Protein sequences were retrieved from UniProtKB/Swiss-Prot (https://www.uniprot.org). The pyrogenic exotoxins of Streptococcus pyogenes (serotype M18) included: SPEA_STRP8 (P62561), SPEC_STRP8 (Q8NKX2), SPEM_STRPY (Q7WYA2), and SPEK_STRPY (A0A5S4TLM8). Human HSPs analyzed were: Endoplasmin (ENPL, P14625), HSP 90-alpha (P07900-1), Endoplasmic reticulum chaperone BiP (Grp78, P11021), HSP70 kDa 1A (HS71A, P0DMV8), and HSP60 (CH60, P10809). Tumor antigens retrieved were Trophoblast glycoprotein (TPBG_HUMAN, Q13641) and carcinoembryonic antigen-related cell adhesion molecule 5 (CEAM5_HUMAN, P06731-1).

The methodology included a multi-step immunoinformatic analysis using common tools for designing epitope-based vaccines (53). As outlined in Figure 1 , the first steps involved multiple sequence and structure alignments of exotoxins, and searching for sequence similarity between HSPs and exotoxins, focusing attention on the SAg region. The HSPs sequences with the best similarity with the common exotoxin region of SAg were analyzed for MHC-I, MHC-II, and B-cell epitopes. Only epitopes with positive class-I immunogenicity values were then considered for comparison with those identified in the SAg region of exotoxins. Matching epitopes would support the hypothesis that HSP sequences, structurally similar to SAgs, represent shared tumor antigens and potential targets of the exotoxin-driven antitumor response.

Protein sequences were analyzed in pairwise and multiple sequence alignments using T-Coffee (https://tcoffee.crg.eu), and Clustal Omega (https://www.uniprot.org/align). Although based on different algorithms, both programs yielded consistent results in cases of high sequence similarity. T-Coffee assigns a maximum consensus score of 1000 (100%). The Expresso function of T-Coffee (https://tcoffee.crg.eu/apps/tcoffee/do:expresso) was used to assess structural similarity. Only uninterrupted segments of at least eight amino acids with the highest consistency scores were considered. BLAST (https://blast.ncbi.nlm.nih.gov) was used to verify whether similar sequences were conserved among HSPs and between HSPs and tumor markers 5T4 and CEA. An expect threshold of E<0.05 was set to determine statistical significance.

Predicted best MHC class I epitopes in exotoxins and HSPs were identified using the IEDB Analysis Resource (http://tools.iedb.org/mhci/), applying IEDB recommended 2020.09 methods: NetMHCpan EL 4.1predicting best MHC-I peptide score and percentile rank, and Ann 4.0 predicting IC50 values. A percentile rank of ≤0.5 and a IC50 value ≤500 nM were both considered indicative of strong binding (54). A representative panel of common HLA alleles was used: A*01:01; A*02:01; A*03:01; A*24:02; B*07:02; B*08:01; B*15:01; B*40:01; B*44:02; B*57:01. Immunogenicity of MHC-I epitopes was evaluated using IEDB’s class I immunogenicity tool (http://tools.iedb.org/immunogenicity/) that predicts immunogenicity of an epitope-MHC complex through the analysis of the location and properties of amino acids. A positive score indicates probable T-cell recognition, while a negative score suggests reduced likelihood (54).

MHC class II epitopes (predicted CD4+ T-cell epitopes) were identified using the IEDB Consensus tool (http://tools.iedb.org/mhcii/), employing a 7-allele reference panel: DRB1*03:01; DRB1*07:01; DRB1*15:01; DRB3*01:01; DRB3*02:02; DRB4*01:01; DRB5*01:01. Peptides with a percentile rank <0.5 and IC50 <50 nM (high affinity) or <500 nM (intermediate affinity) were retained (55).

Linear B-cell epitopes in exotoxins, Grp94, Grp78, and HSP70 were predicted using BepiPred-2.0 sequential B-cell epitope predictor (https://services.healthtech.dtu.dk) (56). A threshold score above the default 0.5 was used to increase specificity and reduce false positives.

