Nanomaterials targeting cancer-associated fibroblasts to overcome stromal barriers in cancer immunotherapy Authors: Zhongsong Zhang, Rong Wang, Long Chen Published: 2026 ABSTRACT AbstractAbstractCancer-associated fibroblasts (CAFs) play a pivotal role in the tumor microenvironment (TME), influencing tumor progression, immune evasion, metastasis, and resistance to therapy. In parallel, emerging evidence underscores the therapeutic potential of targeting CAFs in cancer treatment. While CAFs exhibit significant heterogeneity, with different subtypes contributing to various aspects of tumor biology, including ECM remodeling, immune modulation, and cancer cell proliferation. This complexity presents a challenge in targeting CAFs for effective cancer therapy. Recent advancements in nanotechnology have offered novel strategies for precisely targeting CAFs, thereby modulating the TME to enhance the efficacy of cancer immunotherapy. This review provides a comprehensive overview of CAF biology, the mechanisms through which CAFs contribute to immune suppression and therapy resistance, and the emerging role of CAF-targeted nanomedicines in overcoming these barriers. Key strategies discussed include CAF reprogramming, ECM barrier disruption, and the targeting of immune checkpoints, all aimed at reshaping the TME into a more immune-permissive environment. However, despite promising preclinical results, several challenges remain, such as optimizing nanoparticle delivery, overcoming CAF-related drug resistance, and addressing the complex immunosuppressive networks within the TME. Future research should focus on the development of more targeted and personalized nanomedicine approaches, improving CAF-specific drug delivery systems, and better understanding the dynamic interactions between CAFs and the immune system. Our review highlights the current advances in CAF-targeted nanomedicine and underscores the importance of these strategies in advancing cancer immunotherapy. Ultimately, overcoming the challenges associated with CAF-targeted therapies could significantly improve clinical outcomes and pave the way for more effective, individualized cancer treatments.Graphical Abstract Keywords: Cancer-associated fibroblasts, Nanomedicine, Tumor microenvironment, Cancer immunotherapy, Drug delivery, Stromal targeting Introduction Cancer is a leading cause of death globally, presenting a substantial societal and economic burden [1]. While surgery, radiotherapy, and chemotherapy form the foundation of cancer treatment, their efficacy is often limited against metastatic, recurrent, or treatment-resistant tumors, and they are frequently associated with significant adverse effects [2]. Cancer immunotherapy, which mobilizes the body’s immune system to fight tumors, has emerged as a transformative approach, exemplified by the success of immune checkpoint inhibitors and adoptive cell therapy [3, 4]. However, despite these advances, a substantial number of patients experience primary or acquired resistance to immunotherapy, limiting its broader application. Nonetheless, primary and acquired resistance persists in a considerable proportion of patients, constraining the broader utility of these interventions [5, 6]. A central mechanism underlying this resistance is the immunosuppressive TME, which inactivates infiltrating immune cells and promotes tumor immune escape [7]. As the most abundant stromal cell population, CAFs are critical architects of this immunosuppressive niche. CAFs foster therapy resistance through multiple mechanisms. By extensively producing and remodeling the extracellular matrix (ECM), they create a dense physical barrier that hampers the penetration of therapeutic agents and cytotoxic T cells [8, 9]. Furthermore, CAFs secrete a repertoire of soluble factors (e.g., TGF-β, CXCL12) and exosomes that directly dampen the function of effector immune cells, including NK and CD8 + T cells, while concurrently recruiting immunosuppressive populations such as regulatory T cells and myeloid-derived suppressor cells [10, 11]. The functional heterogeneity of CAFs adds another layer of complexity, with specific subsets, like FAP-positive CAFs, being identified as key drivers of immune evasion [12, 13]. Therefore, given their pivotal role in establishing these stromal barriers, CAFs have emerged as a promising target for therapeutic intervention. However, one of the most significant challenges in targeting CAFs is achieving specificity in modulating their activity without affecting normal stromal functions. This challenge is exacerbated by the functional heterogeneity of CAFs across different tumor types and stages, making it essential to identify CAF subtypes that are most relevant to therapy resistance and tumor progression. However, a significant challenge lies in precisely targeting CAFs within the tumor stroma and modulating their activity. Nanomedicine delivery systems present a promising strategy to overcome this challenge. Engineered nanoscale carriers, including liposomes, polymeric nanoparticles, and exosomes, can be designed for multifunctional integration. These systems protect their payload, enhance its stability and circulation half-life, and, through surface functionalization with targeting ligands (e.g., FAP antibodies, RGD peptides), enable specific recognition and accumulation in targeted CAF subpopulations [14, 15]. A key advantage is their ability to respond to distinctive TME cues (e.g., low pH, elevated matrix metalloproteinase or reactive oxygen species levels) for spatially controlled drug release, thereby maximizing therapeutic efficacy while minimizing off-target toxicity [16, 17]. The integration of such intelligent nanocarriers with immunotherapies enables the design of unified platforms capable of executing dual missions: dismantling the physical barrier to facilitate the infiltration of drugs and immune cells, while locally releasing agents (e.g., small-molecule inhibitors, siRNA) to reverse CAF-mediated immunosuppression [18, 19]. This coordinated approach holds the potential to synergistically reinvigorate anti-tumor immunity and achieve a superior therapeutic outcome [18]. This review discusses the role of CAFs in the TME and their contribution to stromal barriers that hinder cancer therapy. It also explores the design of CAF-targeted nanomedicines, focusing on their mechanisms of action, including receptor-mediated targeting, stimulus-responsive drug release, and their potential to enhance the efficacy of immunotherapies. We highlight the need for precise, subtype-specific targeting of CAFs and the integration of nanomedicines with immunotherapy to overcome CAF-mediated resistance and improve clinical outcomes. Through these advances, CAF-targeted nanomedicines represent a promising strategy for transforming the TME and revolutionizing cancer immunotherapy. This review proposes a mechanistically grounded, barrier-oriented framework for CAF-targeted nano-immunotherapy, in which CAF-mediated resistance is categorized into distinct stromal barrier modules and matched with corresponding nanomedicine design strategies and immunotherapy combinations. Biological characteristics of CAFs CAFs are the most abundant stromal component in the TME [20]. They are essential in tumor development and treatment resistance through various mechanisms, such as ECM remodeling, immunosuppression induction, and angiogenesis stimulation. These diverse functions make CAFs a critical focus in cancer research and the development of targeted therapies. Therefore, understanding their biological mechanisms is crucial for utilizing them in targeted cancer therapy. Origins of CAFs CAFs display significant cellular plasticity and originate from various sources, though their exact origins remain incompletely understood due to the lack of universal CAF-specific biomarkers [20, 21]. A well-known source of CAFs is local tissue-resident fibroblasts [22]. In pancreatic cancer, quiescent pancreatic stellate cells (PSCs) are activated in the TME, adopting a myofibroblast-like phenotype with α-SMA expression, thus transitioning into CAFs [23]. Similar conversion of hepatic stellate cells (HSCs) into CAFs has been reported in hepatocellular carcinoma [24]. CAFs can also originate from bone marrow-derived mesenchymal stem cells (BM-MSCs), as confirmed by in vivo and in vitro lineage-tracing studies [25–28]. Other pathways include cellular reprogramming: epithelial cells can transform into CAFs via epithelial-mesenchymal transition (EMT) [29], while endothelial cells undergo a similar process, known as endothelial-mesenchymal transition (EndMT), expressing fibroblast markers like S100A4 [30]. In cutaneous malignancies, EMT-related markers in CAFs are strongly correlated with disease progression [31]. Macrophages are another source; in non-small cell lung cancer (NSCLC), they differentiate into CAFs through macrophage-to-myofibroblast transition (MMT), a process also observed in renal fibrosis [32, 33]. Additional progenitors include adipocytes[34] and pericytes[35], further emphasizing the heterogeneous origin of CAFs (Fig. 1). Origins, crosstalk and functional roles of CAFs in the TME. CAFs play a central role in shaping the TME through their interactions with various stromal cells, including CD8 + T cells, dendritic cells (DCs), natural killer (NK) cells, and tumor-associated macrophages (TAMs). CAFs secrete cytokines such as IL-6, TGF-β, CXCL12 (also known as SDF-1α), and others, which modulate immune responses, contributing to inflammation, immune suppression, or immune activation depending on the context. These interactions support cancer progression by promoting ECM remodeling and secreting matrix metalloproteinases (MMPs), which enhance tumor growth, invasion, and metastasis. Additionally, CAFs can originate from multiple stromal cell types, including endothelial cells, epithelial cells, mesenchymal stromal cells, adipocyte-derived precursor cells, and normal tissue fibroblasts. This phenotypic heterogeneity influences the distinct functions of CAFs within the TME. CAF activation, triggered by factors such as TGF-β, IL-6, SDF-1, and FGF, drives their transition into activated states that further accelerate tumor progression. Depending on their phenotype, CAFs can be classified into subtypes, including inflammatory CAFs (which promote inflammation), myofibroblastic CAFs (responsible for ECM modulation), and antigen-presenting CAFs (which influence T-lymphocyte responses). These diverse roles of CAFs in regulating the TME and immune responses are crucial for understanding their contribution to cancer progression and resistance to therapy. Adapted from [36], Copyright © 2026 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) Markers of CAFs for targeted nanomedicine CAFs lack a unified specific marker profile. Current identifiers include α-smooth muscle actin (α-SMA), platelet-derived growth factor receptor α/β (PDGFRα/β), fibroblast-specific protein 1 (FSP-1/S100A4), and fibroblast activation protein (FAP). Among these, FAP is selectively expressed in CAFs and pericytes across most human epithelial carcinomas, making it a promising target for stromal-focused therapeutic strategies [37–40]. Experimental depletion of FAP-positive fibroblasts in a transgenic mouse model of lung cancer induced extensive necrosis in both tumor and stromal compartments, highlighting the critical role of this fibroblast subpopulation in tumor progression [41]. In CAF biology, biomarkers are not mere "labels" but represent a decodable interstitial language: they indicate the source, activation intensity, contractility, inflammation, and antigen presentation potential of CAFs, and also determine whether these cells can be targeted and reprogrammed for nanomedicine-based therapies. Therefore, systematically identifying CAF biomarkers and their coupling with nanomedicine is essential for overcoming stromal barriers in cancer immunotherapy [42–44]. The most commonly used CAF biomarkers include FAP, α-SMA/ACTA2, PDGFRβ, PDPN, FSP1/S100A4, COL1A1/COL3A1, POSTN, ITGA11, LRRC15. Recently, the highly expressed MHC II apCAF signature profile and Meflin/ISLR have also gained attention. However, no single molecule can define all CAFs, highlighting the need for CAF research to shift from single marker recognition to multi-marker combined stratification [45–47]. Among these, FAP remains one of the most valuable CAF surface markers for current therapies. It is highly expressed in various solid tumors and closely linked to ECM remodeling, immune suppression, and poor prognosis. Therefore, it serves as both a classic entry point for CAF detection and a prime "drug potential address" for antibodies, ligands, radiation probes, and nanocarriers [48–50]. In contrast, α-SMA/ACTA2 represents a myofibroblast-like contraction program, indicating that CAFs exhibit stronger tensile and matrix compression abilities. These CAFs are often linked to collagen deposition, tissue hardening, and T-cell exclusion. However, α-SMA itself is not tumor-specific, and high α-SMA levels can sometimes contribute to tissue constraint, making its significance more like a "mechanical language" rather than a universal target [45, 51]. PDGFRβ is commonly associated with peripheral vascular support and activated interstitial dilation, while PDPN is linked to migration, invasion, and an inflammatory-like CAF status. When combined with FAP and α-SMA, they can better distinguish functional subgroups. For example, in NSCLC, Akram et al. used a combination of biomarkers to identify five CAF populations, from CAF-S1 to CAF-S5. The newly defined CAF-S5 is characterized by FAP+ PDPN + α-SMA^low/-, exhibiting an inflammatory phenotype rather than a highly contractile one, and is associated with poor survival. This suggests that the immunological significance of FAP-positive CAFs depends on their combination with other biomarkers [46]. Additionally, LRRC15 represents a more myofibrotic and immunosuppressive pathogenic matrix state. Krishnamurty et al. demonstrated that selective clearance of LRRC15^ + CAF can reset tumor matrix settings, reduce its direct inhibition of CD8 + T cells, and enhance anti-PD-L1 therapeutic effects. Therefore, LRRC15 is not only a subtype marker of CAFs, but also a functional indicator of "who is creating immune rejection" [52]. In contrast to these barrier markers, a protective or immune-supportive biomarker spectrum has emerged. For example, apCAF with high MHC II expression is enriched near tertiary lymphoid structures in gastric cancer, enhancing T cell activation, proliferation, and cytotoxicity, and is associated with immune therapy response. This suggests that some CAF markers do not indicate targets for clearance, but rather the anti-tumor niche that should be preserved or utilized [53]. Similarly, Meflin/ISLR is considered a representative of restrictive CAFs (cancer-restraining CAFs). Mizutani et al. found that pharmacologically induced Meflin expression can enhance pancreatic cancer sensitivity to chemotherapy, suggesting that the ‘language’ of CAF markers is not binary, but points to different biological fates: some markers are suitable for ablation, while others are better suited for reprogramming [47]. From the nanomedicine perspective, CAF markers primary value lies in transforming the difficult-to-penetrate matrix into navigable target tissue. For example, FAP-α-responsive prodrug nanoparticles constructed by Sun et al. can be cleaved by FAP in the tumor matrix, releasing DOX and R848, and linking with PD-1/PD-L1 antagonistic peptides. This induces immunogenic cell death, reshapes the TAM phenotype, and enhances T cell responses, demonstrating that CAF markers can serve as ‘trigger switches’ rather than just passive localization [54]. Similarly, FAP-modified polymer nanoparticles developed by Rodponthukwaji et al. showed stronger endocytosis and higher killing effects in CAFs with high FAP expression, suggesting that surface markers can improve nanomedicine selectivity for pathogenic CAF populations, reducing damage to normal stroma [55]. Furthermore, Cheng et al. proposed a bio-responsive nanocomposite that combines CAF deactivation with immunogenic chemotherapy. By delivering dasatinib and paclitaxel, which preferentially accumulate at the tumor edge and weaken the matrix barrier, the nanocomposite not only reduces chemotherapy resistance but also helps transform immune-rejected tumors into a more accessible state for immune cells. This suggests that CAF-targeted nanomedicine does not solely rely on killing CAFs, but can also release stromal barriers by altering their activation program [56]. A further conceptual advancement comes from the mRNA LNP platform, Meng et al. reported that delivering mRNA LNP encoding FAP-specific CAR can instantly reprogram host immune cells in vivo to target the FAP + matrix, leading to stronger tumor regression and immune memory when combined with chemotherapy or immune checkpoint inhibitors in various mouse solid tumor models. This suggests that CAF markers are evolving from nano delivery targets to in vivo cell engineering interfaces [57]. Therefore, in the context of stromal barriers in cancer immunotherapy, the true importance of CAF markers is not to label all fibroblasts, but to distinguish which molecules represent dense ECM, immune rejection, and drug blockade, and which reflect inflammatory synergy or antigen presentation support. This distinction advances nanomedicine from generalized drug delivery to precise interstitial intervention strategies, navigating by subtype, reprogramming by state, and releasing barriers by spatial niche [43, 44, 52, 53]. CAF heterogeneity as the basis of stromal barrier diversity Viewing CAFs solely as activated fibroblasts makes it difficult to explain why different tumors create matrix barriers with distinct properties. Some tumors are marked by collagen deposition and tissue stiffening, which restricts the infiltration of effector T cells. Other tumors primarily rely on inflammatory factors, myeloid cell recruitment, and metabolic remodeling, forming a more subtle yet equally persistent immune suppression network. Additionally, some tumors maintain antigen presentation and immunohistochemical functions, indicating that the "matrix barrier" is not a single, uniform structure but rather a continuous spectrum. This spectrum is driven by CAF heterogeneity and encompasses multiple spatially distinct functional states (Fig. 2) [58–60]. CAF-mediated mechanisms in stromal barrier diversity and tumor progression. This figure illustrates the diverse roles of CAFs in creating distinct matrix barriers within the TME, contributing to tumor progression and resistance to treatment. CAFs mediate chemotherapy resistance by secreting factors like TGF-β, TNFSF4, collagen, and IL-6, which promote drug efflux and hinder drug penetration into cancer cells. They also drive angiogenesis by releasing VEGF, PDGF, and CXCL12, stimulating endothelial cells to form new blood vessels, thus supporting tumor growth and metastasis. Additionally, CAFs promote cancer cell metastasis through cytokines such as CXCL2, CXCL5, and TGF-β, which enhance tumor cell migration and invasion. CAFs further contribute to immune suppression by secreting IL-6, VEGF, TGF-β, and PGE2, which inhibit T cell and NK cell function, creating an immune-evasive microenvironment. They also support tumor cell stemness by secreting factors like POSTN, SPARC, and EGFR, which enhance cancer cell self-renewal and resistance to therapy. Moreover, CAFs play a central role in ECM remodeling, secreting collagen and MMPs that stiffen the tumor tissue and restrict immune cell infiltration. The phenotypic heterogeneity of CAFs allows them to adopt various functional states, which results in diverse stromal barriers across different tumors, shaping immune responses, drug resistance, and metastatic potential. Adapted from [43], Copyright PGE2 Recent pan-cancer single-cell studies have shown that CAFs are not a terminal, consistent cell population, but rather a dynamic one shaped by various origins, differentiation trajectories, and microenvironmental cues. Luo et al. integrated 226 samples from 10 types of solid tumors, showing that CAFs differentiate along various activation states, which are linked to immune regulation, angiogenesis, and EMT-related pathways. Some CAFs also exhibit plasticity, shifting from endothelial or myeloid-related states to stromal states, suggesting that the diversity of the "barrier" originates from the heterogeneity of CAF states [58]. This perspective is further supported at the spatial level, Ma et al. integrated multiple cancer single-cell and spatial transcriptomes, proposing that functional subgroups like mCAF and iCAF exhibit distinct spatial distributions and communication networks. mCAF is more involved in ECM remodeling and angiogenesis, while iCAF is more closely linked to immune suppression and tumor promotion signals. This suggests that although physical and immune suppression barriers often coexist, their dominant CAF subgroups differ[59]. A larger-scale