Development of a multi-epitope mRNA vaccine targeting tumor-associated antigens for esophageal squamous cell carcinoma

Abstract

Esophageal squamous cell carcinoma (ESCC) has limited treatment options post-immune checkpoint inhibitor (ICI) resistance. We developed IPM514, a universal multi-epitope mRNA lipid nanoparticle (LNP) vaccine targeting tumor-associated antigens identified via transcriptomic analysis of TCGA/GTEx datasets and validated in 132 ESCC patients. IPM514 contains 15 fragments from 9 antigens, covering most patients and over half of HLA subtypes. Peripheral blood mononuclear cells (PBMCs) from healthy donors and ESCC patients stimulated with IPM514 effectively expanded specific T cells that exhibited significant cytotoxic activity against ESCC and other squamous cell carcinomas with similar antigen profiles. In HLA-transgenic mouse models, IPM514 suppressed tumor growth, extended survival, and provided durable protection. Importantly, combination therapy with PD-1 blockade augmented antitumor efficacy by promoting immune cell infiltration, upregulating antigen presentation pathways, and reprogramming the tumor microenvironment toward an anti-tumorigenic state. These results demonstrate IPM514 represents a promising novel mRNA vaccine strategy for improving ESCC immunotherapy.

Introduction

Esophageal cancer (EC) represents a significant global health challenge, with ~511,054 new cases and 445,391 deaths reported worldwide in 2022. This malignancy ranks as 11th most common cancer and the 7th leading cause of cancer-related mortality^1,2. Alarmingly, nearly half of these cases occur in China, underscoring a disproportionate burden of disease in this region². EC is primarily categorized into esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC), with ESCC predominating in developing countries like China, where it constitutes over 90% of EC cases³. Unfortunately, the majority of patients are diagnosed at advanced stages, resulting in a five-year survival rate of less than 26%^4,5. While immunotherapy, particularly PD-1 inhibitors combined with chemotherapy, has emerged as a first-line treatment for advanced ESCC^6–12, resistance remains a critical challenge. Approximately 80% of patients experience primary or acquired resistance, significantly limiting their long-term survival benefits of these therapies¹³. Current second-line treatments, including chemotherapy and multi-target tyrosine kinase inhibitors, demonstrate limited efficacy, with a median overall survival of only six months^14–16. This underscores the urgent need for novel therapeutic strategies to overcome resistance and improve outcomes for ESCC patients.

Cancer vaccines have emerged as a promising strategy in oncology, capable of eliciting robust anti-tumor immune responses. Among these, mRNA-based vaccines are particularly noteworthy due to their ability to induce endogenous expression of tumor antigens, thereby triggering potent CD8⁺ T-cell-mediated immunity. mRNA cancer vaccines can be classified into personalized and universal types. Personalized vaccines, such as BNT122 and mRNA-4157, target tumor-specific neoantigens derived from individual patient mutations^17,18. Early clinical trials have shown encouraging results, including extended relapse-free survival in melanoma patients treated with mRNA-4157 combined with immune checkpoint inhibitors (ICIs)¹⁸. However, the high cost, lengthy production times, and challenges associated with accurately predicting neoantigens limit the widespread applicability of personalized vaccines. Universal mRNA cancer vaccines provide an alternative approach by targeting tumor-associated antigens (TAAs) or tumor-specific antigens (TSAs) that are commonly expressed across patient populations. For instance, BNT111 targets melanoma-associated TAAs and has demonstrated safety and efficacy in early-phase trials¹⁹. These universal vaccines address several limitations of personalized approaches, including scalability, cost, and the risk of immune escape associated with single antigen targeting. For highly heterogeneous and aggressive cancers like ESCC, a multi-epitope universal vaccine has the potential to enhance clinical outcomes and accessibility. Specifically, the fragment-based design, rather than full-length antigens, concentrates immunogenicity on validated HLA-restricted epitopes, avoids immunosuppressive non-epitope regions, and enables strategic linker use to prevent neoepitopes—maximizing specific anti-tumor T-cell responses in the target patient population.

In this study, we presented a novel universal mRNA–lipid nanoparticle (LNP) vaccine, designated IPM514, which encoded multiple TAA-derived epitopes restricted by HLA-A*02:01 and HLA-A*11:01 alleles, ensuring broad coverage across diverse patient populations. By targeting a carefully selected panel of nine TAAs that are frequently overexpressed in ESCC, IPM514 aims to address the challenges of immune escape and resistance, potentially providing a breakthrough in immunotherapy for this aggressive malignancy.

Results

Comprehensive identification and validation of TAAs for IPM514 design

IPM514 developed through a rigorous multi-step screening process aimed at identifying TAAs with high specificity and immunogenicity (Fig. 1A and Supplementary Fig. 1A). Initial screenings of transcriptome datasets from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) revealed 4478 genes that were significantly upregulated in ESCC tumor tissues compared to normal esophageal mucosa (Supplementary Fig. 1B). To ensure broad tumor coverage while excluding candidates expressed in normal tissues, candidates were filtered against the Human Protein Atlas (HPA) and Genotype-Tissue Expression (GTEx) normal tissue datasets, retaining only genes with transcripts per million (TPM) < 0.25 across all 50 normal human tissues analyzed (Supplementary Fig. 1C, D). Subsequently, 12 TAA genes with intracellular retention and adequate antigen exposure were further selected.

Fig. 1. Identification of ESCC antigens and design of the IPM514 mRNA vaccine.

A The screening flowchart for identifying the fragments that enrich HLA-A*02:01 and A*11:01-restricted epitopes within candidate TAAs. B Schematic of the IPM514 design, which includes the following components: SP, HLA-B signal peptide; TAA1-TAA9, 30-mer TAA fragments; L1, GS linker (GGSGGGGSGG); L2: GPGPG linker; PADRE, a Pan DR T Helper cell epitope; MITD, MHC class I trafficking domain. Population coverage analysis of the candidate TAAs (C) and the corresponding HLA subtypes (D) in a validation cohort consisting of 132 participants. Created in BioRender. cuiping, A. (2026) https://BioRender.com/nxir12r. E Distribution of epitopes within IPM514, showing epitopes for the seven most prevalent HLA-I alleles predicted in silico (top) and verified through mass spectrometry (bottom). Immunopeptidomics analysis was performed on seven mono-HLA-expressing K562 cell lines 24 h after electroporation with IPM514 mRNA. The identified peptides were derived exclusively from exogenous IPM514 mRNA; untransduced K562 cells showed no detectable presentation of these peptides (data not shown).

A secondary screening prioritized epitopes restricted to high-frequency HLA subtypes, HLA-A*02:01 and HLA-A*11:01. This process incorporated a combination of mass spectrometry validation and The Immune Epitope Database (IEDB) to confirm the presence of immunogenic epitopes. Ultimately, 15 fragments derived from 9 TAAs (ACTL8, BRDT, GNGT1, MAGEA1, MAGEA3, MAGEA4, MAGEA6, MAGEA11, SMC1B) were selected, with each fragment containing at least one validated HLA-A*02:01 and HLA-A*11:01 epitope (Supplementary Table 1). Three TAAs were excluded due to the absence of appropriate fragments. This process incorporated immunopeptidomics (mass spectrometry-based identification of naturally presented HLA-bound peptides) combined with in silico prediction using The Immune Epitope Database (IEDB) to identify peptide sequences with high HLA-binding affinity. These approaches identify candidate epitopes with potential for immunogenicity, which were subsequently validated through functional T-cell assays.

