Introduction

Immune checkpoint inhibitors (ICIs) offer potentially curative cancer treatment by boosting anticancer T-cell immune responses in treated patients. However, their inability to stimulate specific T cells in “cold” tumors significantly limits their efficacy. Therapeutic vaccines could ideally complement ICI treatment in cancers with low mutational burdens and limited spontaneous T-cell responses.1

Despite the potential of therapeutic vaccines, most vaccine trials have failed, largely because of their inability to sustainably stimulate antitumor T-cell responses.2 3 Effective induction of antitumor cytotoxic T cells (CTL) responses initially depends on the selection of appropriate tumor antigens. Typically, tumors present tumor-associated antigens (TAAs) that are shared among patients, offering broad applicability but limited immunogenicity. Tumor-specific antigens or neoantigens, not subject to thymic deletion, are theoretically highly immunogenic; however, their clinical applicability is limited due to patient specificity and rarity in low-mutational burden tumors.4 5 We leveraged two well-described mechanisms to enhance CD8+T cell responses against tumors: peptide molecular mimicry and T cell cross-reactivity. Molecular mimicry involves using peptides from microorganisms that share sequence or structural similarities with tumor antigens, enabling the immune system to target tumor cells like microbial invaders. T cell cross-reactivity allows a single T-cell receptor (TCR) to recognize and respond to multiple peptide-major histocompatibility complexes (MHC), enabling CD8+T cells primed against microbial antigens to also target similar tumor antigens.6 7 We developed a systematic bioinformatic approach to identify sequence similarities between bacterial and TAA peptides. The gut microbiome composition shapes the immune system and impacts vaccination and ICI therapy outcomes.8 Furthermore, the microbiome encodes billions of “foreign” antigens that potentially maintain and trigger a memory T-cell pool at the gut level. Thus, exposure to commensal epitopes might generate rapid induction of a pre-existing pool of cross-reactive T cells, which could be leveraged to maximize the efficacy of peptide-based immunotherapies.9 10 Taking advantage of the widespread presence of these commensal bacteria in the human population, we expected to stimulate a diverse range of memory T cells capable of cross-reacting with TAAs. By employing commensal-derived peptides (CDPs) with high affinity for human leukocyte antigen (HLA) molecules, we addressed the limited immunogenicity of TAA-derived peptides (TAAps) and aimed to activate a robust memory T-cell repertoire maintained to protect against commensal bacteria.

We describe here this approach, termed “oncomimicry”, which enables the discovery of commensal-derived short peptides mimicking TAAps and eliciting potent CTL responses. The selection process for these OncoMimics peptides (OMPs) relies on various criteria, including sequence homologies between OMP and their TAAp counterparts, binding affinities to HLA class I allelic products, predicted cleavage scores and the frequencies of commensal bacterial sources expressing the selected OMPs in the human population. OMP candidates that met these criteria were tested and validated for their immunogenicity and ability to elicit cross-reacting TAAp-specific CTL responses in HLA-A2-humanized mice, resulting in tumor regression. Ex-vivo experiments showed that the identified OMPs stimulated human T-cell proliferation and triggered cytotoxic activity against target cells with homologous TAAs. Finally, initial data from an ongoing clinical trial (NCT04116658) demonstrated that OMPs generate fast, potent and long-lasting immune responses in patients, providing a strong rationale for using CDPs to enhance peptide-based immunotherapy.

Material and methodsIn silico OncoMimics selection

The identification process for OMP is illustrated in figure 1A and refer to the online supplemental methods section for a detailed description.

