Selection of TAAs and MoAs

The paired TAAs and MoAs in the present study were derived from our previous analysis [25]. MAGE-A1, MAGE-A3, MAGE-A3/12, MAGE-A10, MAGE-C1, MAGE-C2, and SSX2 TAAs, together with 3–5 corresponding MoAs derived from Firmicutes and Bacteroidetes phyla were chosen (Table 1). MoAs were selected based on the predicted affinity to the same HLA allele as the corresponding TAA (HLA-A*02:01), with a maximum value lower than 100 nM. Indeed, peptides with such predicted values show 100% confirmation of HLA binding in an experimental setting based on TAP-deficient T2 cells [25, 31,32,33,34]. The overall mean of the predicted affinity values was 13.31 nM and 41 of the 53 peptides were below this value, suggesting very high binding affinity to the HLA-A*02:01 allele. The alignment of MoAs homologous to each TAA confirmed that, despite individual differences, the most predominant aa residues at each position always correspond to those in the TAA sequence (Fig. 1).

Table 1 List of TAAs and paired MoAs selected for the studyFig. 1

SeqLogo analysis of MoAs homologous to TAA: graphical representation of amino acids belonging to the consensus sequences of MoAs aligned to the corresponding homologous TAAs. Amino acid sequences from all the microbiota-derived epitopes with homology to each TAA were piled up to build sequence logos. The height of all aminoacids at each position indicates the sequence conservation at that position, while the height of each symbol within the stack indicates the relative frequency of each aminoacid at that position (https://services.healthtech.dtu.dk/service.php?Seq2Logo-2.0). Different colours indicate different classes of aminoacids

Conformation of selected TAAs and MoAs

Epitope modelling and molecular docking were performed to prove that the sequence homology between the paired TAAs and MoAs was echoed in similar peptide conformations as well as contact areas with the HLA molecule and the TCR. The analysis confirmed that regardless of the position of the amino acid substitution along the peptide sequence, MoAs may have a similar, if not identical, conformation and pattern of contact with the α and β chains of the TCR, as shown by the footprints of the paired peptides (Fig. 2; Suppl. Figs. 2–15). The best examples of identical matching are the following: 1) the KVLEYVTKV peptide derived from Staphylococcus pettenkoferi with a Ile - Thr substitution at position 7 compared to MAGE-A1; 2) the KIAELVHFL peptide derived from Sedimentibacter sp. with a Val - Ile substitution at position 2 compared to MAGE-A3; 3) the FLWGPKALV peptide derived from Chitinophagaceae bacterium with a Pro - Lys substitution at position 6 compared to MAGE-A3/12; 4) the GLYDGMEYI peptide derived from Clostridium kluyveri with a His - Tyr and Leu - Ile substitution at position 8 and 9 compared to MAGE-A10; 5) the KTVEFLAMV peptide derived from Lachnospiraceae bacterium with a Val - Thr and Leu - Val substitution at position 2 and 9 compared to MAGE-C1; 6) the ALSDVEERV peptide derived from Loigolactobacillus jiayinensis with a Lys - Ser substitution at position 3 compared to MAGE-C2; and 7) the KVSEKIFYL peptide derived from Flavobacterium sp. CG_9.1 with Ala - Val and Val-Leu substitutions at positions 2 and 9 compared to SSX2 (Fig. 2). In contrast, other MoAs presented substitutions that slightly or heavily affected the conformation as well as the pattern of contact with the α and β chains of the TCR (Suppl. Figs. 2–15).

Fig. 2

Predicted 3D conformation of TAA and microbiota-derived paired peptides and peptide-TCR interaction. The surface conformation of the most similar paired HLA-A*02:01 restricted TAA and MoAs-derived peptides is shown. Residues in the microbiota-derived epitopes (FIRM = firmicutes; BACT = bacteroidetes) that differs from the TAA sequences are indicated in red color. Red areas = contact points with HLA-A molecule; blue areas = contact points with TCR α chain; Light Blue = contact points with TCR β chain. The images below the peptides show the contact sites with α and β chains of TCR (yellow areas)

Epitope analysis

DNA-barcoded peptide-major histocompatibility complex (pMHC) multimers (HLA-A*02:01) were prepared with seven selected TAAs and 53 homologous MoAs from Firmicutes and Bacteroidetes phyla. In addition, a panel of 64 peptides derived from common viruses was constructed as an overall control of antiviral immune status. PBMCs from HLA-A*02:01 HS (n = 10) and CP (n = 15) patients were purified, incubated with DNA-barcoded pMHC multimers, and stained with a phenotype antibody panel to identify reactive CD8+ T cells [27].

The percentage of CD8+ T cells reacting against MoAs was, on average, higher in CP (95.54%) than in HS (85.18%). In contrast, the percentage of CD8+ T cells reacting to TAAs or cross-reacting with TAAs and MoAs was, on average, higher in healthy individuals (10.75% and 4.07%, respectively) than in cancer patients (2.19% and 2.28%, respectively). This difference was statistically significant for all three comparisons (Fig. 3A). Strikingly, in both groups, three subjects showed a percentage of CD8+ T cells reacting against the TAAs and cross-reacting with TAAs and MoAs well above the average value (Fig. 3A, B).