Four major exotoxins of Streptococcus pyogenes, SPEA, SPEC, SPEM, and SPEK, were analyzed as representative members of the bacterial SAg toxin family. Despite differences in length, multiple sequence alignment using T-Coffee revealed high consensus scores across all toxins, reflecting conserved sequences ( Figure 2A ), consistent with previous observations in other bacterial SAgs (22). The region showing the greatest conservation spanned amino acids 157–221 of SPEA and the corresponding aligned segments in SPEC, SPEM, and SPEK, which included a 12-residue sequence (bold in Figure 2A ) corresponding to the β-strand/hinge/α-helix domain, known to be highly conserved among Staphylococcus aureus and Streptococcus pyogenes SAgs (22). Similar alignment results were confirmed using Clustal Omega (data not shown).

Structural alignment using Expresso (T-Coffee) also revealed strong similarity among exotoxins, particularly within the SAg domain, in both pairwise ( Supplementary Figure S1A ) and multiple alignments ( Supplementary Figure S1B ). The overlap of sequence and structural homology in this domain ( Figure 2A and Supplementary Figure S1 ) supports previous findings that conservation within this region underlies a shared mechanism of action across SAgs (22), indicating potential for similar biological activity (28, 57).

Pairwise alignments of each exotoxin (SPEA, SPEC, SPEM, SPEK) with Grp94, HSP90, Grp78, HSP70, and HSP60 ( Supplementary Figure S2 ) showed that ~90% of SPEA, SPEC, and SPEM sequences aligned with Grp94. SPEM showed the highest coverage with HSP90 (99%), Grp78 (91%), and HSP70 (92%). Each exotoxin sequence was largely covered by overlapping segments from at least two HSPs, especially for SPEA and SPEC ( Supplementary Figure S2A ). Long contiguous HSP segments, often over 30 amino acids, matched exotoxin regions with minimal gaps. To exclude chance similarity, exotoxins were aligned with seven unrelated human proteins (albumin, antitrypsin, beta2 microglobulin, apolipoprotein A, adiponectin, kallikrein, vitamin D-binding protein). In comparison with exotoxins, the control proteins consistently showed lower sequence identity (average ≤60%) and shorter aligned segments (data not shown).

Given the immunological relevance of the SAg domain (22, 25, 27, 57), HSP sequences overlapping this region were examined ( Figure 2B ). Only HSP segments fully aligned with the SAg domain in pairwise alignments ( Supplementary Figure S2 ) were included. Selection criteria were segment length and alignment with multiple exotoxins. Due to the greater length of HSPs relative to exotoxins, matches with common regions of multiple exotoxins supported sequence significance. Grp78 (residues 544–586) and Grp94 (534–570) overlapped various exotoxin segments, often with internal overlap. HSP70 (471–509) aligned with SPEC and partly overlapped its SAg domain. HSP90 and HSP60 showed only sparse, short matches in this region ( Figure 2B ).

Given the high sequence similarity between HSP family members (Grp94/HSP90: 92%; Grp78/HSP70: 96%), homologous segments of aligned HSP sequences were sought. Additional matching segments were thus identified ( Figure 2C ): Grp94 476–505 and 534–570 aligned, respectively, with HSP90 436–456 and 487–523; reciprocally, HSP90 450–468 and 579–596 matched, respectively, Grp94 498–512 and 630–646. HSP70 471–509 corresponded to Grp78 494–532, and Grp78 544–573 aligned with HSP70 521–550. Grp78 519–539 matched HSP70 499–516, while Grp78 573–586 lacked a homolog in HSP70, delimiting alignment to the residue 573. Combined, these homologous sequences in Grp78/HSP70 formed nearly continuous segments (HSP70: 471–550; Grp78: 494–573) separated by a five-residue gap. HSP60 displayed minimal sequence similarity to other HSPs (data not shown), with only two short regions (363–383) aligning with HSP70 (512–521) and Grp78 (590–603) ( Figure 2C ).

These additional homologous HSP sequences aligned with the same exotoxin segments as the originals. The results thus showed that Grp94, and especially Grp78 and HSP70, contained distinct segments with significant similarity to the exotoxin SAg region.