Figure 1

Discovery and selection process of OMPs for cancer immunotherapy. (A) OMPs discovery: A two-step process. (1) A data set of HLA-A2 peptides derived from tumor-associated antigens (TAAs) was generated from public human databases according to several parameters. (2) OMP candidates were identified by scanning each TAA-derived peptide against all peptide sequences of the same length from gut microbiome catalogs using different homology criteria (detailed in online supplemental M&M). (B) TAA-derived peptides distribution based on the number of their commensal-derived peptides: the graph shows the distribution of 224 TAA-derived peptides (TAAps) based on the number of their associated CDPs. (C) Repartition of TAA-derived peptides based on the HLA-A2 affinities of their commensal-derived peptides: the graph represents the repartition of TAA-derived peptides based on two categories: TAAp with strong binder (SB) CDPs≥TAAp or TAAp with no SB strong binder CDPs≥TAAp. In other words, the presence (in red) or absence (in gray) of at least one strong binder CDP, is stronger than its TAAp counterparts. In the legends of figure 1C and E, the symbol “≥” indicates “a stronger predicted affinity than”, while “<” signifies “a weaker predicted affinity than”. (D) Commensal-derived peptide distribution based on HLA-A2 predicted binding affinities: the graph shows the distribution of commensal-derived peptides based on their HLA-A2 predicted binding affinity. Commensal-derived peptides with the highest affinities (lower nanomolar values) are shown on the left side of the graph. We classified the CDPs into three categories based on their %rank (according to NetMHCpan3.0 recommendations): strong binders (SB) were defined as having %rank<0.5, weak binders (WB) with %rank<2 and no binder peptides with %rank>2. (E) Commensal-derived peptide repartition based on their HLA-A2 predicted binding affinity compared with their TAAp counterparts: The graph shows how commensal-derived peptides are categorized into five groups: SB commensal-derived peptide with an HLA-A2 affinity equal to or stronger than the associated TAAp; SB commensal-derived peptide with an HLA-A2 affinity equal to or weaker than the associated TAAp; WB commensal-derived peptide with an HLA-A2 affinity equal to or stronger than the associated TAAp; WB commensal-derived peptide with an HLA-A2 affinity equal to or weaker than the associated TAAp; Commensal-derived peptides that are no binders to HLA-A2. (F) Characteristics of the 10 selected HLA-A2 OMPs in this study and their homologous TAAps: gene ID of the targeted TAA, sequence name of TAAp and corresponding OMP. Amino acid (aa) sequences predicted affinity rank and binding affinity (nM) to HLA-A*02:01 molecule of each pair and the numbers of mismatches are shown. Amino acid substitutions are indicated in red and those in dark bold are the five amino acids in the peptide core. Fold increases in affinity between OMP and corresponding TAAp. CDPs, commensal-derived peptide; HLA, human leukocyte antigen; OMP, OncoMimics peptide.

Measurement of peptide relative binding affinity for HLA-A*02:01

Peptide relative binding affinities for HLA-A*02:01 were measured as previously described with some minor modifications.11 Briefly, T2 cells were incubated with a titration of each tested peptide and 100 ng/mL human β2-microglobulin (Sigma) in serum-free medium (TeXmacs) at 37°C for 20 hours. After incubation, the cells were washed and surface-stained with the HLA-A2 mAb (REA517). HIVpol1,027–1,035 and FlhA645–653 peptides were used as the positive and negative reference controls, respectively. The geometric mean fluorescence intensities of HLA-A2 staining were measured by flow cytometry (MACSQuant analyzers, Miltenyi Biotec) to determine the percentage of HLA-A*02:01 molecules stabilized at the cell surface, as follows:

The relative affinity (RA) of each peptide for HLA-A*02:01 was determined as the ratio of the peptide concentration to the reference peptide concentration (HIV) that stabilized 20% of the HLA-A2 signal at the cell surface. The lower the RA, the stronger the peptide binding to HLA-A*02:01.

Assessment of peptide/HLA-A*02:01 complex stability

The stability of peptide binding to HLA-A*02:01 was measured using T2 cells pulsed overnight with individual peptides and human β2-microglobulin, followed by the indicated chase kinetics performed in the presence of Brefeldin A (Sigma) to inhibit the neosynthesis of HLA-A*02:01, as previously described.11 The dissociation complex at 50% (DC50) was defined as the half-life of the peptide-MHC complex and was calculated as the time required to observe 50% decay of the initially stabilized peptide-MHC complex.

Animal models

All experimental protocols were approved by the local Ethics Committee on Animal Experimentation and followed the guidelines of the EU. For Enterome, approvals were from CEEA N°51, CERFE (D91228107), and APAFIS (#35 017–2021101316078260 v7). At the Université de Franche-Comté, approvals were from the local Animal Ethics Committee (#58) and APAFIS (#2021–004-OA-12PR). Both received validation from the French Ministry of Higher Education, Research and Innovation, and the French Ministry of Agriculture. The previously described HLA-DRB1*01:01/HLA-A*02:01-transgenic mice (A2/DR1 mice), H-2 class I and II knockout animals were obtained from Francois Lemonnier (Institut Pasteur, Paris, France).12 CD8+ and CD4+ T cells were restricted to HLA-A*02:01 and HLA-DR1*01:01, respectively. All the mice used in the described studies were bred and maintained under specific pathogen-free conditions at Charles River Laboratories.

A2/DR1 prime-boost immunizations

Mice were immunized on days 0 (prime) and 14 (boost) with the indicated MHC class I peptides (5, 30, or 95 nmol per mouse) and MHC class II helper peptide (universal cancer peptide 2 (UCP2), 30–100 µg per mouse) emulsified in Montanide ISA 51 VG (ratio 50:50, Seppic). Immunization was performed subcutaneously using 100 µL of the emulsified preparation.

