Fast Co-Delivery of mRNA Neoantigens Targeting Both Immune Helper and Killer Cells Boosts Cancer-Fighting Responses in Colorectal Cancer

In conventional PCV neoantigen design, epitopes are typically crafted by targeting somatic mutations uniquely present in cancer tissues. This study employed an orthotopic model using the MC38 cell line to leverage the distinct genomic differences for inducing tumor‐specific immune responses (Figure1a). Specifically, germline mutations inherent to the mouse strain and cell line used for neoantigen design were targeted. Table S1, Supporting Information details the number of variants identified by the two variant callers, along with those obtained after filtering. Variants exclusive to the tumor were identified by employing two variant callers, resulting in 546 single nucleotide variants (SNVs) and 68 insertions/deletions (InDels) which were selected as consensus candidates for neoantigen design.

ARRIBA and STAR fusion were used to analyze gene fusions in two mRNA‐seq datasets from normal and tumor samples, respectively. This comparison revealed tumor‐specific gene fusions. ARRIBA identified 18 tumor‐specific gene fusions, whereas STAR fusion identified 12, with 8 fusions common to both tools. Further filtering using de novo assembly with Trinity and read support at fusion junctions excluded two gene fusions (Eefsec_Mcm2, Smurf1_Kpna7) due to insufficient junction reads. Consequently, six gene fusions were shortlisted as candidates for neoantigen design.

Following the workflow in Figure 1b, neoantigens derived from the candidate variants and gene fusions were prioritized. The prioritization process considered MHC binding rankings, the expression levels of the respective mutations, and their uniqueness compared with self‐antigens. Therefore, 20 MHC‐I and 12 MHC‐II‐restricted neoantigens were designed and selected, respectively. Tables S2 and S3 (Supporting Information) present a comprehensive list of designed epitopes, along with their target sources and associated genes.

These neoantigen candidates were synthesized as synthetic long peptides (SLPs) to evaluate the immunogenicity of candidate neoantigens from a pool of 32 candidates with high prediction scores for MHC‐I and MHC‐II binding. The 20 neoantigens associated with MHC‐I were synthesized as 8–10 amino acid (aa) peptides, whereas the 12 neoantigens associated with MHC‐II were synthesized as 27 aa peptides. These peptides were pooled, mixed with poly I:C, and administered subcutaneously to C57BL/6 mice once weekly. Splenocytes were harvested and subjected to interferon‐gamma (IFN‐γ) ELISpot analysis 1 week after the last immunization (Figure S1, Supporting Information). Two antigens (CNM‐I‐13 and CNM‐I‐19) in MHC‐I‐associated neoantigens exhibited robust antigen‐specific IFN‐γ secretion (Figure2a). Two MHC‐II‐associated neoantigens (CNM‐II‐01 and CNM‐II‐10) exhibited significant antigen‐specific IFN‐γ secretion (Figure 2a).

Subsequently, mRNA encoding the corresponding neoantigens were constructed and characterized (Figure S2a,b, Supporting Information). Each mRNA was encapsulated in lipid nanoparticles (LNPs) and the mRNA‐LNPs were analyzed by particle size, polydispersity index (PDI), and encapsulation efficiency (Figure S2c–e, Supporting Information). The characterized mRNA vaccines were subcutaneously injected into C57BL/6 mice on days 0, 4, 7, and 14. mRNA vaccines encoding MHC‐II‐restricted neoantigens (CNM‐II‐01 and CNM‐II‐04) significantly increased IFN‐γ secretion; however, IFN‐γ induction did not occur during treatment with mRNA vaccines encoding MHC‐I‐associated neoantigens (Figure 2b). Given that neoantigens length can influence immunogenicity,^[15^] long mRNA (27 aa length) vaccines encoding MHC‐I‐associated neoantigens were constructed (Table S4, Supporting Information). The IFN‐γ ELISpot analysis demonstrated that the long mRNA‐encoded neoantigens were more immunogenic than the short mRNA‐encoded neoantigens (Figure 2c). Overall, 10% of the MHC‐I antigens predicted using the in silico method were immunogenic, whereas 25% of the MHC‐II antigens were immunogenic in our study (Figure 2d).

Flow cytometric analysis was performed to confirm that the predicted pathways for MHC‐I and MHC‐II antigens induced cytokine secretion. MHC‐I neoantigens induced CD8⁺ cell‐specific IFN‐γ and tumor necrosis factor‐alpha (TNF‐α) secretion. In contrast, MHC‐II neoantigens prompted CD4⁺ cell‐specific IFN‐γ and TNF‐α secretion (Figure 2e), suggesting that these identified neoantigens effectively activate the respective T cell pathways through MHC‐I and MHC‐II molecules, thereby validating the reliability of these predictions.