To assess whether the HSP sequences with significant similarity to the SAg region also exhibited immunogenic potential, we analyzed their content of MHC-I epitopes and compared them with those of streptococcal exotoxins. The necessary premise to this was that the exotoxins, like other foreign proteins, are internalized into antigen-presenting cells (APCs) and presented in complex with MHC-I and MHC-II molecules to TCRs (58). Only epitopes with high predicted binding affinity (IC50 <500 nM) were selected from exotoxins and HSPs ( Supplementary Tables S1 , S2 ).

Exotoxin epitopes were predominantly localized in the N-terminal and SAg regions. While isolated epitopes were observed in the N-terminus, a notable concentration of overlapping epitopes was found in the SAg domain: six in SPEA 173–204, six in SPEC 157–187, five in SPEK 182–211, and even thirteen in SPEM 125–188, despite SPEM being the shortest exotoxin ( Table 1 ). All epitopes were allele-specific, except one epitope, each in SPEC and SPEM, which bound two distinct alleles.

Among HSPs, Grp94, HSP90, Grp78, and HSP60 contained numerous epitopes, whereas HSP70 exhibited fewer total epitopes but all concentrated within the SAg-similar regions ( Table 1 and Supplementary Table S2 ). The most frequently targeted alleles included HLA-B40:01 and B15:01, the latter shared with several exotoxin epitopes. Multiple HSP sequences contained overlapping epitopes, particularly within Grp94, HSP90, Grp78, and HSP70 ( Table 1 ).

To determine overlapping between HSP and exotoxin epitopes, MHC-I epitope mapping was performed along previously aligned regions ( Supplementary Figure S2 ). Many HSP epitopes, especially in Grp94 and HSP70, closely matched those of SPEM and SPEA ( Supplementary Figure S3 ). Notably, these epitopes were localized within the same sequences previously aligned to the SAg region, which itself harbored a dense cluster of overlapping epitopes ( Figure 3A ). In contrast, matching epitopes outside the SAg region were fewer and spatially isolated ( Supplementary Figure S3 ), reinforcing the functional significance of the SAg domain in immune stimulation. Importantly, the extended sequences of Grp78 and HSP70, each forming a single long homologous stretch ( Figure 2C ), contained epitopes overlapping with those of all exotoxins, particularly SPEA, SPEC, and SPEM ( Figure 3A ).

While some sequences of Grp94, HSP90, and HSP60 aligned with the SAg domain, not all contained MHC-I epitopes. For instance, Grp94 476–496, aligned with SPEM, contained three epitopes, whereas the longer 534–570 segment, aligning with multiple exotoxins, included only one epitope in the sequence 534–557 aligned with SPEA ( Figure 3A ). HSP90 and HSP60 contributed fewer matching epitopes, two and one, respectively. Collectively, only the Grp78 and HSP70 sequences provided consistent, extensive overlap with the immunodominant SAg regions of all exotoxins.

Beyond binding affinity, T-cell immunogenicity is a critical determinant of functional epitope recognition (54). We therefore assessed the predicted immunogenicity of each HSP epitope with sequence similarity to exotoxins. Only epitopes with positive immunogenicity scores were considered functionally relevant ( Figure 3B ), although all had high MHC-I binding affinity ( Supplementary Figure S3 ).

Among all peptides analyzed, the HSP70 sequence 471–493 and its homologous Grp78 494–516 displayed the highest immunogenicity, containing distinct, strongly immunogenic epitopes ( Figure 3B ). The Grp78 547–571 and the homologous HSP70 524–548 were moderately immunogenic, while the Grp78 556–565 epitope was more potent. Grp94 476–496, with three overlapping epitopes, also showed positive immunogenicity, unlike the Grp94 534–570 segment and all epitopes within HSP90, which were non-immunogenic. HSP60 epitopes displayed weak but detectable immunogenicity ( Figure 3B ).

Thus, only select regions of Grp78 and HSP70, specifically those also aligning with the SAg domains of exotoxins, appeared to possess strong immunogenic responses, supporting their potential to functionally mimic bacterial SAgs.