Restimulation of peptide-specific CD8+ mouse T cells ex vivo postimmunization

To determine the immunogenicity of the analyzed peptides and the cross-reactivity of specific T cells against human TAA homologs, mice were euthanized 21 days post-prime immunization and the number of IFN-γ peptide-specific CD8+ T cells in the spleen of the animals was determined by Enzyme-Linked ImmunoSpot (ELISpot) (Mabtech, 3321-4APT-2 kit) following the manufacturer’s recommendations. Briefly, spleens were harvested and mechanically disrupted and red blood cells (RBCs) were lysed with RBC lysis buffer (Miltenyi Biotec). 2×105 splenocytes were restimulated per ELISpot well using the indicated peptides at a concentration of 10 µM. PMA (Phorbol-myristate acetate, 0.1 µM, Sigma) and ionomycin (1 µM, Sigma) were used as positive controls and EZH2-B2, an HLA-A2:01-restricted peptide, was used as a negative control. Cells were cultured for approximately 20 hours in media (Roswell Park Memorial Institute (RPMI)-1640 (Sigma), 10% fetal bovine serum (VWR), 1% GlutaMAX (Gibco), 1% non-essential amino acid (aa) (Sigma), 10 mM HEPES (Sigma), 1 mM sodium pyruvate (Sigma)+1% penicillin/streptomycin (PenStrep) (Sigma), 50 µM β-mercaptoethanol (β-ME) (Sigma)) and respective peptides at 37°C and 5% CO2. IFN-γ spots were detected according to the manufacturer’s instructions and counted using an iSpot ELISpot Fluorospot reader system (AID). The number of spots obtained for each mouse and each condition was subtracted from the background values (cells cultured in media only) and normalized to the frequency of total T cells from each mouse, resulting in the number of specific IFN-γ-producing T cells/106 T cells.

Tumor protection model

After prime-boost immunization, A2/DR1 mice were subcutaneously injected 21 days post-prime immunization in the right flank with 0.5×106 SARC-A2-hCD20-GFP or SARC-A2-GFP sarcoma cells resuspended in phosphate-buffered saline (PBS). Tumor growth was evaluated twice a week using a caliper and the mice were euthanized when their tumors reached a volume≥300 mm2.

In vivo cytotoxicity in humanized HLA-A2 mice

To test the cytotoxic function of CD8+ T cells that were activated following vaccination, immunized A2/DR1 mice were challenged 6 days post-boost immunization with syngeneic splenocytes pulsed with either a mix of peptides or individual peptides. Unimmunized A2/DR1 donor mice were euthanized and a suspension of syngeneic splenocytes was prepared. To assess the cytotoxic response against peptide pools, splenocytes were split into two fractions and labeled with a cell tracking dye (Carboxyfluorescein Succinimidyl Ester (CFSE) or Cell Trace Violet, Thermo Fisher Scientific) using a low (0.3 µM) or a high concentration (3 µM). To determine the individual contribution of each peptide, refer to online supplemental materials. The bright populations used as target cells were pulsed for 2 hour at 37°C with a peptide pool or a specific peptide at a final concentration of 100 µM before being mixed at an equal ratio with the other populations. Cells were injected intravenously into immunized and naïve mice and the in vivo cytotoxic response was assessed 20 hours post-injection. Antigen-specific lysis was calculated as follows:

Generation of antigen-specific CD8+ CTL in HLA-A*0201 HD PBMCs

Peripheral blood mononuclear cells (PBMCs) were stimulated for 24 hours with OMPs at 10 µM in ImmunoCult medium (STEMCELL). Antigen-specific T cells were enriched by magnetic isolation using a CD137-VioBright-Fluorescein Isothiocyanate (FITC) mAb (REA765, Miltenyi Biotec), anti-FITC microbeads and LS MACS columns (Miltenyi Biotec) as previously described.13 After CD137 enrichment, polyclonal T cell expansion was performed for 8 days with ImmunoCult Human CD3/CD28 T Cell Activator (STEMCELL) in expansion media (ImmunoCult culture medium supplemented with IL-2 (50 U/mL), IL-7 (12.5 ng/mL), IL-15 (12.5 ng/mL) and IL-21 (62.5 ng/mL) (all from Miltenyi Biotec)). Cells were then restimulated every 10 days using peptide-loaded T2 cells (APCs) at a ratio of 10:1 in the expansion medium, as previously described.14 Before co-culture, T2 cells were treated for 1 hour with Mitomycin C (Sigma) at 20 µg/mL, washed and loaded with 100 ng/mL β2-microglobulin (Sigma) and 10 µM peptides overnight. The frequency of antigen-specific CD8+ T cells was assessed by surface staining with CD8 VioGreen mAb (REA734) and fluorescently labeled tetramers. The percentage of cross-reactivity of each healthy donor (HD) (%) was calculated as follows:

Flow cytometry-based cytotoxicity assay in HLA-A*0201 HD PBMCs

T2 target cells were labeled with the CellTrace Far Red Proliferation Kit (Thermo Fisher Scientific) and loaded overnight with 10 µM of the individual peptide and 100 ng/mL β2-microglobulin. Peptide-loaded target cells were co-cultured with peptide-specific CD8+ T cells at 37°C for 24 hours.15 Media and EZH2-B2 irrelevant peptides were used as the negative controls. Antigen-specific CD8+ T cells were stained with VioGreen or PerCPVio700 CD8 mAbs (REA734) and dead cells were identified using the LIVE/DEAD Fixable Violet Dead Cell Stain Kit. Conditions were performed in duplicate and SD were calculated. The percentage of specific cell killing was calculated as follows:

Statistical analyses were performed using the GraphPad Prism (V.8).