single-cell spatial multi-omics study identified four conserved CAF subtypes in 10 types of cancer and over 14 million cells. It demonstrated that their neighborhood composition, immune cell distribution, and patient outcomes varied, directly linking CAF heterogeneity to "spatial histological differences in the matrix barrier" [60]. From a source perspective, CAF heterogeneity is not a random noise that arises solely after activation, but rather carries clear lineage imprints. Helms et al. found through lineage tracing in PDAC that pancreatic stellate cells, traditionally considered the primary source, contribute only a small but functionally unique portion of CAFs. This subset of CAFs is richer in specific ECM programs and forms a matrix niche distinct from other CAF sources, suggesting that CAFs of different origins inherently have the capacity to produce different barrier types [61]. Therefore, the thick-matrix low-infiltration phenotype of certain tumors may not only result from an increased number of CAFs but is more likely due to biased amplification of collagen, fibronectin, basement membrane components, and mechanical tension programs by specific CAF lineages [60, 61]. At the mechanistic level, the barrier dominated by myCAFs resembles the classical physical barrier. This type of CAF highly expresses molecules involved in contraction and matrix assembly. It promotes ECM deposition, cross-linking, and rearrangement, thereby increasing tissue stiffness and forming a dense structure around the tumor nest. As a result, it limits T-cell penetration, alters drug diffusion, and reshapes local vascular perfusion [62, 63]. In contrast, the barrier dominated by iCAFs functions more as a soluble immune barrier. It amplifies paracrine signaling between tumor, stromal, and immune cells through IL-6, CXCL12, and other inflammatory factors. This promotes the recruitment of myeloid cells and their polarization toward immunosuppression. Ma et al. conducted pan-cancer spatial analysis showing that iCAFs are closely associated with the construction of an immunosuppressive microenvironment. Timperi et al. further demonstrated in breast cancer that CAFs can be induced by the CXCL12-CXCR4 axis to promote the formation of STAB1-positive TREM2-high lipid-related macrophages (LAM). These macrophages are enriched in patients resistant to immune checkpoint blockade and directly contribute to immunosuppression. This indicates that some CAF subsets act not as a physical wall but as an immune educator for continuous modulation of myeloid cells [59, 64]. In addition, another type of barrier can be summarized as the metabolic nutritional barrier. Francescone et al. found in pancreatic ductal adenocarcinoma (PDAC) that NetG1-positive CAFs not only support nutrient supply to tumor cells but also synchronously shape the immunosuppressive environment [65]. In other words, this type of CAF does not merely block immune cells via ECM but rearranges local metabolic resources. This makes it easier for tumor cells to survive while making it harder for immune cells to maintain their effector functions, thereby turning nutritional competition into an advantage for immune escape. It is worth noting that CAF heterogeneity is not necessarily synonymous with universal cancer promotion. Instead, it implies that barrier functions have directional differences. For example, single-cell spatial studies have shown that antigen-presenting CAFs (apCAFs) include at least two distinct lineages and niches: mesothelial-like M-apCAFs near cancer cells and fibroblast-like F-apCAFs in lymphocyte-enriched regions. Together, these findings demonstrate that apCAFs are not a uniform entity but can be integrated into different spatial barrier networks [66]. More interestingly, Song et al. observed that apCAFs with high MHC class II expression in gastric cancer are preferentially localized near tertiary lymphoid structures. They can enhance T-cell activation, cytotoxicity, and proliferation, thereby predicting better responses to immunotherapy. This suggests that certain CAF subgroups may even partially dismantle rather than reinforce the matrix barrier, at least in specific cancer types and niches [53]. Therefore, based on current evidence, CAF heterogeneity underlies stromal barrier diversity not only by providing multiple biomarkers but also through systematic differences among CAF subgroups in their origins, spatial localization, ECM assembly capacity, inflammatory secretion profiles, metabolic support mechanisms, and methods of educating myeloid and lymphoid cells. Ultimately, the matrix barrier phenomenon can be categorized into multiple subtypes, including physical rejection, inflammatory inhibition, myeloid reprogramming, metabolic competition, and even context-dependent antigen presentation [67, 68]. This also explains why simply depleting all CAFs often fails to yield stable clinical benefits. The key target for intervention is not the CAF population as a whole but the specific functional subgroups responsible for constructing dominant barriers in particular cancer types and spatiotemporal contexts [69, 70]. The heterogeneity of CAFs, as the biological foundation of matrix barrier diversity, is crucial for understanding how subtypes such as myCAFs (myofibroblast-like), iCAFs (inflammatory-like), and apCAFs (antigen-presenting-like) achieve phenotypic differentiation via signaling pathways including TGF-β/Smad, JAK/STAT3, and NF-κB. This differentiation enables the construction of distinct barriers in various tumor types and stages of progression, providing a theoretical framework for the design of nanodrugs with barrier-matching strategies. The necessity of subtype-aware targeting: harmful vs. protective CAF states CAFs in the TME exhibit remarkable functional duality. Some subtypes act as guardians that limit tumor expansion, whereas others serve as accomplices that accelerate tumor progression and immune escape. This functional heterogeneity provides an essential biological foundation for the development of subtype-specific targeting strategies [63, 71]. The molecular mechanisms and clinical implications of tumor-promoting and tumor-restraining CAF states are critical for understanding the potential adverse effects of non-selective CAF depletion. Such insights offer precise guidance for subtype-specific "barrier-matching" nanomedicine design[72, 73]. Myofibroblast-like CAFs (myCAFs) are characterized by high expression of α-SMA, FAP, and COL1A1. They drive extensive ECM deposition and remodeling via the TGF-β/Smad and YAP/TAZ signaling pathways. This process forms a dense physical barrier while promoting tumor cell invasion and metastasis. In pancreatic ductal adenocarcinoma (PDAC), myCAFs are predominantly localized near tumor cells. They enhance tumor stiffness through mechanical signaling and secrete TGF-β, thereby amplifying local immunosuppression [74, 75]. However, genetic depletion of α-SMA-positive myCAFs unexpectedly resulted in more invasive PDAC tumors, enhanced epithelial-mesenchymal transition (EMT), increased regulatory T-cell (Treg) infiltration, and significantly reduced mouse survival. These findings indicate that certain myCAFs exert tumor-restraining effects during early stages, possibly by preserving tissue structural integrity and supporting immune surveillance [76, 77]. In sharp contrast, inflammatory CAFs (iCAFs), characterized by low α-SMA expression and high levels of IL-6, CXCL12, and LIF, are primarily activated via the IL-1/NF-κB and JAK/STAT3 pathways. They secrete abundant inflammatory factors and chemokines, recruit myeloid-derived suppressor cells (MDSCs), regulatory T cells (Tregs), and M2 macrophages, and establish a soluble immunosuppressive barrier. Consequently, they directly inhibit the function of CD8 + T cells and natural killer (NK) cells. In breast cancer and lung cancer, iCAF enrichment is frequently associated with resistance to immunotherapy [78, 79]. Emerging antigen-presenting CAFs (apCAFs) mimic antigen-presenting cell (APC) function by expressing MHC class II molecules. On one hand, they can activate CD4-positive T cell-mediated anti-tumor immunity. On the other hand, in the absence of co-stimulatory molecules, they tend to induce regulatory T cell (Treg) differentiation, thereby exhibiting a context-dependent dual role [80, 81]. In addition, recent studies have clearly defined remodeling CAFs (rCAFs), which are predominantly enriched during early tumor stages. They limit tumor invasion by secreting inhibitory factors or maintaining normal matrix architecture. Their proportion decreases as tumors progress [82, 83]. Painful lessons from clinical trials further underscore the necessity of subtype-aware CAF targeting. For instance, Hedgehog pathway inhibitors (vismodegib and IPI-926), intended to block myCAF activation, simultaneously damage protective fibroblasts. This results in accelerated tumor progression and loss of survival benefits. The phase III trial of the hyaluronidase PEGPH20 similarly failed to improve overall survival and even exacerbated disease progression in certain patient subgroups [36, 84]. These lines of evidence collectively indicate high plasticity among CAF subtypes. Different subtypes can undergo dynamic transitions driven by TGF-β or inflammatory signals derived from tumor cells. A single “kill-all” strategy can readily disrupt microenvironmental balance and trigger compensatory pro-tumor reprogramming [58, 85]. Therefore, subtype-specific targeting—selectively depleting harmful myCAFs and iCAFs or reprogramming their phenotypes toward a protective state—has emerged as a core direction in CAF intervention strategies. Basic experiments typically employ single-cell RNA sequencing (scRNA-seq) combined with spatial transcriptomics to identify subtype-specific markers, three-dimensional in vitro co-culture systems of CAFs, tumor cells, and immune cells to validate functionality, and conditional knockout mouse models to quantify harmful versus protective effects [86, 87]. Only by accurately distinguishing harmful from protective CAF states can safe and effective CAF-targeted therapies be achieved, thereby overcoming the restrictive effects of matrix barriers on cancer immunotherapy [88]. Interactions between CAFs and various components in the TME Dysregulation of homeostasis within the TME plays a critical role in tumor progression. CAFs, as a key stromal component, create a tumor-promoting niche by forming intricate signaling networks involving cancer cells, ECM components, and various immune cell types. Understanding the mechanisms driving these interactions is crucial for unraveling the pathological process of tumor progression. CAFs and cancer cells The interaction between CAFs and cancer cells primarily drives pro-tumorigenic activities through multiple mechanisms, including regulation of cancer cell proliferation, angiogenesis, invasion, metastasis, therapy resistance, and the establishment of an immunosuppressive microenvironment (Fig. 3). stimulate cancer cell proliferation by undergoing metabolic reprogramming or secreting signaling molecules that supply nutrients and growth signals. In breast cancer, CAF-derived exosomal lncRNA SNHG3 acts as a molecular sponge for miR-330-5p, inhibiting mitochondrial function and enhancing glycolysis in cancer cells, which promotes proliferation [89, 90]. In neuroblastoma xenograft models, pharmacological inhibition of microsomal prostaglandin E synthase-1 (mPGES-1)—a key enzyme in prostaglandin E2 (PGE2) synthesis—significantly reduced tumor cell proliferation, as shown by immunohistochemical analysis [91, 92]. In ovarian cancer, CAF-expressed methyltransferase NNMT promotes cancer cell proliferation through metabolic regulation, underscoring its potential as a therapeutic target [93]. CAFs promote tumor angiogenesis through both VEGF-dependent and independent pathways to form vascular networks. scRNA-seq studies in PDAC reveal CAF overexpression of pro-angiogenic factors, such as VEGFA, PDGFC, and FGF2, which promote angiogenesis by recruiting myeloid and vascular endothelial cells [94]. In lung cancer, CAF–cancer cell contact via the Notch2–Jagged1 pathway promotes vasculogenic mimicry (VM), a process linked to resistance against anti-VEGF therapy. Combined treatment with an anti-VEGF antibody and the Notch inhibitor DAPT (a γ-secretase inhibitor) markedly suppresses lung tumor growth [95]. In colorectal cancer, CAF-secreted CHI3L1 enhances angiogenesis by stimulating IL-8 secretion [96]. CAFs facilitate cancer cell invasion and metastasis mainly by remodeling the ECM or secreting chemokines and cytokines. In prostate and pancreatic cancer models, CAFs align fibronectin matrices through increased contractility and traction forces, promoting directed migration of cancer cells via α5β1 integrin and PDGFRα signaling[97]. In gastric cancer, CAF-mediated activation of the CXCL12/CXCR4 axis supports cancer cell survival and metastasis; this effect is reversed by the CXCR4 antagonist AMD3100, which significantly impairs invasion [98, 99]. In breast cancer, CAF-derived IL32 enhances invasion and metastasis through the integrin β3–p38 MAPK pathway, while IL33 promotes lung cancer metastasis by driving type 2 inflammation [100, 101]. In PDAC, myCAFs stimulate type III collagen production and cancer metastasis via the IL-33–ST2–CXCL3–CXCR2 axis [102]. Crosstalk between CAFs and T cells mediating immunosuppression. This figure depicts the mechanisms by which CAFs suppress T cell function and infiltration. CAFs secrete soluble factors such as TGF-β, CXCL12, IL-6, and lactate, which activate inhibitory signaling pathways (e.g., JAK/STAT3) in T cells, upregulate PD-1/PD-L1 expression, and reduce T cell recruitment. Additionally, CAF-derived ECM remodeling creates a physical barrier that limits T cell penetration into tumor islets. These interactions collectively foster an immunosuppressive niche that enables tumor immune escape. Adapted from [67], Copyright © 2021 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) Mechanisms of CAF-induced therapy resistance include cytokine secretion, exosome-mediated transfer of regulatory molecules, and physical barrier formation. In pancreatic cancer, CAF-derived exosomal miR-106b confers gemcitabine resistance by targeting TP53INP1 [103]. In breast cancer, exosomal miR-22 from CD63 + CAFs induces tamoxifen resistance by binding ERα and PTEN; this can be counteracted using a CD63-neutralizing antibody [104]. In bladder cancer, CAF-secreted exosomal LINC00355 promotes cisplatin resistance via the miR-34b-5p/ABCB1 axis [105]. Additionally, in PDAC, CAFs produce abundant ECM—accounting for 50%–80% of tumor mass, that compresses blood vessels, reduces perfusion, and limits drug delivery, thereby compromising chemotherapy and immunotherapy [106–108]. CAFs also shape an immunosuppressive microenvironment that enables immune evasion. TGF-β and CXCL12 secreted by CAFs inhibit T cell activation and recruitment. CXCL12 further drives Treg differentiation and supports pro-angiogenic macrophages and dysfunctional DCs. In colorectal cancer, a ketogenic diet upregulates HMGCS2, downregulates KLF5, and suppresses CXCL12 in CAFs, thereby attenuating immunosuppression. In pancreatic cancer, apCAFs promote naive CD4 + T cell differentiation into Tregs via IL-1 and TGF-β signaling, impairing CD8 + T cell function [109]. In NSCLC, tumors enriched with MYH11 + αSMA + or FAP + αSMA + CAF subtypes show reduced CD3 + and CD8 + T cell infiltration and are associated with resistance to immune checkpoint blockade [110]. In liver cancer, crosstalk between CAFs and SPP1 + macrophages establishes a tumor immune barrier that excludes CD8 + T cells and limits immunotherapy efficacy [111]. Notably, certain CAF subtypes can enhance treatment sensitivity through specific mechanisms, suggesting novel clinical opportunities. In lung cancer, CD200-positive CAFs increase tumor sensitivity to the EGFR-TKI gefitinib; knockout of CD200 abolishes this effect [112]. In NSCLC, CAF-secreted IGF and IGFBPs correlate with EGFR-TKI sensitivity [113]. In PDAC, CAFs exhibit tumor-restraining properties under hypoxia. Moreover, although tumor-derived Sonic Hedgehog (Shh) promotes fibroblast-rich stroma, Shh knockout in mice unexpectedly increased tumor proliferation and aggressiveness, implying that myCAFs may partially suppress tumor progression via Shh-SMO signaling [114–116]. CAFs and extracellular matrix CAFs act as key regulators of ECM remodeling within the TME. They profoundly influence tumor progression through the synthesis of ECM components, secretion of key remodeling enzymes, and modulation of the ECM’s mechanical properties. CAFs are prolific producers of essential ECM constituents, such as collagen and fibronectin, and a major source of enzymes like MMPs. Increased deposition of ECM components alters tissue architecture, while enzymatic degradation facilitates local invasion. For instance, CAF-derived lysyl oxidase-like 2 (LOXL2) promotes ECM realignment, thereby aiding the migration of prostate and lung cancer cells [117, 118]. In gastric cancer, a distinct subset of ECM-related CAFs (eCAFs), characterized by high periostin (POSTN) expression and elevated ECM remodeling gene levels, drives metastasis and correlates with poor overall survival in patients [119].Additionally, CAF-driven ECM remodeling significantly contributes to therapy resistance by forming a physical barrier and modulating immune responses. In pancreatic cancer, hyaluronic acid (HA) and collagen secreted by CAFs synergize to increase tissue stiffness, compressing tumor vasculature and impeding the efficient delivery of chemotherapeutic agents [120, 121]. In lung cancer, ECM accumulation upregulates PD-L1 expression on cancer cells, fostering an immunosuppressive microenvironment that further compromises treatment outcomes [122]. Moreover, CAF-secreted fibronectin facilitates directed cancer cell migration by enabling increased contractile forces and structural guidance, a process dependent on α5β1 integrin and PDGFRα signaling, as observed in prostate and pancreatic cancer models [97]. CAFs and the immune microenvironment CAFs are crucial "orchestrators" of the tumor immune microenvironment. They suppress antitumor immune responses and create an immunosuppressive niche that offers an "immune sanctuary" for tumor growth through cytokine secretion, immune checkpoint molecule expression, and metabolic competition. Interaction between CAFs and myeloid-derived suppressor cells CAFs actively recruit and functionally activate myeloid-derived suppressor cells (MDSCs) by secreting a variety of chemokines and cytokines, thereby shaping the immunosuppressive microenvironment. CAF-derived chemokines are key mediators of MDSC recruitment. For example, in intrahepatic cholangiocarcinoma (ICC) and lung squamous cell carcinoma (LSCC), CAF-derived CCL2 acts as a crucial signal for recruiting CCR2 + monocytes/MDSCs [123, 124]. Similarly, in hepatocellular carcinoma (HCC), CAFs recruit monocytes to the tumor stroma via the CXCL12/CXCR4 axis [125]. Regarding differentiation and activation, CAFs directly induce monocytes to acquire the immunosuppressive phenotype and functional characteristics of MDSCs. A central mechanism involves IL-6-mediated activation of the STAT3 signaling pathway. For instance, in HCC and esophageal squamous cell carcinoma (ESCC), CAF-secreted IL-6 promotes the differentiation of monocytes into MDSCs with strong T-cell suppressive capacity via STAT3 activation [125, 126]. In ESCC, CAFs further enhance STAT3 activation and promote MDSC generation by delivering microRNA-21 via exosomes, acting synergistically with IL-6 [126]. Functionally, these CAF-educated MDSCs suppress CD8 + T-cell activity through mechanisms such as the robust production of reactive oxygen species (ROS) [127, 128]. The CAF–MDSC interaction has significant clinical relevance, as multiple studies have confirmed that stromal co-infiltration of CAFs and CD11b + myeloid cells is strongly associated with poor patient prognosis [129]. CAF-mediated regulation of macrophages CAFs modulate macrophage polarization, skewing them toward a pro-tumoral M2 phenotype known as tumor-associated macrophages (TAMs). Macrophage colony-stimulating factor (M-CSF/CSF-1) secreted by CAFs is a key driver of monocyte polarization toward M2 TAMs. In pancreatic cancer, for example, CAF-derived M-CSF enhances intracellular ROS production in monocytes, promoting their transition to an M2 phenotype [130]. Additionally, CAFs regulate TAM recruitment and polarization through multiple mechanisms: in lung cancer, lactate-activated CAFs upregulate DESMIN expression and secrete IL-8, facilitating M2 macrophage polarization [131]; in bladder cancer, high NNMT expression in CAFs epigenetically regulates SAA expression, promoting TAM recruitment and contributing to immunotherapy resistance [132]; and in colorectal cancer, CAF-derived CCL2 recruits monocytes and collaborates with them to form an immunosuppressive microenvironment characterized by high CCL2 expression [133]. M2 TAMs can, in turn, enhance CAF activity through factors like TGF-β, forming a pro-tumor feedback loop. In colorectal cancer, for instance, TAMs promote COL10A1 expression in CAFs via the TGF-β/RUNX2 pathway [134]. This "CAF–TAM" regulatory axis has been implicated in tumor progression and resistance to immunotherapy across multiple solid tumors, including pancreatic, lung, and colorectal cancers. The effects of CAFs on natural killer cells CAFs employ a multifaceted strategy to suppress NK cell-mediated antitumor immunity, leveraging soluble factor secretion, metabolic reprogramming, and ECM remodeling. At the cytokine level, CAFs produce immunosuppressive factors such as IL-6, which is secreted by PDGFRβ + CAFs to activate STAT3 signaling and induce NK cell dysfunction, thereby facilitating pancreatic cancer metastasis—a process reversible upon PDGFRβ inhibition [135]. TGF-β is another pivotal mediator that downregulates NK cell-activating receptors such as NKG2D and NKp30, suppresses IFN-γ production, and impairs cytotoxicity through transcriptional repression of DAP12 via miR-183 induction [136–138]. Direct contact between CAFs and NK cells further dampens immune function by reducing surface expression of NKG2D and DNAM-1 [139]. Chemokine-mediated suppression is also evident: CAF-derived CXCL12 inhibits the proliferation of hematopoietic NK cells [140], whereas VEGF fosters an immunosuppressive environment by recruiting MDSCs, which indirectly attenuate NK cell cytotoxicity and IFN-γ secretion [141, 142]. Metabolically, CAFs impair NK cell function through specific metabolites. Podoplanin-positive (PDPN +) CAFs exhibit high IDO1/TDO2 expression and produce L-kynurenine, which interferes with IL-2–dependent regulation of NKp46 and NKG2D receptors [143, 144]. PGE2 secreted by CAFs downregulates key activating receptors—NKp30, NKp44, and CD69—as well as effector molecules like granzyme B and perforin, in a self-reinforcing inhibitory loop [145–147]. Additionally, CAF-upregulated netrin G1 impairs NK cell killing of pancreatic cancer cells via glutamate release [148]. In the context of ECM modulation, CAFs impair NK cell function through proteolytic activity and physical barrier formation. Melanoma-associated fibroblasts (MAFs) secrete MMPs that cleave MICA/B ligands from tumor cells, rendering them resistant to NK cell recognition[149]. Furthermore, ECM components such as collagen I, III, and elastin engage the inhibitory receptor Lair1 on NK cells, directly suppressing degranulation and IFN-γ production [150]. The effects of CAFs on dendritic cells (DCs) As pivotal stromal components of the TME, CAFs regulate DC function primarily through the release of soluble mediators, direct cell-contact signals, and exosome-based communication. These interactions primarily suppress DC maturation and activation while promoting a tolerogenic phenotype, ultimately impairing antitumor immunity [151]. Specifically, CAF-derived cytokines can disrupt DC antigen presentation and polarization capacity. In pancreatic cancer, for example, tumor-derived TNF-α and IL-1β stimulate CAFs to produce thymic stromal lymphopoietin (TSLP), which activates myeloid DCs and drives Th2 polarization. This results in increased Th2 cell infiltration in the tumor stroma, and a high Th2/Th1 ratio correlates significantly with poor postoperative outcomes [152]. CAFs also promote the differentiation of regulatory DCs (regDCs), which have immunosuppressive properties. In hepatocellular carcinoma, CAF-secreted CXCL12 recruits DCs, while IL-6 release activates STAT3 signaling in these cells, upregulating indoleamine 2,3-dioxygenase (IDO) expression. The resulting regDCs have impaired ability to stimulate CD3 + T cell proliferation and actively expand CD4 + CD25 + Foxp3 + Tregs through IDO activity, reinforcing immune tolerance [153]. Exosomal signaling is another CAF-dependent regulatory mechanism. In lung cancer, CAF-derived exosomes carry miR-146a, which downregulates DC maturation markers (e.g., CD80, CD86) and increases secretion of inhibitory cytokines like IL-6 and IL-10, inducing a tolerogenic state. Curcumin counteracts this suppression by reducing miR-146a levels, thereby restoring DC maturation [154]. In esophageal squamous cell carcinoma and colorectal cancer, CAF-secreted WNT2 impairs DC differentiation by suppressing the JAK2/STAT3 pathway, limiting tumor antigen presentation. Targeting WNT2 not only reinstates DC-dependent CD8 + T cell activation but also synergizes with anti-PD-1 therapy [155] (Fig. 4). CAF-mediated regulation of DCs across tumor types. This figure illustrates how CAFs modulate DC function in various cancers (e.g., pancreatic, hepatocellular, lung, esophageal squamous, colorectal cancer). CAFs secrete factors such as TSLP, CXCL12, IL-6, WNT2, and miR-146a (via exosomes) to suppress DC maturation, promote regulatory DC (regDC) differentiation, and drive Th2 polarization. These effects impair antigen presentation, reduce T cell activation, and enhance Treg expansion, contributing to immunosuppression. Conversely, select CAF-derived signals may promote DC immunoactivation in specific contexts. Adapted from [151]. The effects of CAFs on T cells CAFs, as pivotal components of the TME, orchestrate multiple mechanisms—including cytokine secretion, signaling pathway modulation, induction of immunosuppressive cells, and ECM remodeling—that collectively regulate T cell activation, proliferation, effector function, and intratumoral infiltration, thereby promoting tumor immune escape [156]. In terms of T cell functional regulation, CAFs often suppress T cell activity by secreting specific cytokines that trigger downstream inhibitory signaling. In gastric cancer, for example, CAFs enhance IL-8 secretion, which, upon binding to CXCR1/2 receptors, activates P38, JNK, and NF-κB pathways, leading to increased PD-L1 expression on tumor cells. This facilitates the PD-L1/PD-1 interaction with CD8 + T cells, impairing their cytotoxicity; such suppression can be reversed by the IL-8 receptor inhibitor reparixin, which reduces PD-L1 levels and restores CD8 + T cell function [157]. Concurrently, CAF-derived serum IL-8 (sIL-8) in gastric cancer activates the JAK2/STAT3 pathway and downregulates FBXO38, inhibiting PD-1 ubiquitination and degradation. This increases PD-1 expression on CD8 + T cells, intensifying immunosuppression and accelerating lymph node metastasis [158]. In hepatocellular carcinoma, CAF-neutrophil contact induces IL-6 secretion, activating STAT3 signaling in neutrophils and driving PD-L1 release, which engages PD-1 on T cells and disrupts their function [159]. Similarly, in melanoma and colorectal cancer, CAF-derived CXCL5 promotes tumor cell PD-L1 expression through PI3K/AKT pathway activation, establishing an immunosuppressive niche that restrains T cell activity [160]. Beyond direct cytokine-mediated effects, CAFs also employ indirect mechanisms by recruiting and polarizing immunosuppressive cells. In lung squamous cell carcinoma, CAFs use the CCL2/CCR2 axis to recruit CCR2 + monocytes, which in 3D culture systems differentiate into MDSCs capable of strongly suppressing autologous CD8 + T cell proliferation and IFN-γ production—effects reversible by CCR2 inhibition or ROS scavenging [124]. In pancreatic cancer, CAF-monocyte crosstalk induces arginase-I, facilitating M2 macrophage and MDSC differentiation and generating an immunosuppressive environment that suppresses T cell activation and expansion [161]. Regarding T cell infiltration, CAFs modulate T cell trafficking and accumulation through altered expression of chemokine receptors, signaling molecules, or phenotypic switching, although the precise mechanisms are cancer-type dependent. In pancreatic cancer, CAFs downregulate CXCR3 and CCR5 while upregulating CXCR4, thereby limiting T cell migration. However, ligands for CXCR3 and the CCR5 ligand CCL8, enriched in tumor regions, can still support CD8 + T cell infiltration [162]. In colorectal cancer, ASCL2-overexpressing CAFs activate TGF-β signaling, promoting an immune-excluded microenvironment that reduces CD8 + T cell influx. Combining anti-PD-1 therapy counteracts this exclusion, enhancing T cell infiltration and amplifying antitumor immunity [163]. Notably, during immunogenic chemotherapy combined with PD-1 blockade, TGF-β1 neutralization reverses G9a/EZH2-dependent chromatin remodeling at CXCL9/10 promoters in CAFs, restoring CXCL9/10 production and promoting antigen-specific CD8 + T cell recruitment [164]. In oral squamous cell carcinoma, tumor-derived TGF-β1 upregulates AEBP1/ACLP in CAFs, and ACLP levels inversely correlate with CD8 + T cell infiltration, directly impeding T cell entry [165]. Furthermore, Endo180 knockout mice exhibit reduced α-SMA + CAF density, accompanied by markedly increased CD8 + T cell infiltration and heightened sensitivity to immune checkpoint blockade [166]. Distinct CAF-mediated barriers differentially impair immunotherapy modalities Cancer immunotherapy has emerged as a breakthrough in oncology by reactivating antitumor immune responses and overcoming immune escape mechanisms in some patients [167]. Nevertheless, a significant proportion of individuals exhibit limited or no response to treatment. CAFs, as pivotal regulators of the TME, contribute substantially to this resistance by interfering with immunotherapy through multiple pathways. An overview of targeting CAFs for cancer immunotherapy In recent years, cancer immunotherapy has revolutionized cancer treatment by utilizing the immune system’s ability to target and eliminate tumors, leading to durable clinical responses in several refractory solid and hematologic malignancies [168]. Current immunotherapeutic modalities can be categorized into four main types: immune checkpoint inhibitors (ICIs) [169], adoptive cell therapy (ACT) [170], cancer vaccines [171], and oncolytic viruses [172]. Among these, ICIs and ACT have garnered significant attention due to their clinical success and wide applicability. Immune checkpoint inhibitors work by blocking inhibitory signaling pathways such as PD-1/PD-L1 and CTLA-4, thereby reactivating T-cell-mediated antitumor immunity. Clinically approved agents, including PD-1 inhibitors like pembrolizumab and nivolumab, and the CTLA-4 inhibitor ipilimumab, are now standard treatments for melanoma, non-small cell lung cancer, and other cancers [169]. In contrast, adoptive cell therapy involves engineering and expanding a patient’s own immune cells, typically T cells, to enhance their tumor-targeting ability. CAR-T cell therapy has shown remarkable success in hematologic malignancies, such as acute lymphoblastic leukemia [173, 174], while TCR-T cell therapies are showing increasing promise in solid tumors. Despite these advances and the inclusion of immunotherapies in treatment guidelines, resistance remains a major clinical challenge [175]. Some patients exhibit no initial response (primary resistance), while others who initially respond later relapse (acquired resistance) [176, 177]. The underlying mechanisms are multifactorial, involving tumor-intrinsic features, such as low tumor mutational burden or loss of HLA expression [178, 179]—and, more notably, the immunosuppressive properties of the TME [180]. CAFs, the most abundant stromal component in many tumors, are increasingly recognized as key players in driving immunotherapy resistance [180]. Through various mechanisms, including immune cell modulation, ECM remodeling, and secretion of soluble inhibitory mediators, CAFs directly and indirectly impair the efficacy of immunotherapeutic strategies. Therefore, it is crucial to gain a deeper understanding of how CAFs influence different forms of immunotherapy to overcome current therapeutic limitations. Influence of CAFs on ICIs ICIs, including anti-CTLA-4 and anti-PD-1/PD-L1 antibodies, have revolutionized cancer treatment. However, a significant number of patients experience limited benefit, particularly in malignancies with extensive stromal components, such as pancreatic cancer. CAFs, as key stromal components of the TME, modulate local immune status through various mechanisms and play a pivotal role in ICI resistance. Their primary effects are outlined below. CAFs mediate immunosuppression within the TME, compromising the anti-tumor immunity induced by ICIs. A key mechanism involves the secretion of cytokines and chemokines that recruit and activate immunosuppressive cell populations, creating a hostile environment for ICI activity. Notably, CAFs are the primary source of CXCL12 in TME. Signaling through the CXCL12/CXCR4 axis recruits MDSCs and promotes their immunosuppressive polarization. This cascade results in reduced effector T cell infiltration and significantly diminishes the efficacy of anti-PD-L1 therapy. In contrast, inhibition of this pathway reduces MDSC accumulation and can restore ICI responsiveness [181, 182]. Additionally, CAF-derived IL-6 activates the STAT3 pathway in myeloid cells. This activation drives DC differentiation towards a regulatory phenotype, which suppresses T cell activation. In hepatocellular carcinoma models, blockade of CAF-derived IL-6 or myeloid STAT3 reduces immunosuppression and synergizes with ICIs [153]. Furthermore, CAF-secreted LIF upregulates CCL2 expression in TAMs, promoting further myeloid cell recruitment, while suppressing CXCL9 expression, impairing CD8 + T cell recruitment, and intensifying the immunosuppressive landscape [183]. CAFs can also directly impair T cell function, counteracting the primary goal of ICI therapy. They do this through surface molecule expression and abnormal antigen presentation. For example, CAF expression of immune modulators like B7H3, CD73 (an ectoenzyme generating immunosuppressive adenosine), and DPP4 (which inactivates the T cell-attracting chemokine CXCL10) promotes the expansion of CD25 + FOXP3 + regulatory T cells and suppresses CD8 + effector T cells. This suppression often persists even with ICI administration [184, 185]. A distinct subset, apCAFs, expresses MHC-II but lacks essential co-stimulatory molecules and secretes PGE2. Consequently, apCAFs fail to activate CD4 + effector T cells and instead promote Treg expansion, compromising ICI efficacy. Their presence in the pancreatic cancer TME correlates directly with resistance to ICIs [186]. Regulation of immune checkpoint ligand expression is another mechanism of CAF-mediated ICI resistance. CAFs can directly express PD-L1 and PD-L2, which bind to PD-1 on T cells to inhibit their function. In colorectal cancer, for instance, CAF-derived PD-L2 significantly impairs T cell activation by anti-PD-1 antibodies [187]. Alternatively, CAFs can indirectly upregulate PD-L1 on tumor cells through paracrine signals. In melanoma and colorectal cancer, CAF-secreted CXCL5 has been shown to induce tumor cell PD-L1 expression [160]. The formation of a physical barrier by CAFs is a critical factor limiting ICI efficacy in densely fibrotic tumors. Through extensive deposition and remodeling of the ECM, CAFs create a dense physical shield that excludes T cells from the tumor islets. TGF-β signaling drives CAFs to overexpress ECM components, leading to TME stiffness that can reach 20 kPa, far exceeding the 0.05–5 kPa range of normal tissue, and directly obstructing T cell migration [188, 189]. Preclinically, high expression of TGF-β-responsive ECM genes strongly predicts ICI failure. Combining TGF-β blockade with anti-PD-L1 therapy enhances T cell infiltration and improves outcomes [188, 190]. The CXCL12/CXCR4 axis also contributes to T cell exclusion, and its inhibition with agents like NOX-A12 (a CXCL12 inhibitor) can improve T cell infiltration when combined with anti-PD-1 [191]. The functional heterogeneity among CAF subtypes necessitates careful selection of combination strategies. This functional heterogeneity drives the divergent impacts of CAF subsets on ICI response. FAP + CAFs are highly immunosuppressive, characterized by high CXCL12 and CCL2 expression. Therapeutic strategies targeting FAP + CAFs, such as FAP-CAR-T cells or FAP-directed antibody–drug conjugates, show enhanced efficacy when combined with ICIs [192, 193]. In contrast, myofibroblastic CAFs (myCAFs, marked by α-SMA) can exhibit tumor-restraining properties in certain contexts; their nonspecific ablation at early stages may inadvertently exacerbate immunosuppression and reduce ICI effectiveness [94]. Overall, precise targeting of specific immunosuppressive CAF subpopulations is crucial for improving ICI outcomes. In conclusion, CAFs drive ICI resistance through a multifaceted network of mechanisms. Combining ICIs with agents targeting key CAF-associated pathways (e.g., TGF-β, CXCL12/CXCR4) or specific immunosuppressive subtypes (e.g., FAP + CAFs) provides a promising approach to overcoming therapeutic resistance in stroma-rich cancers. Impact of CAFs on ACT ACT, which includes T cells engineered with tumor-reactive T cell receptors (TCRs) or chimeric antigen receptors (CAR-T cells), is a key modality in cancer immunotherapy. While showing remarkable success in hematological malignancies [194], its efficacy in solid tumors is still limited. CAFs, as essential components of the TME, are major contributors to this resistance (Fig. 5). Therapeutic strategies targeting CAFs in cancer treatment. This figure outlines three core therapeutic approaches to target CAFs: 1) Depleting tumor-promoting/immunosuppressive CAF populations (e.g., FAP-targeted CAR-T cells, immunotoxins, or DNA vaccines); 2) Blocking fibroblast recruitment and activation (e.g., inhibitors of TGF-β, FAK, FGFR, or CXCL12/CXCR4 signaling); 3) Shifting the balance of CAF subpopulations or normalizing CAF function (e.g., agonists of vitamin D receptor or angiotensin pathways). Additionally, CAFs or mesenchymal stem cells (MSCs) can be engineered to deliver nanomedicines (prodrugs, drugs, or toxic agents) directly to tumors, overcoming physical barriers and enhancing therapeutic precision. Adapted from[199], Copyright © 2020 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) CAFs present a significant barrier to ACT success through several interconnected mechanisms. First, the CAF-derived and remodeled ECM creates a formidable physical barrier. This dense stroma prevents ACT products, despite their cytotoxic potential, from efficiently infiltrating the tumor core and contacting their target cells [195]. The expression profile of TGF-β-driven ECM genes in CAFs strongly predicts resistance to various immunotherapies, including ACT [188, 196]. Second, CAFs directly inhibit ACT cell function. Notably, CAFs present tumor antigens via MHC-I while expressing surface ligands like PD-L2 and FASL. This combination delivers potent inhibitory and apoptotic signals to antigen-specific CD8 + cytotoxic T cells, including ACT cells targeting the same antigen, leading to their apoptosis [197]. CAF secretion of PGE2 can also suppress NK cell activity, which are emerging as candidates for ACT [146, 147]. Furthermore, CAFs directly express PD-L1/PD-L2, engaging PD-1 on ACT cells, and can upregulate PD-L1 on tumor cells via factors like CXCL5 or exosomal miR-92, amplifying immune checkpoint-mediated inhibition of the infused cells [160]. Beyond direct effects, CAFs impair ACT by fostering a broadly immunosuppressive TME. ApCAFs, which express MHC-II but lack co-stimulatory molecules, fail to prime effector T cells and promote the expansion of CD4 + CD25high Tregs, suppressing local immune responses [186]. By secreting chemokines like CXCL12 and cytokines such as IL-6, CAFs recruit immunosuppressive populations, including MDSCs and TAMs, which further impair ACT cell function and persistence [125, 182]. Targeting CAFs and their associated pathways offers a strategic opportunity to enhance ACT efficacy in solid tumors. Combining TGF-β blockade with anti-PD-L1 therapy improves ACT cell responses [188]. Disrupting downstream TGF-β signaling by inhibiting NOX4 or neutralizing CXCL12 with the aptamer NOX-A12 improves ACT cell trafficking and function within tumors [191]. Moreover, using CAR-T cells designed to target FAP + CAFs disrupts stromal architecture and mitigates local immunosuppression, creating a more permissive TME for other ACT approaches [192, 193]. Based on current findings, these mechanisms suggest that the failure of ACT in desmoplastic tumors results not from a single barrier but from the synergistic effects of physical, soluble, and cellular barriers. Basic experiments typically involve scRNA-seq identification of CAF subtypes, spatial imaging of tumor slices in vivo (α-SMA/CD8/Foxp3 ratio), survival curves, and metastatic foci quantification to assess the extent of barrier damage to ACT [81, 198]. Notably, nonspecific depletion of certain early protective CAF subtypes (such as rCAFs) may exacerbate tumor invasion, emphasizing the need for subtype-specific targeting [73]. In the future, accurately neutralizing these barriers with nanodelivery systems may lift CAF-induced constraints on ACT, transforming fibrotic tumors into "hot tumors" sensitive to adoptive cell