These selected 30-mer fragments were then integrated into the IPM514 vaccine design using GS linkers (GGSGGGGGSGG) to prevent neoepitope formation (Fig. 1B). To enhance vaccine efficacy, a universal PADRE epitope was included to boost T-helper cell responses²⁰, an HLA-B signal peptide (SP) was added at the N-terminal to facilitate antigen presentation, and a C-terminal MHC class I trafficking domain (MITD) was incorporated to optimize MHC class I-mediated presentation²¹.

RNA sequencing analysis of 132 ESCC patient samples from Beijing Cancer Hospital confirmed high expression of at least one selected antigen across all tumor tissues. Notably, over 90% of patients expressed at least three of the identified antigens, with more than 55% showing expressed six or more (Fig. 1C). Furthermore, ~51% of these patients carried either the HLA-A*02:01 or HLA-A*11:01 subtype (Fig. 1D), and 44% of the total patients (58/132) were HLA-matched and expressed at least three of the selected TAAs, underscoring the broad applicability of the selected antigens across diverse populations. Multi-algorithm epitope prediction across all patient-specific HLA class I alleles indicated extensive theoretical peptide–HLA binding coverage within the cohort (Supplementary Data 1). Co-expression analysis using Pearson correlation coefficients revealed strong clustering within the MAGE family (MAGEA1, MAGEA3, MAGEA4, MAGEA6, MAGEA11; 0.51–0.95, FDR < 0.05), consistent with shared epigenetic regulation. ACTL8 and BRDT formed a distinct cluster, while SMC1B and GNGT1 showed more variable expression patterns (Supplementary Fig. 1E). This coordinated expression of multiple TAAs within individual tumors supports the rationale for multi-target vaccination to minimize immune escape. To assess whether thymic epithelial expression of the selected tumor-associated antigens might contribute to central tolerance, we analyzed publicly single-cell RNA-sequencing²², finding the selected TAAs were either undetectable or expressed at very low frequencies (<1% of cells). Stratification by developmental stage and epithelial subpopulation did not reveal broad enrichment within medullary thymic Epithelial Cells (Supplementary Fig. 1F, G).

Interestingly, analyses using RNA-seq data from TCGA revealed that the 9 TAAs included in IPM514 were also overexpressed in several other cancer types, particularly lung squamous cell carcinoma (LUSC) (Supplementary Fig. 2). These findings suggest that IPM514 may offer therapeutic potential beyond ESCC, potentially extending to other malignancies.

Based on the predicted epitope distribution for IPM514, multiple high-density epitope clusters for HLA-A*02:01 and HLA-A*11:01 were identified throughout the IPM514 sequence. Subsequent, we confirmed that these clusters indeed harbor endogenously processed peptides by mass spectrometry-based immunopeptidomics (Fig. 1E). The presented regions effectively avoided linker sequences, thereby minimizing the risk of inadvertent neoepitope formation. Additionally, immunopeptidomic analysis indicated that IPM514 can target five other high-frequency HLA types, highlighting its potential to benefit a broader patient population.

IPM514 mRNA optimization enhances stability, expression, and immunogenicity

To improve the translation efficiency and stability of the IPM514 mRNA vaccine, the sequences were optimized by incorporating natural nucleosides based on GC content, codon adaptation index (CAI), and minimum folding free energy (MFE)²³. Seven candidate mRNA sequences were designed and subjected to a screening process to evaluate structural stability, intracellular half-life, and protein translation levels (Fig. 2A–C). In vitro screening revealed versions 1.2, 1.4, and 1.5 as exhibiting superior protein expression and extended half-life (Fig. 2D). These three candidates were subsequently tested in HLA-A2.1 and HLA-A11.1 transgenic mice to assess their immunogenicity.

Fig. 2. Screening of IPM514 for optimal mRNA sequences in vitro and in vivo.

A Non-denaturing agarose gel characterizing mRNA, showing the correlation of gel mobility and minimum free energy. B Kinetics of intracellular degradation of mRNA were assessed at 0, 3, 6, 12, 24, and 48 h post-transfection in K562 cells, as determined by qPCR. Data are presented from three independent experiments. C Protein expression levels were detected using a cell-free protein synthesis system and quantified based on grayscale intensity values. D A summary of results from in vitro mRNA screening is presented. The criteria for determining protein expression levels are outlined in the Materials and Methods. E A total three doses were weekly injected intramuscularly in HLA-A2.1 or HLA-A11.1 transgenic mice. Spleens were collected to assess the potency of antigen-specific immune responses. 2 × 10⁵ cells per well were stimulated for 16 h with either epitope peptide pools or screening peptide pools. The epitope peptide pools consist of peptides most likely to be presented by HLA-A*02:01 (EP1-3) or HLA-A*11:01 (EP4-6) for each antigen fragment of IPM514. Scanning peptide pools include SP1, SP2, and SP3, which encompass all potential IPM514 antigens. Means (SD) are plotted for each group (n = 5 per group). Statistical analyses were performed using one-way ANOVA with Bonferroni’s correction for multiple comparisons. *P < 0.05, **P < 0.01, and ***P < 0.001. F The IFN-γ ELISPOT assay was used to detect positive responses to individual peptides. The Immunogenicity criteria are detailed in the Materials and Methods. Means are plotted for each group (n = 3 per group). Uncropped and unprocessed scans of the gels are provided in Supplementary Fig. 3.

Mice were immunized intramuscularly with IPM514 once weekly for three weeks. ELISPOT assays conducted on splenocytes demonstrated robust secretion of IFN-γ across all candidates in response to stimulation with epitope or scanning peptide pools (Fig. 2E; Supplementary Tables 2, 3). Notably, IPM514-1.2 induced significantly higher T-cell responses in HLA-A2.1 transgenic mice compared to the other two candidates, and also showed moderately enhanced responses in HLA-A11.1 transgenic mice. Consequently, version 1.2 was identified as the optimal candidate.

To evaluate the immunogenic potential of individual antigen fragments, ELISPOT assays were performed to quantify antigen-specific T-cell responses. Fragments producing spot counts greater than 25 were classified as immunogenic, while those exceeding 150 were categorized as strongly immunogenic. In both HLA-A2.1 and HLA-A11.1 transgenic mice, IPM514 induced measurable immune responses for the majority of antigen fragments, demonstrating its effectiveness in activating antigen-specific T cells (Fig. 2F). Specifically, in HLA-A2.1 mice, 13 of the 15 antigen fragments generated positive immune responses, with TAA3b (MAGEA1), TAA7a (MAGEA4), and TAA7b (MAGEA4) exhibiting strong immunogenicity. In HLA-A11.1 transgenic mice, all 15 fragments elicited positive immune responses, among which TAA1a (ACTL8), TAA3b (MAGEA1), TAA4b (MAGEA6), and TAA6b (SMC1B) were identified as strongly immunogenic. Furthermore, IPM514 was also capable of activating MHC class II-restricted immune responses (Supplementary Fig. 4), highlighting its potential to induce a robust and comprehensive anti-tumor immune response.

IPM514 effectively expands antigen-specific T cells with significant tumor cytotoxicity

To evaluate the biological activity of IPM514, peripheral blood mononuclear cells (PBMCs) from healthy donors expressing HLA-A*02:01 or HLA-A*11:01 subtypes were stimulated in vitro. Dendritic cells (DCs) were electroporated with IPM514 mRNA or co-incubated with IPM514 mRNA-LNP, and subsequently these DCs were co-cultured with autologous T cells. After a 21-day in vitro stimulation (IVS), a significant expansion of antigen-specific CD8⁺ T cells, characterized as TNF-α⁺IFN-γ⁺ dual-positive cells, was observed in both the mRNA group and the LNP group compared to control groups (Fig. 3A, B). All donors in the mRNA group exhibited positive responses, with Donor 5 showing the highest expansion at 2.27%.