Immunomonitoring protocolsTetramer staining assay and memory phenotype analyses in patients

Tetramer staining was performed ex vivo on thawed PBMCs and/or cells that had been subjected to in vitro stimulation (IVS). Cells were washed with Fluorescence-Activated Cell Sorting (FACS) Buffer (PBS supplemented with 2% Fetal Calf Serum (FCS), 0.02% Sodium Azide (NaN3) and 2 mM EDTA) and incubated with two separate tetramer mixes (one containing OMPs and the other containing TAAp-tetramers) each at 2.5 µg/mL for 30 min at room temperature (RT). In some experiments, PBMCs were also incubated with three separate mixes, each containing matched OMP and TAAp tetramers, to assess T cell cross-reactivity. The tetramers were diluted in PBS supplemented with 0.02% NaN3, 2 mM EDTA and 50% FCS for staining. The cells were then washed once with FACS buffer and incubated with mAbs against CD4 (APC-Cy7, BD Biosciences), CD8 (PE-Cy7, Beckman Coulter), CD14 and CD19 (both PerCP, BioLegend) for 20 min at 4°C. A dead live-cell marker (Zombie Aqua, BioLegend) was also included. In the ex vivo setting, mAbs against CD45RA (FITC, BD Biosciences) and CCR7 (BV650, BioLegend) were included in the staining mix to evaluate the naïve/memory/effector phenotypes of the antigen-specific CD8+ T cells. After three washes, the samples were analyzed using an LSRFortessa Cell Analyzer (Becton Dickinson). At least 750,000 and 600,000 cells were acquired in the ex vivo and IVS settings, respectively. The data were analyzed using FlowJo V.10. Tetramer+ T cells were presented as the percentage of cells within living CD8+ lymphocytes. For the analysis of the naïve/memory/effector phenotype, the results are shown as the percentage of cells within tetramer+ CD8+ cells.

IVS of patient PBMCs

PBMCs from patients were expanded for 12 days IVS with the bacterial peptide pool (EO2316, EO2317 and EO2318) before analysis by tetramer staining, IFN-γ ELISpot assay and intracellular staining assay.16 For the expansion procedure, PBMCs were thawed in a thawing Iscove’s Modified Dulbecco’s Medium (IMDM (Gibco) supplemented with 2.5% heat-inactivated human serum (HS, Capricorn), 1% PenStrep solution (Sigma), 50 µM β-ME and 3 µg/mL DNase I (Sigma)), washed (1300 rpm, 8 min, RT), counted and resuspended in dedicated medium (IMDM supplemented with 10% HS, 1% PenStrep and 50 µM β-ME, hereafter referred to as TCM). Cells were seeded either in 24 or 48 well plates at approximately 2.5–3.5×106 cells per well and cultured for 24 hours (37°C, 5% CO2). On day 1, peptides were added to the culture medium (final concentration of 1 µg/mL for each OMP). On day 3, IL-2 was added to the culture medium (final concentration of 2 ng/mL; rhIL-2 reference: R&D 202-IL-010). On day 5, the cells were split by 1/3 and 2 ng/mL IL-2 was added again. On days 7 and 9, the medium was removed from each well and replaced with fresh TCM medium containing IL-2 (2 ng/mL). If necessary, the cells were split in a 1:2 ratio on day 9. On day 12, the cells were collected and their viability and number were assessed with an automated cell counter (Nucleo Counter NC-250) using an Acridine Orange-4′,6-Diamidino-2-Phenylindole (AO-DAPI) staining reagent (solution 18 from ChemoMetec).17

IFN-γ ELISpot assay for patient T cells

For the ELISpot assay, after 12 days of IVS with OMP pools, cells were collected and plated in ELISpot plates (Merck Millipore) pre-coated with a monoclonal anti-IFN-γ mAb (clone 1-D1K purified Mabtech) at a density of 200,000 cells/well. Cells were incubated with the human peptide pool (IL-13RA2, BIRC5 and FOXM1 at 5 µg/mL each) or individual bacterial peptides (EO2316, EO2317, or EO2318, each at 1 µg/mL) for 26 hours. A peptide solvent (10% Dimethyl Sulfoxide (DMSO) in water) was used as the negative control. Phytohemagglutinin (PHA) at 10 µg/mL was used as the positive control. After incubation for 26 hours, the cells were discarded and IFN-γ was detected by adding anti-human IFN-γ biotinylated mAb (clone 7-B6-1, Mabtech) for 2 hours, followed by incubation with ExtrAvidin Alkaline Phosphatase (Sigma) for 1 hour and finally, 5-Bromo-4-Chloro-3-Indolyl Phosphate/Nitro Blue Tetrazolium (BCIP/NBT) substrate (Sigma) was added to the wells. Plates were left to dry for at least 12 hours at RT in the dark prior to plate imaging and spot counting using an Immunospot S6 Universal Analyzer (CTL). The number of spots in peptide-stimulated wells (n=3 for each condition) was compared with those obtained in the negative control (solvent control wells) using a permutation test (distribution-free resampling DFR2x).18 If less than three replicates per condition were available, analyses were performed manually based on positivity criteria from the laboratory (at least twice the spot count in peptide-stimulated wells compared with the solvent control plus a minimum of 6 spots per 100,000 cells seeded).