Fig. 3

CD8+ T cells reacting with TAAs and MoAs. The plots (A) show the % of sorted CD8+ T cells reacting against the MoAs (bacteria + viruses), the TAAs or cross-reacting with TAAs and MoAs (CROSS) (*** = p < 0,0001, * = p < 0,005). B Representative box plot showing the % of sorted reactive CD8 + T cells for a healthy subject (left) and a cancer patient (right)

MoAs pMHC-DNA barcoding evaluation

DNA barcodes with FDR < 0.1% (corresponding to p < 0.001) and Log2FC > 2 over the baseline values for the pMHC library were considered true and significant T-cell responses. The fraction of MoA-reacting T cells showed consistent binding to peptides homologous to MAGE-C2 (C2-BACT1, C2-BACT2, and C2-FIRM3) AND MAGE-A3/12 (A3/12-BACT1) in both tumor patients and HS. Scattered binding to peptides homologous to MAGE-C1 (C1-BACT1, C1-FIRM1, C1-FIRM2) was observed in both groups (Fig. 4A). Interestingly, while binding to such peptides was observed in scattered samples, a completely different predominant pattern was observed in the fraction of cross-reacting T cells. Indeed, the peptides that were more frequently bound were those homologous to MAGE A1 and SSX2 (A1-BACT1, A1-BACT2, A1-FIRM3, A1-FIRM4) (SSX2-BACT1, SSX2-BACT2, SSX2-BACT3, SSX2-FIRM1) with an equal distribution between the two groups (Fig. 4B). The number of peptides bound by cross-reactive T cells was broadly different in both groups, ranging from 0 to 5 in healthy individuals and from 0 to 7 in tumor patients (Suppl. Fig. 16). Binding of peptides homologous to other TAAs was not observed. Consistent binding to the positive control CMVpp65 peptide was observed in of the 21/25 subjects in both groups (data not shown).

Fig. 4

Reactivity to DNA-barcoded pMHC multimer MoAs. The statistically relevant (Log2FC > 2) reactivity to each peptide (column) from each subject (row) is enlighted in red. The different plots show reactivity against MoAs from the single positive (SS) (A) and from the double positive (DS) (B) sorted cell fraction

TAAs pMHC-DNA barcoding evaluation

The analysis of the fraction of double-positive T cells revealed binding to TAAs in both HS and CP, and most of this binding was specific to MAGE-A1. Three (3) HS (H-004, H-007, and H-010) showed binding to the TAAs. Binding values to MAGE-A1 peptide by H-007 and H-010 samples showed a > 2log fold increase, with a high statistical significance (p < 1 × 10–6) (Fig. 5A). In addition, binding values to MAGE-A1 and MAGE-C1 peptides by H-004 sample showed a statistical significance (p < 0.005) with a fold increase nearly reaching the 2log fold increase. Seven (7) CP showed binding to TAAs (T-001, T-003, T-004, T-006, T-010, T-011, and T-015), and some of them to more than a single TAA. Binding values to MAGE-A1 peptide by T-001, T-003, T-006, and T-015 samples showed a > 2log fold increase, with a high statistical significance (p < 1 × 10–6), while T-004 and T-011 samples showed a binding value with a statistical significance (p < 0.005) and a fold increase nearly reaching the 2log fold increase. Binding values to MAGE-C1 peptide by T-001 and T-010 samples, to SSX2 peptide by T-004 and T-011 samples, and to MAGE-A3/12 peptide by T-011 sample showed a > 2log fold increase, with a statistically significant difference (p < 1 × 10–6). In addition, the T-004 sample showed a binding value to A3/12 peptide with statistical significance (p < 0.005) and a fold increase nearly reaching the 2log fold increase. Overall, the samples that reacted with more than single peptides were T-001, which bound MAGE-A1 and MAGE-C1 peptides; T-004, which bound MAGE-A1, SSX2, and MAGE-A3/12 peptides; and T-011, which bound MAGE-A1, SSX2, and MAGE-A3/12 peptides (Fig. 5).

Fig. 5

Reactivity to DNA-barcoded pMHC multimer TAAs. The statistically relevant (Log2FC > 2) reactivity to TAAs from each subject (row) is enlighted in red. The table shows the reactivity to TAA epitopes in both HS and CP

TAAs and MoAs cross-reactivity

According to the pMHC-DNA barcoding evaluation, two samples from the HS and three samples from the tumor patients showed double positivity for binding to homologous TAAs and MoAs with a > 2log fold increase and a highly statistical significance (p < 1 × 10–6). In particular, the H-010, T-001, and T-006 samples showed binding to MAGE-A1 and, all three, the homologous A1-FIRM4, and H-010 bound the A1-FIRM3 (T-006), A1-BACT1 ( T-001), and A1-BACT2 peptides. H-004 and T-001 samples bound to MAGE-C1 and homologous C1-FIRM1. Finally, the T-004 sample bound to SSX2 and the homologous SSX2-BACT1, SSX2-BACT2, and SSX2-BACT3 peptides (Fig. 6). In all such double reactivities, the paired TAAs and MoAs show highly similar, if not identical, conformation and contact areas to both HLA and TCR α and β chains (Fig. 2 and Suppl. Figs. 2–15).