Fig. 3. IPM514 effectively expands antigen-specific T cells in vitro with significant tumor cytotoxic activity.

A, B Autologous T cells from healthy donors were stimulated in vitro with DCs that were either electroporated with IPM514 mRNA or co-incubated with IPM514 mRNA encapsulated LNP. Untreated DCs used as the negative control (NC). Following 21 days of stimulation, cells were analyzed by intracellular cytokine staining (ICS) after stimulation with epitope peptide pools. Representative dot plots are show (A) for Donor 5, while the frequencies of intracellular cytokines are depicted in (B). Each symbol represents a different healthy donor (n = 10). Data shown are mean (SD) compared using Kruskal–Wallis test followed by multiple comparison. Sti, stimulated; Unsti, unstimulated. C–E Cells expanded in vitro from the NC group (DC + T) and the mRNA group (DC-IPM514-mRNA+T) were selected for cytotoxicity assays. Bulk CD3⁺ T cells (after in vitro stimulation) used as effector cells and co-incubated with T2 cells (C, the dotted line indicates T2 target cells, while the solid line represents T2 target cells loaded with IPM514 peptides), ESCC cells (D, the dotted line indicates KYSE-410 wild-type target cells, and the solid line represents KYSE-410 target cells overexpressing HLA-A2.1 or A11.1), and other squamous cell carcinoma tumor target cells (E) at an effector-to-target ratio of 20: 1 for 24 h. The specific lysis was assessed by a luciferase-based cytotoxicity assay. Means (SD) are plotted for each group. Statistical analyses were conducted using two-way ANOVA with Bonferroni’s correction for multiple comparisons. F, G PBMCs from 22 patients with ESCC were incubated with IPM514 peptide pools to induce antigen-specific T cell expansion, while the control group was incubated without peptides. After 17 days of incubation, cells from both groups were analyzed via ICS following stimulation with epitope peptide pools. The Heatmap in (F) and frequencies in (G) of TNF-α⁺, IFN-γ ⁺, and TNF-α⁺IFN-γ⁺ producing T cells are shown. Statistical analyses were conducted using paired T test. ns nonsignificant. *P < 0.05, **P < 0.01, and ***P < 0.001 indicate significance levels.

Cytotoxic assays further confirmed the functional capabilities of the expanded T cells. Compared to control T cells, IPM514-expanded T cells demonstrated significantly enhanced cytotoxicity against T2 cells loaded with IPM514-derived peptides (Fig. 3C). In assessing cytotoxicity against ESCC cell lines, KYSE-410 (HLA-A*24:02) were engineered to overexpress HLA-A*02:01 or HLA-A*11:01. Importantly, IPM514-expanded T cells exhibited significantly enhanced cytotoxicity against KYSE-410 cells that overexpressed the relevant HLA types, with efficacy further improved following peptide loading (Fig. 3D, Supplementary Fig. 5A, B).

To examine antigen-specific cytotoxicity against additional squamous cell carcinoma (SCC) cell lines, SCC-4, SCC-25 (tongue SCC), NCI-H520, and HCC95 (lung SCC) were used (Fig. 3E, Supplementary Fig. 5C, D). The magnitude of cytotoxicity varied across these lines, with NCI-H520 cells showing the strongest responses, followed by HCC95, while SCC-4 and SCC-25 exhibited the weakest killing. This pattern broadly correlated with the number and abundance of IPM514-associated antigens expressed by each line (Supplementary Table 4), although other tumor-intrinsic factors may also contribute to differential susceptibility. These findings demonstrate that IPM514-expanded T cells can recognize and kill multiple SCC types, supporting the potential broad applicability of this vaccine.

In parallel, PBMCs from 22 ESCC patients were stimulated with IPM514 peptide pools. Positive antigen-specific T-cell expansion was observed in 68% (15/22) of patients, as indicated by increased frequencies of TNF-α⁺, IFN-γ⁺, and double-positive CD8⁺ T cells (Fig. 3F, G). Notably, two patients lacking HLA-A*02:01 and HLA-A*11:01 demonstrated antigen-specific T-cell responses. Their HLA alleles (P18: A*02:06/ A*30:01, P19: A*02:03/ A*02:07) belonged to the A2 and A3 supertypes, suggesting that IPM514 may induce T-cell reactivity across HLA superfamilies.

IPM514 demonstrated robust antitumor activity in HLA transgenic mouse models

To evaluate the antitumor effectiveness of IPM514, HLA-A2.1 and HLA-A11.1 transgenic mice were subjected to intramuscular immunization with IPM514 once weekly for three weeks (Fig. 4A). Subsequent analysis of splenocytes from these immunized mice displayed robust antigen-specific responses, characterized by a significantly higher percentages of IFN-γ⁺ CD8⁺ T cells when compared to PBS-treated controls (Supplementary Fig. 6A, B). ELISPOT assays further corroborated these findings, revealing substantial increases in IFN-γ secretion in response to both epitope (EP) or scanning peptide pools (SP) stimulation, indicating strong antigen-specific T-cell activation (Supplementary Fig. 6C, D).

Fig. 4. Immunization with IPM514 induces tumor eradication and significantly prolongs survival.

A Schematic of the luciferase-based ex vivo cytotoxicity assay. HLA-A2.1 or HLA-A11.1 transgenic mice received three intramuscular injections of IPM514 at weekly intervals. Spleens were harvested three days after the third immunization, and CD8⁺ T cells were isolated for co-incubation with tumor cells to assess ex vivo cytotoxicity. Created in BioRender. cuiping, A. (2026) https://BioRender.com/nxir12r. CD8⁺ T cell derived from IPM514 immunized HLA-A2.1 (B) or HLA-A11.1 (C) transgenic mice exhibited specific lysis of MC38-HHD-A2.1-514 or MC38-HHD-A11.1-514 cells, in contrast to MC38-HHD-A2.1 or MC38-HHD-A11.1 cells, respectively. Results are presented as means (SD) for each group (n = 5 per group). Statistical analyses were performed using two-way ANOVA with Bonferroni’s correction for multiple comparisons. Created in BioRender. cuiping, A. (2026) https://BioRender.com/nxir12r. D Treatment schematic for the heterogeneous HLA-A2.1/11.1-HHD-514 tumor model. HLA-A2.1 or HLA-A11:01 transgenic mice were grafted with 1 × 10⁶ MC38-HHD-A2.1-514 or MC38-HHD-A11.1-514 tumor cells, followed by intramuscular immunization with either IPM514 or vehicle four days later. The kinetics of tumor growth until the first mouse reached the humane endpoint are illustrated (E) and overall survival rates (F) for HLA-A2.1 mice treated with various doses IPM514 vaccines. Means (SD) are plotted for each group (n = 8 per group). G Kinetics of tumor growth until the vehicle group reached the humane endpoint are shown for HLA-A11.1 mice treated with varying doses of IPM514 vaccine. Means (SD) are plotted for each group (n = 8 per group). Statistical analyses were performed using one-way ANOVA with Bonferroni’s correction for multiple comparisons at the endpoint. H At the endpoint, ELISpot analysis was conducted on splenocytes stimulated with EP or SP to evaluate IFN-γ production. Means (SD) are plotted for each group (n = 8 per group). Statistical analyses were performed using two-way ANOVA with Bonferroni’s correction for multiple comparisons *P < 0.05, **P < 0.01, ***P < 0.001.