Cytotoxicity assay on glioblastoma cell lines

PBMCs from the three vaccinated patients were stimulated in vitro with OMP peptides (EO2316, EO2317 and EO2318) for 12 days as described above. Cells were then stained with TAAp-tetramers and the CD8+ T cells specific for IL13RA2, BIRC5 and FOXM1 were isolated with a BD Influx Cell Sorter. Sorted cells were polyclonally amplified using ImmunoCult Human CD3/CD28 T Cell Activator (STEMCELL) in expansion media (ImmunoCult culture medium supplemented with IL-2 (50 U/mL), IL-7 (12,5 ng/mL), IL-15 (12,5 ng/mL) and IL-21 (62,5 ng/mL) (all from Miltenyi Biotec)). To perform the cytotoxic assay, U87 (HLA-A2+) and U118 (HLA-A2-) cell lines were labeled with CellTrace Violet (Thermo Fisher Scientific) and co-cultured with the amplified TAAp-specific CD8+ T cells at 37°C for 24 hours in 50% RPMI/50% X-vivo 15 media at different effector-to-target cell (E:T) ratios. T2 target cells were labeled with CellTrace Violet and co-cultured with the amplified TAAp-specific CD8+T cells, peptides (OMPs or irrelevant EZH2 control peptide) and β2-microglobulin at 37°C for 24 hours in 50% RPMI (Gibco)/50% X-vivo 15 (Lonza) media. Media and EZH2-B2 irrelevant peptides were used as negative controls. Specific CD8+T cells were stained using AF700 CD8 (clone SK1, BioLegend) mAb, and dead cells were identified using the LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit. Conditions were performed in duplicate. The percentage of specific cell lysis was calculated as follows:

Statistical analysis

Results are expressed as mean±SEM or SD. The Mann-Whitney U test was used to compare the two groups. Comparisons between tumor growth curves were performed using a two-way analysis of variance test and multiple comparisons were corrected using the Bonferroni coefficient. Statistical significance was determined using Prism software (GraphPad software). Statistical significance was set at p<0.05.

ResultsComprehensive multistep process for discovering and selecting OMPs for cancer immunotherapy

We developed a comprehensive bioinformatic pipeline to identify CDPs that exhibit molecular mimicry (sequence similarities) with TAAps. This approach targets peptides with enhanced affinity for HLA class I molecules, particularly HLA-A2, owing to its prevalence in approximately 49% of the Caucasian population, with a significant majority expressing the HLA-A*02:01 allelic variant.19 Our strategy was to identify immunogenic CDPs capable of initiating cross-reactive CD8+T cell responses against cancer cells presenting these TAAps. We screened 224 HLA-A2 verified TAAps derived from 113 well-described TAAs against a large public database of gut commensal proteins (online supplemental table 1 and online supplemental M&M). This database comprises almost 10 million genes from 1267 individuals, highlighting the potential of the gut microbiome as a source of short peptides that can potentially mimic all described MHC class I tumor peptides.20 Peptide pairs (CDP/TAAp) were selected according to criteria aimed at maximizing aa changes that improve HLA-A2 binding affinity (anchor positions) while maintaining strict identity within the central aa sequence (core positions) crucial for TCR recognition when presented by MHC class I (figure 1A).21

Our screening identified 95,718 CDPs as potential matches, with the number of CDPs per TAAp ranging from 1 to 3,543 (figure 1B, online supplemental table 2). 74.7% of the 9-mers and 61.1% of the 10-mers from the selected TAAps queries returned at least one strong binder (SB) HLA-A2 CDP with enhanced binding properties, demonstrating that the majority of the well-documented HLA-A2 TAAps have at least one SB CDP equivalent (figure 1C). Among all CDPs, 9.2% of 9-mers and 8.0% of 10-mers were predicted to be SB (figure 1D), with 4.6% predicted to have higher affinities than their TAAp counterparts. We designated this high-affinity CDPs subset as OMP candidates, meeting our criteria for sequence homology with TAAps and exhibiting increased HLA-A2 affinities (figure 1E).

