One of the major challenges in the application of T cell-based immunotherapies to solid tumors is the pronounced intra- and inter-tumoral heterogeneity (15–17). Hematologic malignancies and solid tumors differ markedly in both the degree and nature of heterogeneity. The success of CD19 CAR-T therapy in B-cell lymphoma is not solely due to relatively uniform antigen expression, as CD19 is not a tumor-specific antigen, but rather reflects the clinical tolerability of complete B-cell depletion (18). Critically, this reliance on single, uniformly expressed antigens, a core design assumption of most CAR-T therapies, does not apply to solid tumors, and this misalignment directly contributes to clinical failures of CAR-T in solid tumors (19). In contrast, even when lineage- or tissue-restricted antigens in solid tumors show comparatively homogeneous expression, on-target/off-tumor toxicity affecting essential normal tissues often limits safe targeting, as exemplified by early HER2 CAR-T trials in one colon cancer patient (20). As Chen et al. highlighted, preclinical models commonly use antigen-homogeneous cell lines, masking the real-world challenge of clonal escape due to antigen heterogeneity in human solid tumors, which leads to overestimation of CAR-T efficacy (19). Moreover, solid tumors exhibit substantially higher functional heterogeneity driven by spatial constraints and diverse tumor microenvironmental niches, further complicating effective immune targeting (21–23). This heterogeneity complicates the identification of universally expressed, tumor-specific antigens that can be safely and effectively targeted.

To overcome this challenge, tumor-specific antigens (TSAs), particularly neoantigens arising from somatic mutations, have become central to personalized T cell therapies (24). Unlike shared tumor-associated antigens (TAAs), TSAs are exclusively expressed by tumor cells, enabling precise immune targeting with minimal off-target toxicity. By contrast, TAAs are not exclusive to tumor cells but are abnormally overexpressed in tumors while being weakly expressed in normal tissues (e.g., carcinoembryonic antigen, CEA, in colorectal cancer), which may lead to immune tolerance or potential off-target autoimmune reactions due to their expression in normal cells. TSAs, on the other hand, are derived from tumor-specific genetic abnormalities (e.g., KRAS G12D mutation in pancreatic cancer) and thus evade central immune tolerance, inducing more robust and specific anti-tumor immune responses. T cell responses to TSAs are considered a “final common pathway” of effective cancer immunotherapy (25), forming the conceptual basis for individualized approaches such as TCR-T therapy and neoantigen vaccines (12, 26). Therefore, robust and high-throughput strategies for TSA identification are essential. In recent years, integrative pipelines combining genomic and transcriptomic sequencing with functional validation assays have significantly advanced the discovery of immunogenic neoantigens. For instance, tandem minigene libraries, MHC multimer-based sorting, and single-cell T cell receptor (TCR) profiling have enabled the sensitive detection of neoantigen-reactive T cells (27, 28). These platforms are further enhanced by computational tools that prioritize candidate peptides based on MHC binding, antigen processing, and immunogenicity (29). However, despite the large number of non-synonymous mutations found in tumors, only a small fraction gives rise to genuine immunogenic neoantigens, underscoring the need to integrate multi-omics prediction, immunogenicity scoring, and in vivo/in vitro validation (30). This scarcity of immunogenic neoantigens further exacerbates the limitation of single-target CAR-T strategies, as even personalized neoantigen-targeted CAR-T may fail to cover all heterogeneous tumor clones. Resources such as pan-cancer driver mutation databases also offer guidance in identifying clinically meaningful TSAs (31).

Moreover, the sources of TSAs are increasingly recognized as diverse. Clonal somatic mutations, shared by all tumor cells, are preferred due to their stability and therapeutic breadth (32). Yet, noncanonical sources such as aberrant translation products and gene fusions greatly expand the neoantigen landscape (33, 34). TSAs derived from intracellular proteins presented by MHC class I pathways are also gaining attention as “cleaner” targets with less risk of on-target, off-tumor toxicity (35). Additionally, the development of neoantigen-specific TCRs is central to personalized T cell therapies. Multiple strategies have been proposed to identify TCRs with high specificity and affinity. One approach leverages clonal expansion within tumors, as Pasetto et al. showed that highly frequent TCR clonotypes in fresh tumor tissue often reflect antigen reactivity, enabling antigen-agnostic discovery (36). Meanwhile, Strønen et al. demonstrated that naïve T cell repertoires from healthy donors can recognize patient-specific neoantigens, offering an alternative source when autologous TILs are dysfunctional or limited (37). Direct TIL screening remains effective. Tan et al. isolated tumor-reactive TCRs from esophageal cancer TILs and confirmed their in vitro and in vivo function (38). However, as Garcia-Garijo et al. noted, only a small fraction of mutations generate true neoantigens, necessitating integrative pipelines that combine computational prediction, immunogenicity scoring, and functional assays, especially for low–mutational burden tumors (30).