While cytotoxic T lymphocytes are central to the anti-tumor immune response, both B cells and CD4⁺ T helper cells are essential for antigen recognition and the orchestration of adaptive immunity (59). To assess the immunogenic potential of HSP sequences homologous to the SAg domain ( Figure 3B ), the most predictive B-cell and MHC class II epitopes were identified within these regions. Exotoxins were analyzed in parallel for comparison.

Grp94 harbored several long B-cell epitopes, most of which did not overlap with its class I immunogenic sequences ( Figure 4A ). Only two residues, D477 and K479, located within the 476–496 region, served as both B-cell and class I epitopes. K479 (in bold) was the sole residue common to both ( Figure 3 ). In contrast, the sequences 548–559 of Grp78 and its homologous region 524–537 in HSP70 functioned as B-cell epitopes, with partial overlap of class I epitopes ( Figure 4A , underlined and bold), a result that supported the broad immunogenicity of the homologous sequences of Grp78 and HSP70. Compared to HSPs, exotoxins displayed a relatively limited number of B-cell epitopes, with overlapping B-cell and class I epitopes confined to short segments in SPEA, SPEC, and SPEM within the SAg domain.

A search for high-affinity MHC class II epitopes (percentile rank ≤0.50) across a representative panel of seven prevalent HLA class II alleles ( Figure 4B ) identified four Grp94 epitopes within residues 534–551, outside the class I immunogenic region ( Supplementary Table S3 ). Notably, Grp78 and HSP70 contained nine and eleven class II epitopes, respectively. Of the Grp78 epitopes, six were found within residues 496–520, homologous to HSP70 473–498 ( Figure 2C ) and three within the sequence 533–549. Among the eleven HSP70 epitopes, seven mapped to residues 473–498 and four to 520–537, all within the immunogenic 471–550 segment shared with Grp78 544–573 ( Figure 2C ).

Exotoxins, particularly SPEA, SPEM, and SPEK, featured multiple class II epitopes overlapping class I epitopes in the SAg domain (see also Figure 3A ). These epitopes predominantly bound to DRB1*03:01 (nine epitopes) and DRB4*01:01 (seven epitopes), the latter also targeted by epitopes in Grp78 and HSP70. Additional overlaps were found with DRB5*01:01 and DRB3*01:01, shared among both HSPs and exotoxins ( Figure 4B ). Grp78 (residues 505–549) and HSP70 (481–498) featured class II epitopes binding to DRB4*01:01, the same allele recognized by epitopes in the SAg regions of SPEA, SPEC, SPEM, and SPEK. Interestingly, the strongest binding affinity to DRB4*01:01 was observed for HSP70 epitopes within 482–497. Similarly, Grp78 and HSP70 epitopes targeting DRB3*01:01 demonstrated superior binding affinities compared to the corresponding SPEA epitope in the sequence 179–193 ( Figure 4B ).

The homologous 80-residue sequences of Grp78 and HSP70, formed by combining original and cross-matching segments ( Figure 2C ), contained overlapping epitopes of all three classes. Notably, regions Grp78 494–520 and HSP70 471–498 featured near-complete overlap of class I and class II epitopes ( Figure 4C , bold). Additional overlapping B-cell and class II epitopes extended from 526 to 559 in Grp78 and from 504 to 537 in HSP70. Remarkably, within HSP70 517–527, epitopes of all three classes coincided. One particularly immunogenic segment was the B-cell epitope 526–545 in Grp78, which also included three consecutive class II epitopes. This region contained residues 526–539, homologous to SPEA 163–176, encompassing the SAg domain motif ¹⁶⁵TNKKMVTAQELD¹⁷⁶ ( Figure 2B ). This SPEA segment also included two class II epitopes with affinity for the same allele to which also the Grp78 epitopes bound ( Figure 4B ). Another significant region, HSP70 480–511, aligned with SPEM 105–136 in pairwise alignments ( Supplementary Figure S2 ). This SPEM segment, which included the SAg motif ¹³⁰FQNRFVT¹³⁶ ( Figure 3A ), contained class I epitopes that overlapped two class I epitopes of HSP70 480–493 ( Supplementary Table S2 ).