therapy [44]. Translating CAF heterogeneity into intervention strategy Depletion CAFs’ heterogeneity influences the morphology of the matrix barrier, making depletion strategies a true test of our ability to identify specific CAF subtypes to target, rather than broadly eliminating the entire CAF population. This is crucial, as recent spatial omics studies have shown that different CAF subgroups occupy stable and reproducible spatial niches, immune neighborhoods, and exhibit distinct prognostic associations. Therefore, depletion strategies should be based on selecting patients according to dominant barrier types, rather than focusing solely on the overall CAF abundance [51, 60]. Cords et al. reported in a large-scale spatial proteomic study that CAFs do not constitute a single oncogenic population but instead comprise 11 distinct phenotypes with unique spatial distributions and clinical significance. Among these, tumor-like CAFs are significantly associated with poor survival, whereas inflammatory or IFN-responsive CAFs are more frequently linked to an inflammatory TME and improved clinical outcomes. This implies that the premise of effective CAF depletion is to first identify the detrimental subpopulations that dominate immune rejection. Otherwise, generalized removal may inadvertently harm beneficial or neutral stromal components [51]. This conclusion is further supported by large-scale cross-cancer spatial multi-omics analyses. Liu et al. integrated data from 10 cancer types and over 14 million cells, revealing significant differences in immune cell infiltration levels, spatial distribution, and patient survival across conserved spatial CAF subtypes. Their findings suggest that depletion strategies are most appropriate for tumors dominated by CAF subtypes that encapsulate tumor nests, restrict T-cell entry, or cooperate with myeloid suppressor cells to construct barriers, rather than being broadly applied to all CAF-enriched tumors [60]. From a mechanistic perspective, the most compelling evidence for selective depletion currently centers on LRRC15-positive myofibroblastic CAFs. Krishnamurty et al. reported in Nature that TGF-β-driven LRRC15-positive CAFs represent a pathogenic matrix subset that evolves from homeostatic fibroblasts. Selective removal of this population can substantially reduce overall CAF abundance, recalibrate matrix composition toward a more homeostatic fibroblast profile, alleviate direct inhibition of tumor-infiltrating CD8-positive T cells, and thereby markedly enhance the antitumor efficacy of anti-PD-L1 therapy [52]. The key insight from this study extends beyond the utility of LRRC15 as a target. It provides the first functional evidence that, when immunotherapy resistance is primarily driven by a myCAF/LRRC15-positive compressive, embedded, and TGF-β-dependent matrix barrier, subtype-specific depletion is more rational than pan-CAF intervention and is more likely to synergize with immune checkpoint blockade (ICB) mechanisms [52]. It is worth noting that the therapeutic value of targeting LRRC15 extends beyond pancreatic cancer. Further studies in lung cancer conducted have shown that LRRC15-positive CAFs constitute a tumor-specific subgroup. Genetic knockout of Lrrc15 inhibits tumor progression by reducing ECM production by CAFs, suppressing CD206-positive macrophage polarization, and enhancing CD8-positive T-cell activation. This suggests that the benefits of certain CAF depletion strategies derive not only from physical barrier disruption but also from simultaneously weakening the immunosuppressive circuit formed by CAFs and macrophages [200]. A similar mechanism has been observed in glioblastoma. For example, a study found that LRRC15-overexpressing CAF subpopulations are enriched in patients unresponsive to anti-PD-1 therapy. LRRC15 promotes IL-8 expression via the FAK/SRC/NF-κB pathway, thereby driving macrophage infiltration and M2-like polarization. TGF-β1 secreted by M2 macrophages in turn enhances LRRC15 expression in CAFs, forming a positive feedback loop. Therefore, targeting LRRC15 can enhance sensitivity to PD-1 blockade, indicating that selective depletion remains effective even when the barrier core involves not only ECM but also an immunosuppressive field generated by CAF-myeloid cell interactions [201]. Similar to LRRC15, FAP-positive CAFs represent another depletion target closer to clinical translation. However, evidence accumulated over the past five years indicates that FAP is not a universal CAF master switch but rather a context-dependent danger marker that must be interpreted within specific TMEs [202–205]. On the one hand, large-scale tissue and clinical cohort analyses have shown that FAP is highly expressed across various solid tumors, including pancreatic, breast, esophageal, and lung cancers. Moreover, in a meta-analysis of 29 cohorts from 12 atezolizumab-related clinical trials, high FAP expression was consistently associated with worse prognosis. These findings provide a practical basis for using FAP-positive CAF abundance as a screening criterion to identify patients who may benefit most from prioritized depletion strategies [205]. On the other hand, the true potential of FAP targeting is not demonstrated by early non-selective, systemic clearance attempts but rather by more refined, localized, or controllable depletion platforms [202–204]. For example, FAP-targeted near-infrared photoimmunotherapy (NIR-PIT) reported selectively kills FAP-positive CAFs without substantially damaging adjacent normal cells in CAF-rich tumor models. This is accompanied by tumor shrinkage, increased CD8-positive T-cell infiltration, and alleviation of local immunosuppression. This suggests that spatially restricted depletion is more likely to achieve a favorable balance between efficacy and safety than systematic sweeping of the FAP-positive dominant local immune rejection barrier [204]. Similarly, based on the concept of “priming before attacking,” a 2023 study employed TALEN-edited, low-immunogenicity universal FAP-targeted CAR-T (FAP-UCAR-T) cells. This approach achieved CAF depletion, fibrosis reduction, and enhanced tumor infiltration in a triple-negative breast cancer model. When combined with mesothelin-targeted CAR-T cells and anti-PD-1 therapy, this strategy further reduced tumor burden and prolonged survival. These results indicate that, in the context of dense matrix barriers that restrict CAR-T cell infiltration, CAF depletion is more suitable as a pretreatment or priming approach rather than as a standalone therapy [202]. This concept was further validated in a dual-targeting design for pancreatic cancer. Wehrli et al. reported that mesothelin-targeted CAR-T cells engineered to secrete anti-FAP/anti-CD3 T-cell engaging antibody molecules (TEAMs) simultaneously eliminate tumor cells and CAFs in patient-derived and mouse PDAC models. The dual-targeting approach outperformed single-target strategies directed against either tumor cells or stroma alone. These findings emphasize that, in highly heterogeneous tumors with dual physical and immunological barriers, CAF depletion should be integrated into combined modalities that simultaneously target tumor cells rather than pursued as a monotherapy [203]. However, CAF heterogeneity also indicates that not all CAFs should be removed. Kobayashi et al. observed in non-small cell lung cancer (NSCLC) patients and mouse models that Meflin-positive CAFs are associated with improved immune checkpoint blockade (ICB) response, greater CD4-positive T-cell infiltration, and enhanced vascularization. This suggests that Meflin-positive CAFs represent a cancer-restraining or at least immune-supportive subpopulation. Therefore, in tumors with a high proportion of such CAFs, broad-spectrum depletion may inadvertently weaken existing immune-favorable structures. In other words, the primary risk of depletion strategies lies not in insufficient potency but in inaccurate subtype classification [206]. Therefore, when translating CAF heterogeneity into therapeutic decision-making, CAF depletion is not a universal solution but a precise intervention best suited for tumors enriched with detrimental subpopulations, exhibiting clear spatial embedding, harboring pathogenic markers such as FAP or LRRC15, and displaying immunotherapy resistance driven primarily by matrix entry barriers or CAF-myeloid suppressor circuits. In contrast, when protective CAFs such as Meflin-positive subsets or CAF functions oriented toward antigen presentation and vascular support predominate, simple depletion may not be superior to phenotype reprogramming, deactivation, or stratified combination therapies [51, 202–206]. Normalization Unlike depletion strategies that seek to remove CAFs from the tumor ecosystem, normalization aims to restore the imbalanced CAF network toward a homeostatic state. This approach limits excessive fibrosis while facilitating immune cell infiltration without amplifying myeloid-mediated immunosuppression. Therefore, normalization is particularly suitable for tumors exhibiting clear CAF heterogeneity and harboring potential protective or plastic CAF subpopulations [51, 60]. The theoretical foundation for this approach is grounded not in abstract reasoning but in recent spatial and single-cell omics evidence. Cords et al. demonstrated that distinct CAF phenotypes exhibit unique spatial localization, immune neighborhoods, and prognostic significance. Some inflammatory or IFN-responsive CAFs are not linked to the poorest outcomes, whereas highly contractile and tumor-like CAFs are more consistently associated with immune rejection and inferior survival. This indicates that certain CAF components are worth preserving and functionally correcting rather than eliminating [51]. The Meflin-positive remodeling CAF (rCAF) axis best illustrates this principle. Kobayashi et al. confirmed in both mouse models and human samples that Meflin-positive CAFs correlate with improved immune checkpoint blockade (ICB) responses, increased CD4-positive T-cell infiltration, and enhanced vascularization. These findings suggest that certain CAFs are not drivers of barriers but may function as endogenous brakes that limit pathological barrier formation [206]. Building on this understanding, Owaki et al. proposed a “pre-treatment normalization” strategy. The retinoic acid derivative Am80 does not merely reduce CAF numbers but instead promotes expansion of the Meflin-positive rCAF subpopulation, improves antibody delivery, and drives TAMs toward an M1-like phenotype. This enhances the efficacy of anti-PD-L1 therapy in clear cell renal cell carcinoma (ccRCC), urothelial carcinoma (UC), and multiple mouse tumor models. Moreover, this effect requires Am80 administration prior to ICB and is dependent on Meflin expression. These observations indicate that normalization is not a non-specific softening of the matrix but rather a precisely timed intervention that exploits the plasticity window of specific CAF subpopulations [207]. The same principle is evident in the vitamin D/VDR signaling pathway in pancreatic cancer. Kim et al. demonstrated that calcipotriol pretreatment disrupts myofibroblast integrity in three-dimensional PDAC models, enlarges vascular lumens, enhances immune cell infiltration, and produces stronger tumor inhibition with reduced T-cell exhaustion when combined with oncolytic viruses. These results demonstrate that when CAF-driven barriers are dominated by high-density fibrosis and low immune infiltration, it is more effective to first loosen and remodel the matrix—thereby opening a therapeutic window—before integrating immune effector therapies, compared with direct CAF clearance [208]. However, normalization is not always beneficial without trade-offs. Clinical trials of oral paricalcitol combined with gemcitabine plus albumin-bound paclitaxel, as well as a randomized phase II study, indicate that the clinical translation of vitamin D receptor (VDR) agonists does not automatically confer immune benefits [209, 210]. In the latter study, intravenous paricalcitol failed to significantly improve progression-free survival (PFS) or overall survival when combined with pembrolizumab maintenance therapy. The authors highlight insufficient duration of VDR activation as a key factor. This underscores that the clinical success or failure of CAF normalization depends not only on selecting the appropriate strategy but also on adequate drug exposure, intervention timing, and whether the patient harbors a VDR-responsive stromal compartment. [209, 210]. Furthermore, the goal of normalization is not restricted to reverting myCAFs to a quiescent state; it can also involve reprogramming pro-tumor CAFs toward anti-tumor phenotypes. In the study of Cumming et al., them identified a previously unrecognized interferon-responsive CAF (ifCAF) subpopulation in pancreatic ductal adenocarcinoma (PDAC). These ifCAFs exhibit tumor-suppressive properties and can be induced by STING agonists. More importantly, induction of ifCAFs not only suppresses pro-invasive tumor features but also reprograms neutrophils toward an anti-tumor phenotype. This finding suggests that, in myeloid-dominated immunosuppressive barriers, normalization can aim to rebuild beneficial stromal-innate immune interactions rather than simply decreasing fibroblast abundance [211]. Similarly, in a fibrotic gastric cancer model, dual PDGFRα/β blockade did not rely on CAF depletion. Instead, it downregulated the CAF-driven CXCL1/CXCL3/CXCL5/CXCL8-mediated chemotactic program through stromal reprogramming, reduced polymorphonuclear myeloid-derived suppressor cell (PMN-MDSC) accumulation, and restored lymphocyte infiltration, ultimately synergizing with anti-PD-1 therapy. This indicates that when the primary CAF-mediated obstacle is a chemokine-driven myeloid barrier rather than simple physical isolation, the value of normalization lies in reprogramming the cold and fibrotic microenvironment into one that is immunologically recognizable and accessible to the immune system [212]. Therefore, when translating CAF heterogeneity into therapeutic decision-making, normalization is not universally suitable for all CAF-rich tumors but is most appropriate for cases that simultaneously satisfy the following criteria: first, the presence of Meflin-positive, IFN-responsive CAFs, or other protective or plastic stromal subpopulations within the tumor; second, the dominant barriers involve fibrosis-induced compression, restricted infiltration of antibodies and immune cells, or reversible inhibitory circuits mediated by CAF-myeloid cell interactions; third, the therapeutic regimen requires a matrix-priming phase to improve the efficacy of subsequent immune checkpoint blockade (ICB), oncolytic virus therapy, or drug delivery [206, 208, 211, 213]. In other words, normalization is valuable within the framework of CAF heterogeneity not because it is gentler, but because it respects the biological reality that not all CAFs should be eliminated. Often, the most effective approach is to correct dysfunctional CAF states rather than eradicate the entire stromal compartment [206, 208, 211, 213, 214]. Reprogramming Depletion can be compared to demolishing walls, while normalization involves straightening misaligned walls. In contrast, reprogramming entails altering the function of CAFs that once supported tumor "guarding" into a state that can "open the door" or actively facilitate immunotherapy. The effectiveness of reprogramming within the context of CAF heterogeneity lies in the fact that CAFs are not a uniform, terminally differentiated population. Instead, they consist of reversible, plastic functional subsets that can be induced into immune-supportive states [206, 207, 215]. This principle is first illustrated by Meflin-positive CAFs. Miyai et al. reported that a higher proportion of Meflin-positive CAFs in non-small cell lung cancer (NSCLC) correlates with improved patient response to immune checkpoint blockade (ICB) monotherapy, along with increased CD4-positive T-cell infiltration and expanded vascular area. More importantly, forced overexpression of Meflin in CAFs in mouse models enhanced the efficacy of anti-PD-1 therapy. This suggests that certain CAFs do not require elimination but can instead be redirected toward phenotypes that promote immune cell infiltration and tissue perfusion [206]. Building on this logic, Owaki et al. demonstrated that the synthetic retinoic acid derivative Am80 induces pre-conditioning reprogramming of both CAFs and macrophages. It increases the Meflin-positive remodeling CAF (rCAF) subpopulation and drives TAMs toward an M1-like phenotype, thereby enhancing ICB responses in clear cell renal cell carcinoma (ccRCC), urothelial carcinoma (UC), and multiple mouse tumor models. These findings indicate that reprogramming is not limited to single-target inhibition but instead rewrites the broader mesenchymal-myeloid coupling network by exploiting CAF plasticity [207]. More radical functional reprogramming evidence comes from the “turning foes into friends” strategy proposed by Geng et al. In highly fibrotic breast cancer models, they employed spatiotemporally controllable nanosystems to reprogram immunosuppressive CAFs into antigen-presenting cell (APC)-like cells in situ. These engineered CAFs expressed co-stimulatory molecules, promoted antigen-specific CD8-positive T-cell activation, and locally secreted PD-L1 traps to augment checkpoint blockade. Consequently, the approach markedly enhanced tumor control and induced long-term immune memory, effectively converting CAFs from barrier-forming cells into immune-effector amplifiers [215]. In tumors characterized by dense fibrosis and poor T-cell penetration, reprogramming can dismantle the ECM barrier by altering CAF phenotypes without the risks associated with broad-spectrum depletion. Yen et al. developed a gelatinase-responsive nanocarrier that co-delivered an anti-PD-1 antibody and the TGF-β receptor I inhibitor galunisertib. This system not only increased CD8-positive and CD4-positive T-cell infiltration but also reduced collagen deposition and α-SMA expression. Fate-mapping experiments revealed that the treatment inhibited the differentiation of PDGFRα-positive fibroblasts into α-SMA-positive myofibroblasts. These results suggest that the core mechanism involves blocking the progression of fibroblasts toward a pathogenic myCAF phenotype, thereby converting immune-excluded tumors into an immune-accessible state (Fig. 6) [216]. Reprogramming CAFs to dismantle the ECM barrier and enhance immune accessibility. In tumors with dense fibrosis and poor T-cell penetration, reprogramming strategies can dismantle the ECM barrier by altering CAF phenotypes, avoiding the risks associated with broad-spectrum depletion. Yen et al. developed a gelatinase-responsive nanocarrier that co-delivered an anti-PD-1 antibody and the TGF-β receptor I inhibitor galunisertib. This therapeutic system enhanced the infiltration of CD8 + and CD4 + T cells while simultaneously reducing collagen deposition and α-SMA expression, which are markers of fibroblast activation. Fate-mapping experiments demonstrated that the treatment inhibited the differentiation of PDGFRα-positive fibroblasts into α-SMA-positive myofibroblasts. These findings suggest that the core mechanism of this strategy involves preventing the progression of fibroblasts toward a pathogenic myCAF phenotype, thereby converting immune-excluded tumors into immune-accessible states. This approach illustrates a promising strategy for enhancing tumor immunotherapy by reprogramming CAFs and remodeling the TME. Adapted from [216], Copyright © 2025 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) Similarly, the cathepsin B-responsive polymeric dasatinib prodrug developed by Zhang et al. does not eliminate CAFs but instead reprograms them toward a quiescent-like state. This inhibits ECM biosynthesis, enhances deep drug penetration and cytotoxic T lymphocyte (CTL) infiltration, and ultimately increases sensitivity to PD-1 blockade in the 4T1 breast cancer model. These results suggest that, for barriers dominated by dense matrix and low permeability, resetting CAFs to a low-ECM-output state is often more translationally relevant than direct depletion [217]. Another form of reprogramming, closer to niche-level rewriting, targets the PDGF-PDGFR axis. Ishimoto et al. demonstrated in fibrotic gastric cancer that PDGF stimulation upregulates CAF-derived chemotactic programs, including CXCL1, CXCL3, CXCL5, and CXCL8. This drives polymorphonuclear myeloid-derived suppressor cell (PMN-MDSC) accumulation and confers resistance to anti-PD-1 therapy. Dual PDGFRα/β blockade reverses this immunosuppressive ecology via stromal reprogramming at the single-cell and spatial transcriptomic levels, generating synergistic antitumor effects with anti-PD-1 therapy. These findings indicate that when the core CAF-mediated barrier involves myeloid cell recruitment through inflammatory chemotactic programs, effective reprogramming does not merely reduce CAF numbers but rather