TCR source also affects safety and efficacy. Inderberg and Walchli suggested using wild-type TCRs from long-term responders, particularly CD4⁺MHC-II–restricted clones, to balance potency and safety (39). Ahmadzadeh et al. further showed that tumor-infiltrating regulatory T cells (Tregs) harbor unique neoantigen-specific TCRs, offering underexplored sources. Finally, clinical context matters, as TCR-T may suit advanced disease, while vaccines are more appropriate for early-stage tumors (40). Together, these findings shape a multifaceted framework for rational TCR design. Together, these advances underscore the central role of TSAs in overcoming tumor heterogeneity and tailoring immunotherapy to individual patients. Continued progress in multi-omics profiling, antigen processing models, and receptor optimization will further empower TSA-directed strategies to achieve greater precision, efficacy, and durability in the treatment of solid tumors.

The clinical success of T cell–based immunotherapies, including immune checkpoint inhibitors, TCR-T, and CAR-T. critically depends on the efficient infiltration and activation of effector T cells within TME (41). However, many solid tumors exhibit an immune-excluded or “cold” phenotype, characterized by low T cell density and limited access to tumor parenchyma, which compromises therapeutic efficacy (42). Notably, CAR-T engineering has long focused on enhancing effector function (e.g., cytokine secretion, cytotoxicity) while underestimating trafficking and penetration barriers as another key design oversight that drives the gap between CAR-T engineering advances and clinical efficacy in solid tumors (19).

One of the primary barriers to T cell entry is the abnormal structure and function of tumor vasculature. Tumor-associated vessels are often chaotic, immature, and lack hierarchical organization, resulting in poor perfusion, increased interstitial pressure, and erratic delivery of immune cells. Continuous VEGF-driven angiogenesis contributes to this abnormality and promotes the formation of leaky, dilated, and tortuous vessels with disrupted endothelial junctions (43). Furthermore, endothelial cells (ECs) in tumors commonly exhibit a state of “endothelial anergy,” wherein they downregulate adhesion molecules such as VCAM-1 and ICAM-1 and fail to express chemokines like CXCL9 and CXCL10 needed to recruit effector T cells (44). This hyporesponsiveness to inflammatory cues is driven in part by constitutive pro-angiogenic signaling and hypoxia-induced HIF-1α expression (44). However, the role of adhesion molecules can be complex, for instance, in renal cell carcinoma, tumor cells have been shown to escape T cell immunity by paradoxically overexpressing VCAM-1, which normally mediates leukocyte extravasation but, in this context may act to sequester or inhibit infiltrating T cells (45, 46). Strategies that stabilize VE-Cadherin or modulate β-catenin signaling in ECs have been shown to enhance selective T cell transmigration while reducing neutrophil infiltration, suggesting that vascular normalization may selectively benefit antitumor immunity (47). Yet, the clinical benefit of such strategies remains limited because CAR-T design does not integrate trafficking-enhancing modifications (e.g., chemokine receptor engineering) to complement vascular normalization, reflecting a disconnect between engineering priorities and clinical requirements. Beyond vascular obstacles, the extracellular matrix (ECM) serves as a physical barrier that impedes T cell motility and tissue penetration (48). Many tumors exhibit a dense and fibrotic ECM composed primarily of collagen and secreted by activated cancer-associated fibroblasts (CAFs) (49). High collagen density not only blocks T cell infiltration but also modulates immune responses by engaging inhibitory receptors or restricting cytokine diffusion (43, 49). Tumor cells may also form tightly packed nests via increased intercellular adhesion, and lack of lymphatic drainage further isolates the tumor core from immune surveillance.

Chemokine dysregulation further contributes to immune exclusion. CAF-derived CXCL12 can form gradients that sequester T cells in stromal regions, preventing their access to malignant cells (50). Additionally, tumors lacking Batf3-lineage CD103⁺dendritic cells fail to produce CXCL9 and CXCL10, which are essential for recruiting cytotoxic T cells into tumor nests and thereby generating a non-T cell-inflamed microenvironment that resists both endogenous immune responses and adoptive T cell therapies (51). Beyond dendritic cells, tumor-intrinsic activation of the WNT/β-catenin pathway has also been shown to suppress CCL4 expression, which further inhibits Batf3⁺dendritic cell recruitment and impairs initial T cell priming (52). In parallel, tumor-associated macrophages (TAMs) also play a pivotal role in chemokine-mediated immune exclusion. Under hypoxic and fibrotic conditions, TAMs adopt an immunosuppressive phenotype and secrete CCL2 and CCL5, which recruit suppressive myeloid cells and divert effector T cells away from the tumor core (53, 54). Importantly, when defining TAMs based on immunosuppressive functions, emerging evidence suggests that they cannot be adequately explained by a simplistic M1/M2 dichotomy, but rather reflect a spectrum of context-dependent functional states shaped by metabolic stress, spatial localization, and microenvironmental cues (55). Furthermore, TAM-derived VEGF and MMP9 promote aberrant angiogenesis, indirectly disrupting T cell infiltration by contributing to vascular disorganization (56), consistent with recent efforts to redefine TAM functional classification beyond the traditional M1/M2 framework (55). Compounding this chemokine dysfunction, recent evidence from breast cancer models indicates that the migratory deficiency may also be intrinsic to antitumor T cells themselves. Specifically, cytotoxic CD8⁺T cells and Th1 cells were found to exhibit impaired chemotaxis relative to Tregs and Th2 cells, a defect mechanistically linked to IFN–STAT1–driven upregulation of regulator of G protein signaling 1 (RGS1). RGS1 acts to suppress chemokine receptor signaling, thereby limiting effector T cell trafficking to the tumor site. Mechanistically, type I interferons, such as IFN-β, have been shown to robustly induce the expression of regulator of G protein signaling 1 (RGS1), which functions as a negative regulator to inhibit G-protein signaling (57). This IFN–driven upregulation of RGS1 acts to suppress chemokine receptor signaling, thereby limiting effector T cell trafficking to the tumor site. Notably, genetic or pharmacologic targeting of RGS1 restored intratumoral migration and significantly improved therapeutic responses (58). Beyond physical and chemokine barriers, the metabolic deregulation of the TME constitutes a formidable chemical barrier, particularly characterized by acidosis. Tumor cells frequently exhibit elevated aerobic glycolysis, leading to the accumulation of lactate and protons, which creates a low-pH microenvironment. This acidic milieu significantly impairs T cell motility, proliferation, and cytotoxic function, effectively acting as a mechanism of immune evasion (59, 60). A recent study highlighted the critical association between pH regulation and the immunological state of the tumor, demonstrating that bicarbonate transporters (such as SLC4A7) on tumor cells regulate intracellular pH while acidifying the extracellular space, thereby maintaining immunological dormancy and excluding T cell infiltration (61). Strategies targeting these pH-regulating mechanisms may therefore restore a favorable environment for anti-tumor immunity (60).