Figure 5 illustrates more clearly how the sequences of Grp78 and HSP70 are composed almost entirely of overlapping epitopes from different classes. While the left regions of both HSP sequences contain nearly identical types of epitopes of similar lengths, the right regions show some differences. Specifically, the HSP70 segment (residues 517–545) includes peptides that function as both MHC-II and B-cell epitopes and also overlap more extensively with MHC-I epitopes compared to the corresponding Grp78 segment (residues 540–568).

Similarly, the left regions of both HSP sequences exhibit similarity exclusively with specific sequences from SPEC and SPEM. In contrast, the right regions share similarities with a broader range of exotoxin sequences. Notably, it is the Grp78 sequence—in contrast to HSP70—that shows a more extensive resemblance to all exotoxin sequences, many of which contain multiple MHC-I epitopes.

A careful inspection of the multiple exotoxin sequences aligned with the HSP sequences also shows how broad is the overall similarity coverage by both Grp78 and HSP70 of SPEA (163–217), SPEC (129–179) and SPEM (105–159). This similarity fully encompasses the shared SAg domain of these exotoxins (see Figure 2A ). Less extensive similarity is instead observed between the HSP sequences and SPEK (185–206), although its SAg region is partially covered, mostly by Grp78 ( Figure 5 ).

The preceding findings supported the notion that the immunogenic sequences of Grp78 and HSP70 possess intrinsic capacity to trigger a complete immune response, comparable, or even superior, to that elicited by streptococcal SAg exotoxins. These sequences, when presented as tumor Ags, may therefore have the potential to activate effective anti-tumor immunity. Despite extensive experimental evidence demonstrating their role as tumor Ags, HSPs have not yet been translated into clinical use as diagnostic or prognostic biomarkers. To explore this possibility further, we examined whether the immunogenic regions of Grp78 and HSP70 exhibited sequence similarity with tumor-associated antigens currently used in clinical practice, namely 5T4 and CEA. The trophoblast glycoprotein 5T4 is a cell surface antigen overexpressed across a broad spectrum of malignancies, with limited expression in normal tissues (60). Although 5T4 has been evaluated as a target for tumor immunotherapy (18), clinical outcomes have remained modest (35, 36, 61).

No sequence similarity was observed between 5T4 and the immunogenic region of Grp94 476–496. In contrast, both Grp78 and HSP70 exhibited high-scoring sequence alignments with 5T4. Pairwise alignments conducted using both T-Coffee and Clustal Omega revealed a high degree of similarity, with the strongest alignment occurring between residues 493–504 of Grp78 and 470–479 of HSP70 and the 283–292 region of 5T4 ( Figure 6A ). Multiple sequence alignment further confirmed this observation, yielding a consensus score exceeding 90% when the full-length sequences of Grp78 and HSP70 were aligned with 5T4 ( Supplementary Figure S4 ), a result not replicated with any other HSP.

Next, sequence similarity was assessed with CEA, a well-characterized glycoprotein of the immunoglobulin superfamily and a widely used biomarker in various cancers (62). As with 5T4, Grp94 did not show any significant similarity to CEA. However, homologous immunogenic sequences of Grp78 and HSP70 did display notable alignment with CEA. Unlike 5T4, where the similarity was confined to a single segment, CEA exhibited multiple, discrete regions of similarity. This is consistent with the genomic architecture of the CEA family, which comprises a series of tandem gene duplications encoding repeated amino acid motifs (62).

Grp78 sequences 494–517 and 547–568, both within the broader 494–568 region, aligned with CEA sequences, with which also the HSP70 471–493 segment showed similarity ( Figure 6B ). Additionally, the HSP70 region spanning residues 524–545 aligned with three other distinct CEA sequences. When mapped sequentially, the regions of CEA showing similarity to HSP sequences revealed that HSP70 aligned with a broader set of CEA segments ( Figure 6C ). Notably, while the left portions of the HSP sequences (Grp78 493–504 and HSP70 470–479) were more similar to 5T4 ( Figure 6A ), their right portions (Grp78 547–568 and HSP70 524–545) demonstrated greater similarity to CEA ( Figure 6B ).

These findings reinforce the immunogenic relevance of the Grp78 and HSP70 sequences and support their potential involvement in tumor immunity through cross-reactivity with clinically recognized tumor antigens.