10 OMPs were selected to assess their biological potential. Six OMP-matched TAAps derived from overexpressed tumor antigens (IL-13RA2, BIRC5, FOXM1, UBE2C, CDC20 and KIF2C) prevalent in solid tumors such as glioblastoma, lung, breast and colorectal cancers have been identified as key tumor drivers.22–27 While these six TAAs are associated with a range of solid tumors, IL13RA2, BIRC5, and FOXM1 are ideal targets for immunotherapy in patients with glioblastoma. IL13RA2, BIRC5, and FOXM1 are overexpressed in glioblastoma tissue and play critical roles in tumor progression. All three proteins are found overexpressed in a large fraction of glioblastomas (60–85%) and contribute to the survival of tumor cells, being associated with poor prognosis. The remaining four OMPs matched TAAps derived from CD22, CD37, BAFF-R and CD20, which are B cell surface markers, used as targets in B cell lymphoma therapies.28 These 10 OMPs were selected from a comprehensive pool of candidates, distinguished as the most promising based on several parameters. This included sequence similarity to TAAps to retain molecular mimicry, high HLA-A2 binding affinity, prediction of potent immunogenicity, likelihood of proteasomal processing and prevalence within the gut microbiome to augment the probability of a responsive T cell repertoire (online supplemental table 2and online supplemental M&M for details). The selected OMPs presented two or three aa mismatches and exhibited significantly enhanced predictive affinities, ranging from 1.8-fold to 400-fold higher than those of their TAAp counterparts (figure 1F). This selection process aims to predict whether these OMPs can trigger strong immune responses in a wide population segment.

OMPs induce cross-reactive T-cell responses in vivo

To validate the in silico HLA-A2 affinity predictions, we used a TAP-deficient HLA-A*02:01-positive T2 cell line. These cells provided an established model to assess peptide-HLA-A2 binding (online supplemental figure 1A). Our in vitro binding assays revealed that each tested OMP exhibited enhanced binding affinity compared with the corresponding TAAp (figure 2A, online supplemental figure 1B,C). Peptide-HLA-A2 stability assay further highlighted that OMP/HLA-A2 complexes were significantly more stable than their TAAp/HLA-A2 counterparts, with dissociation half-life complex (DC50) values ranging from 2 hours to more than 24 hours (figure 2B, online supplemental figure 1B and D). This enhanced stability suggests prolonged antigen presentation, potentially improving T cell recognition and response. The immunogenicity capacity and ability of OMPs to induce cross-reactive T cells were subsequently investigated in vivo using humanized HLA-A2-DR1 transgenic mice. On OMP immunization, antigen-specific IFN-γ T cell responses were evaluated (figure 2C). Robust T cell responses, defined as exceeding 1,000 IFN-γ SFU (Spot-Forming Unit) per 10⁶ total T cells, were observed for all OMPs tested, while no activation was detected in naïve mice. The induction of cross-reactive T-cell responses was further confirmed by the capacity of OMP-induced T cells to produce IFN-γ on ex vivo restimulation with the corresponding TAAps (figure 2D, online supplemental figure 1E). The number of T cells responding to OMPs and TAAps restimulation on OMPs immunization in individual mice strongly correlates, with a Pearson coefficient ranging from 0.42 for the least correlated peptide pair, OMP10/UBE2C, to 0.99 for the most correlated pair OMP66/BAFF-R. 7 out of 10 peptide pairs showed strong correlations (≥0.80) (online supplemental figure 1F). Cross-reactivity levels, defined as the percentage of OMP-induced T cells that recognize corresponding TAAps, ranged from 13.8% (OMP16/IL-13RA2) to 82.2% (OMP66/BAFF-R) (figure 2D). Based on OMPs capacity to induce TAAp-specific responses, representing the ratio between OMP immunogenicity and cross-reactivity capacity, we observed the OMP65/CD37, OMP66/BAFF-R and OMP17/BIRC5 induce high TAAp-specific IFN-γ responses (>5,000 SFU) and cross-reactivity levels (≥75%). OMP12/KIF2C, OMP11/CDC20, OMP72/CD20 and OMP10/UBE2C induce lower TAAp-specific IFN-γ responses (2,000–5,000 SFU) and cross-reactivity levels ranging from 30% to 80%. Lastly, OMP64/CD22, OMP18/FOXM1 and OMP16/IL-13RA2 induce lower TAAp-specific responses (<2,000 SFU) and cross-reactivity levels≤20% (except OMP64). Overall, these results reveal the capability of OMPs to induce strong IFN-γ-producing T cell responses, recognizing both peptides (OMPs and TAAps). Mice were then immunized with either OMPs or TAAps and T cells were restimulated ex vivo with either peptide (figure 2E). T cell responses against KIF2C, IL-13RA2, BIRC5, CD37 and BAFF-R peptides were weaker when mice were immunized with TAAps than with their mimic OMPs, demonstrating superior immunogenicity for OMP12, OMP16, OMP17, OMP65 and OMP66 over their homologous TAAps (figure 2F, top panel). For CDC20-specific, CD20-specific and UBE2C-specific T-cell responses, the magnitudes were similar, regardless of the initial vaccination (figure 2F, middle panel). Only CD22-specific and FOXM1-specific T-cell responses were weaker after OMP immunization than after TAAp immunization (figure 2F, bottom panel). Additionally, OMP-specific responses induced by OMP immunization were consistently stronger than those induced by TAAp immunization, except for the OMP64/CD22 pair, which showed similar levels (figure 2F).