A luciferase-based ex vivo assay was utilized to quantify the cytotoxic activity of CD8⁺ T cells derived from IPM514-immunized mice. These T cells exhibited marked lysis of target cells expressing both HLA and IPM514 antigen (MC38-HHD-A2.1-514 and MC38-HHD-A11.1-514), with cytotoxicity increasing in a dose-dependent manner across effector-to-target ratios of 5:1, 15:1, and 45:1 (Fig. 4B, C). Notably, minimal lysis of antigen-negative parental tumor cells confirmed the specificity of the immune responses induced by IPM514. A strong correlation between IFN-γ secretion and cytotoxic activity were observed, further supporting the functional efficacy of IPM514-induced immune responses (r = 0.7375, P = 0.0149; Supplementary Fig. 6E).

To evaluate the in vivo anti-tumor effects, tumor-bearing HLA transgenic mice were treated with IPM514 following the subcutaneous inoculation of MC38 tumor cells expressing HLA and IPM514 antigens (Fig. 4D). Mice received IPM514 at varying doses of 0.5 μg, 2 μg, 5 μg, or 10 μg. At treatment initiation on Day 4 post-inoculation, the mean tumor volume was ~20–30 mm³. In the HLA-A2.1 transgenic model, administration of IPM514 at 2 μg and 10 μg doses resulted in measurable tumor growth inhibition compared to vehicle controls, although tumor volume differences at individual time points did not reach statistical significance due to inter-animal variability(Fig. 4E). Importantly, survival analysis demonstrated significant extension of median survival in the 2 μg (45 days), 5 μg (48 days), and 10 μg (52 days) groups compared to vehicle controls (34 days) (Fig. 4F). At the study endpoint, 25% of subjects in the 2 μg, 5 μg, and 10 μg groups achieved complete tumor regression or minimal residual tumors, confirming therapeutic efficacy despite variable growth kinetics.

In the HLA-A11.1 model, IPM514 demonstrated dose-dependent efficacy, as evidenced by significantly reduced tumor volumes in the 5 μg and 10 μg treatment groups compared to controls (Fig. 4G). ELISPOT analysis revealed dose-dependent increases in IFN-γ secretion in response to SP and EP stimulation, with the 10 μg group showing the most pronounced effect (Fig. 4H). This increase in IFN-γ levels corresponded with enhanced CD8⁺ T-cell activation and effective tumor control.

Long-term antitumor protection and durable T cell memory induced by IPM514

The efficacy of IPM514 in preventing tumor formation, recurrence, and inducing long-term memory T cell responses was evaluated using a prophylactic tumor model in HLA-A2.1 transgenic mice. Mice were immunized intramuscularly with either 1 μg or 4 μg of IPM514 on three times at weekly intervals, followed by a tumor challenge on Day 17 and a rechallenge on Day 122 (Fig. 5A). All control mice developed tumors reached the humane endpoints between Days 25 and 37, while IPM514 treated groups remained tumor-free throughout the study (Fig. 5B, C). Upon tumor rechallenge, the control group exhibited rapid tumor progression, while mice treated with IPM514 maintained 100% protection rate against rechallenges (Fig. 5D).

Fig. 5. IPM514 elicits long-term antitumor protection and establishes durable T cell memory.

A Schematic of the prophylactic vaccination and rechallenge model. HLA-A2.1 transgenic mice were vaccinated weekly with three doses of IPM514 prior to subcutaneous inoculation with MC38-HHD-A2.1-514 tumor cells. After 105 days, tumor-free mice were rechallenged with MC38-HHD-A2.1-514 tumor cells, while s a control group of age-matched mice received vehicle injections. Created in BioRender. cuiping, A. (2026) https://BioRender.com/nxir12r. Survival curves (B) and tumor volume following the 1st (C) or 2nd (D) tumor inoculations are shown. E At designated time points (D17, D76, D122, D140 and D150) following the initial vaccination, the proportion of MAGEA1-specific CD8⁺ T cells in the peripheral blood of the mice was detected using flow cytometry. MAGEA1 is the single antigen eliciting the strongest immune response, characterizing the overall immune response to IPM514. Evaluation of the T cell immune response is shown through intracellular cytokine staining (F) and ELISpot analysis (G). Each symbol represents an individual mouse (n = 8). Error bars are SD. Statistical analyses were performed using two-way ANOVA with Bonferroni’s correction for multiple comparisons *P < 0.05, **P < 0.01, ***P < 0.001. H Data are presented as mean ± SD (n = 8 per group). Statistical significance was determined by one-way ANOVA with Bonferroni’s correction for multiple comparisons. ***P < 0.001 compared to vehicle control.

Analysis of peripheral blood revealed robust antigen-specific CD8⁺ T cell responses following vaccination. The mean frequencies of CD8⁺T cells on Day 17 were 9.62% in the 1 μg group and 10.93% in the 4 μg groups (Fig. 5E). Although these frequencies declined by Day 122, they showed significant rebound upon rechallenge, with fold increases of 3.9 for 1 μg group and 5.32 for 4 μg group by Day 140, correlating with effective tumor control.

Further analysis of splenic immune responses confirmed the enhanced activation of T cells. The 4 μg group exhibited a higher frequency of IFN-γ⁺ CD8⁺ T cells compare to the 1 μg group (Fig. 5F). ELISPOT assays demonstrated a 2.16-fold increase in IFN-γ secretion in the 4 μg group relative to the lower dose (Fig. 5G). Moreover, IPM514 facilitated T-cell differentiation into memory phenotypes, with significant increases in CD8⁺ effector memory T cells (T_EM, CD62L^⁻CD44^⁺) in peripheral blood (1 μg: 32.53%, 4 μg: 34.73%; ***P < 0.001 vs. vehicle: 7.22%) and central memory T cells (T_CM, CD62L^⁺CD44^⁺) in spleens (1 μg: 71.71%, 4 μg: 59.52%; **P < 0.01 vs. vehicle: 42.83%) (Fig. 5H). Statistical analysis was performed using one-way ANOVA with Bonferroni’s correction for multiple comparisons.

Synergistic enhancement of immune response and tumor lymphocyte infiltration in combination with PD-1 inhibitor

Next, we assessed the synergistic effects of IPM514 combined with anti-PD-1 therapy in HLA-A2.1 transgenic mouse model. Mice were immunized intramuscularly with IPM514 on Days 4, 11, and 18 post-tumor inoculation, while anti-mouse PD-1 antibodies or isotype controls were administrated starting on Day 11 (Fig. 6A). At the endpoint, tumor growth inhibition rates (TGI_TV) were greater in all treatment groups compared to the control group, with the combination therapy group exhibiting significantly enhanced (Fig. 6B).

Fig. 6. Synergistic effect of IPM514 with PD-1 inhibitor enhance immune response and tumor lymphocyte infiltration.

