Neoantigen Targeting as a Novel Approach for Therapy-Resistant Tumors

Abstract

Neoantigens have emerged as central targets in the development of individualized cancer vaccines, as they are recognized by the immune system as foreign and can elicit potent anti-tumor responses. Advances in computational algorithms and machine learning have improved the accuracy of neoantigen prediction, and clinical trials have reported promising results of vaccines incorporating validated neoantigens into peptide, RNA, or dendritic cell platforms. Despite this progress, the identification of immunologically effective neoantigens remains a complex multi-step process. Current attention is directed not only to conventional missense and indel mutations but also to non-coding RNA-derived neoantigens. A critical challenge is the establishment of reliable systems to verify whether candidate neoantigens can activate cytotoxic T lymphocytes. Moreover, neoantigen expression and immune responses are influenced by tumor-intrinsic and therapeutic factors. High programmed death-ligand 1 expression is known to suppress immune recognition, while ARID1A mutations, associated with tumor progression, can enhance neoantigen expression during chemoresistance, suggesting that drug resistance may be accompanied by new immunogenicity. This work provides an overview of current methodologies for neoantigen identification, advances in prediction and validation strategies, and the dynamic interplay between tumor-intrinsic and immune-related factors that regulate neoantigen expression. We also highlight recent clinical insights as well as novel analytical approaches and discuss challenges and future directions in this rapidly evolving field, emphasizing the potential of neoantigen-based therapies to transform cancer immunotherapy.

Key Points

Neoantigens can arise from various changes in gene sequence and expression, including missense mutations, short insertions and deletions (indels), gene fusions, aberrant splicing events, virus-derived sequences, and non-coding regions of the genome.

Neoantigens have been identified as highly specific and potent targets capable of inducing tumor-specific T-cell responses. Currently, research focuses on identifying and characterizing neoantigens with strong immunogenicity.

Neoantigens have been clinically exploited through various modalities, including peptide vaccines, messenger RNA platforms, and T-cell receptor-engineered T-cell therapies.

Introduction

Cancer remains a major global health challenge, causing approximately 10 million deaths annually, and ranking as the second leading cause of mortality worldwide. Breast, lung, colorectal, and prostate cancers are the most frequently diagnosed types, highlighting the urgent need for innovative therapeutic strategies. In addition to conventional chemotherapy and radiotherapy, immunotherapy has emerged as a transformative paradigm-shifting approach in cancer treatment. Immunotherapy can achieve durable tumor control by modulating the tumor microenvironment and harnessing the immune system of the patient. The advent of immune checkpoint inhibitors (ICIs) has marked a breakthrough in cancer treatment, although their efficacy is not robust across patients.

Recent advances in cancer immunotherapy encompass a broad spectrum of modalities, including cytokine therapy, monoclonal antibodies, bispecific T-cell engagers, chimeric antigen receptor-T, and T-cell receptor-engineered T (TCR-T) cell therapies, and therapeutic cancer vaccines. Among these, antigen-specific therapeutic approaches have received considerable attention. Tumor antigens are categorized into tumor-associated antigens, which are the protein products of genes aberrantly expressed in tumors but silenced in normal tissues, and tumor-specific antigens, which arise from somatic mutations or viral infections and are highly tumor specific [1]. Among tumor-specific antigens, neoantigens derived from tumor-specific mutations are particularly attractive targets for cancer immunotherapy because they escape central tolerance and elicit strong immune responses [2].

Tumor-derived neoantigens, which are theoretically recognized by the immune system as non-self, similar to pathogen-derived antigens, possess both high specificity and immunogenicity [1]. Active immune editing during tumor evolution influences the presentation of neoantigens, shaping tumor-immune interactions [2]. Clinically, the observation that patients with a high tumor mutational burden (TMB) often respond well to ICIs underscores the biological and therapeutic significance of neoantigens [3, 4]. However, because TMB does not always correlate with clinical outcomes, qualitative aspects of neoantigens have emerged as critical determinants of therapeutic efficacy. Technological advances such as next-generation sequencing, mass spectrometry, and bioinformatics have significantly accelerated the identification of tumor-expressing neoantigens. Nevertheless, predicting the immunogenicity of candidate neoepitopes remains challenging, compounded by factors such as human leukocyte antigen (HLA) polymorphisms and mosaic antigen expression within tumors. Moreover, neoantigen-targeted therapies are limited by factors, such as limited T-cell receptor (TCR) affinity, immune evasion mechanisms, and heterogeneous treatment responses.

This review summarizes the current understanding of neoantigen biology, computational analyses, and vaccine-related clinical responses, highlighting how these insights may shape the future of personalized cancer immunotherapy. Continued progress in neoantigen research holds promise for advancing cancer treatment to the next frontier.

Programmed Death-Ligand 1 Expression Affects Neoantigen Presentation

A key concept in understanding anti-tumor immunity is the cancer-immunity cycle, as described by Chen and Mellman [5]. In this cycle, tumor antigens released into the tumor microenvironment are captured and processed by dendritic cells (DCs), which then migrate to the draining lymph nodes to prime tumor-specific T cells. Activated T cells enter circulation and exert cytotoxic effects at the tumor site, completing the cycle. Therapeutic strategies, such as DC-based vaccines, leverage this mechanism by loading DCs with tumor antigens to induce robust anti-tumor responses. Dendritic cell-based immunotherapy traditionally involves generating DCs from hematopoietic stem cells or monocytes, pulsing them with tumor antigens, or adding peptide antigens to induce antigen-loaded DCs, which are then administered to patients. The selection of highly immunogenic antigens is the key to this strategy. Among these, neoantigens, personalized cancer antigens derived from tumor-specific mutations, have attracted significant attention because of their strong immunogenicity. However, unless these neoantigens are efficiently taken up and presented by DCs, they cannot elicit effective anti-tumor immunity.

One way to assess whether highly immunogenic tumor antigens functionally contribute to immune responses is to evaluate immune checkpoint blockade (ICB) therapy. In the tumor microenvironment, T cells are often inactivated by programmed death-ligand 1 (PD-L1) signaling from tumor cells. By administering anti-programmed cell death protein 1 (PD-1) antibodies, inactivated T cells can be reactivated to exert their anti-tumor effects. Indeed, the PD-1 blockade has demonstrated strong clinical efficacy in non-small cell lung cancer, gastric cancer, and melanoma [6–8].