Generally, the MHC‐I pathway is activated by peptides that are typically 8 to 10 amino acids in length, while the MHC‐II pathway is known to be activated by longer peptides, ranging from 11 to 14 amino acids in length. To determine the minimal epitopes, we designed 10‐mer peptides (CNM‐I‐13‐01 to CNM‐I‐13‐10; 10 peptides) within MHC‐I antigen (CNM‐I‐13) and 14‐mer peptides (CNM‐II‐04‐01 to CNM‐II‐04‐13; 13 peptides) within MHC‐II antigen (CNM‐II‐04), including each mutation (Figure3a; Table S5, Supporting Information). C57BL/6 mice were immunized with mRNA‐based MHC‐I neoantigens (CNM‐I‐13), MHC‐II neoantigens (CNM‐II‐04), or co‐administration of both at 0, 4, 7, and 14 days. One week after the last immunization, the splenocytes were harvested from immunized mice, stimulated with each peptide, and evaluated for immunogenicity using IFN‐γ ELISpot.

The synthesis of CNM‐I‐13‐01 and CNM‐I‐13‐02 peptides failed, which appears to reflect a limitation of peptide‐based vaccines due to variations in physicochemical properties depending on amino acid sequences.

Some minimal epitopes of MHC‐I and MHC‐II antigens were found to be non‐immunogenic or weakly immunogenic, even if they contained tumor‐specific mutations. Subsequently, CNM‐I‐13‐05, ‐06, ‐07, and ‐08 in MHC‐I restricted antigen were highly immunogenic (Figure 3b) and CNM‐II‐04‐09, ‐10, ‐11, ‐12, and ‐13 in MHC‐II restricted antigens induced antigen‐specific secretion of IFN‐γ in MHC‐II antigen strongly (Figure 3c), suggesting that not all mutant antigens are presented; only certain of peptide complexes with MHC and TCR to function as antigens.

To next confirm antigen‐specific immune responses elicited by individual antigens and their combination, T cell mediated cytokine levels were evaluated using splenocytes isolated from mice vaccinated with MHC‐I/II antigen, or a combination of both. In the IFN‐γ ELISpot assay, antigen‐specific IFN‐γ secretion by CD8⁺ T cells was significantly higher in the MHC‐I/II co‐administration group compared to the MHC‐I‐only vaccination group (Figure 3d), whereas there were no significant differences in IFN‐γ secretion by CD4⁺ T cells between the MHC‐II vaccination group and the MHC‐I/II co‐administration group (Figure 3e). These patterns were also observed in flow cytometric analysis, which showed that co‐administration with MHC‐I and MHC‐II vaccines significantly increased the secretion of IFN‐γ or TNF‐α in CD8⁺ T cells, but not in CD4⁺ T cells (Figure 3f,g; Figure S3, Supporting Information). These results demonstrated that co‐administration of MHC‐I and MHC‐II not only triggers respective antigen‐specific immune responses but also that CD4⁺ T cell activation stimulated by MHC‐II antigens contributes to the activation of antigen‐specific CD8⁺ T cells. Additionally, spleen cells were collected and stained with PE‐H‐2D^b tetramer conjugated with CNM‐13‐06 (10‐mer) to detect MHC‐I antigen‐specific T cells using flow cytometry. It showed that there was inter‐individual variation; however, some mRNA‐vaccinated individuals exhibited tetramer binding. (Figure S4a−c, Supporting Information).