halts the continuous emission of erroneous chemotactic signals by CAFs [217]. It is worth noting that reprogramming does not always revert CAFs to a normal fibroblast state; in some cases, it actively induces novel anti-tumor CAF phenotypes. Huang et al. identified interferon-licensed CAFs (ilCAFs) characterized by high IRF1 expression in rectal cancer. These ilCAFs are enriched in tumors that respond favorably to radiotherapy and recruit T cells and DCs via CCL4 and CCL5. Activation of IFN-γ/STING signaling further promotes polarization of these CAFs and augments radiotherapy-induced anti-tumor immunity. This suggests that CAF reprogramming can be strategically designed to redirect originally neutral or pro-tumor CAFs toward an immune-permissive state [218] (Fig. 7). Similarly, Cumming et al. analyzed the heterogeneity of FAP-positive cells in pancreatic ductal adenocarcinoma (PDAC) and identified a previously underappreciated interferon-responsive CAF (ifCAF) subpopulation. This subtype exhibits tumor-restraining properties and can be induced by specific stimuli. These findings further demonstrate that the primary clinical insight from CAF heterogeneity is not whether CAFs should be targeted but whether pro-tumor CAF programs can be reprogrammed into interferon-driven, immune-supportive, or even tumor-suppressive programs [211]. Therefore, when translating CAF heterogeneity into therapeutic decision-making, reprogramming is most suitable for three scenarios: first, tumors harboring favorable CAF states that can be amplified or induced, such as Meflin-positive CAFs, ifCAFs, and ilCAFs; second, the dominant barrier stems not from CAF abundance but from aberrant functional programs, such as myofibroblastic ECM assembly, TGF-β-driven immune suppression, or CXCL1/CXCL3/CXCL5/CXCL8-mediated myeloid recruitment; third, the therapeutic objective extends beyond enabling T-cell entry to actively engaging the matrix in the orchestration and amplification of anti-tumor immunity [207, 211, 215, 216]. In other words, reprogramming offers a more precise, “surgical” approach to CAF intervention than depletion. It acknowledges the pathogenic potential of CAFs while leveraging their plasticity to reshape the matrix from an immunotherapy barrier into a supportive scaffold that actively facilitates and amplifies anti-tumor immune responses [219, 220]. Mechanism of CAF reprogramming and immune response enhancement in rectal cancer. The figure illustrates the role of interferon-licensed CAFs (ilCAFs) in promoting radiotherapy (RT)-induced anti-tumor immunity in rectal cancer. Before treatment (Pre-treatment), tumor-associated myofibroblastic CAFs (myCAFs), inflammatory CAFs (iCAFs), and immunoregulatory CAFs (ilCAFs) interact with the TME. During RT and STING agonist treatment, the activation of IFN-γ/STING signaling within ilCAFs results in the increased expression of IRF1 and phosphorylation of STAT1, leading to the secretion of CCL4 and CCL5. These chemokines recruit CD8 + T cells and dendritic cells (DCs), promoting the immune response within the TME. In post-treatment (Post-treatment), the recruited immune cells, particularly CD8 + T cells, release Granzyme B, further enhancing tumor cell elimination. The reprogramming of CAFs, especially through STING agonist activation, shifts the TME to a more immune-permissive state, augmenting the overall effectiveness of radiotherapy. Adapted from [218], Copyright © 2020 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) Barrier-matched design principles for CAF-targeted nanomedicine The TME is a complex and dynamic ecosystem, where CAFs simultaneously establish physical, soluble, and cellular immune barriers that hinder effective anti-tumor immunity. As reviewed above, these barriers interact and reinforce one another, creating a formidable obstacle to immunotherapy and conventional drug delivery. Overcoming them requires a comprehensive understanding of CAF heterogeneity, spatial distribution, and functional plasticity. Barrier-matched design principles provide a conceptual framework for CAF-targeted nanomedicine, emphasizing precise alignment of nanoplatforms with the specific barrier type—physical, soluble, or cellular—present in a given tumor. By integrating targeted delivery, responsive release, and multifunctional payloads, these principles aim to convert “cold” and treatment-resistant tumors into immunologically “hot” and therapeutically accessible states. In the following sections, we discuss four design objectives that embody these principles: overcoming the physical barrier, neutralizing soluble-factor-mediated suppression, rewiring the cellular immune barrier, and enabling subtype-aware CAF depletion, normalization, or reprogramming. Together, these objectives provide a roadmap for rational nanomedicine design that addresses the multi-layered defense established by CAFs and guides translation from preclinical models to clinical applications (Table 1). Table 1Barrier-matched design principles for CAF-targeted nanomedicine: strategies, mechanisms, and applicationsDesign strategyCore MechanismKey features / advantagesExample materials / carriersTypical applicationsRefCAF Targeting (Active targeting)Utilizes CAF-specific surface markers for selective binding and internalizationEnhances targeted uptake, reduces off-target effects; increases accumulation in CAF-rich areasFAP antibody-modified NPs, α-SMA recognitionTargeting CAFs, reducing immunosuppression[36]Physical Barrier Penetration DesignAdjusting particle size/surface modifications to enhance ECM penetrationSmaller size or ECM degradation improves penetration, facilitates drug and immune cell entrypH-sensitive nanoparticles, enzyme-responsive carriersEnhance deep penetration capability[87]CAF Depletion StrategyDirect killing of CAFs or gene silencingDecreases ECM formation, relieves immune suppressionsiRNA/drug-loaded carriersReduces CAF-mediated barriers[221, 222]CAF ReprogrammingChanging CAF phenotype to non-tumor-promoting/immune-supportive typeCorrects CAF function, avoids harmful side effects from total depletionDNA/miRNA carriers, signal pathway inhibitorsGenerating immune-friendly microenvironment[221, 223]ECM Degradation AssistanceNanocarriers carry ECM-degrading enzymes or signal blockersDegrades dense ECM, overcomes physical barriersCollagenase carriers, MMP-responsive systemsImproves cell/drug infiltration[221, 223, 224]Immunomodulatory Drug DeliverySimultaneously delivers immune-activating molecules with drugsSynergistic CAF-mediated immunosuppression reversal, enhanced immunotherapy efficacyImmune checkpoint inhibitors + CAF carriersCombination therapy for improved response[225, 226] Barrier-matched design principles for CAF-targeted nanomedicine: strategies, mechanisms, and applications Overcoming the physical barrier In solid tumors dominated by CAF, the physical barrier is a real therapeutic bottleneck rather than an abstract concept. It consists of collagen, hyaluronic acid, contractile myofibroblast-like CAF, and a resulting high-density, high-tension, low-perfusion matrix network. This barrier limits deep delivery of nanomedicine and effector T cells and amplifies T cell exhaustion through mechanical compression, spatial isolation, and ligand-dependent inhibitory signals. Therefore, transforming the CAF-related matrix from a tumor-protective layer into a passable, perfusable, and immune-infiltratable microenvironment is the primary objective in designing CAF-targeted nanomedicine. This transformation is also essential for subsequent ICB, CAR-T, and in situ immune activation to be effective [52, 166, 227, 228]. Recent studies indicate that this goal should not be interpreted as mere CAF removal. Instead, it requires a barrier-matched engineering strategy. By targeting dominant matrix components and physical limitations in different tumors, a combined approach—ECM degradation, solid stress relief, CAF reprogramming rather than complete ablation, and synchronous immune amplification—should be employed to avoid metastasis or tissue damage from excessive matrix removal [52, 166, 227]. For example, Zhang et al. developed a lipid/polymer composite nanoparticle, LPPR, to deliver the relaxin plasmid in a triple-negative breast cancer model. The core combines positively charged polymeric methaformin and pRLN, while stability is enhanced by an outer layer of lipid-polyglutamic acid. The particle is approximately 205.5 nm in size, demonstrating good penetration in 4T1 luc/CAFs tumor spheres. It reduces expression of profibrotic factors, attenuates CAF activation, and limits interstitial deposition in vivo. Ultimately, it increases cytotoxic T cell infiltration in the tumor by 2.2-fold and enhances the therapeutic effect of aPD-L1. This work demonstrates that loosening the dense CAF-driven matrix before immunotherapy sensitization aligns better with treatment strategies for matrix-rich tumors than simply delivering immune drugs into the tumor [229] (Fig. 8). Nanoparticle-mediated CAF reprogramming and enhancement of anti-tumor immunity in triple-negative breast cancer (TNBC). (A) Schematic illustration of the preparation of LPPR nanoparticles (PGA/PolyMet-pRLN) by encapsulating the relaxin plasmid and using DOTAP and Chol to modify the surface, followed by intraperitoneal injection into a mouse model. (B) Activated CAFs are targeted with relaxin plasmid-loaded nanoparticles, which reprogram these cells and reduce their suppressive activity. The nanoparticles further enhance the efficacy of αPD-L1 therapy by promoting tumor cell recognition and CD8 + T cell activation. (C) In aggressive TNBC, CAFs contribute to a desmoplastic, immune-suppressive microenvironment. However, the combination of LPPR nanoparticles and αPD-L1 therapy reshapes the TME, reducing desmoplasia and promoting immune responsiveness. This approach enhances T cell infiltration, reshapes the TME by decreasing immunosuppressive factors such as TGF-β, FAP-α, and α-SMA, and increases immune responses (IFN-γ, TNF-α) and anti-tumor immunity. The resulting tumor is more responsive to immune therapy, providing a promising strategy for treating TNBC. Adapted from [229], Copyright Another example is the pH/MMP-2 dual-sensitive polymer/calcium phosphate hybrid nanocarrier developed by Xiao et al. for hepatocellular carcinoma. This system co-delivers hyaluronidase (HAase), IL-12, and anti-PD-L1 antibodies. Tumor acidity triggers CaP disintegration, releasing HAase and IL-12 to degrade hyaluronic acid-rich ECM and promote CTL expansion. MMP-2 mediates the release of α PD-L1 within the tumor, linking matrix dismantling, T cell expansion, and checkpoint blockade in a sequential process. Basic experiments showed that this design enhances CTL infiltration and effectively inhibits mouse HCC growth. For tumors enriched in HA and resistant to T cell penetration, pre-positioning an enzyme-catalyzed barrier removal module is a valuable nanodesign principle [230]. Hou et al. further advanced this perspective to the mechanical barrier itself. They constructed a fibrin nanocomposite hydrogel, LOS, for in-situ injection in residual tumor models after chemoimmunization, co-delivering losartan and iron oxide-related nanocomponents (FeOX@Gel). Sustained release of losartan reduced ECM deposition and associated solid stress, improving drug entry, enhancing OX-induced immunogenic cell death, and ultimately overcoming chemoresistance in immunotherapy. This indicates that for highly hardened tumors with limited perfusion and compressive physical barriers, simply targeting CAF cells is insufficient; stress unloading and mechanical normalization must be incorporated into nanomedicine design [231]. In addition, Li et al. provided indirect support through studies on metastatic colorectal cancer. Using ^68 Ga FAPI PET/CT, they dynamically monitored CAF burden and observed that high FAPI uptake correlated with reduced immune cell infiltration and poor immunotherapy response. In a mouse peritoneal/liver metastasis model, the TGF-β receptor inhibitor SB525334 first reduced CAF burden, and subsequent PD-L1/CTLA-4 double blockade with KN046 significantly increased T cell infiltration and efficacy. Notably, the authors suggested that a lead CAF/TGF-β inhibition of approximately seven days can achieve subsequent ICB sensitization. This implies that overcoming physical barriers does not require prolonged, strong CAF ablation but can be achieved via short-term, monitored, and reversible barrier modulation [232]. Under the principle that reprogramming is superior to one-size-fits-all removal, targeted nanoplatforms using CAF homology or FAP orientation have gained attention. Cheng et al. reported PMNP-SAB@RTM, which mixes red blood cell membranes with tumor-associated fibroblast membranes to coat porous magnetic nanoparticles, loaded with salvianolic acid B. This design enables local enrichment under an external magnetic field. Its advantages are not merely drug delivery, but selective reduction of TAF-related ECM generation, promotion of T cell infiltration, transition of TNBC from cold to hot, and enhancement of PD-1/PD-L1 small molecule inhibitor BMS-1 efficacy. This biomimetic approach demonstrates that physical barrier removal can be achieved through targeted CAF functional reprogramming rather than direct matrix degradation, potentially improving safety [233]. Similarly, Shen et al. developed a CAF-targeted PLGA nanoemulsion for colorectal cancer, co-delivering DOX with siRNA targeting HGF. Aptamer modification improved CAF selectivity, inducing partial CAF apoptosis and reducing ECM deposition while inhibiting residual CAF HGF expression. This disrupted CAF self-amplification and pro-invasive feedback after chemotherapy, ultimately enhancing tumor permeability and inhibiting proliferation, migration, and invasion. In tumors with high CAF secretion, dual-mode strategies combining ECM reduction and paracrine inhibition may better maintain an open tumor stroma than simple fibrosis-targeted approaches [234]. Although FAP-CAR-T is not strictly a nanomedicine, Liu et al. demonstrated its methodological relevance in pancreatic cancer. Sequential infusion of FAP-targeted CAR-T followed by CLDN18.2 CAR-T showed that CAF reduction decreased MDSC recruitment and increased accumulation of CD8 + T cells and CAR-T cells within tumors. This indicates that the physical barrier and immunosuppressive barrier in matrix-rich tumors are coupled by CAFs, forming a compound barrier. Therefore, CAF-targeted nanomedicine should evaluate not only collagen reduction but also T cell infiltration, immunosuppressive cell regression, and subsequent immunotherapy amplification [235]. In summary, CAF-targeted nanomedicine design for overcoming the physical barrier has evolved from simple fibrosis reduction to multi-layered strategies. First, relaxin, losartan, or TGF-β axis inhibitors reduce CAF activation, collagen deposition, tissue hardening, and compressive stress [229, 231, 232]. Second, enzymatic components such as HAase preferentially disassemble the hyaluronic acid/ECM network, opening penetration pathways for nanomedicine and T cells [230]. Third, FAP or CAF membrane homology enables selective delivery, limiting barrier modulation to the lesion microenvironment and increasing safety [233, 235]. Fourth, barrier removal is coupled temporally with ICB, immune activation, or cell therapy, translating physical pathway opening into immune benefits [232, 234, 235]. Current research highlights that precise, controllable, and temporal intervention in CAF-ECM mechanical networks is the key to transforming barrier tumors into treatable tumors. Neutralizing the soluble-factor barrier In the TME, CAFs form an invisible immunosuppressive barrier by continuously secreting soluble factors such as TGF-β, IL-6, and CXCL12. This soluble-factor barrier not only directly inhibits CD8⁺ T cell activation and infiltration but also amplifies tumor immune evasion by recruiting regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs), representing a major bottleneck in solid tumor immunotherapy [221, 236]. The second design objective for CAF-targeted nanomedicine focuses on neutralizing this soluble-factor barrier. Its core purpose is to deliver inhibitors, siRNA, or epigenetic modulators precisely to block CAF-secreted signaling pathways or reprogram CAF phenotypes. This approach relieves the chemical exclusion imposed on immune cells, lays the foundation for synergistic enhancement of nanodelivery systems and immune checkpoint blockers (ICBs), and ultimately improves the conversion of cold tumors to hot tumors [36, 86]. Recent studies have developed multiple barrier-matched nanostrategies. FAP-targeted PLGA nanoparticles (FAP-CAR-CM@PLGA-AB NPs) employ a dual mechanism: FAP-scFv mediates precise recognition and ABT-263 clears senescent CAFs, while nintedanib inhibits CAF activation. This system significantly downregulates immunosuppressive cytokines, including TGF-β, IL-6, and IL-10. In a mouse breast cancer model, it achieved a tumor inhibition rate of 86.7%, promoted extensive CD8⁺ T cell infiltration, and effectively reversed radiotherapy resistance [237]. Similarly, E8R8-modified liposomes (E8R8-Lip) target both FAP⁺ and α-SMA⁺ CAFs. By downregulating CAF-derived CXCL12, IL-6, and TGF-β, these liposomes alleviate tumor hypoxia, inhibit CAF–tumor cell crosstalk, and completely block lung metastasis in a 4T1 breast cancer model, while maintaining good biocompatibility [36]. Targeting IL-6 specifically, a liposomal system co-delivering nilotinib (CAF activation inhibitor) and doxorubicin blocked the JAK/STAT pathway in vitro, reducing IL-6 secretion by over 70% as measured by ELISA, thereby enhancing tumor cell sensitivity to chemotherapy. In vivo, this formulation inhibited peritoneal metastasis and prolonged survival in a gastric cancer model, demonstrating the direct therapeutic effect of neutralizing soluble-factor barriers [238]. Another representative design is the pH/heparinase-responsive nanocomposite (R(D)/H(S) NPs) loaded with SB431542 (TGF-β receptor inhibitor) and DOX. In the acidic TME, SB431542 is preferentially released to inactivate CAFs and reduce collagen I expression, followed by DOX penetration into the tumor core. Preclinical studies in pancreatic cancer xenografts showed a 50% reduction in collagen deposition and a threefold increase in CD8⁺ T cell infiltration, enhancing chemoimmunotherapy efficacy [239]. Moreover, the carrier-free nanoremodeling agent QN NP, a self-assembled quercetin-based system, normalizes CAFs by inhibiting the TGF-β/Smad2/3 pathway. In triple-negative breast cancer models, it significantly reduced TGF-β and IL-6 secretion, remodeled the ECM, and enhanced anti-PD-1 therapy response, achieving a tumor growth inhibition rate more than twice that of the control group [239]. To further address CAF heterogeneity, TET@ACNP, PEG-PCL biomimetic nanoparticles loaded with tetrahydroberberine, employ CAF-specific ligands for precise delivery. By simultaneously inhibiting TGF-β/Wnt/β-catenin and IGF2 signaling, they achieve multidimensional neutralization of the soluble-factor barrier. In a TNBC mouse model, this system enhanced docetaxel penetration, reduced MDSC proportions by 40% and Tregs by 35%, and ultimately achieved complete tumor regression under combined chemoimmunotherapy [240]. These examples collectively demonstrate that the barrier-matched design of CAF-targeted nanomedicines must balance targeting specificity (FAP/α-SMA ligands), responsiveness (pH/enzyme/photothermal triggers), and multifunctionality (co-delivery of inhibitors with chemotherapy or gene drugs). Preclinical evaluation typically includes cytokine ELISA and flow cytometry in vitro, tumor immunofluorescence analysis of CD8⁺/Foxp3⁺ ratios in vivo, and survival and metastasis assessments to quantify neutralization efficacy [87, 241]. In the future, integrating AI-assisted multi-omics screening with preclinical multi-organ chip models will further accelerate the translation of these nanostrategies from laboratory research to clinical applications, potentially overcoming the constraints imposed by soluble-factor barriers on cancer immunotherapy. Rewiring the cellular immune barrier In the TME, CAFs act as hidden "immune conductors," continuously recruiting and polarizing regulatory T cells (Tregs), myeloid-derived suppressor cells (MDSCs), and M2 TAMs through the CXCL12-CXCR4 axis, TGF-β/Smad signaling, and the IL-6/STAT3 pathway. Simultaneously, they directly inhibit infiltration and effector functions of CD8⁺ T cells and NK cells, forming a dynamic cellular immune barrier. This barrier exacerbates tumor immune escape and significantly reduces the response rate to immune checkpoint blockers (ICBs) in solid tumors [242, 243]. The third design objective of CAF-targeted nanomedicine focuses on rewiring this cellular immune barrier. Its primary goal is to achieve CAF depletion, reprogramming, or phenotype transformation through precise delivery of siRNA, metabolic inhibitors, or photodynamic agents. This approach reverses the recruitment of