLAST results were processed and shared peptides were identified from ungapped hit alignments including 8 or more residues with ≥ 80% identity. BLAST searches and processing of BLAST results were performed using an ad-hoc PERL script that will be provided by Dr. Reche upon writing request.

Peptides were assessed as potential T cell epitopes by predicting their binding to major histocompatibility complex (MHC) molecules. Peptide binding to MHC class I (MHC I) molecules was predicted using standalone versions of RANKPEP (36, 37) and NetMHCpan (38, 39). RANKPEP and NetMHCpan prediction models were selected to match the size of the peptides if they include 8 or 9 residues. The binding of peptides with more than 9 residues to MHC I molecules was assessed by evaluating that of all nested 9mer peptides. Binding of a peptide to a given MHC I molecule was considered to occur at a 2% Rank cutoff given by both RANKPEP and NetMHCpan, which allows selecting weak and strong binders % Ranks of test peptides are obtained by ranking their predicted binding affinity or binding scores compared to a large set of random peptides. The use of binding thresholds based on % Rank removes bias of certain molecules towards higher or lower predicted affinities and facilitates comparing and combining predictions by distinct methods. Binding of peptides to MHC class II (MHC II) was predicted using NetMHCIIpan (40) using a 10% Rank cutoff, which allows detecting strong and weak binding peptides. Human MHC II molecules targeted for predictions included HLA-DR, HLA-DQ, and HLA-DP molecules.

B cell epitopes were predicted using an standalone version of BepiPred1.0 (41). BepiPred reports antigenicity values per residue (a_i ), and a global B cell epitope score (B) was computed as indicated elsewhere (34, 42) consisting of the average a_i value. Peptides with B value ≥ 0.4 were considered as antigenic or potential B cell epitopes.

χ² tests were used as reported previously (43) to assess whether the distribution of the cross-reactive epitope sequences in proteins is proportional to the size of the proteins. The χ²-statistics value was computed using equation 1.

In this equation, k is the number of protein antigens, O_i the number of observed epitopes in antigen i, and E_i the number of expected epitopes in antigen i if they were distributed proportionally to the size of the proteins. The null, H₀ , hypothesis considers that epitopes are distributed proportionally to the size of proteins and it is rejected when the computed χ²-statistic value is above the χ²–distribution value at k – 1 degrees of freedom and a given α value.

The percentage of the world population that could respond to CD8 and CD4 T cell epitopes (population coverage) was computed after their MHC binding profiles using a command line version of EPISOPT (44) and the IEDB PPC tool at http://tools.iedb.org/tools/population/iedb_input (45), respectively, considering the relevant allele expression for the entire world population. The presence of a C-terminus in potential CD8 T cell epitopes compatible with cleavage by the proteasome was predicted from the relevant antigens using PCPS at http://imed.med.ucm.es/Tools/pcps/ with default settings (46, 47). Ectodomains of hemagglutinin (HA), neuraminidase (NA) and matrix protein 2 (M2) from IAV (A/Puerto Rico/8) were identified from UNIPROT accession H2KIW3_9INFA, H2KIW6_9INFA and H2KIW4_9INFA, respectively. PyMOL Molecular Graphics System Version 2.4.1 Schrödinger, LLC was used to map peptide sequences in tertiary structure of influenza A virus hemagglutinin (HA) (PDB: 1RU7) and to generate molecular renderings. Relative solvent accessibility (RSA) of peptide residues mapping in HA and M2 protein was calculated using NACCESS (48) and average solvent accessibility (ASA) of peptides was computed upon them as reported elsewhere (42, 49). Venn diagrams were generated using the nVennR package version 0.2.3 (50).