Multiple non-T cell populations within the TME contribute to immunosuppression and impede effective T cell responses. Tumor-associated natural killer (NK) cells, despite their cytotoxic potential, often adopt a dysfunctional state. Tang et al. identified a subset of RGS1⁺NK cells enriched across cancers, characterized by impaired antitumor function and resistance to immunotherapy, potentially regulated by LAMP3⁺dendritic cells (62). Expanding on the mechanism of TuNK-mediated immunosuppression, a recent study demonstrated that TuNK cells can actively drive tumor immune tolerance (63). Specifically, TuNK cells were found to promote the accumulation and suppressive function of myeloid-derived suppressor cells (MDSCs) through the IL-6/STAT3 signaling axis, thereby facilitating tumor progression (63). Similarly, B cells can exert either immunosuppressive or anti-tumor functions depending on the tumor microenvironment, with regulatory B-cell subsets contributing to immune suppression in certain contexts. Ma et al. revealed that extrafollicular ITGAX⁺B cells are associated with poor prognosis and immunotherapy resistance by promoting a T cell-suppressive program through glutamine-derived metabolic–epigenetic crosstalk (64). In early-stage lung adenocarcinoma, IgA⁺plasma cells were found to co-localize with Tregs, suggesting a collaborative role in dampening antitumor immunity (65). CAFs also participate in immune evasion; Krishnamurty et al. demonstrated that LRRC15⁺myofibroblasts, induced by TGF-β signaling, suppress CD8⁺T cell function and impair responses to PD-L1 blockade (66). Additionally, Wang et al. identified CD300ld on neutrophils as a key mediator of tumor-induced immunosuppression by enhancing PMN-MDSC recruitment and activity via the STAT3-S100A8/A9 axis, ultimately restricting T cell activation (67).

Beyond neutrophils, Myeloid-derived suppressor cells (MDSCs) represent another formidable barrier to effective T cell immunotherapy within the solid tumor microenvironment. These pathologically activated immature myeloid cells accumulate during tumor progression and exert potent immunosuppressive effects that directly impede antitumor immunity (68). MDSCs inhibit T cell proliferation and cytotoxic function through multiple mechanisms, such as the depletion of essential amino acids (e.g., L-arginine via Arginase 1 and iNOS expression) and the production of reactive oxygen and nitrogen species (ROS and RNS) that disrupt T cell receptor signaling (68). Furthermore, MDSCs can facilitate the recruitment of regulatory T cells (Tregs) and promote tumor angiogenesis (69), collectively creating a hostile environment that limits the infiltration and efficacy of therapeutic T cells (70). Notably, this suppressive effect of MDSCs can override the enhanced function of engineered CAR-T cells, as they target the same signaling pathways modified by CAR engineering— a key mechanism underlying CAR-T clinical failure in solid tumors that is often underappreciated in preclinical development.

MDSCs are broadly classified into two major subsets based on their phenotypic and morphological features: granulocytic or polymorphonuclear MDSCs (G-MDSCs or PMN-MDSCs) and monocytic MDSCs (M-MDSCs) (68, 71). While both subsets contribute to T cell paralysis, they utilize different dominant suppressive mechanisms.

G-MDSCs, which are morphologically similar to neutrophils, primarily suppress T cells through the production of reactive oxygen species (ROS) mediated by increased NADPH oxidase (NOX2) activity. The accumulation of ROS creates an oxidative environment that induces the loss of the T cell receptor (TCR) -chain, thereby impairing T cell signaling and proliferation (68). In contrast, M-MDSCs, resembling monocytes, exert their suppressive effects predominantly through the upregulation of inducible nitric oxide synthase (iNOS) and Arginase 1 (ARG1) (68). This leads to the generation of nitric oxide (NO) and downstream reactive nitrogen species (RNS).