MC38 tumor‐bearing mice were administered the MHC‐I vaccine, MHC‐II vaccine, or both vaccines to investigate the anti‐cancer efficacy of the mRNA vaccines (Figure4a). Compared with the phosphate‐buffered saline (PBS) administration group, the mRNA vaccine administration group exhibited a significant decrease in the growth and volume of colon tumors (Figure 4b–d; Figure S5a and S5b, Supporting Information). The group co‐administered the MHC‐I and MHC‐II vaccines exhibited markedly improved the number of tumor‐free mice (3/6) compared to the group administered the MHC‐I (1/6) or MHC‐II vaccine (0/6) alone. Furthermore, spleen cells from vaccinated mice were isolated to analyze the T cell population (Figure S6, Supporting Information). Flow cytometry data showed that CD4⁺/CD8⁺ T cells were significantly high in the MHC‐II vaccinated group (Figure 4e). Additionally, co‐administration of the MHC‐I and MHC‐II vaccines induced more proliferation of CD8⁺ short‐lived effector T cells (SLECs, CD127^‐ KLRG1 ⁺ of CD8⁺ cells) than only the MHC‐I vaccine (Figure 4f), suggesting that MHC‐II antigen‐specific CD4⁺ T cells help proliferation and differentiation of short‐lived effector CD8⁺ T cells. The proliferation of regulatory T cells (T_reg), which typically indicate the inhibition of immune responses, was significantly reduced in all vaccinated groups (Figure 4g). Antigen‐specific IFN‐γ and TNF‐α secretion from CD8⁺ T cells was significantly increased in the MHC‐I vaccinated group (Figure 4h, upper). On the other hand, immunization with the MHC‐II vaccine increased IFN‐γ and TNF‐α levels in CD4⁺ T cells (Figure 4h, lower). Interestingly, antigen‐specific CD8⁺ T cells secreted significantly higher levels of IFN‐γ and TNF‐α in the group receiving both MHC‐I and MHC‐II vaccines than MHC‐I‐only vaccination group in the tumor‐bearing mouse model, whereas there were no significant differences in the IFN‐γ and TNF‐α secretion by CD4⁺ T cells between the MHC‐II vaccination group and the MHC‐I/II co‐administration group. These results showed similar patterns to the results from healthy mice (Figure 3d–g), suggesting that co‐administration of these vaccines can induce an effective immune response and enhance anti‐cancer efficacy. To further validate neoantigen‐specific immune responses induced by mRNA‐based neoantigens in the tumor‐bearing mouse model, IFN‐γ ELISpot analysis was performed.

The results showed that the MHC‐I vaccine elicited MHC‐I antigen‐specific immune responses and the MHC‐II vaccine elicited MHC‐II antigen‐specific immune responses (Figure S7a−c, Supporting Information). Furthermore, the group that received co‐administration of the MHC‐I and MHC‐II vaccines showed significantly increased MHC‐I antigen‐specific immune responses than the group receiving only the MHC‐I vaccine, but not MHC‐II‐related immune responses. These results suggest the MHC‐II antigens not only stimulate CD4⁺ T cells to induce a type 1 helper T cell response but also activate CD8⁺ T cells to function as helper T cells.

To test the effects of the mRNA vaccine on preventing tumor recurrence, tumor‐bearing mice were treated using tumor resection surgery and injected with PBS or co‐administered the MHC‐I and MHC‐II mRNA vaccines (Figure5a). Compared to the PBS‐treated group, the mRNA vaccine group exhibited a significant reduction in tumor recurrence (Figure 5b). The antigen‐specific IFN‐γ was secreted by spleen cells isolated from vaccinated mice (Figure S8, Supporting Information). These results show that mRNA vaccines encoding neoantigens can reduce the recurrence burden of colon cancer by antigen‐specific T cell responses. Additionally, the anti‐cancer efficacy of prophylactic vaccination was tested by administering PBS or MHC‐I/MHC‐II vaccines once per week for 3 weeks before tumor inoculation (Figure S9a, Supporting Information).

Groups receiving the mRNA‐based neoantigen vaccine exhibited significantly reduced tumor growth and a higher number of tumor‐free mice compared to the control group (Figure, Supporting Information). Additionally, IFN‐γ ELISpot analysis confirmed the induction of neoantigen‐specific IFN‐γ secretion, indicating a robust immune response following prophylactic vaccination (Figure, Supporting Information). S9b S9c

The efficacy of anti‐cancer vaccines varies based on the stage of cancer development.^[16^] C57BL/6 mice were inoculated with MC38 cells ranging from 1 × 10⁵ cells to 3 × 10⁵ cells to investigate the efficacy of the mRNA vaccines at different stages of cancer development (Figure 5c–f). Tumor volume and weight were significantly inhibited in the vaccine‐treated group compared with the PBS‐treated group, and the anti‐tumor effect was greater in the group transplanted with a low low‐dose cancer cells (Figure 5c–h; Figure S10a, Supporting Information). In particular, two of six mice in the mRNA vaccine group were tumor‐free (Figure 5e,h), indicating that anti‐cancer vaccine administration was more effective in earlier cancer stages.