Figure 2

OMPs selected in vitro for their HLA-A2 binding and stability properties and in vivo for their immunogenicity are capable of eliciting potent cross-reactive responses in HLA-A2/DR1 mice. (A, B) Binding and stability of TAA-derived peptides and their OMP counterparts. Comparison of HLA-A2 binding (A) and stability (B) of TAA-derived peptides (red), their OMP counterparts (blue) and the HIV reference peptide (green) in the T2-binding assay. The data represent four to seven independent experiments. Symbols represent the mean and error bars indicate the SEM. (C) Schematic representation of the in vivo experimental setup used to assess OMPs immunogenicity and CD8+ T cell-dependent cross-reactive responses against TAAps. (D) In vivo immunogenicity and cross-reactivity. The frequency of peptide-specific T cells that produce IFN-γ was determined by ELISpot analysis of splenocytes from mice vaccinated with the indicated peptides. The negative control, OMP and TAAp are shown in gray, blue and red, respectively. Below each graph, the percentages of cross-reactivity (XR) between OMP and its TAA counterpart are displayed. (E) Schematic representation of the in vivo experimental set-up used to compare the capacity of OMPs and TAA-derived peptides to induce a TAAp-specific response. (F) OMP-induced and TAAp-induced cross-reactive responses. Comparison of OMP efficacy in inducing a TAAp-specific response to that of their TAA counterparts in A2/DR1 humanized mice assessed on splenocytes from mice immunized with the indicated peptides by IFN-γ ELISpot assay. Dark blue, red, light blue and orange indicate OMP vaccination and restimulation conditions, OMP vaccination and TAAp restimulation, TAAp vaccination and OMP restimulation conditions and TAAp vaccination and restimulation conditions, respectively. The data shown in (D) and (F) are from one to five independent experiments; symbols indicate individual mice (n=5–25 mice) and bars represent min and max values. Statistical comparisons were performed using an unpaired non-parametric test (Mann-Whitney). *p<0.05, **p<0.001, ****p<0.0001, ns: non-significant. ELISpot, Enzyme-Linked ImmunoSpot; HLA, human leukocyte antigen; OMP, OncoMimics peptide, TAA, tumor-associated antigen.

OMPs trigger CTL responses with antitumor activity

The in vivo cytotoxicity of the OMP-triggered T cells was investigated using a fluorescence-based CTL assay. Following a prime/boost immunization of A2/DR1 mice with the OMP pools (OMP17, OMP18, OMP10, OMP11, OMP12) or (OMP64, OMP65, OMP66, OMP72), a 1:1 ratio of syngeneic splenocytes loaded with the corresponding OMPs (bright labeling) and unloaded (dim labeling) were injected into immunized or control animals (figure 3A). Flow cytometry analysis, conducted 20 hours post-injection, demonstrated that OMP-induced T cells exhibited substantial cytotoxic activity towards OMP-loaded target cells without affecting control cell viability, indicating that OMPs elicit robust OMP-specific CTL responses (figure 3B). Individual OMP-specific CTL activity was observed against T2 cells loaded with most of the OMPs tested, although the CTL activity against OMP64 was lower and the one against OMP72 was negligible (online supplemental figure 2A,B). More importantly, OMP-specific CTL responses induced by OMPs were TAAp-specific. Splenocytes loaded with matched TAAps pool (BIRC5, FOXM1, UBE2C, CDC20 and KIF2C) or (CD22, CD37, BAFF-R and CD20) were efficiently killed, demonstrating the OMP/TAAp cross-reactivity of the OMP induced T cells in vivo (figure 3C).