Critically, the interplay between ROS (from G-MDSCs) and NO (from M-MDSCs) can generate peroxynitrite (PNT), a highly reactive nitrogen species. PNT causes the nitration of tyrosine residues in the TCR-CD8 complex, interfering with the conformational flexibility required for antigen recognition and causing T cells to become unresponsive to antigen-specific stimulation (72). Collectively, these oxidative and nitrosative stress pathways orchestrated by M-MDSCs and G-MDSCs form a rigid metabolic barrier that limits the efficacy of adoptive T cell therapies.

As summarized in Figure 2, these vascular, stromal, and immunologic mechanisms cooperate to exclude or inactivate effector T cells, forming a multi-layered defense that protects tumors from immune attack. Overcoming these barriers requires rationally designed combination therapies that normalize blood vessels, modulate stromal architecture, reinvigorate T cell metabolism, and block immunosuppressive circuits. Recent preclinical and clinical data suggest that targeting these pathways in concert may convert “cold” tumors into “hot” ones and improve responses to T cell–based therapies across diverse cancer types. Specifically, strategies to facilitate this conversion focus on three key aspects: priming the immune system, normalizing the tumor stroma, and removing immunosuppressive brakes. Approaches such as radiotherapy, photothermal therapy (PTT), and oncolytic viruses can induce immunogenic cell death (ICD) and release tumor antigens, thereby recruiting innate immune cells. Concurrently, vascular normalization agents can repair chaotic tumor vessels to enhance T cell trafficking, while STING agonists trigger type I interferon production to reshape the inflammatory landscape. Integrating these strategies with adoptive cell transfer holds the promise of overcoming the immune-excluded phenotype.

T cell–based immunotherapies in solid tumors are frequently limited by profound T cell dysfunction and functional heterogeneity within the tumor microenvironment (TME). These dysfunctions encompass multiple non-mutually exclusive states, including exhaustion, senescence, and acquisition of suppressive or terminal effector phenotypes, which collectively impair T cell infiltration, persistence, and effector function. Understanding how these states arise and how they associate with therapeutic success or failure is therefore essential for elucidating the major biological obstacles to effective T cell–based therapies. For CAR-T therapy, the failure to account for TME-induced T cell dysfunction represents a critical design flaw that even exhaustion-resistant CAR-T cells engineered to downregulate PD-1 can succumb to redundant suppressive signals in solid tumor TME, leading to clinical failure.

Single-cell RNA sequencing (scRNA-seq), often combined with TCR clonotype tracing, has revolutionized our understanding of tumor-infiltrating T cell heterogeneity, developmental trajectories, and functional states. This high-resolution approach enables systematic dissection of dysfunctional T cell states within tumors and provides mechanistic insights into why T cell–based therapies succeed in some contexts but fail in others.

Across multiple cancer types, including hepatocellular carcinoma (73), non-small cell lung cancer (74), and colorectal cancer (75), scRNA-seq studies have revealed a spectrum of tumor-reactive CD8⁺T cell subsets, notably precursor exhausted T cells (Tpex), tissue-resident memory T cells (Trm), and CXCL13⁺CD8⁺T cells. Specifically, Duhen et al. identified that the co-expression of CD39 and CD103 marks a distinct population of tumor-infiltrating CD8+ T cells (CD8+ TILs) that are highly enriched for tumor reactivity in both primary and metastatic solid tumors (76). Furthermore, other phenotypes correlated with favorable prognosis and better therapeutic responsiveness have been characterized, including TCF1+ stem-like T cells (77, 78), which are essential for sustaining immune responses, and CD69+CD103+ resident memory phenotypes associated with enhanced local cytotoxicity (79, 80). These populations are often clonally expanded and associated with favorable clinical outcomes and improved responses to immune checkpoint blockade (ICB) (75, 81–84). In contrast, enrichment of terminally exhausted CD8⁺T cells or immunosuppressive subtypes such as TGF-β1⁺CD4⁺T cell, CD8⁺Temra, IL1R2⁺Tregs and KIR⁺CD8⁺T cells has been linked to poor prognosis and resistance to therapy (74, 84, 85). Preclinical models often use immunodeficient mice or syngeneic models with mild immunosuppression, failing to recapitulate the full extent of human TME-induced T cell exhaustion, which leads to overoptimistic predictions of CAR-T efficacy (19). Together, these findings highlight how distinct dysfunctional or transitional T cell states shape therapeutic responsiveness in solid tumors.

In the context of CAR-T therapy, single-cell profiling has enabled the identification of highly proliferative, memory-like T cell subsets that are predictive of therapeutic efficacy (86). For instance, TIGIT⁺CD62L^lowCD27⁻ T cells were found to give rise to the most potent CD19-specific CAR-T cells in pediatric B-ALL (87), while CD8-fit T cells with robust metabolic and migratory profiles were strongly associated with clinical responses in large B cell lymphoma (88). These findings support the prioritization of functionally competent T cell subsets during CAR-T manufacturing. Furthermore, IL-10–secreting CAR-T cells demonstrated superior metabolic resilience and durable antitumor activity in solid tumor models, offering a promising avenue to overcome exhaustion (89).