A Schematic of the combination therapy involving IPM514 and mPD-1 inhibitor. HLA-A2.1 transgenic mice were grafted with 1 × 10⁶ MC38-HHD-A2.1-514 tumor cells and immunized intramuscularly four days later with three doses of either 2 µg IPM514 or empty LNP. When the average tumor volume reached ~60 mm³ (D11), mice were administered isotype antibodies (Ab) or anti-mPD-1 (clone RMP1-14, BioXCell) at a dose of 1.6 mg/kg via intraperitoneal injection (i.p.) every four days for a total of four doses. Created in BioRender. cuiping, A. (2026) https://BioRender.com/nxir12r. B Kinetics of tumor growth are displayed until the control group reached the humane endpoint. Means (SD) are plotted for each group (n = 6 per group). Statistical analyses were performed using one-way ANOVA with Bonferroni’s correction for multiple comparisons at the endpoint. Evaluation of the T-cell immune response is demonstrated through intracellular cytokine staining (C) and ELISpot assay (D). Data are shown as means (SD) for each group (n = 6 per group). Statistical analyses were performed using two-way ANOVA with Bonferroni’s correction for multiple comparisons. E Analysis of immune cell subsets in tumor microenvironment (TEM) using multiplex immunohistochemistry (IHC). Multiplex IHC showing the distribution of CD3⁺ cells (pink), CD4⁺ cells (gray), CD8⁺ cells (green), PD-1⁺ cells (yellow), and DAPI-stained nuclei (blue). Scale bar = 100 μm. Quantification is provided in Supplementary Fig. 7.

Then we observed enhanced antigen-specific T cell responses in the combination treatment group. Intracellular cytokine staining (ICS) revealed a marked increase in the frequency of IFN-γ⁺ CD8⁺ T cells in mice receiving combination therapy, compared to those treated with monotherapies, indicating heightened T cell activation (Fig. 6C). ELISPOT assays further validated these findings, demonstrating significantly elevated IFN-γ secretion from splenocytes in the combination therapy group, which exhibited the highest levels among all treatment conditions (Fig. 6D). These results underscore that the joint application of IPM514 and PD-1 blockade synergistically amplifies antigen-specific T cell responses.

Analysis of tumor microenvironment revealed increased infiltration of effector immune cells within the combination therapy group. Utilizing multiplex immunohistochemistry (IHC) staining, we observed elevated levels of CD3⁺, CD4⁺, and CD8⁺ T cells, as well as CD3⁺CD8⁺PD-1⁺ T cells, in tumor tissues from mice treated with the combination therapy compared to those receiving monotherapies (Fig. 6E and Supplementary Fig. 7). These findings underscore the effective recruitment of effector immune cells facilitated by the combination therapy.

Single-cell transcriptomic analysis reveals enhanced immune landscape induced by combination therapy

To elucidate the mechanisms underlying the enhanced antitumor efficacy of IPM514 combined with anti-PD-1 therapy, we performed single-cell RNA sequencing (scRNA-seq) on tumor samples collected at the endpoint. Clustering analysis successfully classified the cells into distinct populations, including cancer cells, macrophages, monocytes, and various T cell subsets (Supplementary Fig. 8A). Notably, we observed significant enrichment of CD8⁺ memory T cells, proliferative T cells, and activated T cells were in the combination therapy group compared to the control and single-agent groups (Fig. 7A).

Fig. 7. Comprehensive single-cell transcriptomic analysis reveals enhanced immune landscape induced by combination therapy.

A The distribution preference of each treatment group across all cell clusters, measured by the Ratio of Enrichment (Ro/e). B Dot plot illustrating expression patterns of antigen and HLA genes in cancer cells from each treatment group. C Gene set enrichment analysis (GSEA) using MSigDB Hallmark processes on genes from all tumor cells, ranked by log2 fold change between each treatment group and control. The normalized enrichment score (NES) is provided, with all pathways filtered using a significance threshold of P < 0.05. D The UMAP visualization of lymphocyte cell subsets and GSEA on functional CD8⁺ T cells genes, ranked by log2 fold change between each treatment group and control. NES values are included, and all pathways are filtered with a significance threshold of P < 0.05. E TCR clonal proportion plot comparing different treatment groups, with cells grouped by the frequency of clonotypes. F Visualization of ligand-receptor cross-talk from cancer cells to CD8⁺ T cells. Dot sizes represent p-value significance (only those with p < 0.01 are displayed) and are colored according to communication probability. G UMAP visualization of myeloid cell subsets, showing the preference of each treatment group within these subsets, measured by Ro/e. H Differences in communication strength from M1-like macrophages to CD8⁺ T cell subsets between the combination treatment and control. A red line indicates increased interaction, while a blue line indicates decreased interaction; line thickness represents the magnitude of difference between groups.

Analysis of antigen and HLA expression within tumor cells demonstrated that IPM514 and combination therapy effectively reduced the levels of HLA and tumor antigen expression, as well as the proportion of antigen-presenting tumor cells (Fig. 7B and Supplementary Fig. 8B). This indicates that IPM514 elicits a robust antigen-specific cytotoxic response, which is further enhanced by PD-1 blockade. However, the reduction or loss in antigens may also indicate a potential mechanism of immune escape.

Differential gene expression analysis of tumor cells revealed significant enrichment of multiple immune-related pathways across the treatment groups (Fig. 7C). Notably, pathways associated with allograft rejection and complement activation were markedly enriched, underscoring enhanced immune-mediated tumor destruction.

Further analysis of CD8⁺ T cells within the tumor microenvironment showed significant activation and proliferation across all treatment groups, with combination therapy eliciting the highest levels of interferon response pathway activation (Fig. 7D). TCR sequencing revealed substantial clonal expansion of CD8⁺ T cells following treatment, with newly expanded clones predominantly exhibiting proliferative and cytolytic T cell phenotypes (Fig. 7E and Supplementary Fig. 8C). These findings underscore the critical role of these expanded clones in recognizing tumor antigens and mediating antitumor activity.

Interactions between tumor cells and immune cells were notably strengthened following treatment. Combination therapy markedly increased both the number and intensity of antigen presentation-related interactions between tumor and immune cells, exceeding those observed in single-agent treatments (Fig. 7F and Supplementary Fig. 8D).

Macrophages and monocytes in the combination therapy group showed a pronounced shift toward an anti-tumor phenotype, characterized by increased expression of M1-like markers (CD80 and IL1a) and reduced presence of immunosuppressive M2-like macrophages (Fig. 7G and Supplementary Fig. 8E). Moreover, interactions between M1 macrophages and T cells were significantly enhanced, with the combination therapy showing the most pronounced effects thereby supporting a reprogramming of the tumor microenvironment to favor antitumor immunity (Fig. 7H and Supplementary Fig. 8F).

Discussion

The advent of mRNA-based vaccines has revolutionized vaccine development, establishing this innovative platform as a promising tool for combating not only infectious diseases but also a broad range of malignancies^24,25. In the past decade, an increasing number of preclinical and early clinical investigations have provided compelling evidence that mRNA-based cancer vaccines can induce robust and antigen-specific T cell responses, particularly when combined with other immunotherapeutic strategies^26–28. While notable successes have been achieved with mRNA vaccines in melanoma and various other tumor types^17–19, the clinical application of mRNA vaccines in ESCC remains largely unexplored. Here, we report a universal multi-epitope mRNA-LNP vaccine specifically designed to address critical challenges associated with ESCC treatment, and we provide a comprehensive preclinical evaluation of its immunogenicity, antitumor efficacy, and translational potential.