immunosuppressive cells, restores immune cell activity and spatial distribution, provides a key breakthrough for synergy between nanocarrier systems and ICBs, and ultimately transforms cold tumors into hot tumors responsive to immunotherapy [86, 221]. For example, some studies have developed several highly matched nanostrategies. FAP-targeted ferritin nanoparticles (α-Fe₂O₃, FAP-Z@FRTs) selectively kill CAFs and induce immunogenic cell death via FAP-specific single-chain antibody-mediated photodynamic therapy (PDT). In a 4T1 breast cancer mouse model, this system significantly reduces ECM deposition, promotes dendritic cell maturation and CD8⁺ T cell infiltration, and triggers anti-CAF-specific CTL responses, achieving ablation of distal tumors and synergizing with anti-PD-1 therapy [244]. Similarly, FAP-responsive lipid nanoparticles (FAP-LNPs) co-loaded with HK2 and MTCO1 siRNA block glycolysis and mitochondrial respiration in CAFs through metabolic reprogramming. In an RM-1 in situ prostate cancer model, combining FAP-LNPs with αPD-1 significantly reduced MDSC and Treg proportions by more than 40% and increased CD8⁺ T cell infiltration over threefold, achieving a tumor inhibition rate of 70% [245]. Biomimetic nanoparticles targeting the cellular immune barrier in triple-negative breast cancer (TNBC), such as TET@ACNP, use a PEG-PCL backbone loaded with tetrahydroberberine and surface-modified with anisamide ligands and CAF membranes. Homologous targeting of CAFs allows precise delivery and phenotype reshaping by simultaneously inhibiting TGF-β/Wnt/β-catenin and IGF2 signaling. In a TNBC mouse model, this system reduced recruitment of Tregs and MDSCs by 35–40% (assessed by IHC and flow cytometry), enhanced CTL infiltration and activation, and achieved complete tumor regression when combined with docetaxel [240]. Another representative design is carbon dot-based responsive nanocomposites (Pep-APCDs@Fe/DOX-LOS), in which a FAP peptide-sensitive shell dissociates at the tumor site, releasing losartan to degrade collagen and DOX/Fe to induce immunogenic cell death. In a breast cancer model, this system achieved an 84.8% tumor growth inhibition rate. ELISA and multicolor flow cytometry confirmed reduced Treg and MDSC proportions, increased T cell and NK cell infiltration, and complete prevention of lung metastasis [36]. Additionally, PLGA-based dual-targeted nanoparticles (FAP-CAR-CM@PLGA-AB NPs) integrate FAP-scFv recognition with ABT-263-mediated clearance of senescent CAFs, while nintedanib synergistically inhibits CAF activation. In a radiation-resistant breast cancer model, this system downregulated TGF-β, IL-6, and IL-10, enhanced CD8⁺ T cell infiltration, and achieved a tumor inhibition rate of 86.7% [36]. Collectively, these examples highlight that the design of CAF-targeted cellular barrier remodeling requires integrating targeting ligands (FAP/α-SMA/CAF membrane), stimulus responsiveness (enzyme/pH/light), and multifunctional payloads (siRNA, chemotherapy, or photosensitizers). Preclinical evaluation typically includes cytokine ELISA and flow cytometry analysis of in vitro CAF-immune cell co-cultures, immunohistochemistry or immunofluorescence of tumor tissues (CD8⁺/Foxp3⁺/CD11b⁺Gr-1⁺ ratios), as well as survival curves and metastatic nodule quantification to comprehensively assess immune reprogramming effects [87, 241]. Some strategies further enhance biocompatibility and reduce off-target toxicity through CAF membrane coating or homologous biomimetic design [246]. Moreover, CAF subtype-specific nano-delivery guided by spatial transcriptomics and validation in humanized mouse models is expected to accelerate clinical translation, overcoming the limitations imposed by the cellular immune barrier on cancer immunotherapy [87, 241]. Subtype-aware depletion, normalization, or reprogramming In the design of CAF-targeted nanomedicine, what truly needs to be eliminated is often not the CAF population itself but the specific pro-tumorigenic functional states that are tightly coupled with immune rejection, dense matrix formation, and therapeutic resistance. Therefore, “matching barriers by subtype and selecting interventions by function” should constitute the core principle of this design goal. On one hand, multi-omics studies have shown that CAFs do not constitute a single uniform population but can stably differentiate into distinct lineages, including immune-regulatory, mechanically responsive/matrix-remodeling, and homeostatic subsets. These lineages remain largely conserved across species and tissue origins. This conservation suggests that nanomedicine should abandon “pan-CAF clearance” as the default strategy and instead focus on specific barrier-driving mechanisms, such as excessive collagen deposition, chemotactic factor secretion, antigen-presentation bias, or spatial immune cell exclusion [247]. Correspondingly, a growing body of evidence indicates that indiscriminate CAF depletion can produce unintended adverse consequences. For instance, depletion of Endo180/MRC2-positive matrix-remodeling CAFs has been shown to significantly inhibit tumor growth and metastasis, highlighting that certain matrix-remodeling subpopulations represent actionable “danger nodes” [248]. On the other hand, the ZEB1-dependent myCAF program exerts a dual role in colorectal cancer by both restraining inflammatory spread and actively shaping physical and immunosuppressive barriers. Although Zeb1 deficiency enhances lymphocyte infiltration and improves immune checkpoint inhibitor (ICI) sensitivity, it simultaneously exacerbates inflammation-driven tumor progression. This duality underscores that decisions regarding CAF removal or retention must be guided by a triple stratification based on subtype, disease stage, and tumor molecular background [249]. This also explains why normalization and reprogramming have garnered increasing attention in recent translational research compared with simple depletion. In rectal cancer, for example, iCAFs are strongly associated with resistance to neoadjuvant chemoradiotherapy. IL-1-driven inflammatory polarization and senescence programs represent key mechanisms underlying this resistance. Inhibiting IL-1 signaling, blocking CAF senescence, or promoting matrix repolarization can restore therapeutic sensitivity. This evidence strongly supports prioritizing the design of nanocarriers capable of delivering modulators of the IL-1 axis, NF-κB pathway, or senescence-related programs rather than simply aiming to reduce overall CAF numbers [250]. It is worth noting that apCAFs should not be uniformly regarded as detrimental. A recent study demonstrated that apCAFs expressing high levels of MHC class II in gastric cancer are preferentially localized near tertiary lymphoid structures, where they enhance T-cell activation and predict improved immunotherapy responses [53]. However, spatial single-cell studies indicate that apCAFs comprise at least two distinct lineages: mesothelial-like (or epithelial-like) and fibroblast-like subpopulations. Certain subpopulations promote primary tumor formation, peritoneal metastasis, and therapeutic resistance via SPP1-related signaling programs. In non-small cell lung cancer (NSCLC) patients receiving neoadjuvant immunotherapy, IFN-γ-induced expansion of apCAFs has been linked to poor therapeutic efficacy via the JAK1/2–STAT1–IFI6/27 axis [66, 251]. These observations indicate that future nanomedicines should avoid the simplistic logic of “clear all apCAFs upon detection.” Instead, they should employ sophisticated, combination-targeting strategies capable of recognizing spatial localization, secretion profiles, and context-dependent signaling states. From a design principle perspective, subtype-aware CAF-targeted nanointerventions should satisfy at least three key requirements. First, in tumors dominated by a physical barrier driven by myCAFs or matrix-remodeling CAFs, priority should be given to weakening excessive collagen and fibronectin deposition as well as relieving solid stress, while preserving the minimal matrix support necessary for tissue homeostasis. Second, in tumors dominated by immune rejection driven by iCAFs, immunosuppressive CAFs, or specific oncogenic apCAFs, the focus should be on suppressing key immunosuppressive signals, including CXCL12, IL-6, TGF-β, IGF-2, complement components, and SPP1. Third, in CAF states exhibiting plasticity windows, priority should be given to reprogramming activated CAFs back toward a quiescent or immune-supportive phenotype rather than irreversibly depleting the entire population. This approach better aligns with the biological reality of CAF heterogeneity and reversible phenotypic transitions [53, 66, 247, 249]. Recent nanomedicine examples have begun to embody this principle. Wang et al. developed a CREKA-modified hydroxyethyl starch nanoplatform that specifically recognizes fibronectin highly expressed by CAFs. In triple-negative breast cancer models, this system achieves deeper drug penetration through localized CAF modulation, relief of mechanical compression, and improved perfusion and permeability. Its therapeutic value lies not merely in reducing CAF numbers but in preferentially dismantling the physical barrier driven by myCAF-like programs [252]. Fei et al. integrated CAF reprogramming and vascular normalization into a single nanosystem. In liver cancer models, this platform reversed activated CAFs to a quiescent state while simultaneously normalizing aberrant vasculature and improving oxygen delivery, thereby sequentially potentiating photodynamic therapy. This approach represents a paradigm in which normalization proves superior to pan-CAF clearance [253] (Fig. 9). Sequential therapy using AEAA-MSNS@RA/GA and MSNs@Ce6 for tumor microenvironment (TME) regulation. (A) Schematic illustration of the sequential therapy strategy. First, AEAA-MSNS@RA/GA nanoparticles are injected into the mouse model to target activated CAFs and deliver therapeutic agents. Afterward, MSNs@Ce6 nanoparticles are introduced, and photodynamic therapy (PDT) is applied using a laser to activate the Ce6 agent, enhancing the treatment effect. (B) TME regulation by the sequential therapy approach, where AEAA-MSNS@RA/GA nanoparticles target and inactivate activated CAFs in the TME. This results in a shift from an immunosuppressive environment, characterized by activated CAFs and immune-suppressive cells, to a more immunostimulatory environment, with a reduction in collagen deposition and activation of immunostimulatory cells, thereby improving anti-tumor immunity. The combination of targeted therapy and PDT effectively reprograms the TME, enhancing the overall therapeutic response. Adapted from [253], Copyright © 2023 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) A more subtype-aware example has emerged from recent studies on stromal tumors. Cheng et al. developed magnetic nanoparticles coated with a hybrid red blood cell membrane and tumor-associated fibroblast (TAF) membrane to deliver salvianolic acid B. This system inhibited TAF-mediated ECM production and promoted T-cell infiltration in triple-negative breast cancer (TNBC) models. When combined with PD-1/PD-L1 inhibitors, the treatment converted “cold” tumors into immunologically “hot” tumors. This outcome essentially represents functional reprogramming directed against CAF-dominated immune rejection barriers [233]. Moreover, Xian et al. reported a CAF cell membrane-biomimetic nano-regulator (TET@ACNP) that not only inhibits CAF activation and collagen secretion while promoting cytotoxic T-cell infiltration but also alleviates chemotherapy resistance and immunosuppression by suppressing Wnt/β-catenin and IGF-2–related signaling programs. This suggests that future nanosystems could simultaneously address both stromal and secretory barriers by reading multiple CAF functional outputs [240]. Going further, recent studies in pancreatic ductal adenocarcinoma (PDAC) have begun to use inhibition of specific pro-tumorigenic CAF subtypes as a direct design endpoint. Immunotherapy nanoparticles reported in Nano Today in 2025 simultaneously reduce the proportion of pro-tumorigenic subpopulations such as myCAFs and apCAFs, enhance drug penetration, increase CD8-positive T-cell infiltration, and decrease regulatory T cells (Tregs), thereby converting subtype-specific regulation into measurable immune benefits [254]. At the same time, inhibition of nicotinamide N-methyltransferase (NNMT) has been shown to attenuate CAF-driven myeloid-derived suppressor cell (MDSC) recruitment and restore immune checkpoint blockade (ICB) efficacy. These findings suggest from another perspective that the most promising CAF-targeted nanomedicines of the future will not simply identify a CAF marker and eliminate the entire population. Instead, they will integrate spatial localization, subtype recognition, functional readout, and reversible remodeling into a closed-loop strategy: selectively depleting pathogenic subpopulations, normalizing plastic but not inherently carcinogenic populations, and preserving or even amplifying the anti-tumor functions of subpopulations that already contribute to immune synergy. This integrated approach represents a more mature and biologically rational path to overcoming the stromal barriers that limit cancer immunotherapy [254, 255]. Design failure modes and translational constraints If the ideal outcome of CAF-targeted nanomedicine is dismantling barriers, opening pathways, and enabling sensitization, the most common challenge once this strategy advances to clinical translation is rarely whether the intervention is effective. Instead, the critical questions become which specific barrier is being dismantled, which CAF subpopulation is being targeted, and whether the tumor ecosystem will mount a compensatory response after disruption. Therefore, the purpose of this section is to advance CAF-targeted nanointerventions from conceptually feasible strategies to a more rigorous, clinically controllable level. The first major failure mode in CAF-targeted design is the erroneous treatment of highly heterogeneous, plastic, and conditionally protective CAF subpopulations as a uniform pathogenic target. Multi-omics studies have demonstrated that CAFs can be stably classified into at least three conserved categories: steady-state-like, mechanically responsive/matrix-remodeling, and immunomodulatory subsets. However, the functional roles of these subpopulations vary substantially depending on tumor type, disease stage, and even spatial location within the same lesion. This implies that broad-spectrum “pan-CAF” depletion strategies driven by single markers such as FAP, α-SMA, or fibronectin inherently carry a high risk of biological mismatch [247]. This risk is not merely theoretical but has been repeatedly confirmed in preclinical studies. In PDAC models, disruption of the PKN2 axis—which drives differentiation of pancreatic stellate cells into myCAF-like phenotypes, attenuates matrix contraction and fibroblast invasion. However, it paradoxically induces a switch from myCAFs to inflammatory CAFs, thereby promoting local tumor invasion. These findings indicate that even myofibroblast-like CAFs can give rise to more oncogenic alternative states upon perturbation [256]. Similarly, in pancreatic cancer, EGFR/ERBB2-activated myCAFs have been shown to drive metastasis, whereas LRRC15-positive myofibroblastic CAFs directly suppress CD8-positive T-cell function and limit the efficacy of PD-L1 blockade. Collectively, this positive and negative evidence underscores that the central challenge is not whether CAFs can be targeted but which specific CAF subtypes should be modulated, rather than treating the entire fibroblast ecosystem as a homogeneous entity [52, 257]. The second major failure mode arises from the mismatch between nanoparticle delivery physics and tumor histology. Although CAF-targeted nanomedicine often begins with the assumption that matrix-rich tumors are ideal candidates for nanosystems, the opposite is frequently true: tumors that depend heavily on dense matrix and elevated interstitial fluid pressure to sustain barriers are typically less permissive to nanoparticle extravasation, deep diffusion, and uniform intratumoral distribution. Current large-scale nanodesign strategies still rely on the EPR effect as the default premise. However, EPR is highly heterogeneous across patients, cancer types, and even between primary and metastatic lesions in the same individual [258–260]. In recent years, the concept of active transport and retention (ATR) has highlighted that clinical failures often stem not from insufficient targeting ligands but from an incomplete understanding of the actual mechanisms governing tumor endothelial translocation, matrix retention, and lymphatic efflux. Consequently, many “passive accumulation plus active recognition” strategies that perform well in murine models fail to translate reliably to human tumors [259]. The third failure mode occurs after nanoparticles enter the body: the protein corona fundamentally alters the biological identity of the particles, thereby modifying their circulation half-life, RES)clearance rate, cellular uptake preferences, and effective ligand exposure. Consequently, a nanosystem that achieves efficient CAF uptake in vitro through FAP targeting or CAF membrane homology recognition may fail to function according to its intended design in vivo. This discrepancy largely explains why preclinical results of CAF-targeted nanomedicines are frequently difficult to reproduce across different models [261, 262]. The fourth limitation stems from failure in model extrapolation. Many current CAF-targeted nanomedicines demonstrate positive outcomes primarily in subcutaneous xenograft tumors, immunodeficient mouse models, or relatively homogeneous syngeneic tumors. These models fail to recapitulate the vascular heterogeneity and organ-specific fibrosis characteristic of human tumors, nor can they replicate the dynamic, evolving interactions among CAFs, T cells, myeloid cells, and therapeutic pressure. Consequently, even when systems such as CREKA-modified hydroxyethyl starch nanomedicine, SAB-loaded biomimetic particles, or CAF membrane-biomimetic nano-regulators demonstrate favorable outcomes—including CAF reduction, enhanced infiltration, and improved ICI sensitivity—in TNBC models, these results cannot be assumed to translate with the same magnitude or direction in heterogeneous clinical populations [233, 240, 252]. The fifth constraint lies at the pharmaceutical and industrial level. The greater the pursuit of multi-responsive mechanisms, dual ligands, membrane coating, co-delivery, and temporally controlled release, the more complex the formulation becomes. This complexity renders it increasingly difficult to define and control critical quality attributes, maintain batch-to-batch consistency, and achieve scalable, cost-effective manufacturing under good manufacturing practice (GMP) conditions. Ultimately, this creates a substantial gap between laboratory precision and GMP-compliant manufacturability—a key reason why most tumor nanomedicines remain stalled at the proof-of-concept stage for extended periods [263, 264]. In addition, FAP) the most widely used clinical imaging target for CAFs, is not tumor-specific. Recent systematic reviews have shown that FAPI-PET exhibits substantial uptake in infections, inflammatory conditions, fibrotic diseases, and various benign lesions. This observation indicates that positive FAPI-PET signals do not necessarily reflect tumor-specific CAF enrichment. Consequently, all nanodelivery systems, radioligand therapies, and companion diagnostics based on FAP must account for the risk of off-target accumulation and patient stratification bias arising from non-tumoral fibrotic tissues [265]. Therefore, from a translational perspective, the most critical issue to avoid in future CAF-targeted nanomedicine development is not insufficient drug potency but four systemic shortcomings: imprecise biological definitions, overly simplistic assumptions about delivery mechanisms, inadequate model predictive power, and insufficient early consideration of manufacturing and regulatory requirements. The truly viable pathway forward involves patient stratification according to CAF subtypes and spatial distribution, optimization of particle engineering informed by actual in vivo delivery mechanisms and protein corona behavior, rigorous efficacy validation in immunocompetent and organ-specific models that more closely mimic human disease, and simultaneous integration of chemistry, manufacturing, and controls (CMC), scalability, and companion diagnostics from the earliest design stage—rather than attempting translational modifications only after obtaining promising animal data [258–260, 265]. Current nanotechnology in enhancing cancer immunotherapy and targeting CAFs As tumor immunotherapy continues to develop, recent advances in nanotechnology have opened new avenues for enhancing cancer immunotherapy, particularly through the targeting of CAFs within