The presence of neoantigens is considered to be one of the reasons for the success of PD-1 therapy. Cancer types vary in TMB, that is, tumors with a high TMB, such as melanoma, colon, esophageal, and gastric cancers, tend to respond better to ICB [9]. In contrast, tumors with a low TMB, such as neuroblastomas, thymic tumors, small intestinal cancers, myeloid leukemia, and pituitary adenomas, often show poor responses. Another report highlighted that ICB is particularly effective in tumors with mismatch repair deficiencies, implicating neoantigen involvement [10].

Reportedly, a higher clonal TMB correlates with stronger immunogenicity [11]. In cases with high complete response or partial response rates, clonal TMB was detected along with CD8⁺ T cells and CXCL9 expression in the tumor microenvironment. These findings support the existence of a triangle of interactions among TMB, cytotoxic T lymphocytes (CTLs), and PD-L1 expression, which together determine the response to checkpoint blockade.

Although PD-L1 expression is considered a surrogate for CTL response, clinical outcomes indicate a more complex relationship. For example, while patients with non-small cell lung cancer with >50% PD-L1 expression often respond well to PD-1 therapy, those with 1–49% expression show minimal benefit [6, 12] (Table 1). In some melanoma cases, higher PD-L1 expression correlates with better outcomes [13, 14]; however, reports exist where PD-L1^low or PD-L1^negative tumors respond strongly to ICB [15]. Our analysis using TCGA data showed that patients with colon cancer who had PD-L1 ^low and cytolytic activity (CYT)^high profiles achieved the highest survival rates [15]. Moreover, patients with PD-L1^low and high microsatellite instability tumors exhibited prolonged survival. These findings support the subdivision of tumors into four clinical groups based on PD-L1 and CYT expression. Despite the clear relationship between TMB and CTL infiltration, the association between PD-L1 levels and the actual anti-tumor response is not straightforward. Thus, the interplay between PD-L1 expression, TMB, and CTL responses is critical for determining the efficacy of ICB (Fig. 1). Although PD-L1 expression can be induced by interferon-γ (IFN-γ) secreted from neoantigen-specific T cells, it is not solely determined by antigen load; intrinsic PD-L1 expression and defects in IFN-γ signaling may lead to immune evasion even in high-TMB tumors. In clinical practice, tumors exhibit heterogeneous PD-L1 expression; while some tumors with high PD-L1 expression respond well to PD-1 blockade, cases exists in which tumors with little to no PD-L1 expression show a favorable response to PD-1 inhibitors [6, 12–25].

Table 1.

Differences in clinical response to ICB therapy associated with PD-L1 expression

PD-L1 dependence	Cancer type	PD-L1 validation	References
Independent	Cervical cancer	IHC	[16]
	Non-small-cell lung cancer	IHC	[17–19, 25]
	Vulvar squamous cell carcinoma	IHC	[20]
Positive	Breast cancer	IHC	[21]
	Melanoma	IHC, mRNA	[13, 14, 22]
	Non-small-cell lung cancer	IHC	[6, 12, 22, 23]
	Oral squamous cell carcinoma	IHC	[24]
Low or Negative	Colon adenocarcinoma	mRNA	[15]
	Uterine corpus endometrial carcinoma	mRNA	[15]

Fig. 1.

Triangular relationship among programmed death-ligand 1 (PD-L1) expression, tumor mutational burden (TMB), and cytotoxic T lymphocyte (CTL) response. The responsiveness to an immune checkpoint blockade depends on the relationship between PD-L1 expression, TMB, and CTL response. a In tumors where malignant cells do not intrinsically express PD-L1, but a high TMB facilitates the induction and intra-tumoral infiltration of CTLs, interferon-gamma secreted by activated CTLs can induce PD-L1 expression on tumor cells. This phenomenon, termed acquired immune resistance, is generally susceptible to a therapeutic blockade with anti-programmed cell death protein 1 (PD-1) antibodies. b Conversely, in the context of innate immune resistance, whereby tumor cells constitutively express PD-L1 as a consequence of genetic alterations or aberrant activation of oncogenic signaling pathways, clinical responsiveness to anti-PD-1 antibodies is limited, even in the presence of a high TMB, if antigen-specific T-cell priming is absent and lymphocytic infiltration into the tumor microenvironment does not occur. Moreover, in tumors with intrinsically low PD-L1 expression, anti-tumor immunity may remain effective provided that CTLs are successfully elicited; however, when accompanied by major histocompatibility complex deficiency or alternative immune evasion mechanisms, therapeutic efficacy cannot be anticipated

Based on the results of preclinical and clinical studies, we address an important issue. Notably, a high TMB level leads to IFN-γ-induced PD-L1 expression, enhancing the likelihood of ICB efficacy. However, some PD-L1^positive tumors exhibit innate resistance to anti-PD-1 therapy by suppressing neoantigen-specific CTL induction. These contradictory results suggest that the relationship between PD-L1 expression and neoantigen-specific CTL activity is critical. Clinically, PD-L1^low, CYT ^high, and high microsatellite instability may be predictive of a favorable prognosis in colorectal cancer.

Neoantigen Patterns

Neoantigens can be classified as shared and personalized. Shared neoantigens arise from frequently mutated genes, such as driver mutations, across multiple types of tumors, with examples including TP53 and KRAS. Mutations associated with specific tumor types also cover shared neoantigens and have great potential as off-the-shelf vaccines for preventing tumor development. However, these hot mutation-derived, highly immunogenic neoantigens are easily eliminated during the early phase of tumor development. Therefore, palpable tumors contain fewer common mutations, resulting in few shared neoantigens [26]. Conversely, personalized neoantigens are unique to each patient and arise primarily from passenger mutations. Owing to the potential for simultaneous targeting by vaccines to avoid tumor immune escape, these private antigens have been widely utilized and shown to be functional in clinical trials [27, 28]. In addition to the traditional neoantigens uncovered by short-read high-throughput sequencing, long-read sequencing has enabled the identification of new types of neoantigens. Both tumor tissues and circulating plasma DNA can be used to detect mutations [29, 30]. Moreover, RNA within plasma extracellular vesicles has also been shown to have potential use in liquid biopsies for predicting neoantigens, in addition to conventional circulating DNA [31]. Although mass spectrometry-based major histocompatibility complex (MHC) ligandome analysis is the most reliable method to identify actual neoantigens present in tumors, its limited throughput and the relatively advanced skills required for ligandome purification make its application in clinical settings challenging [32, 33]. Therefore, many MHC affinity prediction algorithms have been developed and optimized with high accuracy using the latest deep learning and trained models based on HLA ligandomes [34–36]. Their accuracy improved significantly with the input of unique datasets [37–42]. Despite this progress, a significant proportion of predicted neoantigens are not actually present on tumors or are not highly immunogenic owing to different neoantigen-coding gene expression levels and HLA polymorphisms in each patient [43, 44]. Several neoantigens predicted to induce CD8⁺ T-cell responses also elicit CD4⁺ T-cell responses [45, 46]. However, a key aspect of neoantigens is the synergistic activation of CD4⁺ and CD8⁺ T cells by multiple neoantigen vaccines, which is critical for inducing strong anti-tumor immune responses [47–49]. Further optimization is required for accurate prediction. To date, validating immunogenicity by corresponding TCRs and creating a library should be facilitated for the scalable application of neoantigen therapy, although it requires significant effort [50–53].