By leveraging a multi-step screening strategy that integrated large-scale transcriptomic databases (TCGA, GTEx, HPA, and GEPIA), we identified multiple TAA candidates that display broad and preferential overexpression in ESCC tumor tissues compared to healthy esophageal mucosa. This rigorous selection process ensured that the final vaccine construct targeted antigens with robust tumor specificity and minimal expression in normal tissue, thereby potentially reducing on-target, off-tumor toxicity. The selected epitopes were designed to accommodate two widely prevalent HLA subtypes, HLA-A*02:01 and HLA-A*11:01, thus maximizing the potential patient population for the IPM514 vaccine. Remarkably, these antigens also exhibit overexpression in several other malignancies, particularly LUSC, suggesting the broad applicability of IPM514 across other tumor types.

Consistent with previous studies showing frequent co-expression of MAGE family members and other TAAs (e.g., ACTL8, SMC1B) in ESCC and other squamous cell carcinomas (SCCs)^29,30 the 9 TAAs selected in IPM514 exhibited high co-expression in 132 ESCC patient samples, with over 90% of patients expressing at least three of these antigens, which lays a foundation for the broad anti-tumor effect of IPM514. While N1-methylpseudouridine (m1Ψ) has been extensively utilized in mRNA vaccine to enhance mRNA stability and mitigate innate immune activation, recent studies indicate that natural nucleosides may elicit more robust T-cell responses³¹. Nonetheless, m1Ψ modifications can mitigate +1 ribosomal frameshifting, thereby preserving translational fidelity³². To address these competing considerations, we optimized the mRNA sequence of IPM514 by incorporating natural nucleosides to enhance translational efficiency while also extending the mRNA’s half-life, thereby striving for an ideal balance between stability and immune responsiveness.

A persistent challenge in evaluating multi-epitope tandem vaccines is the lack of specific antibodies capable of detecting the full-length construct in vitro. To address this limitation, we employed a rabbit reticulocyte lysate–based cell-free protein synthesis (CFPS) system for comparative assessment of IPM514 expression. Although interspecies variations might restrict the direct applicability of these findings to human cell contexts, the CFPS platform nonetheless provided a valuable framework for guiding mRNA construct optimization, highlighting those with optimal expression levels and translational properties. Importantly, the in vitro findings closely mirrored subsequent in vivo immunogenicity assays, resulting in the identification of IPM514 version 1.2 as the most effective construct. Moreover, extensive immunogenicity analyses demonstrated that nearly every antigenic fragment encoded by IPM514 elicited measurable T-cell responses in vivo, underscoring the broad and potent immune activation induced by IPM514. This comprehensive immune reactivity reinforces translational promise of IPM514 for both preclinical and clinical applications.

In vitro stimulation of PBMCs remains a widely adopted approach in vaccine performance evaluation, enabling researchers to assess immune responses and identify optimal antigen formulations prior to clinical testing^33–35. In our study, IPM514 induced marked antigen-specific T-cell expansion in PBMCs derived from both healthy donors and ESCC patients. When DCs were electroporated with IPM514, they elicited significantly higher frequencies of IFN-γ⁺ TNF-α⁺ CD8⁺ T cells, underlining the potency of this vaccine in activating cellular immunity. Notably, the mRNA-transfected DC group generated a more pronounced amplification of antigen-specific T cells compared to the LNP delivery group in our in vitro system. It is possible that electroporation exhibited higher transfection efficiency than LNPs under in vitro conditions. However, LNPs play a crucial role in the in vivo setting by ensuring mRNA stability, enabling targeted delivery to antigen-presenting cells, and providing intrinsic adjuvanticity via activating innate immune signaling. Thus, the LNP formulation is essential for the in vivo efficacy and clinical translation of IPM514.

Furthermore, cytotoxicity assays demonstrated that T cells expanded by IPM514 were highly effective in lysing T2 cells loaded with IPM514-derived peptides and HLA-matched ESCC and other SCC lines. The observed differences in cytotoxic capacity corresponded closely with the levels of IPM514-specific T cell expansion and the tumor antigen expression profiles of target cells, further underscoring the specificity of immune responses elicited by IPM514.

HLA transgenic mice provide a robust platform for in vivo evaluation of immunotherapies^36,37. In both HLA-A2.1 and HLA-A11.1 transgenic mouse models, IPM514 exhibited dose-dependent tumor growth inhibition and significantly prolonged survival. Notably, prophylactic administration of IPM514 completely abrogated tumor engraftment and provided durable protection against tumor rechallenge. Moreover, IPM514 not only elicited robust antigen-specific CD8⁺ T-cell responses but also promoted the formation of memory T-cell indicating a sustained immunological defense against tumor recurrence.

The formation of cold tumors is driven by an immunosuppressive tumor microenvironment (TME), which hinders immune cell infiltration and activity, rendering immunotherapies ineffective. Reprogramming the TME is a critical strategy to enhance cancer treatment outcomes³⁸. IPM514 overcomes these barriers by activating tumor-specific T cells, which exhibit enhanced efficacy in T cells infiltrating and eradicating tumors once PD-1-mediated inhibition is alleviated. Additionally, single-cell transcriptomic analyses revealed profound reprogramming of the TME, characterized by enhanced antigen presentation, increased infiltration of memory and effector T cells, and a polarization of myeloid cells toward anti-tumor phenotypes, all of which are critical drivers of therapeutic success^38,39. While this reduction in antigen and HLA expression confirms that IPM514 elicits robust antigen-specific cytotoxic responses, particularly when combined with PD-1 blockade, it also reveals potent immune selective pressure that may ultimately drive immune escape through antigen loss or HLA downregulation. Therefore, follow-up strategies to overcome such immune escape will be necessary.

These findings suggest that combining checkpoint blockade with a multi-epitope mRNA vaccine can significantly enhance therapeutic efficacy. The vaccine primes a broader and more robust T-cell repertoire, helping to overcome the limitations commonly associated with monotherapies.

We acknowledge that treatment was initiated at an early tumor stage (Day 4), necessitated by the rapid growth of murine tumors, which outpace human malignancies and leave a narrow window for T-cell priming. This suggests that the optimal clinical application of IPM514 may be in the adjuvant setting, where it targets minimal residual disease rather than bulky tumors. Beyond these model-specific considerations, further investigations remain necessary. Although our preclinical models were meticulously designed to mimic human HLA-restricted antigen presentation and immune responses, the inherent complexity, genetic heterogeneity, and diverse immune microenvironments of human tumors surpass the capacity of mouse models and in vitro assays to replicate these dynamics fully⁴⁰. The next critical step involves conducting clinical trials, which will be essential for validating the therapeutic efficacy of IPM514, identifying predictive biomarkers of response, and optimizing immunization protocols and treatment sequencing to maximize clinical benefit.

In summary, IPM514 represents a next-generation, universal mRNA cancer vaccine strategically designed to address critical challenges in the immunotherapy of ESCC. Through multi-epitope targeting and optimized antigen presentation, IPM514 elicits robust and durable antigen-specific T cell immunity. Notably, combining IPM514 with PD-1 blockade demonstrates potential to reprogram the TME, thereby enhancing anti-tumor immunity and overcoming the therapeutic resistance commonly observed in ESCC. Our comprehensive preclinical findings provide a strong rationale for the development of more effective and broadly accessible immunotherapies. By extending the clinical applicability of mRNA vaccine technology, this strategy has the potential to improve outcomes in ESCC and other malignancies.