In addition to tumor-infiltrating lymphocytes, peripheral blood T cell phenotypes have emerged as complementary indicators of therapeutic competence. Patients with a higher proportion of naïve-like CD8⁺T cells in circulation exhibited better clinical outcomes in CLDN18.2-targeted CAR-T trials (90). Combined with tumor microenvironmental features such as antigen density and MYC-mediated immunosuppression, single-cell data are shaping a multi-parametric framework for responder stratification and therapy customization. In summary, single-cell technologies provide mechanistic underpinnings for therapeutic success and failure, informing the refinement of TCR-T and CAR-T therapies. By integrating transcriptional, clonal, and phenotypic data, this approach is accelerating the development of more effective, durable, and personalized T cell-based immunotherapies.

Summarily, the efficacy of T cell–based immunotherapies in solid tumors is hindered by a series of interconnected challenges, including the lack of universal tumor-specific antigens, physical and immunologic barriers that restrict T cell infiltration, and functional exhaustion of T cells within the TME. For CAR-T therapy, these challenges act synergistically to override engineering advances, and the core cause lies in the misalignment between CAR-T design assumptions (e.g., single-antigen targeting, neglect of trafficking, underestimation of TME suppression) and the biological reality of solid tumors— a mechanistic disconnect that directly drives clinical failures. From a mechanistic perspective, the inset of Figure 2 illustrates the multiple factors that limit T cell infiltration, activation, and therapeutic efficacy within the solid tumor microenvironment. Tumor vasculature is often abnormal and poorly organized, with downregulation of key adhesion molecules such as VCAM-1 and ICAM-1, restricting T cell trafficking. The extracellular matrix forms dense physical barriers, while insufficient recruitment of CXCL9/CXCL10-producing CD103⁺dendritic cells further impair T cell homing. Tumor antigen heterogeneity complicates the identification of universal neoantigens, reducing the applicability of broadly targeted T cell therapies. In parallel, the tumor microenvironment harbors multiple immunosuppressive cell types, including regulatory T cells (Tregs) (91), ITGAX⁺B cells (64), M2 macrophages (92), CD300ld⁺neutrophils (67), ISG^lo+neutrophils (93), and RGS^hi NK cells (62), which release inhibitory cytokines such as IL-10 and TGF-β, collectively suppressing T cell function, proliferation, and persistence. These interconnected factors create a hostile milieu that severely limits the clinical efficacy of adoptive T cell–based therapies in solid tumors, necessitating innovative strategies to enhance therapeutic responses.

T cell-based vaccines (Tvax) represent an emerging class of cancer immunotherapy platforms that leverage the patient’s own T cells not only as effectors but also as antigen-delivering agents. In contrast to conventional vaccine approaches that rely on dendritic cells or peptide formulations, Tvax utilizes autologous T cells genetically engineered to express tumor-associated antigens and immunostimulatory molecules. These modified T cells traffic to lymphoid tissues, where they promote antigen cross-presentation and amplify endogenous T cell responses, effectively functioning as cellular adjuvants.

A landmark study by Veatch et al. established the principle of Tvax by engineering patient-derived T cells to co-express tumor neoantigens alongside immunomodulatory factors such as CD80, IL-12, and GM-CSF. This approach resulted in robust CD4⁺and CD8⁺T cell activation through dendritic cell engagement and antigen cross-presentation. Notably, the combined expression of IL-12 and GM-CSF led to the most potent T cell priming effects, highlighting the potential of Tvax to bridge innate and adaptive immunity (94, 95). Building upon this concept, recent efforts have focused on equipping therapeutic T cells, including TCR-T and CAR-T cells with additional “vaccine-like” functionalities to sustain their activity in the tumor microenvironment. For example, synthetic cytokine receptors such as engineered IL-21R allow T cells to maintain long-term STAT3 activation, thereby enhancing memory formation and resistance to exhaustion (96). Similarly, co-expression of membrane-tethered IL-15 and IL-21 in engineered T cells has demonstrated broad efficacy in epithelial tumor and neuroblastoma pre-clinical models, establishing a modular strategy for T cell armoring (97). Furthermore, LAG-3Ig, TCR cognate peptides, and tumor neoantigen-engineered T cell vaccines exerted a synergistic boosting effect on TCR-T cells, thereby enhancing their antitumor activity against heterogeneous MC38 colorectal carcinoma in murine models (98).

Chemokine and cytokine engineering strategies have also proven effective in improving the tumor accessibility and persistence of therapeutic T cells. For instance, T cells modified to express the chemokine receptors CCR2b or CXCR2 respond to tumor-derived CCL2 or IL-8, significantly enhancing tumor homing and local cytotoxicity in malignant pleural mesothelioma and melanoma preclinical model (99, 100). Moreover, CAR-T cells co-expressing IL-7 and CCL19 have been shown to reconstitute lymphoid-like niches within the tumor microenvironment, thereby promoting dendritic cell and T cell infiltration, improving CAR-T cell survival, and establishing durable immune memory in B16F10 melanoma cells and Lewis Lung Carcinoma cells mouse models (101, 102).