Missense Mutations

Missense mutations caused by single-nucleotide variants are typical sources of neoantigens (Fig. 2). Owing to the spread of high-throughput sequencing and well-established analysis pipelines [54–56], researchers can more easily identify missense mutation-derived neoantigens in human patients and experimental mouse models [57, 58]. Frequently detected mutations, such as KRAS G12D in HLA-C*08:02 [59], NRAS Q61K/Q61R in HLA-A*01:01 [60], TP53 R175H in HLA-A*02:01 [61], H3.3 K27M in HLA-A*02 [62], U2AF1 Q157R in HLA-A*33 [63], EGFR L858R in HLA-A*11:01, T790M in HLA-C*15:02 [64], PIK3CA in HLA-A*03[65], and RET M918M in HLA-DPB1*04:01/02 [66], are immunogenic in patients with specific alleles. These epitopes are expected to use shared neoantigens. However, while they are easiest to detect, missense mutation-derived neoantigens sometimes show slight differences from the original WT peptide sequence, resulting in insufficient immunogenicity. With respect to the similarity of the original sequence, the characterization of epitope specificity and immunogenicity is important before clinical use, if possible. Moreover, the mutation position, which regulates the affinity to the MHC complex or corresponding TCR, is critical for predicting missense-mutated neoantigens [67].

Fig. 2.

Mechanisms underlying the induction of novel immune responses to neoantigens. In cancer, missense mutations causing single amino acid substitutions or indel mutations introducing multiple amino acid changes can occur in genes encoding self-antigens. These altered peptides enter the endoplasmic reticulum, bind to major histocompatibility complex class I molecules, and are transported to the cell surface, where they are recognized by CD8⁺ T cells. As a result, self-derived antigens that have acquired mutations can be strongly recognized by autologous T cells. aa amino acids

Short Indel Mutation

Insertions and deletions (indels) are typically associated with deleterious effects on the original protein structure (Fig. 3). These frameshift mutations can generate new open reading frames. Previous clinical studies have demonstrated that mismatch repair-deficient and high microsatellite instability tumors are vulnerable to host immunity [68]. Additionally, these immune-sensitive cells accumulate a higher number of indel mutations as they proliferate [69]. As indel-derived neoantigens have not naturally occurring peptide counterparts, the value of indels is higher than that of missense mutations [70]. Although the biological functions of many indel mutations in tumorigenesis remain unclear, they are frequently detected in mouse models and patients. An intriguing example is mutant NPM1 in AML. Mutant NPM1 caused by four base pair insertions generates AML-specific C-terminal 11 amino acid sequences. This new open reading frame was processed and presented as a neoantigen on HLA-A* 02:01 and HLA-A* 11:01. Specific TCR-T therapies are under development for the future application of these neoantigens [71, 72]. Furthermore, not limited to specific indel-derived neoantigens, peptides derived from recurrent frameshifts can be utilized in shared neoantigen cocktails to treat mismatch repair-deficient tumors [73, 74].

Fig. 3.

Sources of neoantigens in cancer. Genetic alterations that can give rise to neoantigens in cancer include missense mutations, short insertions or deletions (indels), fusion genes, structural variants, abnormal splicing events, antigens derived from oncogenic viruses, and antigens originating from endogenous retroviruses. ORFs open reading frames

Fusion Genes

Chromosomal translocations, deletions, and inversions are major causes of tumorigenesis (Fig. 3). Fusion proteins gain aberrant functions or lose their original function, causing abnormal cell proliferation. Some fusion genes have been reported to acquire immunogenicity, particularly in the peptide sequence of their fusion domains [75]. Specifically, ETV6-RUNX1 [76], DEK-AFF2 [77], CBFB-MYH11 [78], SYT-SSX [79], EWS-FLI1 [80], and DNAJB1-PRKACA [81] have been reported to induce T-cell responses in a fusion domain-specific manner. Fusion genes are considered to be less immunogenic because of evolutionary selection [75]. Advances in analytical approaches will facilitate the development of highly immunogenic neoantigens derived from fused genes [82, 83].

Abnormal Splicing

Tumors have alternative splicing patterns that are not detected in normal tissues [84]. Previous research has also validated the potential use of chemical compounds to enhance neoantigen generation by modifying the Ser/Arg-rich splicing factor 6 (SFRS6) function [85] and RNA-binding motif protein 39 (RBM39) stability [86]. Moreover, several splicing abnormalities that are conserved across patients with tumors have been identified as shared neoantigens [87, 88]. Recently, multiple software packages have been developed to detect the splicing neoantigen burden [89–91].

Endogenous Retroviral Sequence

Oncogenic viral infections, such as Epstein–Barr virus, human T-cell leukemia virus, human papillomavirus, hepatitis B virus, and hepatitis C virus, cause several types of malignancies, including lymphoma, cervical cancer, and liver cancer. These viral components elicit T-cell-mediated host immune responses [92–96]. In addition to exogenous viruses that infect only patients, previous studies have suggested the clinical potential of endogenous retroviral sequences that infect the ancient Homo sapiens genome [97, 98]. A recent study revealed that endogenous retroviral elements are transcribed and present as immunogenic epitopes by ribosomal profiling and mass spectrometry analysis [99, 100]. Proposedly, the transactivation of these retroviral elements can be triggered by DNMT and HDAC inhibitors used in clinical treatment, which augment the immunogenicity of specific tumors [101]. As their sequences are well conserved in humans, researchers expect the effective use of endogenous retrovirus-derived peptides as shared neoantigens [102]. Pipelines to detect such immune-activating endogenous retroviral elements are also being developed [103].