Cross-reactive immunity is much more likely to happen between antigens with high sequence similarity. However, because B and T cells recognize small regions in protein antigens (epitopes), cross-reactivity is better predicted by the volume of peptides (peptidome) that are shared between antigens (34). Thereby, we determined the shared peptidomes between 8 common respiratory viruses (IAV, IBV, HRV A, B and C, RSV A and B, and SARS-CoV-2) and 7 bacteria species, including BCG and those in MV130. To that end, we obtained the relevant proteomes and through a BLAST based approach (details in Methods) identified all unique peptides with identity ≥ 80% and length ≥ 8 in common between viruses and bacteria. The results of this analysis are summarized in Table 2. As noted by previous works, the volume of shared peptides increases with the size of the proteomes (34, 51). Thus, of all the studied bacteria and viruses, K. pneumoniae (KPN) followed by Bacille Calmette-Guérin (BCG) share with SARS-CoV-2 (SARS) the largest number of peptides, 138 and 102, respectively.

Interestingly, the peptides that are shared between the selected respiratory viruses and each bacterium in MV130 are largely distinct, as highlighted by the Venn diagrams depicted in Figure 1. Within the set of peptides shared by any of the viruses and a particular MV130 bacterium, only a handful coincides with those shared with another bacterium. An exception occurs in the case of S. aureus (SAU) and S. epidermidis (SEP), whose shared peptidomes with viruses are highly overlapping, as one could expect for they are closely related bacteria. As a result, the number of peptides that are shared between MV130, as a whole, and the different viruses is close to the sum of peptides that are shared by each bacterium in MV130 (Table 2). This effect is far from trivial as indicates that combining different bacteria, as those in MV130, increases the chance for cross-reactive immunity. Moreover, it supports the noted view (34) that a diverse microbiota helps to fight viral infections (26) by enhancing cross-reactive immunity.

The large size of the peptidome shared between MV130 and the studied respiratory viruses is indicative that MV130 could be a major source of cross-reactive immunity against these viruses. However, for the realization of cross-reactive immunity these peptides must be recognized by the adaptive immune system. Here, we explored such realization for IAV, since it causes respiratory infections prevented by MV130 (10, 11) and MV130 confers resistance to lethal IAV challenges in mice (12). To that end, we predicted the potential of the peptides shared between MV130 and IAV to represent B and T cell epitopes. Briefly, peptides predicted to bind class I and/or class II MHC molecules (in human HLA molecules) were considered CD8 and CD4 T cell epitopes, respectively (details in Methods). Moreover, for fulfillment of T cell cross-reactivity we required that both, the IAV peptide and the equivalent MV130 peptide, bind to the same MHC/HLA molecule. Because B cell reactivity is somewhat less predictable than T cell reactivity, B cell cross-reactivity was considered when either the IAV peptide or the equivalent MV130 peptide had a B cell epitope score ≥ 0.4 (details in Methods). In Table 3, we summarize the results of this analysis. Many more cross-reactive CD8 than CD4 T cell epitopes were predicted. This is the expected result, as CD4 T cell epitopes are longer than CD8 T cell epitopes and there are few peptides in the shared peptidome with the size required to be CD4 T cell epitopes. Nonetheless, we realize that we identified fewer cross-reactive T cell epitopes than in a previous work (33, 34) using Immune Epitope Database (IEDB) MHC-binding models through the RESTful interface (52).

As expected, the number of distinct cross-reactive B and T cell epitope peptides correlated with the number of shared peptides and hence with the size of the proteomes of the bacteria in MV130 (Table 2 and Supplementary Dataset 1). These cross-reactive peptide epitopes were private, differ between bacteria, and so the potential cross-reactive immunity of the MV130 formulation is the sum of each individual bacterium. Hence, MV130 appears to be a truly enhanced source of cross-reactive immunity. We next examined in detail the predicted cross-reactive epitopes to evaluate to what extent they could provide protective immunity against IAV.

MHC molecules restricting T cells are highly polymorphic in humans (HLA), bind/present distinct sets of peptides that can nonetheless be overlapping, and are expressed at different frequency in the population depending on ethnicity and geography (53, 54). Subsequently, the immunogenicity of any given T cell epitope varies between individuals, as it is contingent on the expression/presence of the specific MHC molecule restricting the epitopes. In Table 4, we show all the potential cross-reactive T cell peptide epitopes along with their predicted HLA binding/presentation profiles (details in Methods). We also show the MHC molecules expressed by C57BL/6 and BALB/c mice strains that can present these same peptide epitopes (peptides that were only predicted to bind to mouse MHC molecules are not shown). Given the key role of CD8 T cells in clearing and containing viral infections we will pay particular attention to cross-reactive CD8 T cell epitopes.