A cytotoxicity assay was performed to evaluate the cytotoxic function of lymphocytes. TILs isolated from mice administered with PBS or the mRNA vaccine were mixed with firefly luciferase‐transduced MC38 colon cancer cells (MC38‐luc). The viability of MC38‐luc cancer cells was significantly decreased in the mRNA vaccine‐treated group (Figure 5i). These results show that the mRNA vaccine can activate the cytotoxic function of lymphocytes. Furthermore, Splenocytes and TILs from all mice administered with the mRNA vaccine exhibited antigen‐specific T cell responses (Figure S10b,c, Supporting Information). As tumor‐infiltrating CD8⁺ T cells are a crucial factor in cancer immunotherapy, TILs were isolated from vaccinated mice to analyze the tumor‐infiltrating T cell population. The results of the flow cytometry analysis showed significantly more CD8⁺ T cells in the mRNA vaccination group than in the PBS group (Figure S11a, Supporting Information). Additionally, the mRNA vaccine significantly reduced the proliferation of suppressor T cells (CD25^highCD4⁺) within tumor tissues (Figure S11b, Supporting Information).

Tumor tissues were subjected to hematoxylin and eosin (H&E) staining, and immunohistochemical analysis was performed to assess histological differences in tumors between the PBS‐treated and mRNA‐treated groups. Consequently, the necrotic areas in the H&E staining were equally immunostained using a mouse anti‐CD8 antibody in the immunohistochemistry (IHC) staining, suggesting that CD8⁺ T cells infiltrated the tumor and influenced the death of tumor cells (Figure 5j).

A previous study has shown that anti‐cancer efficacy may vary depending on the timing of cancer vaccine treatment.^[17^] The primary mRNA vaccination was administered to MC38‐bearing mice at different time points (1, 4, 7, and 11 days after inoculation of tumor cells) to confirm the efficacy of the cancer vaccine based on the vaccination time (Figure6a). Compared with other treatment groups, the earliest mRNA vaccine group exhibited significant tumor growth inhibition and tumor disappearance in seven out of eight animals (Figure 6b,c).

IFN‐γ ELISpot analysis was performed using TILs to confirm the induction of antigen‐specific T cell responses based on vaccination timing. Mice in the late vaccination group, where antigen‐specific immune responses were induced in TILs, exhibited smaller tumor sizes (Figure 6d). Furthermore, the correlation between antigen‐specific T cell‐related IFN‐γ secretion and anti‐tumor activity was assessed using Pearson's correlation coefficient, which showed a negative correlation (r = −0.835, p < 0.001) between antigen‐specific IFN‐γ secretion and tumor growth (Figure 6e). These findings indicate that the infiltration of neoantigen‐specific T cells into tumor tissues is a critical factor for improving anti‐cancer efficacy.

Despite the excellent immune activation achieved through cancer vaccine treatment, T cell exhaustion remains an obstacle to successful anti‐cancer immunotherapy.^[18^] The results of flow cytometric analysis for T cell exhaustion markers showed that the early immunized group exhibited lower expression of PD‐1 and Tim‐3 in CD4⁺or CD8⁺ T cells (Figure 6f). The expression of Tim‐3 in CD4⁺ or CD8⁺ T cells appeared to gradually increase with delay in the vaccination period. These results suggested that early vaccination with mRNA cancer vaccines and combination therapy with ICIs are necessary for efficient anti‐cancer treatment.

Cancer vaccines can prevent tumor recurrence or metastasis by inducing the differentiation of memory T cells.^[19^] To determine whether the mRNA vaccine confers protection by inducing persistent immune memory against cancer antigens, tumor‐free mice were re‐challenged with colon cancer cells following the initial mRNA vaccination (Figure7a). In the mRNA vaccine‐treated group, all mice survived up to 30 days post‐rechallenge, whereas the PBS‐administered control group exhibited significantly lower survival rates (Figure 7b). Isolated splenocytes and TILs were subjected to ELISpot analysis for IFN‐γ secretion 54 days after the last vaccination to confirm the persistence of antigen‐specific immune responses induced by mRNA vaccines (Figure 7c). Additionally, mRNA vaccine administration increased the population of CD4⁺ or CD8⁺ T cells in both the spleen and TILs (Figure 7d and 7e). In the spleen of mice treated with the mRNA vaccine, a decrease in T_reg cells (FoxP3⁺CD25⁺CD4⁺ cells) and a concomitant increase in effector memory CD8⁺ T cells (CD44⁺CD62L^‐ CD8⁺ cells) were confirmed at 54 days after last vaccination (Figure 7f,g), indicating that the mRNA vaccines maintained long‐term immune responses by generating memory T cells. Furthermore, the neoantigen‐specific T cell responses in splenocytes were still present 94 days after the last vaccination (Figure 7 h). Consequently, mRNA vaccination induced persistent MHC‐I and MHC‐II antigen‐specific immune responses in splenocytes and TILs, respectively.