Long Non-coding RNA

Amino acid sequences not only from conventional coding sequences but also from RNAs previously considered to be non-coding, such as long non-coding RNA (lncRNAs), are an alternative source of neoantigens. Proteomic data analysis showed that tumor cells express more peptides from previously predicted non-coding RNAs than normal cells do [104]. Their expression levels are regulated by PRMT5, and they are overexpressed in many types of tumors. Additionally, some lncRNA-derived peptides induce anti-tumor immune responses [105]. The expression levels of lncRNA regions as predictive biomarkers of ICB response are obscure because not all lncRNAs are translated and presented on the tumor [106]. Nevertheless, further identification of lncRNA-derived neoantigens facilitated by useful tools is expected to uncover full immune surveillance and potential clinical applications [107, 108].

Structural Genomic Variants

Genomic rearrangements with insertions longer than 50 bp, deletions, duplications, amplifications, copy number alterations, or translocations and inversions are typical structural variants. Only long-read and mate-pair sequencing can overcome insufficient human genome resolution [109–112]. These structural variants have been comprehensively analyzed and integrated as resources [113, 114]. In addition, viral infection and endogenous retroviral activation affect structural variants and drive tumorigenesis [115]. These tumor-specific alterations have great potential as sources of new open reading frames that give rise to neoantigens [116].

In summary, all tumor-specific peptide sequences are potential immune targets. If these uniquely expressed sequences are appropriately processed and presented on MHC-I and MHC-II cells, they would represent attractive therapeutic targets. In addition to genomic alterations in tumors, several post-translational modifications, such as glycosylation and phosphorylation, have been reported to induce T-cell responses [117–119]. As discussed above, conventional TMBs and neoantigen load are indicators of missense or short-indel mutations. Many exceptions exist derived from a new type of neoantigen, including cases in which immunotherapy is effective despite a low number of mutations [77]. Accurately predicting immunotherapy responses remains challenging, considering that it is systemically regulated by various factors, such as PD-L1 expression, the tumor microenvironment, and inflammation parameters, in addition to genomic variants. Therefore, establishing a neoantigen library as a strong indicator of ICB and neoantigen-related therapeutic responses is critical and helpful for future medicine. Notably, tumoricidal activity is induced not only through host immunity, which can be boosted by vaccines, but also through third-party-derived TCR-T cells or designed TCR-mimic antibodies and their derivative chimeric antigen receptor [65, 120]. Recently, researchers have established a large-scale TCR database, which can be utilized for isolating pairs of tumor antigens and respective TCRs [121, 122].

The identification of clinically relevant neoantigens is time consuming. However, this strategy remains promising, with the potential to facilitate personalized medicine in future clinical settings.

Neoantigen Biology

Although the whole process from resection of tumor tissues to the design of neoantigen targets is well established, substantial optimization is still required to prepare clinically applicable vaccines, especially for rapidly developing tumors, in a timely manner. Furthermore, combination therapy is a critical and stringent predictive biomarker for extending the clinical benefits. Herein, we discuss neoantigens and their biological functions (Table 2).

Table 2.

Clinical trials of personalized neoantigen vaccines

Trial ID	Modality/cancer	Phase (patients)	Immune response	References
NCT01970358	Long peptides/melanoma	Phase I (6)	Durable CD4⁺/CD8⁺T cell responses (68%/59%), with long-term clonal persistence and tumor infiltration (> 4 year)	[45, 196]
NCT04161755	mRNA-lipoplex/pancreatic cancer	Phase I (16)	Neoantigen-specific CD8⁺T cells in 50% of patients, with clones showing ~ 7.7-year lifespan and TRM-like phenotype with tumor infiltration	[198, 199]
NCT03289962	mRNA-lipoplex/advanced solid tumors	Phase I (213)	Immune responses in 71% of patients (CD4⁺and/or CD8⁺), with peripheral CD8⁺T cells up to 23% (median 7.3%) and TIL CD8⁺T cells up to 7.2%, persisting for ~ 23 months	[200]
NCT03313778	mRNA-LNP/non-small cell lung cancer, melanoma	Phase I (16)	De novo induction and enhancement of pre-existing CD8⁺and CD4⁺T cells, with responses detectable up to ~ 30 weeks	[201]
NCT03897881	mRNA-LNP/melanoma	Phase II (157)	Combination therapy reduced recurrence or death risk by 49% (HR 0.51), with 2.5-year RFS of 74.8% vs 55.6% and improved DMFS (HR 0.38) showing a favorable trend in OS	[202]
NCT03639714	Prime(DNA)–boost(mRNA)/colorectal cancer, non-small cell lung cancer, gastroesophageal adenocarcinoma, urothelial cancer	Phase I (14)	Neoantigen-specific CD8⁺T cell responses in all patients (13/13), with long-term persistence (> 52 weeks), induction of polyfunctional CD8⁺T cells, and 1.6–6.4-fold increase in tumor infiltration of reactive clones	[203]
NCT02950766	Long peptides/renal cell carcinoma	Phase I (9)	Predominantly CD4⁺(≈ 99%) polyfunctional memory responses (IFNγ, IL-2), with durable expansion of vaccine-specific TCR clones recognizing autologous neoantigens, persisting up to 3 years	[204]
NCT03359239	Long peptides/urothelial carcinoma	Phase I (10)	Neoantigen-specific T cell responses in all patients, with 55% of peptides eliciting immunity (CD8⁺46%, CD4⁺54%) and polyfunctional cytokine production	[205]
NCT03953235	Prime(DNA)–boost(mRNA)/non-small cell lung cancer, CRC, PDA, etc.	Phase I/II (19)	KRAS-specific T cell responses ex vivo in 31% of patients (durable up to 6 months) and TP53-specific responses in 83%	[206]
NCT03970382	Neoantigen-specific TCR-T/solid tumors	Phase I (16)	Dose-dependent peripheral expansion of neoantigen-specific T cells (up to 15%, 37% with IL-2) and tumor infiltration with infused neoTCRclones detected post-treatment	[207]
NCT04625205	Neoantigen-specific T cell/melanoma	Phase I (9)	Neoantigen-specific CD8⁺(3–9) and CD4⁺(2–7) responses per patient, with de novo induction in most cases and post-infusion detection in blood and tumor; polyclonal and polyfunctional cytotoxic T cells	[208]
NCT03412877	Neoantigen-specific TCR-T/colorectal cancer	Phase II (7)	Polyfunctional neoantigen-specific responses (IFNγ, GM-CSF, IL-2, Granzyme B) observed. TCR-transduced cells persisted ≥ 10% at 1 month (n = 5) and > 2 years in a responder	[209]

Chemotherapy and Neoantigens

Chemical compounds that suppress tumor proliferation and induce tumor cytotoxicity are among common first-line options. Chemotherapy is currently the most prevalent treatment in terms of accessibility and availability. DNA damage-inducing agents have long been used as anti-tumor drugs [123]. DNA alkylating agents, platinum agents, and anthracyclines directly induce DNA double strands by chemically crosslinking both strands [124, 125]. Topoisomerase poisons induce single-stranded and double-stranded breaks by inducing cleavage complex persistence [126]. Antimetabolites, which affect the nucleotide pool and nucleotide replacement, are another type of anti-tumor drugs [127, 128]. These compounds induce severe DNA damage in all cell types, including non-tumoral cells, which causes undesired adverse events such as fatigue, mucositis, hair loss, vomiting, and insufficient hematopoiesis with anemia and leukopenia.