DCs play a vital role in processing antigens and priming T cells for an effective anti-tumor immunity, leading to tumor regression. Notably, accumulating evidence has demonstrated that CAR-T therapy can actively promote DC recruitment to the tumor microenvironment. Specifically, Ma et al. reported in a landmark study published in Cell that vaccine-boosted CAR-T cells significantly enhanced DC recruitment to tumors, increased tumor antigen uptake by DCs, and further elicited the priming of endogenous anti-tumor T cells (103). This process is critically dependent on IFN-γ secreted by CAR-T cells, and genetic amplification of IFN-γ expression in CAR-T cells can further strengthen this effect, thereby enhancing the control of heterogeneous tumors. Depletion experiments in murine tumor models confirmed that DCs are indispensable for the enhanced anti-tumor efficacy induced by vaccine-boosted CAR-T cells, highlighting the crucial crosstalk between CAR-T therapy and DC recruitment in amplifying anti-tumor immunity.

In addition to enhancing infiltration, strategies have also been developed to counteract the suppressive signals within the tumor milieu. A notable example is the use of inverted cytokine receptors (ICRs), which convert inhibitory signals such as TGF-β into stimulatory signals that promote T cell survival and expansion. One such ICR, TB15, incorporates the extracellular domain of TGF-βRII fused with the intracellular signaling domain of IL-15Rα, allowing engineered CAR-T cells to thrive in TGF-β–rich environments by activating pro-survival pathways (104). These innovative signaling rewiring strategies represent a major advance in enabling engineered T cells to overcome immune suppression and persist in mouse colon cancer.

Furthermore, engineered T cells secreting Flt3L have been shown to expand intratumoral dendritic cells and initiate de novo immune responses, acting as self-adjuvanting cellular vaccines (105). Importantly, clinical observations have also confirmed that mutation-specific CD4⁺T cells can mediate tumor regression and broaden epitope recognition, further supporting the concept of leveraging engineered T cells as therapeutic immunization agents (106). Altogether, Tvax offer a powerful paradigm for immunotherapy, uniquely positioned to combine targeted tumor killing with systemic immune education. By integrating vaccine functions directly into engineered T cells, Tvax platforms provide a promising solution to overcome immunosuppression, enhance immune memory, and reduce antigen escape in solid tumors. This concept stands in contrast to conventional, non–T cell–based vaccines and represents a new frontier in adoptive cellular immunotherapy.

In parallel with T cell-based therapies, non–cell-based cancer vaccines have emerged as a versatile and increasingly validated strategy to elicit tumor-specific immune responses. By delivering tumor-derived antigens directly in vivo, these platforms aim to stimulate endogenous T cell immunity with greater flexibility and scalability than cellular therapies. Among these, peptide vaccines, mRNA vaccines, and virus-based vectors have shown notable clinical progress in recent years.

Peptide-based neoantigen vaccines constitute one of the earliest forms of personalized cancer vaccines. These vaccines are typically composed of multiple long peptides derived from patient-specific tumor mutations, identified through next-generation sequencing and HLA binding prediction algorithms (107). Notably, NEO-PV-01, developed by Neon Therapeutics, demonstrated encouraging immune activation when combined with PD-1 blockade and chemotherapy in non–small cell lung cancer, reinforcing the potential of personalized peptide vaccines in first-line combination settings (108). Building on this, the NeoVax study in melanoma patients showed that vaccination induced durable memory T cell responses and epitope spreading, even four years post-treatment, suggesting possible long-term protective immunity (109). Importantly, similar peptide-based strategies were also applied in glioblastoma, where neoantigen vaccines successfully elicited intratumoral immune responses in a typically immune-excluded tumor context (110). In the landmark NeoVax studies, DCs play a pivotal role—they capture and cross-present NeoVax’s neoantigen peptides to naive CD4⁺/CD8⁺T cells, triggering durable tumor-specific immune responses with the help of adjuvant poly-ICLC that enhances DC activation and antigen presentation capacity. Currently, a Phase Ib study (NCT04930783) is investigating the combinatorial strategy of NeoVax with CDX301 and Nivolumab/Pembrolizumab for melanoma. As a recombinant Flt3 ligand, CDX301 regulates DC proliferation and activity to strengthen antigen presentation, while Nivolumab/Pembrolizumab reinvigorates exhausted T cells. This synergy amplifies NeoVax-induced immune responses and improves therapeutic efficacy. These findings collectively highlight the feasibility, safety, and immunogenicity of long-peptide vaccines across diverse tumor types.

While peptide vaccines provide precise epitope delivery, mRNA-based vaccines offer enhanced flexibility, rapid manufacturing, and multi-epitope encoding in a single formulation. One of the earliest clinical demonstrations came from Sahin and colleagues, who designed personalized RNA mutanome vaccines targeting multiple patient-specific mutations in melanoma. This approach successfully induced broad poly-specific T cell responses, with two patients showing objective tumor regression and one achieving complete remission when combined with anti–PD-1 therapy. Tumor biopsy analysis further confirmed that vaccine-induced T cells infiltrated the tumor and recognized autologous targets, thus validating the concept of RNA-based neoantigen vaccination in humans (111). Building on this foundation, early-phase trials of personalized mRNA vaccines encoding patient-specific neoantigens or recurrent driver mutations (e.g., KRAS G12D) have shown that they can induce T cell responses in gastrointestinal cancers, with favorable safety profiles (112). Going a step further, the BNT211 trial combined CLDN6-targeted CAR-T cells with an mRNA vaccine (CARVac), achieving not only high CAR-T expansion and engraftment, but also a 33% objective response rate in heavily pretreated solid tumor patients (113). In the adjuvant setting, mRNA vaccines are also gaining ground. For instance, the KEYNOTE-942 trial showed that mRNA-4157 (V940) combined with pembrolizumab significantly prolonged recurrence-free survival in resected high-risk melanoma compared to pembrolizumab alone (114). These results collectively suggest that mRNA vaccines can function both as standalone immunotherapies and as amplifiers of other modalities, such as CAR-T or immune checkpoint blockade.