Methods

Ethics statement

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Yicon (Beijing) Medical Technology Inc. (approval no. PZ-AUP-S-20240408-02) and performed in accordance with institutional and state guidelines. HLA-A2.1 transgenic mice (C57BL/6N-B2m^{tm2(B2M/HLA-A2.1/H2-D)Bcgen}/Bcgen) and HLA-A11.1 transgenic mice (C57BL/6JNifdc-B2m^{tm1(B2M/HLA-A11.1/H2-D)Bcgen}/Bcgen) were purchased from Biocytogen Pharmaceuticals (Beijing) Co., Ltd., maintained in a specific-pathogen-free facility at Yicon (Beijing) Medical Technology Inc., and used at 6-12 weeks of age. For human studies, tumor, adjacent non-tumorous tissues, and peripheral blood samples from patients with esophageal squamous cell carcinoma (ESCC) were collected at Beijing Cancer Hospital after informed consent had been obtained. The human study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Peking University Cancer Hospital (approval no. 2023KT101). Written informed consent was obtained from all participants prior to sample collection.

Tumor-associated antigen identification and epitope selection

To identify ESCC-associated tumor antigens, RNA-seq data from 80 ESCC tumors in TCGA and 555 normal esophageal mucosa samples from GTEx were analyzed using DESeq2. Differentially expressed genes were defined by adjusted P < 0.01 and log2 fold change > 2, and were further restricted to genes expressed in at least 20% of ESCC samples with TPM > 1. Candidate antigens were cross-referenced with curated tumor-associated antigen databases and filtered against GTEx and HPA normal tissue expression data to prioritize targets with minimal normal tissue expression (TPM < 0.25 across 50 normal tissues). T-cell epitopes within selected antigens were predicted using NetMHCpan, NetMHC, NetMHCcons, PickPocket, SMM, SMMPMBEC, MHCflurry, and MHCnuggets for HLA-A*02:01 and HLA-A*11:01. Candidate epitopes were prioritized by binding affinity (IC50 < 500 nM), percentile rank, immunogenicity metrics, and support from IEDB. To estimate patient coverage, antigen positivity was defined using transcript-per-million thresholds derived from control tissues, and expression frequencies were evaluated in the ESCC cohort and across additional TCGA/GTEx cancer datasets.

Whole-exome sequencing and HLA genotyping

Tumor and matched adjacent normal tissues from 132 patients with ESCC were snap-frozen and stored at −80 °C. Genomic DNA from normal tissue was extracted using the TIANamp Genomic DNA Kit. Whole-exome libraries were prepared with the Agilent SureSelect Human All Exon V6 kit and sequenced on an Illumina NovaSeq 6000 platform to an average depth of 200× for tumors and 100× for matched normal tissues. Germline-derived WES data from normal tissues were used for HLA genotyping. Reads were aligned to GRCh38 using BWA-MEM, HLA-related reads were extracted, and HLA-A, HLA-B, and HLA-C alleles were inferred with OptiType v1.3.1, with concordance confirmed across repeated runs with different random seeds.

RNA constructs, in vitro transcription, and LNP formulation

IPM514 was designed from a codon-optimized sequence as described previously²³. Seven candidate IPM514 mRNA (IPM514-1.0 to IPM514-1.6 sequences were deposited in GenBank under accession numbers PZ155111–PZ155117) constructs were cloned into optimized mRNA production plasmids, linearized, and transcribed in vitro with T7 RNA polymerase (Yeasen) without modified nucleosides. mRNA integrity was assessed by agarose gel electrophoresis and samples were stored at −80 °C. For lipid nanoparticle (LNP) preparation, ALC-0315, ALC-0159, DSPC, and cholesterol (AVT) were mixed in ethanol at a molar ratio of 46.3:1.6:9.4:42.7, and mRNA was dissolved in 25 mM sodium citrate (pH 4) at 0.2 mg/mL. The aqueous and ethanol phases were mixed at a 3:1 ratio in a NanoAssemblr Ignite microfluidic system, diluted in Tris buffer (pH 7.4), and concentrated by 100-kDa ultrafiltration to 0.5 mg/mL. Particle size, polydispersity index, and zeta potential were measured on a Zetasizer Ultra.

qPCR, cell-free translation, and immunopeptidomics

To compare the seven candidate mRNA constructs, K562 cells were electroporated with 75 μg mRNA (1 ×10⁷ cells per reaction) and harvested at 0, 3, 6, 12, 24, and 48 h. Total RNA was extracted with the FastPure Cell Tissue Total RNA Isolation Kit (Vazyme), reverse-transcribed with PrimeScript RT Reagent Kit with gDNA Eraser (Takara), and analyzed by qRT-PCR with TB Green Premix Ex Taq on an Applied Biosystems 7500 system. Primers targeting the 3’UTR region were Forward, CACCACCTCTGCTAGTTCCAG, and Reverse, TGTGGCTAGGCTAAGCGTTTT. Relative expression was calculated by the delta-delta Ct method; Ct values > 35 were excluded. In vitro translation was assessed using nuclease-treated rabbit reticulocyte lysate (Promega), followed by SDS-PAGE and ImageJ-based densitometry. To validate HLA presentation, mono-HLA-expressing K562 cells were electroporated with 300 μg IPM514 mRNA (4 ×10⁷ cells per reaction), incubated for 24 h, lysed, and HLA complexes were immunoprecipitated with anti-FLAG M2 affinity gel. Eluted peptides were analyzed by LC-MS/MS on a timsTOF Pro2 instrument (Bruker) in data-dependent PASEF mode.

Cell lines

K562 cells were used to generate mono-HLA-expressing lines carrying Flag-tagged HLA-A*02:01, HLA-A*11:01, HLA-A*24:02, HLA-B*40:01, HLA-B*46:01, HLA-C*01:02, or HLA-C*07:02. The ESCC cell line KYSE-410 (HLA-A*24:02) was engineered to express HLA-A*02:01 or HLA-A*11:01. Additional tumor cell lines included T2, SCC-4, SCC-25, HCC95, and NCI-H520. MC38 cells were engineered to express HLA-A*02:01/Db or HLA-A*11:01/Db chimeric molecules, and derivative lines expressing IPM514 fragments were established for in vivo studies. Luciferase-expressing derivatives were generated where indicated for killing assays. Cell HLA types are listed in Supplementary Table 5. Cells were maintained in RPMI 1640 supplemented with 10% FBS, GlutaMax, and penicillin-streptomycin.

In vitro stimulation of PBMCs

PBMCs from healthy donors (Milestone Bio) and patients with ESCC (Beijing Cancer Hospital) were thawed and cultured in X-VIVO 15 medium. For healthy donor experiments, CD14+ monocytes were isolated with EasySep Human CD14 Positive Selection Kit II and differentiated into dendritic cells with IL-4 (20 ng/mL) and GM-CSF (100 ng/mL). Mature dendritic cells were loaded with IPM514 by electroporation (10 μg mRNA; 1300 V, 10 ms, 3 pulses) or mRNA-LNP treatment, and then cocultured with autologous T cells isolated using EasySep Human T Cell Isolation Kit at a 1:4 dendritic cell:T-cell ratio in the presence of IL-7 (5 ng/mL), IL-15 (5 ng/mL), and IL-2 (10 ng/mL). Antigen-specific stimulation was performed with IPM514 peptide pools (1 μg/mL per peptide) plus poly(I:C) (20 μg/mL). For ESCC patient PBMCs, thawed PBMCs were cultured directly with IL-4 and GM-CSF, stimulated on day 4 with peptide pools plus poly(I:C), and expanded from day 6 onward in medium containing human AB serum, IL-7, IL-15, and IL-2. Negative controls were cultured identically without peptide.