DNA damage, including single-stranded DNA breaks and double-stranded DNA breaks are sensed and repaired by the base excision repair, nucleotide excision repair, mismatch repair, homologous recombination repair, and non-homologous end joining pathways, depending on the type of damage [129]. Specifically, single-stranded DNA breaks are repaired by poly(ADP-ribose) polymerase (PARP), whereas double-stranded DNA breaks are repaired via BRCA1/2-mediated HRR with high accuracy or via Ku70/80-mediated NHEJ [123]. Mutations in DNA repair-related genes, such as POLE, MLH1, MSH2/6, and PMS2, are associated with an enhanced immune response [130, 131]. Some cells fail to repair and stop proliferating, which leads to cell death. However, other surviving cells tend to harbor excess mutations as annotated by the mutation signature [132]. Although such mutagenesis does not always improve the immunotherapy response, these mutations could become primary sources of chemotherapy-induced neoantigens, as they can be induced by in vitro and in vivo treatments [133, 134]. PARP inhibitors are expected to show synthetic lethal effects by blocking single-stranded DNA repair, particularly in tumors with defective HRR. Cells harboring HRR-related gene mutations or deficiencies are sensitive to PARP inhibitors, whose efficacy varies depending on the mutation type [135]. This effect was also augmented by targeting lncRNA MALAT1 [136]. As PARP inhibitors potentiate further DNA damage, previous research has explored their synergy with ICB therapy. Treatment with PARP inhibitors induces tumoral PD-L1 expression [137]. A preclinical study showed a promising synergistic effect of the combination of a PARP inhibitor and ICB [138, 139]. Several clinical trials have shown the clinical benefits of these combinations in several types of tumors, especially in those with BRCA1/2 alterations [140, 141]. In contrast, other trials showed comparable clinical responses or limited efficacy compared with those of monotherapy, indicating a challenge to versatile application [142, 143]. Although this combination is reasonable and attractive for augmenting the ICB response, further optimization and patient selection by more solid biomarkers are still needed [144]. For this, computational algorithms to detect homologous recombination deficiency are being developed [145]. Additionally, the mechanism of resistance to PARP inhibitors should be carefully determined because PARP inhibitor treatment does not significantly change DNA mutation profiles within a few months in the experimental setting [146].

Molecular Targeted Therapy and Neoantigens

Molecular targeted therapy is a relatively new class of anti-tumor drugs (Fig. 4). As many tumors also acquire driver mutations that confer abnormal growth signaling, these inhibitors are designed to block tumor-specific oncogenic signaling. Compared with genotoxic agents, their action is more tumor selective and does not impair immune cell function [147]. Moreover, several targeted therapies affect the tumor microenvironment and immune cells, enhancing their anti-tumor activity [148–150]. Hence, combination therapy with ICB is expected to enhance the clinical response [151, 152]. However, dual treatments should be carefully administered because of the excessive risk of adverse events in some cases [153]. Although their mechanism of action is not directly associated with DNA mutagenesis, surviving tumor cells also acquire gain-of-function or loss-of-function mutations in specific genes to circumvent and compensate for the original cellular signaling inhibited by targeted therapy. For example, in epidermal growth factor receptor (EGFR) inhibitor resistance, the binding pocket of EGFR inhibitors is mutated to reduce their affinity. To enhance EGFR signaling, cells acquire other mutations in specific exons as well as gene amplification. Moreover, other growth signaling pathways, such as ERK and PI3K/AKT, are augmented by bypassing signaling pathways, such as NRAS, KRAS, BRAF, PTEN, AXL, MET, IGF1R, and HER2 [154]. Although targeted therapy shows a strong tumor-suppressing effect in the early phases of treatment, only a small population that shows persistence eventually acquires a more aggressive behavior, leading to relapse and recurrence. These compensatory pathways have also been applied to other types of targeted therapy resistance because the functions of oncogenic mutations are conserved [155, 156]. A recent study also proposed a unique mechanism involving abnormal splicing, in which a specific mutation in U2AF1 conferred chemoresistance by suppressing global translation and, conversely, inducing an integrated stress response [157].

Fig. 4.

Identification of neoantigens (NeoAg) and therapeutic factors affecting their emergence. Mutations that may give rise to NeoAg were identified through genomic and transcriptomic analyses, followed by computational prediction tools. The final step is the immunological validation of predicted NeoAg. Certain therapies can influence NeoAg. For example, acquired drug resistance to chemotherapy or radiation therapy can induce mutations leading to the emergence of new NeoAg. CTC circulating tumor cell, CTL cytotoxic T lymphocytes, DAMPs damage-associated molecular patterns, EGFR epidermal growth factor receptor, HLA human leukocyte antigen, MS mass spectrometry, RNA-seq RNA sequencing, TKI tyrosine kinase inhibitor, WES whole exome sequencing

As one molecular targeted therapy tends to rapidly generate resistance by introducing another oncogenic driver mutation, combinatorial treatments to block other potential bypassing signaling pathways have been well studied [158–160]. To eradicate tumor cells entirely without conferring any opportunity for recurrence, new combination treatment strategies have been proposed. Combinations of EGFR and SYK inhibitors [161]; EGFR and YAP1 inhibitors [162]; BRAF and HSP90 inhibitors [163]; BRAF, MEK, and CDK inhibitors [164]; and HER2, AKT, and PI3K inhibitors [165] are typical examples. Despite these multiple targeting efforts, tumors still become drug resistant and relapse. Notably, the combination of immunotherapy and targeted therapy improves patient survival [166]. A recent phase III clinical trial of patients with BRAFV600-mutant metastatic melanoma investigated combinations of nivolumab/ipilimumab and dabrafenib/trametinib, reporting a superior response to ICB as first-line treatment compared with targeted therapy and suggested the better treatment order of ICB followed by targeted therapy [167]. Another phase III clinical trial of patients with non-small cell lung cancer with EGFR tyrosine kinase inhibitor-resistant tumors reported that ivonescimab, a bispecific antibody against PD-1 and VEGF, plus chemotherapy showed significantly improved progression-free survival compared with chemotherapy alone [168]. As neoantigen vaccines can evoke direct immune response compared with ICB, this evidence encourages the potential combined usage of neoantigen vaccines and targeted therapies even at the disease progress stage after resistance. Additionally, genomic analysis focusing on patients with EGFR, BRAF^V600E, KRAS^G12C, and HER2 showed that chromosomal instability and enhanced mutagenesis rates result in shorter responses to targeted therapy [169]. To overcome this limitation and the variable clinical response across patients, assuming that tumor-specific genomic alterations can be targeted by the immune response, the application of neoantigen-based therapy derived from such an excessive genomic alteration in a chromosomally unstable tumor holds huge potential. Furthermore, predictive biomarkers will enable more realistic application of this approach.