In addition to peptide and mRNA vaccines, oncolytic viruses (OVs) represent a unique class of non-cell-based immunotherapy that combines direct tumor cell lysis with in situ antigen release and immune priming. Modern OV designs increasingly incorporate immunomodulatory payloads to boost therapeutic efficacy. For example, a recombinant HSV-1 virus encoding a cetuximab-CCL5 fusion protein (OV-Cmab-CCL5) was shown to suppress EGFR-positive glioblastoma growth by recruiting innate and adaptive immune cells into the tumor microenvironment (115). Moreover, clinical experience with agents such as T-VEC and ongoing trials with new OV candidates underscore both the therapeutic potential and the translational challenges of this platform, including immune evasion and delivery efficiency (116). Nevertheless, their compatibility with other immunotherapies makes OVs a promising component of multi-modal treatment strategies.

Taken together, these advances in peptide, mRNA, and virus-based cancer vaccines underscore the increasing relevance of non–cell-based platforms in the immunotherapy landscape. While each modality has distinct strengths, their integration with T cell therapies, checkpoint inhibitors, and tumor-targeted approaches may offer synergistic benefits. Moving forward, improvements in neoantigen prediction, delivery vectors, and combination regimens will be pivotal in optimizing their clinical utility across both early-stage and metastatic cancers.

Despite the remarkable success of CAR T cell therapy in hematologic malignancies, its efficacy in solid tumors remains limited, largely due to poor T cell infiltration. Solid tumors are often characterized by physical barriers such as dense extracellular matrix (ECM), an immunosuppressive TME, and chemokine mismatch, all of which collectively hinder T cell trafficking, survival, and antitumor activity within the tumor core (117, 118).

Various strategies have been developed to overcome these challenges and redefine the TME (119). One key approach involves engineering T cells to express chemokine receptors that match tumor-secreted chemokines, thereby improving homing and accumulation within the tumor (117). For example, forced expression of CXCR2 or CCR2b on CAR-T cells enables response to CXCL8 or CCL2, respectively, significantly enhancing migration into the tumor microenvironment (120). Moreover, remodeling the extracellular matrix is another promising tactic. Overexpression of MMP-7 or SPP1 (osteopontin) in GD2-CAR-T cells has been shown to improve extravasation and interstitial migration through ECM-dense regions in neuroblastoma models, resulting in significantly enhanced tumor control without increased off-tumor toxicity (121). Co-stimulatory receptor optimization has also shown efficacy. A study revealed that OX40 co-expression enhances CAR-T infiltration by promoting heparan sulfate-mediated adhesion, leading to improved tumor localization and persistence (122). Similarly, CD93 blockade was found to upregulate vascular adhesion molecules, facilitate effector T cell infiltration, and enhance CAR-T efficacy in melanoma models (123).

In addition to receptor-level modifications, some researchers have focused on intercellular cooperation. A novel DNA nanolinker platform connects T cells with tumor-homing monocytes, leveraging monocyte infiltration pathways to transport T cells into tumor sites. This “monocyte hitchhiking” strategy significantly enhanced T cell accumulation and intratumoral recruitment in breast cancer and melanoma models (124). Synthetic receptor systems also offer precise control over T cell function. For example, STAR-T cells, which integrate antibody-based antigen recognition with native TCR signaling machinery, have demonstrated enhanced antigen sensitivity, resistance to exhaustion, and superior antitumor effects in multiple solid tumor models (125). Synthetic circuits such as synNotch further enable context-dependent cytokine release and spatially restricted activation, thereby indirectly supporting infiltration and persistence (126).

Furthermore, advances in cytokine engineering have contributed to overcoming TME-induced dysfunction. Strategies include integrating cytokine genes (e.g., IL-7, IL-15, IL-18) or synthetic cytokine receptors to enhance T cell persistence and resistance to suppression (120). Combination therapies, such as CAR-T plus checkpoint inhibitors or anti-fibrotic agents, are also being explored to further disrupt physical and immunological barriers (127). Together, these intrinsic engineering approaches aim to enhance T cell access to tumor cores while maintaining functional durability.

Meanwhile, approaches originating from other scientific disciplines are being actively investigated to address existing barriers and enhance the efficacy of T cell-based immunotherapies. For example, photodynamic therapy (PDT) and photothermal therapy (PTT) have been explored to modulate the TME, especially the immunosuppressive cell populations (128). For instance, a design of PDT/PTT nanoplatform loads catalase (CAT) and an anti-GITR antibody (DTA-1) onto PDA–ICG photothermal–photosensitizer nanoparticles to target the constitutive expression of GITR on Tregs (129), while another design further enhances PDT/PTT efficacy to elicit potent tumor-specific T cell responses and overcome immunological resistance through synergistic activation of the cGAS–STING pathway and immunogenic cell death (ICD) (130). ICD also helps reverse immunosuppression and enhance antitumor immune responses when combined with gemcitabine (GEM)–mediated depletion of myeloid-derived suppressor cells (MDSCs) (131). In a further example, an NIR-II–responsive degradable pseudo-conjugated polymer (PSP)–based PDT system co-delivers regorafenib, which is released under 808 nm laser irradiation to normalize tumor vasculature, alleviate hypoxia, and enhance reactive oxygen species (ROS) generation for antitumor activity (132), which can be amplified by integration with polymer-encapsulated carbonized hemin nanoparticles (P-CHNPs) (133).