Cytotoxicity assays

For T2-based assays, target cells were incubated with luciferase mRNA-LNP and peptide pools, then plated in white 96-well plates. Effector cells were added at effector:target ratios of 20:1 or 100:1 and luminescence was measured after 24 h. Similar luciferase-based killing assays were performed with KYSE-410, KYSE-410-A2.1, KYSE-410-A11.1, SCC-4, SCC-25, NCI-H520, and HCC95 target cells. Specific lysis was calculated as 100× [(experimental killing − spontaneous luminescence)/(maximal killing − spontaneous luminescence)]. For ex vivo killing assays, splenic CD8+ T cells isolated from immunized mice using EasySep Mouse CD8+ T Cell Isolation Kit were cocultured with MC38-HHD-A2.1, MC38-HHD-A2.1-514, MC38-HHD-A11.1, or MC38-HHD-A11.1-514 luciferase-labeled target cells at effector:target ratios of 5:1, 15:1, and 45:1 for 24 h before luminescence measurement.

Mouse immunization and tumor studies

For immunogenicity studies, HLA-A2.1 or HLA-A11.1 transgenic mice received intramuscular injections of IPM514 mRNA-LNP or PBS once weekly for three doses. In construct-screening experiments, 4 μg was administered per dose (n = 8 per group). For ex vivo cytotoxicity assays, mice received three weekly immunizations of 4 μg IPM514 mRNA-LNP or PBS (n = 5). In therapeutic studies, mice were inoculated subcutaneously with 1 ×10⁶ MC38-A2.1-HHD-514 or MC38-A11.1-HHD-514 cells. Four days post-inoculation, tumors were palpable but not measurable by caliper (estimated diameter <2 mm), mice were randomized by body weight to receive vehicle or IPM514 (0.5, 2, 5, or 10 μg) three times weekly (n = 8). In prophylactic and rechallenge studies, HLA-A2.1 mice were vaccinated weekly for three weeks with 1 or 4 μg IPM514 mRNA-LNP or vehicle before tumor challenge; tumor-free mice were rechallenged on day 105. In combination studies, HLA-A2.1 mice bearing MC38-HHD-A2.1-514 tumors received intramuscular IPM514 (2 μg) or empty LNP on days 4, 11, and 18, together with intraperitoneal anti-mPD-1 (clone RMP1-14, BioXCell) or isotype control (1.6 mg/kg every four days for four doses) beginning on day 11. Tumor size was measured every 2–3 days using calipers and volume was calculated as 0.5 × length × width². Mice were anesthetized with isoflurane (5% induction, 2% maintenance) for invasive procedures and euthanized by controlled CO₂ exposure followed by cervical dislocation when tumors exceeded 2000 mm³, health deteriorated, or the monitoring period ended.

ELISpot, flow cytometry, and immunohistochemistry

ELISpot assays were performed with freshly isolated splenocytes using a mouse IFN-gamma pre-coated kit (Dakewe). Cells (2 ×10⁵ per well) were stimulated with peptide pools (2 μg/mL per peptide) for 16–24 h; PMA/ionomycin and medium served as positive and negative controls. Responses were considered positive when counts exceeded 25 spots per 2 ×10⁵ cells after background subtraction and were at least twofold over the control group. For flow cytometry, human PBMCs after in vitro stimulation and mouse splenocytes were stained with antibodies against CD3, CD4, CD8a, CD44, CD45, and CD62L, with viability dye and, where indicated, HLA-A*02:01 MAGE-A1 tetramer. Intracellular cytokine staining for IFN-gamma and TNF-alpha was performed after peptide restimulation in the presence of transport inhibitor for at least 6 h. Data were acquired on a DxFLEX flow cytometer and analyzed with CytExpert. For immunohistochemistry, formalin-fixed paraffin-embedded tumor sections (3 μm) were stained with antibodies against CD3, CD4, CD8a, and PD-1, imaged on the TissueFAXS SL system, and quantified with StrataQuest.

Single-cell RNA-seq and TCR-seq

Single-cell transcriptome and paired TCR libraries were prepared from tumor samples using the SeekOne DD Single Cell Immune Profiling Kit and sequenced on an Illumina HiSeq X-Ten. Raw data were processed with SeekSoul Tools v1.2.2, aligning to the mm10 reference and human HLA sequences, assigning barcodes, and generating UMI matrices. Downstream analysis was performed in Seurat v5.0.0, retaining cells with 400–8000 genes, 500–30,000 UMIs, and <10% mitochondrial UMIs. Ten Seurat objects were merged, normalized by LogNormalize, and variable genes identified. Batch effects were corrected with Harmony, and clustering and UMAP visualization used the top 35 PCA components. Cell types were annotated using canonical markers, and immune compartments were reclustered for high-resolution analysis. TCR reads were processed with SeekSoul Tools, assigning high-confidence productive chains to each cell. Clonotypes were defined by identical paired TCRs, and clonotype analysis was done with scRepertoire.

For preferential cell accumulation, the ratio of observed to expected cell numbers (Ro/e) was calculated using chi-square testing. Differentially expressed genes were identified using Seurat’s FindMarkers, with gene lists analyzed by GSEA in clusterProfiler. Cell-cell communication between cancer cells and CD8+ T cells or myeloid cells was analyzed with CellChat. Thymic epithelial cell expression was analyzed from the human thymus atlas data, considering genes with log-normalized expression > 0.

Statistical analysis

Statistical analyses were performed using the GraphPad Prism software v.8.0. Multiple groups were compared using one- or two-way analysis of variance (ANOVA) with Bonferroni’s correction for multiple comparisons. survival times of animals between groups using the Kaplan–Meier method. Appropriate methods are indicated in the legends, with significant differences marked in all figures. Values and error bars are expressed as means (SD). Degrees of significance were defined as P < 0.05 (*P < 0.05, **P < 0.01, and ***P < 0.001). Single-cell RNA-Seq data statistical analyses were conducted using R software (version 4.2.0), with significant findings validated across multiple computational approaches to ensure reproducibility and robustness of the results.

Author contributions

H.L. and L.Q. conceived and designed the study. H.L., L.Q., T.D., C.L., Y.Z., Z.W., Z.W., and Q.Z. developed the methodology. H.L., T.D., C.L., Y.Z., Z.W., Z.W., P.Y., Q.Z., X.Z., J.Y., X.J., A.Y., and Y.H. conducted the investigation. H.L., Y.Z., A.Y., and X.J. contributed to visualization. H.L. wrote the original draft of the manuscript. L.Q., Z.L., and H.Z. reviewed and edited the manuscript. H.L. and C.A. managed project administration. H.L., L.Q., N.S., Z.L., and H.Z. supervised the study. N.S., Z.L., and H.Z. acquired funding. All authors read and approved the final manuscript.

Data availability

The mRNA sequences of IPM514-1.0 to IPM514-1.6 have been deposited in GenBank under accession numbers PZ155111-PZ155117. The datasets generated and analyzed during the current study are not publicly available due to patient privacy considerations and institutional restrictions on human genomic data sharing, but are available from the corresponding author on reasonable request.

Competing interests

N.S., H.Z., H.L., C.L., Y.Z., C.W., Z.W. and L.Q. are inventors on a pending patent application entitled “A cancer vaccine and its application” (PCT/CN2024/090619). The patent applicant is Beijing Immupeutics Medicine Technology Limited. The application is currently pending. The patent relates to the design and application of the multi-epitope cancer vaccine described in this manuscript. The other authors declare no competing financial or non-financial interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41541-026-01503-y.