ARID1A-Deficient Tumors Acquire Immunogenic Neoantigens During the Development of Resistance to Targeted Therapy

The mammalian switch/sucrose non-fermentable (SWI/SNF) chromatin remodeling complex is an attractive tumor target for immunotherapy [170]. The SWI/SNF complex is classified into BAF, PBAF, and GBAF complexes based on subunit composition. These complexes are involved in various physiological processes by regulating the structural accessibility of chromatin. In addition to epigenetic modification-mediated gene expression, DNA repair and post-translational modifications are also regulated because of their global functions [171, 172]. Subunits of the SWI/SNF complex are frequently mutated in tumors, suggesting their role as tumor suppressors. A multivariate Cox regression analysis revealed that patients with SWI/SNF complex-related gene alterations showed a better clinical response after ICB treatment in many cohorts [173]. In contrast, not all patients with SWI/SNF mutations showed a better clinical response to ICB therapy, indicating that it is not a robust predictive biomarker [174].

ARID1A is a frequently mutated SWI/SNF complex component and a potential predictive biomarker for ICB therapy [175, 176]. Importantly, ARID1A mutations resulted in poor survival without ICB treatment, but this tendency was reversed by ICB treatment [177]. This indicates a difference in clinical efficacy between conventional therapy and ICB treatment. To address this issue, the function of ARID1A in tumors was investigated and characterized. A previous study has shown that ARID1A directly interacts with MSH2, a mismatch repair enzyme, and facilitates MMR. Hence, ARID1A deficiency resulted in microsatellite instability, leading to an ICB response [178]. Reportedly, ARID1A facilitates the recruitment of DNA repair enzymes by keeping open the chromatin structure in the damaged DNA locus. In the absence of ARID1A, insufficient repair results in the accumulation of micronuclei leading to anti-tumor inflammation mediated by the activation of the cGAS-STING pathway (Fig. 5) [179]. Moreover, ARID1A acts as a tumor suppressor by regulating the tumor microenvironment. Loss of ARID1A promotes polymorphonuclear myeloid-derived suppressor cell infiltration into tumors via excessive nuclear factor-κB activation. [180].

Fig. 5.

Therapy-induced neoantigen expression in DNA repair-deficient tumors. Following chemotherapy, DNA repair genes, including those involved in nucleotide excision repair, mismatch repair, base excision repair, homologous recombination, and nonhomologous end-joining may be impaired. Alterations in these pathways, together with mutations in genes regulating tumor cell death, oncogenic transformation, and hypermutation during tumor progression, contribute to the development of resistance to cancer therapies and immune checkpoint inhibitors. ICB immune checkpoint blockade, PARP poly(ADP-ribose) polymerase, SWI/SNF switch/sucrose non-fermentable

To stratify ARID1A deficiency in more detail, we attempted to consolidate the concept of drug resistance to targeted therapy. Our study compared a BRAF inhibitor-resistant BRAF^V600E melanoma model with and without ARID1A expression. Comparison of the genomic status revealed that BRAF inhibitor-resistant melanoma, in the absence of ARID1A, gained excessive passenger mutations in addition to common resistance-associated mutations such as NRAS and KRAS. These accumulated mutations conferred higher immunogenicity in ARID1-deficient melanoma through drug resistance-specific neoantigen presentation compared with that of ARID1A-deficient drug-sensitive and ARID1A-intact drug-resistant and drug-sensitive cells (Fig. 5) [181]. Although BRAF inhibitor resistance was observed in vitro, the results indicated that drug resistance by itself unfortunately results in a poor response to immunotherapy. However, by acquiring drug resistance, ARID1A-mutated tumors may become responsive to immunotherapy. Based on these observations, patients with ARID1A-deficient tumors can benefit from combination therapy with targeted therapy and ICB, as well as neoantigen vaccines. Furthermore, as targeted therapy resistance arises from common driver mutations, as discussed above, prophylactic treatment with such a shared neoantigen vaccine would synergize to prevent any drug-resistant tumors. To address this point, a comprehensive analysis of drug resistance in vivo should be carefully conducted with a particular focus on the immune response and associated mutagenesis, based on the concept of immune editing. ARID2 deficiency also evokes cytotoxic T-cell responses and confers ICB sensitivity [182]. Recently, chemical compounds designed to inhibit or degrade the SWI/SNF complex have been shown to exert tumor-suppressive effects by dampening intrinsic oncogenic gene expression programs [183, 184]. Although the effect of these drugs on the DNA repair machinery of the SWI/SNF complex has not been sufficiently clarified, co-treatment with SWI/SNF inhibitors and ICB is an interesting strategy because ARID1A and the SWI/SNF complex are closely linked to DNA repair in a direct and indirect manner [178, 185, 186]. Caution should be exercised regarding the risk of forced mutagenesis, leading to more aggressive intractable tumors.

Radiation Therapy and Neoantigens

Ionizing radiation therapy (RT) is another treatment option for tumors. Low linear energy transfer radiation, such as X-rays and gamma rays, generates reactive oxygen species and indirectly induces DNA damage. In contrast, high linear energy transfer radiation, such as electron, proton, and carbon ion beams, directly destabilizes the DNA structure [187]. Although several studies have shown that RT treatment induces an immunosuppressive environment by accumulating regulatory T cells and monocyte-derived myeloid-derived suppressor cells, especially in RT-resistant tumors [188, 189], previous studies have demonstrated that the RT combination improves the response to ICB therapy [190, 191]. Experimentally, neoantigens that emerge after in vitro exposure to RT are involved in tumor rejection and can be used as vaccines [192, 193]. Considering that RT induces strong DNA damage, neoantigen vaccines could be effective even in RT-resistant tumors, especially in those with SWI/SNF complex mutations [186, 194, 195].