Given that mRNA vaccines increased the expression of immune checkpoint molecules, a combination treatment of ICIs and cancer vaccines was performed in a colon cancer model. C57BL/6 mice were transplanted with 1 × 10⁵ cells of MC38 colon cancer cells in the right flank, then administered with 200 µg of ICIs (anti‐mouse PD‐1 and Tim‐3 antibody) on 3, 8, and 11 days or with 10 µg of mRNA vaccine on 7 and 10 days, or both (Figure8a). Notably, tumor growth was significantly reduced in the mRNA vaccine and ICIs‐treated groups compared to the control group (Irrelevant mRNA). The combined treatment group exhibited a more pronounced reduction in tumor growth than either the mRNA vaccine or ICIs alone (Figure 8b,c). Moreover, in the combination therapy group, tumors completely regressed in three out of eight mice, highlighting a synergistic effect on tumor growth inhibition (Figure 8d,e). To further investigate the impact of combination therapy on immune cell infiltration, flow cytometric analysis of tumor‐infiltrating lymphocytes (TILs) was conducted. The results revealed that treatment with both the mRNA vaccine and ICIs significantly increased the presence of immune cells, including CD45⁺, CD3⁺, CD8⁺, and CD4⁺ T cells, within the tumor microenvironment (Figure 8f). This suggests that the combination therapy promotes immune cell infiltration into the tumor, thereby enhancing anti‐tumor immunity. Furthermore, to assess the safety of the mRNA‐based cancer vaccine, histopathological analysis was performed on major organs, including the heart, kidney, liver, lung, and spleen (Figure S12a, Supporting Information). All groups, including PBS, MHC‐I, MHC‐II, and MHC‐I/II vaccine groups, showed no significant changes or pathological findings. Specifically, no evidence of pericarditis, calcification, muscular degeneration, immune cell infiltration, or abnormal heart diameter was observed in the heart. Similarly, the kidney, liver, lung, and spleen showed no signs of inflammation, tissue structural changes, fibrosis, or other pathological alterations (Figure S12b, Supporting Information). These findings indicate that the mRNA‐based cancer vaccine is safe and does not cause detectable tissue damage or toxicity in evaluated organs.

Each neoantigen candidate was chemically synthesized as a peptide (Peptron, Daejeon, Korea). The peptides were synthesized in lengths of 8–10 mers or 27 mers. SM‐102, distearoylphosphatidylcholine (DSPC), polyethylene glycol (PEG), and cholesterol for LNP formulation with mRNA were purchased from Avanti Polar Lipid, Inc. (Birmingham, AL, USA). Mouse anti‐CD4 APC‐Cy7, fluorescein isothiocyanate antibodies (Cat.100414, 130 308), anti‐CD8‐BV421 antibody (Cat.100738), anti‐CD45‐BV650 antibody (Cat.103151), anti‐TNFa‐fluorescein isothiocyanate antibody (Cat.506304), anti‐CD127‐APC antibody (Cat.135012), anti‐KLRG1‐BV605 antibody (Cat.138419), anti‐CD62L‐PE Cy7 antibody (Cat.104418), anti‐PD‐1‐PE antibody (Cat.135206), anti‐PD‐1‐APC‐Cy7 antibody (Cat.135224), anti‐CTLA‐4‐PE antibody (Cat.106306), and anti‐Tim‐3‐BV421 antibody (Cat.119723) were purchased from Biolegend (San Diego, CA, USA). Mouse anti‐CD8‐Percp‐Cy5.5 (Cat.45‐0081‐82), anti‐CD8‐PE‐Cy7 antibody (Cat.25‐0081‐82), anti‐CD45‐Percp‐Cy5.5 antibody (Cat.45‐0451‐81), anti‐IFN‐γ‐PE antibody (Cat.12‐7311‐82), and anti‐IFN‐γ‐APC antibody (Cat.17‐7311‐82) were purchased from Invitrogen (Waltham, MA, USA). Mouse anti‐FoxP3‐PE antibodies (Cat.560408) were purchased from BD Biosciences (San Diego, CA, USA). The epitope (SYALILRTIL, CNM‐13‐06)‐specific tetramer was produced as described previously.^[33^] In brief, recombinant H2‐D^b protein obtained from Escherichia coli BL21 (DE3) was refolded in the presence of the SYALILRTIL peptide and human β2m, biotinylated with biotin‐protein ligase BirA, and then purified with Superdex‐75 gel filtration column (GE Healthcare Life science). Purified monomer was tetramerized with R‐phycoerythrin conjugate (SAPE; Thermo Fisher Scientific, USA).