MV130 encompasses 37 unique peptide sequences consisting of cross-reactive CD8 T cell epitopes with IAV that are distributed through all IAV antigens but M2 (Tables 3, 4). These epitopes are not distributed proportionally to the size of the IAV antigens as revealed by χ² statistics (p< 0.005). The largest contributions to χ² statistics are found in non-structural protein 2 (NS2), polymerase PA and M1 (Figure 2A). In particular, while NS2 and M1 bear more cross-reactive T cell epitopes than the expected, PA includes fewer than the expected (Figure 2B). The uneven distribution of cross-reactive CD8 T cell epitopes throughout the IAV proteome supports the specificity of T cell cross-reactivity. In fact, it is worth noting that 24 of the 37 cross-reactive peptides coincide with IAV-specific CD8 T cell epitopes deposited in the IEDB (Table 4).

The majority of potential MV130-IAV cross-reactive CD8 T cell epitopes can also be presented by more than one HLA I molecule and 5 of them can also be presented by HLA II molecules (Table 4). The combined phenotypic frequency in the population of all HLA I molecules restricting these cross-reactive CD8 T cell epitopes would imply that MV130 could elicit cross-reactive CD8 T cell immunity to IAV will in ≥ 95% of the population the regardless of their genetic background (details in Methods). Bacteria eliciting memory CD8 T cells capable of recognizing antigens displayed by virally infected cells can occur through antigen cross-presentation (55). This mechanism enables professional antigen presenting cells to redirect antigens taken from the extracellular milieu for presentation in the context of HLA I molecules and prime CD8 T cells against extracellular antigens (55). There are distinct cross-presentation pathways, including some that are proteasome dependent, just like the classical class I antigen presentation pathway (56). In this regard, 29 out of 37 shared peptides defining potential cross-reactive T cell epitopes have a C-terminus that is compatible with cleavage by the proteasome (Table 4, details in Supplementary Dataset 1). Vaccines consisting of inactivated virus or viral antigens can likely induce protective anti-viral CD8 T cell memory thanks to this same mechanism.

Since only 5 potential cross-reactive CD4 T cell epitopes were detected (Tables 3, 4), it could be argued that MV130 may not induce enough T helper (Th) cells to support the whole antiviral potential of cross-reactive CD8 T cells. However, the percentage of the world population that could respond to any of these 5 cross-reactive CD4 T cell epitopes is actually about 33.7% as computed using the frequency of the relevant HLA II molecules (see Methods). Moreover, it is likely that there are many more cross-reactive CD4 T cell epitopes than those detected through our methodology. In fact, CD4 T cells are in general more cross-reactive than CD8 T cells and can recognize many distinct peptides despite sharing little sequence similarity (57, 58). In this context, our findings explain the increase (~ 30-fold) in the influenza virus-specific CD8 T cell response, following MV130 treatment in patients with RRTI (5).

In this study, we trusted the detected T cell cross-reactivity between MV130 and IAV on the predicted binding of shared peptides to the same HLA alleles. However, binding of peptides to MHC molecules is not enough to guarantee T cell reactivity in vivo. Thus, it has been shown that peptides with high binding affinity for MHC molecules in vitro can nonetheless be excluded from T cell recognition in vivo, very likely due to a lack of appropriate antigen processing (59). Hence, the actual realization of the predicted T cell cross-reactivity from MV130 to IAV is contingent on the appropriate processing of antigens. This processing will involve, on the one hand, the uptake of bacteria by antigen-presenting cells, processing of bacteria antigens and presentation of peptide antigens by HLA molecules to prime T cells. On the other hand, it will require that IAV infected cells and/or antigen-presenting cells that have captured IAV antigens do also process the antigens and present the counterpart peptides by the same HLA molecules. These antigen processing events were not taken in consideration in this study because of their complexity and because they are less predictable than binding to MHC molecules. Thereby, in vivo studies are required to confirm the predicted T cell cross-reactivity of MV130 to IAV. Given that there are cross-reactive CD8 T cell epitopes restricted by both human and mouse MHC molecules (Table 4), mouse infectious disease models could be used to identify cross-reactive immunity relevant to humans and contribution to IAV protection.