Clinical Trials of Personalized Neoantigen Vaccines: Recent Advances and Outcomes

The first-in-human personalized neoantigen vaccine trial (NCT01970358) in patients with melanoma, conducted at the Dana-Farber/Harvard Cancer Center, evaluated NeoVax, a long-peptide vaccine targeting tumor-specific neoantigens, and provided foundational evidence that such individualized vaccines could safely elicit durable and polyclonal T-cell responses in patients with cancer [45, 196]. In this phase I study, up to 20 neoantigen peptides, each 15–30 amino acids in length, were selected per patient based on whole-exome and RNA sequencing of tumor and normal tissues, and predicted for high-affinity HLA class I binding. These peptides were grouped into four pools and co-administered subcutaneously with the TLR3/MDA5 agonist poly-ICLC. Among the ten high-risk patients with cutaneous melanoma enrolled, six completed the vaccine regimen; vaccination was well tolerated, with only mild flu-like symptoms, injection-site reactions, and fatigue reported. Immunomonitoring revealed a robust induction of de novo neoantigen-specific CD4⁺ and CD8⁺ T-cell responses, with a predominance of CD4⁺ T-cell responses, as measured by ELISpot and intracellular cytokine staining. At a median follow-up of 25 months, four patients remained disease free, and two experienced recurrences but subsequently achieved complete remission following anti-PD-1 therapy (pembrolizumab), which coincided with the emergence of new vaccine-specific T-cell responses [45]. These findings suggest that the combination protocol with ICB enhances T-cell responsiveness and contributes to improved therapeutic efficacy.

With a longer follow-up in this cohort (median, 55 months), all patients remained alive, and six showed no evidence of disease [196]. Longitudinal profiling demonstrated that vaccine-induced CD4⁺ T cells underwent phenotypic maturation from naive-like to cytotoxic, and ultimately memory-like states, accompanied by the development of a diverse and durable TCR repertoire. Vaccine-primed responses persisted for up to 4.5 years. In some cases, pembrolizumab treatment further broadens the TCR repertoire, enhances the durability of T-cell responses, and promotes epitope spreading [196]. Taken together, the molecular analyses and clinical outcome data demonstrate the feasibility and safety of personalized neoantigen vaccines for melanoma. Furthermore, they validate the long-term immunogenicity of neoantigens derived from somatic mutations, supporting their integration into adjuvant treatment strategies and combination therapies with ICIs.

Autogene cevumeran, an individualized neoantigen vaccine formulated with uridine messenger RNA (mRNA)-lipoplex nanoparticles, has emerged as a promising strategy for eliciting durable and polyfunctional T-cell responses against tumor-specific mutations, particularly in pancreatic ductal adenocarcinoma (PDAC), which is historically resistant to immunotherapy. Notably, even in PDAC, which is generally characterized by low PD-L1 expression, vaccine-targetable neoantigens were successfully identified. This finding aligns with the results of a large-scale analysis by Yarchoan et al. [197], which demonstrated that TMB and PD-L1 expression are independent biomarkers and do not necessarily correlate. It supports the notion that neoantigen identification is feasible regardless of PD-L1 expression status.

In a phase I trial (NCT04161755) involving patients with resected PDAC, a treatment regimen combining atezolizumab (anti-PD-L1), Autogene cevumeran (targeting up to 20 neoantigens per patient), and modified FOLFIRINOX chemotherapy demonstrated its feasibility, tolerability, and robust immunogenicity. Neoantigen-specific CD8⁺ T-cell responses were induced in 8 of 16 patients (50%), with responses targeting multiple epitopes in half of these individuals. Using high-sensitivity tracking methods of the TCR repertoire, such as CloneTrack, somatic mutation or short indel-derived neoantigen-derived vaccine-expanded T cells have been shown to reach up to 10% of circulating T cells, re-expand after booster vaccination, and persist as long-lived polyfunctional effector cells. Importantly, these T-cell responses correlated with improved clinical outcomes; at the 18-month median follow-up, recurrence-free survival was significantly prolonged in responders (not reached) compared with that in non-responders (13.4 months; P = 0.003) [198]. These results clearly indicate that the magnitude of neoantigen-specific T-cell responses is closely associated with clinical efficacy. An updated analysis with a median 3.2-year follow-up confirmed this benefit (P = 0.007) and revealed that vaccine-induced T-cell clones exhibited multi-year persistence, with an average estimated lifespan of 7.7 years—some exceeding the expected lifetime of the host—and assumed a cytotoxic, tissue-resident memory-like phenotype up to 3 years post-vaccination [199]. In a separate multi-center phase I study (NCT03289962) involving 213 pretreated patients with advanced solid tumors, Autogene cevumeran, either as monotherapy or in combination with atezolizumab, was found to be safe and immunogenic, inducing polyepitopic CD4⁺ and CD8⁺ T-cell responses in 71% of patients. These responses were mostly undetectable at baseline, persisted for up to 23 months, and were detectable in both peripheral blood and tumor tissue, where they constituted up to 7.2% of tumor-infiltrating lymphocytes. Objective responses have been observed in patients with immunotherapy-refractory disease [200]. Together, these findings highlight the feasibility, durability, and clinical potential of individualized mRNA-lipoplex neoantigen vaccines, such as Autogene cevumeran, particularly for immunologically cold tumors, such as PDAC.

Messenger RNA-4157 (also known as V940) is a personalized mRNA-based neoantigen vaccine that has demonstrated safety, immunogenicity, and clinical benefits in combination with pembrolizumab. In an earlier phase I study (NCT03313778), mRNA-4157 was evaluated in resected non-small cell lung cancer (monotherapy) and melanoma (combination cohort), and demonstrated safety and strong immunogenicity. Neoantigen-specific T-cell responses, both de novo and pre-existing, were induced and sustained for up to 30 weeks post-treatment, with evidence of cytotoxic T-cell proliferation [201]. In a phase II trial (KEYNOTE-942, NCT03897881) involving patients with resected high-risk stage IIIB-IV cutaneous melanoma, combination therapy reduced the risk of recurrence or death by 49% compared with pembrolizumab treatment alone (hazard ratio 0.510, P = 0.019), with a 2.5-year recurrence-free survival rate of 74.8% versus 55.6%. Distant metastasis-free survival also significantly improved (hazard ratio 0.384, P = 0.0154) and a favorable trend was observed in overall survival. A recurrence-free survival benefit was observed across several subgroups, including those with a high TMB and PD-L1 positivity, and the safety profile remained favorable without increased immune-related adverse events [202]. These findings support the potential use of mRNA-4157 as a safe and effective adjuvant immunotherapy for solid tumors. The incorporation of ICB into treatment protocols markedly enhances T-cell responses and substantially improves clinical outcomes.