MC38 murine colon adenocarcinoma cells were purchased from Sigma‐Aldrich (St. Louis, MO, USA). MC38 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA), penicillin (100 IU mL⁻¹), streptomycin (100 µg mL⁻¹), and amphotericin B (0.25 µg mL⁻¹) and grown in a humidified incubator with 5% carbon dioxide (CO₂) at 37 °C. All animal experimental procedures were approved by the Institutional Animal Care and Use Committee of the Catholic University of Korea (CUK‐IACUC‐2023‐034). Female C57BL/6J mice (6–8 weeks old) were purchased from DBL Co., Ltd. (Chungcheongbuk‐do, South Korea). They were housed in a temperature‐controlled room and maintained on a 12‐h light/dark cycle with free access to food and sterile water under specific pathogen‐free conditions.

Sequencing on normal colon tissue and cancer cells was performed to identify TSAs unique to cancer tissues through comparative analysis with normal tissues and design effective epitopes. Exome and mRNA‐seq libraries were constructed using Agilent's SureSelectXT Library Prep Kit and the TruSeq Stranded mRNA Library Prep Kit, respectively. Two biological replicates were prepared for mRNA‐seq. Sequencing was conducted on the Illumina Novaseq 6000 platform to generate 101‐bp paired‐end reads following the manufacturer's protocol. The read data were processed with Trimmomatic to remove low‐quality bases and sequencing artifacts using the parameters TruSeq3‐PE‐2.fa:2:30:10:2 HEADCROP:5 TRAILING:20 MINLEN:75. TSAs were identified as inducers of specific immune responses in tumor tissues based on SNVs and small InDels from the Exome data and gene fusion information from the RNA‐seq data. Figure 1b illustrates the overall process of identifying cancer‐specific variants. The reference genome GRCm39 from the Ensembl database was used for TSA discovery, annotation, and expression profiling.

For TSA identification using Exome data, the trimmed and filtered reads were mapped to GRCm39 using BWA‐MEM2,^[34^] and PCR duplicates were removed using the MarkDuplicate module in Picard tools (http://broadinstitute.github.io/picard/↗). As the tumors were generated using a distinct cancer cell line and not typical somatic mutations, germline variant callers GATK HaplotypeCaller^[35^] and Samtools^[36^] were employed for variant calling. The identified variants were filtered using VCFtools^[37^] to exclude those with variant quality < Q30, missing data, multi‐allelic loci, and read depth < 30. Gene‐related information for the variants was obtained using snpEff^[38^] and Variant Effect Predictor^[39^] with Ensembl 108 annotation. From the annotation data, non‐synonymous mutations present exclusively in tumor tissues and absent in normal tissues, identified by both variant callers, were selected. These variants were considered as candidate TSAs for epitope design.

For gene fusion identification using RNA‐seq, reads from the two biological replicates were mapped with STAR aligner^[40^] and identified gene fusions using ARRIBA^[41^] and STAR‐Fusion.^[42^] Gene fusions were consistently identified by both replicates, and both tools were selected. These gene fusions were matched to the assembly results obtained using Trinity^[43^] to confirm their presence in hisat2.^[44^] Confirmed gene fusions, SNVs, and InDels identified from exome sequencing were used as candidate TSAs for epitope design.

Antigen prioritization was conducted to design epitopes capable of inducing specific immune responses against tumor tissues from the identified candidate TSAs. Initially, featureCounts from the subread package^[45^] on the mRNA‐seq alignment results obtained with the STAR aligner were used to select TSAs with sufficient expression levels. Expression levels in counts per million were normalized and ranked using edgeR,^[46^] after which candidate TSAs were associated with genes in the bottom 10% of expression levels. Furthermore, to exclude candidate TSAs whose aa sequences are similar to those of self‐genes, potentially failing to induce an immune response, TSAs indistinguishable were filtered out from self‐gene aa sequences based on Ensembl 108 annotation. For the remaining candidate TSAs, netMHCpan^[47^] and netMHC‐IIpan^[47^] were used to assess binding affinity to the MHC alleles H‐2‐Dd, H‐2‐Kd, H‐2‐Kb, and H‐2‐Db in the C57BL/6 mouse strain for both MHC‐I and MHC‐II. Strong binders with a binding strength rank within the top 1% were filtered out. Using NetChop,^[48^] the number of cleavage points within the designed epitopes was also examined. Considering the individual characteristics and sources (SNV, InDel, and gene fusion) of the candidate epitopes, the final candidate epitopes for MHC‐I and MHC‐II, which were expected to potentially induce specific immune responses against the tumor, were manually selected. In contrast to MHC‐I epitope design, the approach for designing MHC‐II epitopes entails systematically extracting the 13‐mer flanking sequences on either side of the target coordinate, which is then utilized as the epitope.