Preexisting protective cross-reactive immunity to virus is more often linked to T cells (21, 60). However, cross-reactive antibodies between viruses and bacteria have been reported (61, 62) and we have previously shown that the spike protein of SARS-CoV-2 includes potentially neutralizing B cell epitopes that are shared with bacteria targeted by diphtheria-tetanus vaccines (33, 34). Thereby, we investigated MV130 cross-reactive B cell epitopes mapping on the ectodomains of IAV proteins that are known be targeted by antibodies hampering viral entry. These proteins are HA, NA and M2. We found 8 of such cross-reactive B cell epitopes in HA, one in M2 and none in NA (Table 5). The average solvent accessibility (ASA) of these cross-reactive B cell epitopes (details in Methods) is greater than 25% (Table 4), indicating that they are readily accessible to antibodies. Moreover, we verified that 5 of these cross-reactive B cell epitopes coincide with experimentally determined B cell epitopes deposited at IEDB, including LLTEVETP, which matches a known B cell epitope in M2 annotated as neutralizing in IEDB (targeted by neutralizing antibodies). M2 is a proton-selective transmembrane ion channel located in the viral envelope required for the efficient release of the IAV genome into host cells (63, 64). Interestingly, only the N-terminal region of M2 (residues 1-22) surfaces the virion membrane and it is in this precise region that lays LLTEVETP (residues 3-10). This region is extremely conserved across all reported influenza A viruses and hence cognate antibodies could provide heterotypic influenza immunity. Although our approach did not yield any cross-reactive T cell epitope in M2, CD4 and CD8 T cell epitopes in M2 ectodomain have been reported that can mediate protective immunity to IAV (65, 66). Whether these T cell epitopes could have been predicted as cross-reactive using less stringent criteria of similarity is something to consider. The main target of neutralizing antibodies against IAV is however HA, as this protein dominates the surface of IAV and facilitates viral entry into host cells (57). Hence, we investigated the neutralizing potential of HA cross-reactive B cell epitopes by mapping them on the 3D-structure of HA and examining relevant structural information.

The mature HA includes 3 HA1 and 3 HA2 subunits –derived from the same polypeptide chain – that fold into a trimeric structure depicting a globular head domain and a stem domain (Figure 3A) The globular domain is made of HA1 subunits and includes the receptor binding domain (RBD), which attach sialic acid in various membrane proteins, facilitating viral entry (67). This globular domain, particularly the vicinity of the RBD, is hence the subject of recognition by many neutralizing antibodies (68). Most of the selected cross-reactive B cell epitopes map in the surface of the globular domain, relatively close to the RBD (Figure 3A). Thus, one could speculate that antibodies recognizing these B cell epitopes could impede IAV attachment to host cells and block viral entry. This effect can be readily visualized for the cross-reactive B cell epitope PKESSWPN (HA residues 120-127) (Figures 3A, B), since this epitope is adjacent to the 130-loop (residues number 134-142) which takes part of RBD (67). To a lesser extent, the stem domain can also be the target of neutralizing antibodies, which generally interfere with conformational changes required for IAV membrane fusion (67, 68). Interestingly, one of the cross-reactive B cell epitopes in the stem domain of HA is AIAGFIEG (Figures 3A, C), which coincides with a known B cell epitope recognized by neutralizing antibodies (Table 5) (69). Thereby, we can foresee that the neighboring cross-reactive B cell epitope QNAINGITNK could also be neutralizing (Table 5 and Figure 3A).