Neoantigen-based clinical trials have expanded the scope of cancer immunotherapy. Palmer et al. developed a personalized neoantigen vaccine using a heterologous prime-boost regimen with chimpanzee adenovirus (ChAd68) and self-amplifying mRNA and administered it in combination with multiple ICIs. In patients with advanced metastatic solid tumors, the vaccine induced durable neoantigen-specific CD8⁺ T-cell responses and demonstrated a potential overall survival benefit in patients with microsatellite-stable colorectal cancer [203]. Braun et al. reported that a personalized, somatic, mutation-derived long-peptide vaccine induced durable memory T-cell responses and complete recurrence-free survival at a 40-month median follow-up in nine patients with high-risk renal cell carcinoma [204]. Saxena et al. evaluated the somatic, mutation-derived, personalized peptide vaccine PGV001 with atezolizumab in urothelial carcinoma and reported neoantigen-specific T-cell responses in all patients and objective responses in metastatic cases [205]. In a phase I/II study targeting shared neoantigens in solid tumors, Rappaport et al. used a chimpanzee adenovirus prime and a self-amplifying mRNA-boost vaccine encoding KRAS and TP53 driver mutations in combination with ICIs; although the clinical efficacy was limited, strong CD8⁺ T-cell responses against TP53 neoantigens were detected, underscoring the challenge of immunodominance in shared antigen vaccines [206].

Clinical Trials of Personalized Neoantigen-Responding TCR-Gene Transduced Adoptive T-Cell Therapy: Recent Advances and Outcomes

In addition to active immunization strategies, such as vaccines, adoptive immunotherapy using TCR-T cells engineered with neoantigen-specific TCRs represents an important modality that directly exploits neoantigen specificity. From a scientific perspective, TCR-T therapy plays a pivotal role as a “proof of concept,” demonstrating not only the immunogenicity of identified neoantigens but also their potential to drive tumor regression. Although technical complexity, manufacturing costs, and the need for individualized production currently limit broad clinical application, the potential of this approach remains promising. In fact, objective responses in patients with advanced solid tumors have already been reported in multiple clinical trials targeting both shared and personalized neoantigens. For example, Tran et al. demonstrated tumor regression using TCR-T cells targeting the KRAS G12D mutation [59]; Kim et al. validated the safety and immunogenicity of TCR-T cells targeting TP53 [61]; and Chandran et al. did so for PIK3CA [65], all of which are major driver mutations. Yonezawa et al. also established an automated manufacturing process for TCR-T cells targeting NPM1 mutations, indicating the potential for expanded clinical applications, including in hematological malignancies [71].

On the personalized medicine front, Foy et al. conducted a first-in-human phase I trial in which autologous T cells were engineered with up to three neoantigen-specific TCRs using non-viral CRISPR-Cas9 precision editing, and infused into patients with refractory solid tumors. Although the therapy demonstrated successful T-cell trafficking and intratumoral persistence, clinical responses were limited, with stable disease observed in 5 of 16 patients [207]. Borgers et al. demonstrated the feasibility of BNT221, a personalized neoantigen-specific T-cell therapy, in ICB-refractory melanoma, achieving disease stabilization in most patients and inducing highly specific, polyfunctional CD8⁺ and CD4⁺ T-cell responses [208]. Furthermore, Parkhurst et al. reported that personalized TCR-T therapy induced objective partial responses in three patients with metastatic colorectal cancer, a population typically resistant to immunotherapy, strongly supporting the promise of this approach in epithelial solid tumors [209]. These diverse approaches reinforce the versatility of neoantigen-targeting strategies across tumor types and treatment modalities.

Outstanding Issues for Further Development

A diverse array of neoantigen sources have also been identified. To maximize clinical impact, optimal selection of neoantigens and design of immune strategies to exploit them are essential. In particular, we need a prospective evaluation of treatment timing (neoadjuvant vs adjuvant vs metastatic settings); relationship with tumor debulking; sequence and pairing with radiotherapy, chemotherapy, or targeted agents; and predictability of therapy-induced neoantigen generation. Additional priorities include generalizing prediction algorithms across HLA diversity, integrating immunopeptidomics (mass spectrometry) with computational prediction. Of note, protein language models can effectively improve prediction performance compared with conventional prediction models [210, 211]. Moreover, improving manufacturing throughput and turnaround time (including Good Manufacturing Practice and quality testing), ensuring safety (off-target autoreactivity/epitope spreading), and standardizing immune monitoring (e.g. TCR tracking and liquid biopsy) are critical in real clinical settings.

Conclusions and Perspective

The concept that non-viral tumors harbor tumor-specific antigens dates back to the early 20th century and has evolved with advances in genomic analysis, leading to the identification of personalized neoantigens and their clinical testing. To expand the promise of cancer immunotherapy, optimizing quality-based antigen selection (considering clonality, expression, presentation, and MHC class I/II), determining appropriate treatment timing and combinations, establishing rapid and reproducible manufacturing processes, and implementing standardized immune monitoring, are essential.

Recent clinical trials have demonstrated that neoantigen-specific T-cell responses correlate clearly with therapeutic efficacy. Furthermore, even when personalized neoantigen vaccines alone are insufficient, combination with ICB significantly enhances T-cell responses and clinical outcomes. These findings indicate that neoantigen immunogenicity is often independent of PD-L1 expression, supporting antigen discovery even in low-PD-L1-expressing tumors such as PDAC. In addition, TCR-T cells that target neoantigens serve as critical proof-of-concept tools to validate the direct anti-tumor potential of identified neoantigens. As more evidence accumulates comparing TCR-T and neoantigen vaccines, optimal strategies will emerge based on tumor type and patient context. With these components in place, neoantigen-targeted therapies are poised to become a central pillar of personalized cancer immunotherapy — from the adjuvant setting to advanced treatment-refractory disease.

Declarations

Funding

This work was supported by the Japan Society for the Promotion of Science KAKENHI Grant number JP 25K10457.

Conflicts of interest/competing interests

Masahiro Okada, Satoru Yamasaki, Go R. Sato, Kanako Shimizu, and Shin-ichiro Fujii have no conflicts of interest that are directly relevant to the content of this article.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and material

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.

Code availability

Not applicable.

Authors’ contributions

MO, SY, and SF: writing original draft, reviewing, editing, and visualization. KS and GS: reviewing, editing, and visualization. SF: conceptualization, supervision, writing original draft, reviewing, editing, and visualization. All authors have read and approved the final version of the article