the TME [266, 267]. Nanomaterials, with their unique ability to deliver therapeutic agents precisely to CAFs, are being explored to modulate immune responses and overcome barriers to effective immunotherapy. This section will discuss the current strategies leveraging nanotechnology to target CAFs, improve immune cell infiltration, and enhance the efficacy of existing cancer treatments (Table 2; Fig. 10). Table 2Comparative summary of nanomedicine-based CAF-targeting strategies to enhance cancer immunotherapyTarget (CAF Focus)Cargo / functional payloadNanomaterial classTumor modelImmune endpoints assessedPK / toxicity evaluationLimitations & evidence qualityRefCAF cell membrane + sigma receptor on CAFsTetrandrine as nano-regulator + docetaxelCell membrane-based / anisamide-modified co-nanoparticlesTNBC mouse model↑ CD8⁺ T cells infiltration; ↓ immunosuppressive cytokinesBasic safety reported; limited biodistributionSingle tumor model; lacks longitudinal immune memory data[240]Broad CAF modulation (depletion / reprogramming)Multiple (metformin, relaxin, TGF-β modulators)Polymer / metal / mesoporous / cell membrane NPsTNBC, BRCA, gastric, pancreatic↑ CD8⁺ TILs; ↓ collagen; ↑ vessel patencySparse PK; limited toxicity profilingDiverse systems; variable endpoints; limited standardization[36]CAF inhibition / depletionVarious anti-CAF drugsNano-drug delivery systemsPreclinical models summaryImmune microenvironment remodelingPredominantly preclinical; limited PK dataNot quantitative cross-study comparison[268]CAF-targeted immunomodulationNanocarriers for immunomodulatorsStimuli-responsive nanocarriersMultiple models (reviewed)Enhanced immunotherapy responseDescribed conceptuallyNot experiment-level PK/toxicity[221]Nanomaterials targeting CAFs in immunotherapyFunctional nanomedicines (multiple payloads)Polymers / metal / inorganic NPsModels vary among studiesEnhanced CTL infiltration; tumor regressionLimited clinical translational dataPreclinical focus; varied methodologies[44]Reprogram CAFs via nanoparticlesMetal NPs modulating CAF activityMetal (Au, Ag, Fe₃O₄) NPsOrthotopic breast / HCC models↑ CD8⁺ T cell infiltration; ↓ ECM barriersSome survival data, limited toxicityHeterogeneous platforms; inconsistent controls[36, 269] Comparative summary of nanomedicine-based CAF-targeting strategies to enhance cancer immunotherapy Nanotechnology-based strategies for targeting CAFs and enhancing cancer immunotherapy. (A) Nanoparticles (NPs) are designed to penetrate the TME and accumulate at the tumor site. By targeting fibroblast activation protein (FAP) on CAFs, these nanoparticles deplete activated CAFs, reducing the tumor’s immune-suppressive environment and promoting therapeutic efficacy. (B) Engineering CAFs involves modifying their function to transform them from immunosuppressive cells into immune-activating "friends." This process enhances the recruitment of immune cells such as DCs, natural killer (NK) cells, and T cells to the tumor site, thus improving the anti-tumor immune response. (C) Blocking CAF function through nanoparticle delivery targets signaling pathways within CAFs, such as those involved in ECM production. This approach reduces ECM matrix proteins and overcomes cancer chemoresistance, enhancing the overall effectiveness of immunotherapy and chemotherapy. These strategies illustrate the potential of CAF-targeting nanomedicines in reshaping the TME, improving immune cell infiltration, and boosting the efficacy of existing cancer therapies. Adapted from [36], Copyright © 2026 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) Nanomedicine delivery systems targeting CAFs In recent years, nanodrug delivery systems have been widely studied in cancer treatment and immunotherapy. Their various types offer great options for precise delivery, controlled release, and combination therapy [87, 241]. Common types of nanodrug delivery systems include metal nanoparticles, polymer nanoparticles, nanocapsules, nanoparticles, and nanocarbon materials. Additionally, liposomes and other new nano systems have shown strong potential for use [270–272]. Below, we will introduce several main types of nanomedicine delivery systems in detail (Table 3). Table 3Summary of nanodrug delivery systems in immunotherapy: types, characteristics, and applicationsDelivery system typeKey characteristicsRepresentative materialsApplication examplesRefMetal NanoparticlesUnique optical/magnetic properties; easy to functionalize; suitable for imaging and photothermal therapyGold nanoparticles, Silver nanoparticles, Iron oxide nanoparticlesTargeted drug delivery; photothermal therapy; combined immunotherapy[272–274]Polymeric NanoparticlesExcellent biocompatibility; controllable degradation; high drug loading capacityPLGA, PEG, PolycaprolactoneDelivery of chemotherapeutic drugs, antigens, and immunomodulators; long-term controlled release systems[275, 276]LiposomesMimic the natural cell membrane; low toxicity; encapsulate both hydrophilic and lipophilic drugsPhospholipid bilayer liposomesDelivery of chemotherapeutic drugs (e.g., Doxil); vaccine delivery; nucleic acid therapy[277, 278]Nanocapsules/NanoparticlesIdeal for large molecule drugs; excellent controlled-release performance; protect drug stabilityPolymeric capsules, nanoemulsions, solid lipid nanoparticlesGene therapy; vaccine delivery; protein delivery[279, 280]Carbon NanomaterialsHigh surface area; multifunctional; excellent mechanical and thermal propertiesCarbon nanotubes, Graphene, FullerenesDelivery of immunomodulatory molecules and antigens; serve as adjuvants in photothermal therapy[281, 282]DendrimersHighly branched structure; numerous terminal functional groups; easy modification and drug loadingPAMAM dendrimersDelivery of nucleic acids; chemotherapeutic drugs; immunomodulators[283, 284]NanogelsHigh water content; responsive controlled-release; excellent softnessCrosslinked polymer hydrogels, Protein-based nanogelsDelivery of proteins, antibodies, nucleic acids, and immunomodulatory molecules[285, 286]ExosomesNaturally derived; low immunogenicity; serve as natural delivery vehiclesCell-secreted exosomesDelivery of drugs, genes, and immunomodulatory molecules; enhancement of intercellular communication and immune responses[287, 288] Summary of nanodrug delivery systems in immunotherapy: types, characteristics, and applications Nanoparticles are crucial for drug delivery because they can carry large amounts of drugs, remain stable, and target specific areas of the body. Metal nanoparticles, such as gold, silver, and ferrite, are widely studied for their unique optical properties, which are beneficial for cancer treatment [289, 290]. These nanoparticles can enhance imaging (e.g., contrast agents) and deliver drugs directly to cancer cells [291]. For example, gold nanoparticles are used in photothermal therapy, where they convert light into heat to selectively destroy tumor cells [292]. Polymer nanoparticles, such as polylactic acid-hydroxyacetic acid copolymer (PLGA) and PEG (polyethylene glycol)-modified nanoparticles, are stable, biocompatible, and can release drugs in a controlled, sustained manner [289, 293]. They are commonly used in chemotherapy to protect drugs while circulating in the body. Additionally, nanoparticles can be modified with specific ligands, such as antibodies or small molecules, to help them accumulate more drugs in tumor regions [294]. This ability to target tumors enhances drug delivery precision and reduces side effects [289]. Modifying nanoparticles to target tumor markers makes them highly effective for cancer treatment while minimizing harmful effects [295, 296]. Nanocapsules or nanoparticles are effective drug delivery systems that protect drugs from degradation and control their release by encasing them in core–shell structures [297, 298]. This makes nanocapsules especially useful for delivering large-molecule drugs or vaccines, such as mRNA vaccines and antibody drugs [299]. Liposomes, a common type of nanocapsule, exhibit excellent biocompatibility and can mimic cell membrane characteristics [300]. As a result, they are widely used in drug delivery and vaccine administration [301]. Liposomes improve drug stability and enhance targeting within the body, ensuring drugs are released at the right site and improving treatment outcomes [277]. Thus, nanocapsules and liposomes improve treatment efficiency and reduce drug side effects, making them essential tools in modern drug delivery [302]. Nanocarbon materials, such as carbon nanotubes and graphene, have recently become a key focus in nanomedicine research [303]. These materials have excellent biocompatibility, adjustable surface properties, and high drug-carrying capacity [303]. Carbon-based nanomaterials can also enhance drug targeting and boost the effectiveness of immunotherapy through surface modifications, such as functionalization and adding functional groups [304]. Graphene, with its unique nanostructure, is also used to deliver immunostimulatory molecules and antigens [305]. However, the potential for carbon-based nanomaterials to induce toxicity or cause an inflammatory response in the body remains a concern. Their long-term biocompatibility and clearance from the body require further investigation before they can be applied [306]. Nanogels can deliver chemotherapy drugs, preventing their release in healthy tissues, which reduces toxic side effects and ensures drugs are more focused on tumor sites [307]. Nanogels are also commonly used to deliver vaccines, antibodies, or immune-boosting molecules [308]. In immunotherapy, nanogels can carry immune-stimulating molecules, such as TLR agonists, to enhance the immune system’s response [308]. Extracellular vesicles have become widely used as nanodrug delivery systems in areas such as tumor therapy, gene therapy, and vaccine delivery [309]. For example, they can carry chemotherapy drugs, siRNA, mRNA, or antibodies to target tumor or immune cells [310]. Extracellular vesicles also play a key role in immunotherapy by delivering tumor antigens or immune-boosting molecules to activate specific immune cells, such as dendritic cells, enhancing anti-tumor immune responses [311]. In gene therapy, exosomes deliver gene-editing tools, such as CRISPR/Cas9, for gene repair or knockout [312]. Because they are derived from natural sources and are biodegradable, extracellular vesicles have a rapid clearance rate in the body, reducing the risk of toxicity and immune reactions. As a result, they hold great potential for future use [313, 314]. Targeting immune suppressive cells to enhance anti-tumor immune response in TME The combination of nanomaterials and immunotherapy, especially in cancer treatment, shows great promise [315]. Nanomaterials can enhance the immune response and improve treatment effectiveness by modulating the tumor environment, promoting immune cell infiltration, increasing immune cell anti-tumor activity, and directly targeting tumor cells [316, 317]. A key aspect is addressing the immunosuppressive features of TME. The TME contains immunosuppressive cells, such as Tregs, M2 macrophages, and CAFs, which reduce immune cell activity by releasing suppressive factors like TGF-β, IL-10, and VEGF [318, 319]. Recent studies, such as those by Xu et al., have provided deeper insight into the receptor-ligand interactions between nanoparticles and TME components, emphasizing specific molecular recognition pathways [320]. Nanoparticles can modulate the TME by targeting immune-suppressing cells, reducing their numbers, or blocking their function, which helps immune cells penetrate and work more effectively [321]. For example, Peng et al. demonstrated that functionalized nanoparticles, such as those carrying anti-CD25 antibodies, selectively target Tregs, leading to a significant reduction in their infiltration into tumors and enhancing the anti-tumor immune response [322]. The surfaces of some nanoparticles can be modified with specific ligands, such as anti-CD25 antibodies or anti-CTLA-4 antibodies, to target Tregs and reduce their infiltration into tumors [323]. Studies have shown that functionalized nanoparticles, such as polymer nanoparticles, delivering anti-TGF-β antibodies significantly decrease the number of Tregs [324]. This weakens their inhibitory effect on effector T cells and enhances tumor immunotherapy effectivenes.[266, 324]. As a result, tumor immune escape is reduced, and immune cells can more easily enter the tumor area, strengthening the effect [325]. Additionally, research has elucidated the targeting of immune-suppressive M2 macrophages via specific ligands, such as CD163 antibodies, showing that liposomes reduce M2 macrophage numbers and enhance immune activity [326]. CAFs are key immune-suppressing cells in the TME that promote tumor growth and spread by releasing factors like TGF-β [327]. Recent advances in receptor-mediated targeting have identified specific receptors on CAFs, such as FAP and PDGFRβ, critical for precise nanoparticle targeting [328]. Nanoparticles target CAFs by delivering small-molecule inhibitors or antibodies to block their activity [329]. For example, nanoparticles modified with anti-FAP antibodies specifically target CAFs, reduce their numbers, and weaken their immunosuppressive effects [328]. Studies have shown that this approach improves the tumor’s immune environment and boosts the effectiveness of immunotherapy [330]. By improving tumor blood vessel permeability and enhancing drug distribution in tumor areas, nanoparticles promote the infiltration of effector immune cells, such as CD8 + T cells and NK cells [331]. Studies on nanoparticle modifications highlight the significant role of surface charge and size in overcoming endothelial cell barriers and improving immune cell penetration [332]. Nanoparticles can enhance accumulation in tumor tissue by altering the vascular barrier, allowing immune cells to better recognize and attack tumor cells [333]. For example, polymer nanoparticles widen the gaps between tumor endothelial cells, helping immune cells, especially effector T cells, penetrate tumor tissue more effectively [272]. Some preclinical studies show that nanoparticles improve immune cell penetration into tumors, enhancing the effectiveness of immunotherapy [334]. Nanomaterials not only promote immune cell infiltration but also activate their functions. By delivering cytokines (such as IL-2, GM-CSF) or immune enhancers (such as STING agonists), nanoparticles stimulate immune cell activity and boost their anti-tumor response [335]. For instance, a study showed that nanoparticles delivering IL-2 enhancers significantly increased T cell infiltration into tumors and improved T cell killing activity [336]. Furthermore, nanomaterials can efficiently deliver immune enhancers, cytokines (such as IL-2, GM-CSF), and tumor antigens, activating anti-tumor immune responses and boosting immune cell function [337]. Recent work by Yu et al. demonstrated the clinical relevance of nanoparticle-delivered immune enhancers, particularly using STING agonists to enhance tumor antigen presentation and T cell reactivity in preclinical studies [331]. For example, nanoparticles delivering STING agonists stimulate anti-virus immune responses, activate immune cells, and improve tumor antigen presentation, enhancing T cell reactivity [334]. Additionally, nanoparticles stimulate dendritic cell activity, increasing tumor-specific immune responses [272]. Nanoparticles enhance their ability to kill tumor cells when combined with immunotherapy agents. Nanoparticles not only directly destroy tumor cells, but also pair with immune checkpoint inhibitors (like PD-1 and CTLA-4 antibodies) to boost immune cell anti-tumor activity [338]. For example, in combination therapies like photothermal and photodynamic therapies, nanomaterials such as gold nanoparticles and carbon nanotubes generate local heat or reactive oxygen species under laser light, directly killing tumor cells [339]. Research has shown that combining gold nanoparticles with immunotherapy in photothermal therapy destroys tumor cells and activates immune cells through local thermal effects, boosting immune responses. Additionally, nanoparticles deliver chemotherapy drugs, targeting drug release to tumor areas and killing tumor cells through drug toxicity [340]. For instance, chemotherapy drugs like paclitaxel, loaded into nanoparticles, are delivered at high local concentrations to tumor cells, while also delivering immune enhancers to improve immune system effectiveness [341]. Nano carrier targeting CAFs in tumor immunotherapy Specific recognition mechanism The specific recognition mechanism involves nanocarriers selectively targeting CAFs by binding to surface molecules unique to them [342]. CAFs play a crucial role in the TME, influencing processes such as tumor cell growth, movement, and immune escape [343]. Because CAFs are essential for tumor development, growth, and metastasis, targeting them precisely with nanocarriers can enhance treatment effectiveness and minimize damage to normal cells [330]. As a result, specific recognition mechanisms have become a key strategy for improving the accuracy of cancer treatment [344]. CAFs typically express specific receptors on their surface that play crucial roles in the TME, making them ideal targets for nanocarriers [345]. Common receptors on CAFs include TGF-β receptors and CD90 [328]. These receptors allow nanocarriers to precisely target CAFs by binding to specific ligands. Nanomaterials can enhance their binding strength with ligands such as peptides, antibodies, or nucleic acids through surface modification [346]. A notable example is nanoparticles modified with anti-CD90 antibodies, which specifically target CAFs and enhance therapeutic effectiveness [347]. As CD90 is a key receptor on CAFs involved in cancer progression, researchers have developed targeted drug delivery systems based on CD90 to enhance the precision and effectiveness of treatments [348, 349]. In addition to receptor-mediated recognition, CAFs express other specific surface markers, providing additional targeting options for nanocarriers [350, 351]. Some small-molecule drugs or nanomaterials target CAFs by binding to carbohydrate structures on their surface [352]. This mechanism provides additional targeting opportunities for nanocarriers, especially in complex TMEs, facilitating better recognition and accumulation in CAF regions [353]. A key marker for CAFs is FAP, which is rarely found in normal tissues but is highly expressed in CAFs [354]. Studies have shown that nanocarriers achieve targeted delivery by modifying FAP antibodies (such as sibrutuzumab) or FAP-targeting peptides (such as FAP-228) [355]. Nanosystems can selectively target CAFs through antigen–antibody binding or ligand-receptor interactions [349]. For instance, FAP-targeted gold nanoparticles showed significant accumulation in the CAF region in a pancreatic cancer model, demonstrating the effectiveness of this targeting strategy [353]. Additionally, PDGFR β is a receptor commonly overexpressed in CAFs and plays a role in matrix remodeling and tumor progression [356]. As a key receptor on CAFs, PDGFR β is involved in the interaction between tumor cells and the matrix and is closely linked to tumor invasiveness and metastasis [357]. Targeted therapies aimed at PDGFR β can specifically bind to CAFs, regulate their microenvironment, and disrupt tumor growth and metastasis [358]. Moreover, nanoparticles further enhance CAF targeting by modifying PDGFR β inhibitors (such as imatinib) or targeting peptides [359]. This approach not only improves drug targeting but also regulates CAF function and inhibits tumor progression by blocking PDGFR β [358, 359]. Members of the integrin family, such as αvβ3 and αvβ5, are highly expressed in CAFs, especially during tumor progression, where integrins mediate the interaction between cells and the matrix [360, 361]. Through these receptors, CAFs form strong connections with the ECM, promoting tumor cell migration and invasion [357, 362]. RGD peptides (arginine-glycine-aspartic acid) are commonly used to target integrins [363]. These peptides specifically bind to integrin family members, enabling targeted delivery to CAFs [349]. Recent studies show that nanocarriers modified with RGD peptides effectively bind to CAFs, deliver drugs precisely, enhance therapeutic effectiveness, and reduce tumor metastasis and invasion [364, 365]. Functional interaction mechanism The interaction between nanosystems and CAFs plays a crucial role in cancer nanotechnology. These mechanisms enhance targeted effects and improve therapeutic efficacy through the physical and chemical interactions between nanosystems and CAFs [366, 367]. Nanoparticles induce receptor-mediated endocytosis by binding to specific receptors on CAF cell surfaces, promoting their internalization [368, 369]. Once internalized, nanoparticles release drugs or therapeutic molecules, such as RNA interference agents or gene therapy vectors, to inhibit or modify CAF function [367, 368]. For instance, some studies indicate that nanoparticles