The neoantigen sequences were inserted into the pCUK3‐1 mRNA platform backbone, as described in a previous study.^[49^] Additionally, the neoantigen sequences were located between a signal peptide and an MITD sequence to facilitate efficient antigen presentation, as in previous studies.^[11, 15^] The epitopes were designed in the form of pentamer or hexamer for each mRNA construct, and each epitope was designed by connecting with a glycine‐serine aa linker. The plasmid deoxyribonucleic acid was purified using an endotoxin‐free purification kit (Nucleo bond Xtra Midi Plus EF kit, MACHEREY‐NAGEL, Düren, Nordrhein‐Westfalen, Germany). It was prepared using a linearization process using Not I endonuclease (Enzynomics, Daejeon, South Korea). The mRNA was produced using an EZ T7 High‐Yield in vitro Transcription kit (Enzynomics) and a Cap 1 capping analog (SMARTCAP, ST PHARM, Gyeonggi‐do, South Korea). Wild‐type uridine was replaced with N1‐methyl‐pseudouridine‐5′‐triphosphate (m1ΨTP; TriLink, USA). The IVT mRNA mixture was purified via LiCl precipitation after DNase I (Enzynomics) treatment. dsRNA in the mRNA mixture was removed using the cellulose resin.^[50^]

All lipid components (molar ratio of 50:10:38.5:1.5, Ionizable cationic lipid:DSPC:Cholesterol: DSPE‐PEG) were dissolved in chloroform‐methanol solution (1:1, v/v) and then dried under pressure to form a thin film for long‐term storage. The lipid mixture and mRNA were dissolved in absolute ethanol and citrate buffer (pH 4.0), respectively.^[51^] Subsequently, the lipid mixture and mRNA were formulated at a ratio of 1:3 (v/v) using a microfluidic channel (Encell master, Enparticle, Busan, South Korea). The mRNA‐LNP solution was serially centrifuged for buffer exchange via ultrafiltration (30 kDa, Merck Millipore, Burlington, MA, USA).

RNA quality was confirmed via agarose gel electrophoresis. mRNA was quantified using absorbance at 260 nm, and the purity was confirmed using the Abs.260/Abs.280 ratio. Product‐related dsRNA was measured using an anti‐dsRNA‐based dot blot assay.^[50^] The sizes and distributions of mRNA‐LNPs were measured using dynamic light scattering (Zetasizer 2000, Malvern Panalytical Ltd., Malvern, UK). The encapsulation efficiency was analyzed using the agarose gel electrophoresis method and Ribogreen assay (Quanti‐iT RiboGreen RNA assay kit, Invitrogen, Waltham, MA, USA), following the manufacturer's instructions.

Female 6–8‐week‐old C57BL/6 mice were subcutaneously vaccinated with 20 µg of mRNA‐LNP on days 0, 3, 7, and 14 and 100 µg of peptides with 50 µg of poly (I:C) (Sigma‐Aldrich) on days 0 and 7. The vaccinated mice were euthanized 7 days after the last vaccination, and splenocytes were harvested to confirm immunogenicity using ELISpot and flow cytometric analysis.

Tumor‐bearing mice were subcutaneously vaccinated with 10 to 20 µg of mRNA‐LNPs. ICIs, such as anti‐mouse PD‐1 antibody (InVivoMAb, Clone: RMP1‐14, BioXcell, NH, Lebanon) and anti‐mouse Tim‐3 antibody (InVivoMAb, Clone: B8.2C12, BioXcell), were intraperitoneally administered at a dose of 200 µg per injection on days 3, 8, and 14 after tumor inoculation. Tumor sizes were measured using an electronic vernier caliper at 2‐ to 4‐day intervals, and tumor volume was calculated using the equation V (mm³) = (length × width²) ÷ 2.

Tumors were harvested, cut into pieces of appropriate size, and chopped with sterile scissors. The chopped tissue was digested in serum‐free RPMI containing 1 mg mL⁻¹ Collagenase D (Sigma‐Aldrich) and 0.1 mg mL⁻¹ DNase I (Sigma‐Aldrich) at 37 °C for 1 h. Subsequently, they were grounded through a 70 µm cell strainer and centrifuged at 1000 rpm for 5 min. The supernatant was removed, and 1 mL of ACK lysis buffer was added to the pellet. The ACK lysis buffer was neutralized with RPMI1640 containing 10% fetal bovine serum and centrifuged. The pellet was resuspended in RPMI1640 medium.