Fig. 6

Double reactivity to DNA-barcoded pMHC multimer TAAs/MoAs. The statistically relevant (Log2FC > 2) reactivity to TAAs and MoAs from each subject (row) is enlighted in red. The table shows the reactivity to homologous coupled peptides (TAA and MoAs) in both HS and CP

Validation of CD8+ T cell cross-reactivity

To confirm that the double positivity corresponded to true T cell cross-reactivity to the paired TAAs and MoAs, tetramer-staining analyses were performed.

Because of the limited availability of stored samples, the analysis was performed on sample T-004 only, which showed a broad reactivity against the SSX2 TAA and SSX2-BACT1, BACT2, and BACT3 MoAs during the DNA-barcoding screening.

Tetramer staining showed the cross-reactivity of CD8+ T cells with the SSX2 peptide and each of the paired MoA. Unstimulated PBMC were analyzed by flow cytometry using paired fluorescent HLA-A2/peptide tetramers and CD8-specific Abs. SSX2-, MoA-, and cross-reacting T cells were detected in the unstimulated PBMC of T-004 tumor patients. The SSX2-reacting CD8+ T cells were approximately 0.01%, and the MoA-reacting CD8+ T cells were approximately 0.004% for BACT1 and BACT2 and 0.036% for BACT3 pepide. Different levels of cross-reacting CD8+ T cells were observed in all three comparisons. The percentage of such CD8+ T cells was directly correlated with that observed for single peptide reactivity. The highest percentage of cross-reacting CD8+ T cells (0.011%) was observed in the SSX2/BACT3 comparison (Fig. 7A). The presence of circulating primed T cells specific for the MoAs, homologous to TAAs, was assessed by an ex vivo immunization of PBMCs. Isolated PBMCs were stimulated with SSX2-BACT2 and SSX-BACT3 peptides for 5 days and CD8+ Tcell cross recognition of both MoAs and homologous SSX2 derived epitopes was assessed via tetramer staining. The results showed an increase in T cell recognition of both BACT2/3 and SSX2 epitopes, together with an increasing frequency of cross-reacting CD8+ Tcells as showed in Fig. 7B.

Fig. 7

A Cross-reactive CD8 + T cells in pMHC tetramer staining. Unstimulated PBMCs from sample T-004 were incubated with pMHC tetramers loaded with TAA and MoAs homologous peptides. Dot plots show the reactivity against the SSX2 peptide (PE-A) and SSX2-BACT1 (APC-A), SSX2-BACT2 (PE-CF594-A) and SSX2-BACT3 (PE-Cy7-A). B Cross-reactive CD8+ T cells in pMHC tetramer staining after in vitro pre-immunization. Cross-reactivity against the paired SSX2-BACT2/BACT3 peptides was evaluated after an in vitro pre-immunization. PBMCs were stimulated with MoAs derived epitopes and tetramer staining was used to assess the cross-recognition of homologous TAA peptide

IFN-γ release after peptide stimulation

The activation of antigen-specific CD8+ Tcells was evaluated by the production of IFN-γ after stimulation with SSX2, SSX2-BACT2 and SSX2-BACT3 peptides.

The flow cytometric analysis of IFN-γ release revealed a relevant production in all subjects (Fig. 8), with an average fold increase against the non stimulated controls of 6.52 for SSX2, 6.48 for SSX2-BACT2 and 6.07 for SSX2-BACT3.

Fig. 8

IFN-γ release by CD8+ T cells. Secreted IFN-γ was detected using the secretion assay on viable IFN-γ+ CD8+ Tcells. The graph shows the increase in the IFN- γ production after the peptide stimulation

Cross-reactive CTL activity

The final proof of cross-reactivity was provided by assessing the cytotoxic activity of PBMCs stimulated ex vivo with the MoA-derived peptide BACT3 on TAP-deficient T2 cells loaded with SSX2, SSX2-BACT2 or SSX2-BACT3 peptides.

The results showed a cross-reactive killing activity of activated PBMCs on T2 cells presenting each of the three peptides. Interestingly, the average percentage increase of CTL activity did not reach the statistical difference in the three settings, suggesting the comparable efficient targeting of T2 cells presenting one of the paired antigens (Fig. 9).

Fig. 9

CTL activity. Cytotoxic activity of PBMCs was assessed by fluorimetric assay in TAP-deficient T2 cells loaded with the indicated peptides. A Average percentage increase of CTL activity over T2 control cells; B example of layout result in a single individual, showing increase in 7AAD fluorescence intensity