Figure 3

OMP-based vaccines elicit functional cytotoxic T cells in mice that are cross-reactive with human TAA-derived peptides in vivo. (A) Schematic representation of the experimental setup employed to assess T cell cytotoxicity elicited in vivo by OMPs. HLA-A2-DR1 mice were vaccinated with OMPs and challenged post-vaccination with syngeneic splenocytes labeled with cell tracking dye. Target T cells (red) were pulsed with a pool of OMPs or TAA-derived peptides and mixed at an equal ratio with unpulsed control cells (black) before being adaptively transferred to immunized mice. (B, C) In vivo cytotoxic activity of OMP-induced T cells against OMP-pulsed or TAAp-pulsed target T cells. Percentages of in vivo specific lysis of splenocytes pulsed with the indicated pool of OMPs (B) or TAA-derived peptide counterparts (C) (challenging peptides) after immunization with the indicated peptide pool or control peptides (immunizing peptides). Data shown in (B) and (C) are from two to four independent experiments, symbols indicate individual mice (n=10–20 mice) and bars represent min and max values. Statistical comparisons were performed using an unpaired non-parametric test (Mann-Whitney). ***p<0.001, ****p<0.0001. (D) Flow cytometry analysis performed on SARC-A2 GFP (left panel) and on SARC-A2-GFP-CD20 (right panel) sarcoma cells showing post-transduction expressions of GFP and hCD20 proteins. (E) Schematic representation of the experimental setup used to evaluate the antitumor effect of OMPs. Mice were immunized with OMP72 using a prime-boost administration regimen. 21 days post-prime immunization, mice were inoculated with hCD20-GFP-expressing or GFP-expressing SARC-A2 sarcoma cells and tumor growth was monitored. (F) Tumor kinetics on individual mice over time for each group. Tumor size in A2/DR1 naïve or OMP vaccinated mice engrafted with 0.5×106 SARC-A2-hCD20-GFP or SARC-A2-GFP tumor cells (n=8 mice) was measured (mm2) over 30 days. (G) hCD20 expression assessment in the vaccinated group. Flow cytometry dot plots showing the expression of human CD20 and GFP in SARC-A2 cells extracted from whole tumors on day 30 in animals still bearing tumors (n=4). (H) Vaccine-specific induced T cell responses. The frequency of peptide-specific T cells producing IFN-γ per million splenic T cells was determined by ELISpot on day 30. ELISpot, Enzyme-Linked ImmunoSpot; HLA, human leukocyte antigen; OMP, OncoMimics peptide; TAAp, tumor-associated antigen-derived peptide.

To further explore the OMP-induced TAAp-specific CTL activity in vivo, we established a tumor protection model using the syngeneic HLA-A2+ sarcoma cell line (SARC-A2), as it is the only compatible tumor cell line available for the A2/DR1 mouse model.29 Attempts to develop SARC-A2 constructs with ectopically expressed TAAs were unsuccessful due to rapid TAA expression loss or spontaneous tumor rejection, limiting our preclinical studies. However, SARC-A2 engineered to express the human CD20 antigen as well as GFP (SARC-A2-GFP-hCD20) demonstrated sufficient stability and did not undergo spontaneous rejection, allowing short-term implantation experiments. We established a tumor protection model using SARC-A2-GFP-hCD20 and the GFP-alone control cell line (SARC-A2-GFP) (figure 3D). This approach allowed us to assess the capacity of OMP72 to trigger CD20-targeted CTL-mediated tumor cell killing. After immunization with OMP72, half of the A2/DR1 mice engrafted with SARC-A2-GFP-hCD20 tumor cells exhibited tumor protection compared with the control groups (figure 3E,F). Interestingly, in mice in which tumor control was not achieved, we observed a loss of CD20 surface expression in 99% of the tumor cells, suggesting a potential escape mechanism from OMP72-specific T-cell surveillance (figure 3G). Supporting this hypothesis, comparable levels of OMP72-specific IFN-γ-producing cells were measured in both responsive and non-responsive mice, indicating no significant difference in the immune response induced in either group (figure 3H). Altogether, these results demonstrate the ability of OMPs to induce OMP-/TAAp-specific CTL activity in vivo.

OMP-specific T cells represent a prevalent pool of T cells in the human population

The ability of OMPs to induce efficient TAAp-specific cross-reactive CTL responses was evaluated in human PBMCs isolated from HLA-A2+ HDs and stimulated in vitro in the presence of OMPs (figure 4A). Flow cytometry analysis using peptide-MHC tetramers revealed that all OMPs induced cross-reactive OMP-/TAAp-specific CD8+ T cells (online supplemental figure 3A–C). Within the same OMP/TAAp pair, the extent of cross-reactivity varied among different HD PBMCs, ranging from 10% to 100%. On average, some OMP/TAAp pairs showed cross-reactivity levels below 33% (OMP10, OMP11, OMP18, OMP64), some between 33% and 66% (OMP16, OMP65, OMP66, OMP72) and some above 66% (OMP12, OMP17), compared with the total elicited OMP-specific T cell responses (online supplemental figure 3D). This pattern emphasizes variability not only within individual responses to the same OMP/TAAp pair, but also across different OMP/TAAp pairs. OMP-amplified human T cells effectively recognized and destroyed OMP-loaded T2 target cells, whereas control T2 cells (pulsed with irrelevant peptides or unloaded) were untouched, confirming the specificity of the killing activity. More importantly, OMP-expanded human T cells also killed T2 cells loaded with matched TAAps, demonstrating the TAAp-specific CTL activity of these T cell clones. Reduced and variable cytotoxic values were observed for IL-13RA2, likely because of the low affinity of the IL-13RA2 peptide and its poor stability when loaded onto T2 cells (figure 4B). In most cases, killing occurs at a low E:T ratio of 1:10, strengthening the