Mice treated with PBS, peptides, irrelevant mRNA, and mRNA vaccines were euthanized, and splenocytes and tumor lymphocytes were isolated. Subsequently, 2.0 or 5.0 × 10⁵ cells were seeded into each well of anti‐IFN‐γ coated 96‐well plates (Cat.MSIPS4510, Millipore, Burlington, MA, USA) and stimulated with MHC‐I or MHC‐II predicted peptides for 24–48 h in a humidified incubator at 37 °C, 5% CO₂. After washing the plates, an ELISpot assay was performed following the manufacturer's protocol (Cat. 3321–4APT, Mabtech, Nacka Strand, Sweden). The spots were counted using an ELISpot reader (ImmunoSpot Software, C.T.L corporation, Beaverton, OR, USA). The immunogenicity of neoantigens was evaluated by performing IFN‐γ ELISpot analysis repeatedly.

MC38 cells expressing firefly luciferase (MC38‐Luc) were seeded into 96‐well plates at a density of 5 × 10⁴ cells per well. These cells were then treated with TILs obtained from the tumor tissues of mice treated with PBS or the mRNA vaccine. The control wells were treated with RPMI 1640 medium containing 10% fetal bovine serum and 1% penicillin/streptomycin. The ratio of TILs (Effector cells) and MC38‐Luc cells (Target cells) was 5:1, and the co‐culture was incubated for 18 h at 37 °C in an atmosphere containing 5% CO₂. After incubation, the cells were washed twice with PBS. Luminescence was measured using a Luciferase Assay System (E4550, Promega Corp., Madison, WI, USA), following the manufacturer's instructions.

Tumor tissues were sectioned into appropriately sized pieces and fixed overnight in a 4% paraformaldehyde solution at room temperature for 24 h. They were then transferred to 70% ethanol before the paraffin‐embedding process. Tissue blocks were sectioned to a thickness of 10 µm, and the sections were mounted on glass slides. H&E staining for pathological analysis was performed by KP&T Technology in South Korea.

IHC for CD8 was performed by staining cells with an anti‐mouse CD8 antibody (Clone EPR21769↗, Abcam, Cambridge, UK). Following the manufacturer's instructions (HRP/DAB detection IHC kit, ab64264, Abcam), the binding of the CD8 antibody was visualized using 3,3′‐diaminobenzidene (DAB) tetrahydrochloride hydrate.

Splenocytes and TILs were seeded into 96‐well plates at densities of 2 × 10⁵ or 1 × 10⁶ cells per well and treated with Fc block (Cat. 553 141, BD Biosciences) to block non‐specific Fc receptor‐mediated fluorescent antibody binding after washing with wash buffer (0.5% fetal bovine serum in PBS). For surface marker analysis, the cells were stained with fluorescence‐labeled mouse antibodies against cell surface markers (including anti‐CD45, anti‐CD4, anti‐CD8, anti‐PD‐1, anti‐CTLA‐4, and anti‐Tim‐3) for 30 min at 4 °C. To assess antigen‐specific cytokine secretion, the cells were stimulated with 200–500 ng/well MHC‐I and MHC‐II peptides at 37 °C in an atmosphere containing 5% CO₂ for 12 h and stained using the same immune cell population analysis protocol. Subsequently, the cells were permeabilized using Cytofix/Cytoperm (Cat. 554 714, BD Biosciences) and stained with fluorescence‐labeled mouse anti‐IFN‐γ and anti‐TNF‐α antibodies for 30 min at 4 °C. For transcription factor analysis, such as FoxP3, the cells were permeabilized using the FoxP3/Transcription Factor Staining Buffer Set (Cat. 00‐5523‐00, eBioscience) and stained with fluorescence‐labeled mouse anti‐FoxP3 antibodies for 30 min at 4 °C. Flow cytometry and data analysis were performed using a Beckman CytoFLEX (Beckman Coulter, Brea, CA, USA).

Statistical analysis of data was performed using GraphPad Prism (version 9.0; GraphPad, La Jolla, CA, USA) with significance defined as non‐significant (ns; p > 0.05) or significant (p ≤ 0.05 [*], p ≤ 0.01 [**], and p ≤ 0.001 [***]). All results from independent experiments are presented as the mean ± standard error of the mean (SEM), unless otherwise stated. Tumor growth was determined using a two‐way analysis of variance, followed by post hoc testing with Dunnett–Turkey's multiple comparison test. The survival rate was determined using the log‐rank test.

The data that support the findings of this study are available from the corresponding author upon reasonable request.