Besides directly intervening in the microenvironment of solid tumors, PDT and PTT also have the potential to enhance T cell infiltration by softening the tumor ECM (134, 135) and improving blood flow and vascular permeability (136). Notably, PDT and PTT have distinct drawbacks: PDT can exacerbate intratumoral hypoxia, whereas PTT upregulates heat shock protein (HSP) expression (137). Consequently, combining the two modalities is actively being explored, with PTT alleviating hypoxia and PDT decomposing HSPs, thereby mitigating each other’s limitations (138).

The integration of immune checkpoint blockade (ICB) with adoptive cell therapies such as TCR-T or CAR-T has emerged as a promising yet complex approach to enhancing antitumor immunity, particularly in solid tumors. Immune checkpoints including PD-1, CTLA-4, LAG-3, TIM-3, and TIGIT exert distinct regulatory effects across T cells, NK cells, and dendritic cells, necessitating precise selection and timing in combinatorial strategies (139). Moreover, FGL1 has recently been identified as a key LAG-3 ligand involved in immune escape and PD-1 resistance, offering a potential new axis for combination therapy (140).

Accumulating evidence indicates that ICB most effectively amplifies endogenous, vaccine-primed antitumor T cell responses, whereas its ability to reinvigorate already-dysfunctional engineered T cells remains inconsistent. Several studies have shown that ICB can synergize with cancer vaccines. Liu et al. demonstrated that combining neoantigen vaccines with PD-1 blockade expanded cytotoxic CD8⁺T cell subsets with low inhibitory receptor expression and enhanced chemokine signaling (141). Puig-Saus et al. showed that durable responses to PD-1 therapy in melanoma were associated with persistent, polyclonal neoantigen-specific CD8⁺T cells identified using single-cell and CRISPR-based profiling (142). In hepatocellular carcinoma, Yarchoan et al. found that a DNA neoantigen vaccine combined with IL-12 plasmid and pembrolizumab achieved a 30.6% objective response rate and induced robust CD4⁺and CD8⁺T cell responses in 86.4% of patients (143). Collectively, these studies support the notion that ICB primarily enhances de novo or vaccine-induced T cell immunity rather than reversing established dysfunction.

However, the efficacy of ICB in combination with engineered T cells remains less predictable. Stromnes et al. found that TCR-T cells targeting pancreatic tumors rapidly developed transcriptional features of exhaustion that were refractory to checkpoint blockade, suggesting that T cell dysfunction was rooted in both in vitro programming and suppressive tumor environments (144). Davies et al. further demonstrated that PD-1 blockade improved tumor control mainly by stimulating endogenous T cells rather than reversing the dysfunction of transferred TCR-T cells, indicating a lack of true synergy in such contexts (145).

Concurrently, ICB can also be enhanced by interdisciplinary approaches, including the aforementioned PDT method. For instance, He et al. demonstrated that PDT could contribute to an immunogenic environment in the tumour, which significantly enhanced PD-L1 checkpoint blockade therapy by generating systemic antitumor immunity (146). Such physical modulation strategies also offers possibility of increasing applicability and operability through miniature optical devices (147).

Another notable example of ICB enhancement is its combination with dietary intervention, which can modulate the gut microbiome and thereby influence therapeutic outcomes (148–151). Short-chain fatty acids (SCFAs) have been linked to ICB efficacy, though both positive (152) and negative (153) associations have been reported. Dietary fiber is another key modulator of the gut microbiome, with lower fiber intake observed to correlate with reduced ICB response (154). A phase II clinical trial (NCT04645680) is currently evaluating the impact of a controlled high-fiber dietary intervention on ICB efficacy in melanoma patients, with its findings expected to provide important translational insights into the role of dietary interventions in cancer immunotherapy (154, 155).

These results collectively emphasize that ICB can markedly enhance vaccine-induced endogenous T cell responses but may have limited efficacy in reinvigorating already-dysfunctional engineered T cells. Notably, integrating ICB with other physical modalities such as PDT has shown the potential to potentiate antitumor immunity. Future strategies should consider temporally staged or context-specific ICB administration, possibly in combination with synthetic modifications to TCR/CAR constructs and adjunctive therapies like PDT/PTT, to fully unlock the combinatorial potential in solid tumor immunotherapy.

Significant progress has been made in translating T cell-based immunotherapies to solid tumors through improved target identification, cell engineering, and clinical trial design. Although response rates remain variable across cancer types, emerging clinical evidence suggests that, in specific contexts, adoptive T cell therapies can induce durable remissions. This section highlights recent developments in several major solid tumor types, as summarized in Table 1, with the corresponding overall response rates (ORRs) visualized in Figure 4.