improve endocytosis by binding to specific integrin receptors on CAFs, thereby enhancing drug delivery efficiency. This mechanism enhances drug targeting to CAFs, providing therapeutic benefits to the TME [368]. Furthermore, certain nanomaterials can respond to external stimuli like pH, temperature, or light [370]. Under specific conditions, these materials can interact with CAF cells, changing their activity and increasing the specificity and effectiveness of treatment [368, 371]. For example, pH-sensitive nanomaterials can release drugs or therapeutic molecules in the acidic conditions typical of the TME, specifically targeting CAFs [352, 368]. The acidic environment in tumor tissue is often a marker of CAF enrichment [372]. Using this external stimulus–response mechanism allows drugs to be released specifically in the TME, enhancing treatment efficacy and minimizing toxicity to healthy cells [373, 374]. Nanoparticle size and surface charge significantly influence their adsorption, endocytosis, and transport across CAF cell membranes [375]. Smaller nanoparticles (less than 100 nm) typically exhibit higher cell adsorption capacity and more efficient endocytosis, facilitating easier absorption by CAFs [376]. For instance, 100 nm liposomes demonstrated high permeability in a pancreatic cancer model, enabling the nanosystem to penetrate the TME and accumulate in the CAF region [377]. Moreover, negatively charged nanoparticles often enhance binding to CAF cells through charge interactions, facilitating easier membrane interaction and improving endocytosis [378]. For example, cationic polymer-modified nanoparticles significantly increase CAF cell uptake efficiency, enhancing therapeutic effects [379]. Responsive release strategy The responsive release strategy enhances the specificity and efficacy of targeted therapy by designing nanosystems that release drugs in response to specific biological conditions [380]. In cancer treatment, the TME exhibits distinct physical and chemical properties, such as acidity, redox state changes, and elevated enzyme activity, making it ideal for developing responsive nanosystems [381]. CAFs in the TME secrete matrix-degrading enzymes, such as matrix metalloproteinases 2 and 9 (MMP-2/9), which degrade the ECM and promote tumor cell invasion and metastasis [382]. In response, nanoparticles can be engineered to be enzyme-responsive, releasing drugs specifically in CAF-enriched regions [383]. For example, studies show that MMP-2-responsive nanoparticles release TGF-β inhibitors in CAF-enriched regions, effectively reversing the pro-tumor phenotype of CAFs [384]. This strategy enables the nanosystem to exploit the excessive MMP-2 activity in the TME, selectively release therapeutic drugs, and inhibit the immunosuppressive effects of CAFs [385]. The TME is weakly acidic (pH 6.5–6.9), providing a solid basis for designing pH-responsive nanoparticles [386]. pH-responsive nanosystems can trigger drug release under acidic conditions, accurately targeting tumor areas and minimizing the impact on healthy tissues [387]. For instance, pH-sensitive nanoparticles release CAF inhibitors, such as pirfenidone, in the acidic TME, effectively inhibiting CAF activation and reducing their support of tumors [388]. This mechanism allows nanoparticles to release therapeutic drugs under specific conditions in the TME, enhancing targeting and treatment efficiency [389]. The TME typically contains high levels of reactive oxygen species (ROS), which significantly affect cellular physiology [390]. Redox-responsive nanosystems are designed to release drugs in environments with high ROS levels, enabling targeted therapy [391]. For example, ROS-responsive nanoparticles release nitric oxide (NO) donors in the TME, inducing apoptosis in CAFs [392]. This strategy exploits the redox state of the TME to precisely release therapeutic drugs, reduce CAF-mediated immunosuppression, and enhance tumor treatment [393]. Research on nanosystems targeting CAFs is advancing, demonstrating significant clinical potential (Table 4) [87]. First, developing a multifunctional nanosystem that targets multiple sites enables simultaneous targeting of both CAFs and tumor cells, enhancing drug targeting and therapeutic efficacy [399]. This approach enhances treatment efficiency and addresses tumor immune escape [400]. Second, AI-driven design provides new insights into targeted therapy by using machine learning to predict the affinity of CAF-targeting ligands and nanoparticle behavior in the body, accelerating the development of new therapeutic carriers [401]. Lastly, as research progresses, developing a standardized evaluation system will facilitate the transition of nanosystems from the lab to clinical use [328]. These innovative strategies establish a solid foundation for the application of nanosystems in targeted CAF therapy, providing more precise and effective treatment options for cancer [214]. Table 4Mechanism of targeted CAFs by nanomedicine systemMechanism categoryDetailsTargets or moleculesMode of actionRefCell Surface Receptor RecognitionNanoparticles recognize specific receptors on CAFs such as CXCR4, etcCXCL12/CXCR4, etcEnhances nanoparticle binding to CAF cells via receptor-mediated recognition[87, 269, 394]CAF Secreted Signaling Molecule RecognitionNanoparticles interact with CAF-secreted cytokines or chemokines (e.g., TGF-β, IL-6) for targetingTGF-β, IL-6, SDF-1, etcPromotes selective nanoparticle uptake through interaction with secreted signaling molecules[36, 87, 221]Receptor-Mediated EndocytosisNanoparticles with surface ligands (e.g., RGD peptides) bind to receptors (e.g., integrins) on CAFs and facilitate endocytosisIntegrins, receptor peptidesNanoparticles enter CAF cells through receptor-mediated endocytosis[86, 224, 395]Surface Charge and Size EffectsThe size and surface charge of nanoparticles affect their binding capacity to CAFs, with particles smaller than 100 nm being more easily internalizedN/AParticles with negative charge enhance targeting efficiency through electrostatic interactions[395–397]Extracellular Matrix InteractionsNanoparticles interact with ECM components (e.g., collagen, fibronectin) produced by CAFs, affecting their distribution and targeting efficiencyCollagen, fibronectinNanoparticles interact with ECM components, influencing localization and targeting efficiency[36, 86, 398]Drug Delivery SystemNanoparticles serve as drug carriers to deliver therapeutic agents to CAFs, modifying their functions or inhibiting their support for tumor growthN/ADrugs are delivered to CAFs via nanoparticles, exerting anticancer effects[221, 224, 241] Mechanism of targeted CAFs by nanomedicine system CAF-targeting and immune regulation via nanotechnology in cancer immunotherapy CAFs not only weaken immune cell function by releasing immunosuppressive substances, but also create physical barriers that limit drug penetration, further reducing the effectiveness of immunotherapy [402, 403]. Therefore, targeting CAFs has become an important focus in tumor immunotherapy [404]. Nanomedicine delivery systems help target CAFs in two ways: first, they improve drug delivery [405]; second, they enhance immunotherapy by reversing the immune suppression caused by CAFs [405, 406] (Fig. 11). Targeting CAFs to Enhance Immunotherapy Using Nanomedicine Delivery Systems. This figure illustrates how nanomedicine delivery systems help target CAFs in two critical ways to improve cancer immunotherapy: 1. Improved Drug Delivery: Nanomedicine systems enhance drug penetration by overcoming the physical barriers created by CAFs, allowing for more effective drug delivery to tumor sites. 2. Reversing Immune Suppression: Nanoparticles help reverse the immune suppression caused by CAFs by targeting and modulating their activity, thus enhancing the immune response and improving the efficacy of immunotherapy. Adapted from [407]. Copyright © 2025 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) Improving the impact of CAFs on drug delivery CAFs create a dense barrier by releasing components such as collagen and hyaluronic acid and tightening the cytoskeleton. These factors limit drug and immune cell penetration, reducing tumor treatment effectiveness [379, 406]. To address this, nanomaterials enhance drug delivery by improving permeability and targeting specific barriers. One method to improve drug delivery is by breaking down the ECM barrier [408, 409]. Nanoparticles can carry enzymes such as collagenase, hyaluronidase, or MMPs, which break down ECM components produced by CAFs [410]. This reduces the physical barrier’s density and increases drug permeability in the tumor [411]. For example, in a pancreatic cancer model, collagenase-loaded nanoparticles reduced ECM density, enhancing the delivery of chemotherapy drugs such as gemcitabine [412, 413]. Another strategy is to increase drug accumulation around CAFs through targeting approaches. By enhancing the permeability and retention effect (EPR effect) and utilizing active targeting strategies, such as FAP and PDGFR β ligand modification, nanoparticles bring drugs closer to CAFs and increase local drug concentrations [379, 414]. For example, in a PDAC cancer model, FAP-targeted liposomes increased drug accumulation around CAFs, enhancing the effectiveness of immunotherapy, such as anti-PD-1 antibodies [415]. Finally, nanoparticles enhance immunotherapy by facilitating more effective T cell infiltration into the tumor [328]. Nanosystems can increase tumor blood vessel permeability and promote cytotoxic T lymphocyte (CTL) infiltration by breaking down the ECM and preventing CAF contraction [416]. Studies show that ROS-responsive nanoparticles release NO donors, relax CAF matrix contraction, and significantly enhance CTL infiltration into the tumor [417, 418]. These strategies enable nanosystems to overcome drug delivery barriers posed by CAFs, improving immunotherapy. This approach enhances drug delivery to tumors and strengthens immune cell activity, offering new solutions for cancer treatment [330]. Reversing the immunosuppressive effect of CAFs to improve immunotherapy resistance CAFs form a strong immune barrier in the TME, promoting tumor escape from immune surveillance and exacerbating resistance to immunotherapy [403, 406]. Therefore, overcoming this immune barrier and reversing the immunosuppressive effects mediated by CAFs are key to improving the efficacy of tumor immunotherapy (Fig. 12) [404]. As an emerging therapeutic strategy, nanosystems can reverse the immunosuppressive functions of CAFs through multiple mechanisms, significantly enhancing the efficacy of immunotherapy [329, 405]. CAF-mediated resistance mechanisms in chemotherapy and immunotherapy. CAFs form a strong immune barrier in TME, contributing to both chemotherapy and immunotherapy resistance. (A) In chemotherapy resistance, CAF-derived miR-34b-5p-loaded exosomes (LINCO035s EVs) increase the expression of ABCB1 efflux pumps in cancer cells, promoting drug resistance by expelling chemotherapy agents such as cisplatin. CAFs also secrete HJURP, which enhances glutamine metabolism and TCA cycle activity, contributing to resistance by boosting mechanisms like doxorubicin (Dox) resistance. IL-6 secretion from CAFs activates STAT3 signaling in cancer cells, further promoting chemotherapy resistance, while CAF-secreted MK induces expression of drug resistance regulators (MRP1, ABCC2), leading to reduced chemotherapy efficacy. (B) In immunotherapy resistance, CAFs form a physical barrier by interacting with T cells via PD-1/PD-L1 interactions, inhibiting T cell activation through the MHC pathway. CAFs also secrete immunosuppressive factors such as IL-6, VEGF, TGF-β, and CXCL12, which impair T cell activation and promote immune evasion. Additionally, CAFs induce metabolic alterations in T cells (via IDO, ARG2), promoting T cell apoptosis and hindering immune responses. CAFs further recruit immunosuppressive cells, including Tregs, MDSCs, and M2 macrophages, into the TME, further promoting immune suppression. Finally, CAFs modulate immune checkpoint molecules like PD-L1, reducing T cell activity and enhancing cancer cell immune resistance. These mechanisms highlight the role of CAFs in promoting both chemotherapy and immunotherapy resistance within the tumor microenvironment. Adapted from [72], Copyright © 2025 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) CAFs maintain an immunosuppressive state in the TME by releasing large amounts of factors such as TGF-β, IL-6, and VEGF [419]. To reverse this effect, nanoparticles can deliver small-molecule inhibitors (such as TGF-β inhibitors or STAT3 inhibitors) or RNA interference (siRNA) to block the release of these immunosuppressive factors [358, 420]. This approach can effectively reverse immune escape in the tumor environment [421]. For example, TGF-β siRNA-loaded nanoparticles reduced TGF-β secretion in CAFs, successfully reversing the immunosuppressive state in the tumor and improving the effectiveness of anti-PD-1 antibodies [422]. The main benefit of this approach is its ability to specifically target CAFs, suppress their immunosuppressive effects, and minimize side effects on healthy tissues [386]. CAFs usually have a pro-tumor effect in the tumor environment, which helps tumors grow and escape immune attack [423]. Nanoparticles can change the behavior of CAFs, turning them into anti-tumor cells by delivering immune modulators like STING agonists and TLR agonists [424, 425]. By activating the type I interferon pathway, nanoparticles can boost the immune function of CAFs, promote the development of DCs, and activate cytotoxic T lymphocytes (CTLs), improving immunotherapy [426, 427]. For example, STING agonist-loaded nanoparticles can enhance dendritic cell antigen presentation and cytotoxic T cell activity by activating the type I interferon signaling pathway in CAFs, greatly improving the response to immunotherapy [428]. This strategy provides a new approach to shift the TME from immune suppression to immune activation. In addition to blocking the release of immunosuppressive factors and reprogramming CAFs, nanoparticles can selectively remove immunosuppressive CAFs by delivering pro-apoptotic drugs (such as Triptolide) or gene editing tools (such as CRISPR-Cas9) [429, 430]. This approach reduces the number of immunosuppressive cells in the TME and helps immune cells recognize and attack tumors more effectively [430]. For example, CRISPR-Cas9 nanoparticles can knock out the FAP gene in CAFs, significantly decreasing Treg infiltration and enhancing the effectiveness of anti-CTLA-4 antibodies. In this way, the nanosystem not only removes immunosuppressive cells but also creates a more favorable environment for immunotherapy [431, 432]. CAFs recruit immunosuppressive cells (such as Tregs and MDSCs) by releasing chemokines (such as CXCL12), which worsen the immune escape mechanism [386, 433]. Nanoparticles can deliver CXCL12/CXCR4 axis inhibitors (such as AMD3100), blocking the interaction between CAFs and immunosuppressive cells [434]. This reduces the accumulation of Tregs and MDSCs in the tumor and enhances the effectiveness of immunotherapy [435]. For example, nanoparticles with CXCR4 inhibitors significantly reduce Tregs and MDSCs in the tumor, enhancing the anti-tumor immune response of anti-PD-L1 antibodies [386, 436]. This strategy enables nanosystems to block the recruitment of immunosuppressive cells, improving immune cell infiltration and tumor control [434]. The combined application of nanotechnology and multimodal immunotherapy Immunotherapy has made significant advances in tumor treatment, particularly with ICIs and cell therapies such as CAR-T and TCR-T cell therapy [437, 438]. However, single treatment approaches still face challenges such as tumor immune escape, immune tolerance, and low local drug concentration [439, 440]. Combining nanotechnology with immunotherapy can greatly improve therapeutic effectiveness by enabling precise drug delivery, targeted effects, and enhanced TME mechanisms [441, 442]. In this article, we will explore in detail the two primary strategies for combining nanotechnology with immunotherapy.. The combined application of nanotechnology and immune checkpoint inhibitors Nanoparticles can enhance drug concentrations at tumor sites and improve therapeutic efficacy by precisely delivering immune checkpoint inhibitors, such as pd-1 and pd-l1 antibodies [443, 444]. For example, delivery systems such as polymer nanoparticles and liposomes can target tumor cells or immune cells through surface modifications, releasing immune checkpoint inhibitors directly at the tumor site [445, 446]. Research shows that nanoparticles delivering immune checkpoint inhibitors can increase local drug concentrations, enhance the anti-tumor activity of immune cells, and significantly reduce immune escape compared to conventional injection methods [447, 448]. Studies have shown that liposomes delivering pd-1 antibodies can achieve higher local concentrations in the tme, effectively overcoming the tumor’s immune tolerance [449, 450]. The combination of pd-1 antibodies delivered by liposomes and chemotherapy drugs has exhibited significant tumor-suppressive effects [451]. The tme contains immune-suppressive cells (such as tregs and mdscs) and factors (such as tgf-β and il-10), which can diminish the efficacy of immunotherapy [452, 453]. Nanotechnology can target these immune-suppressive cells and factors to modify the tme, enhancing the effects of immune checkpoint inhibitors [454]. For example, nanoparticles can carry immune-boosting molecules, such as sting agonists and tlr agonists, to activate dcs and effector t cells, working synergistically with immune checkpoint inhibitors [455]. In one study, nanoparticles delivered pd-1 antibodies and tlr agonists (such as cpg dna), which activated dcs and enhanced t cell anti-tumor activity, significantly improving the efficacy of pd-1 inhibitors [456]. Nanocarriers can improve the stability and extend the half-life of immune checkpoint inhibitors, reduce non-specific drug distribution, and minimize side effects, thereby enhancing therapeutic efficacy [454, 457]. For example, peg-modified nanoparticles can increase the circulation time of drugs in the body, allowing immune checkpoint inhibitors to reach tumor sites more effectively [453]. In one study, peg-modified gold nanoparticles were used to deliver ctla-4 antibodies, successfully extending their in vivo half-life and significantly enhancing tumor suppression efficacy [458]. Nanomaterials as carriers for CAR-T cell or TCR-T cell therapy CAR-T (chimeric antigen receptor T cell) and TCR-T (T cell receptor T cell) therapies are advanced immunotherapies, particularly effective for hematological cancers such as leukemia and lymphoma [459]. While these therapies are effective in some patients, their efficacy in solid tumors is often limited by factors such as immune suppression in the TME, T cell exhaustion, and immune escape mechanisms [460, 461]. Nanomaterials can play a crucial role in CAR-T and TCR-T cell therapies by enhancing the targeting, persistence, and effectiveness[462, 463] of these cell therapies [464]. So far, a key challenge for CAR-T and TCR-T cell therapies remains ensuring these cells effectively reach tumor sites and overcome the tumor microenvironment’s immune barriers [465]. However, nanomaterials offer potential solutions through surface modifications that enhance targeting precision [466, 467]. By attaching specific ligands, antibodies, or other targeting molecules, these engineered materials help guide therapeutic cells to tumors. This approach improves cellular recognition and promotes accumulation at disease sites [462, 463]. By modifying the antibodies on nanoparticles (such as anti-CD19 or anti-HER2), immune cells can bind more effectively to tumor cells, improving their targeting ability. This modification enables CAR-T and TCR-T cells to recognize and bind more accurately to tumors, enhancing the success of immunotherapy.[468]. Studies show that when polymer nanoparticles are connected to CAR-T cells, they guide the cells to the tumor site, improving the immune cells’ ability to recognize tumor markers [469]. For instance, adding CD19 antibodies to nanoparticles has been shown to improve the ability of CAR-T cells to target and treat tumors [470, 471]. Nanoparticles can carry immune-boosting substances to enhance immune cell function, particularly in the challenging TME [472]. The tumor environment often has substances (like TGF-β, IL-10) that reduce immune cell activity and weaken immunotherapy. Nanoparticles can carry molecules, such as cytokines and immune boosters, to help immune cells counter these factors and function more effectively in the tumor [315, 473]. For example, IL-15 helps CAR-T cells survive longer and proliferate faster, even under tumor pressure [474]. Delivering IL-15 using nanoparticles can improve CAR-T cell function against the tumor [385]. A study showed that nanoparticles carrying IL-15 helped CAR-T cells grow and work better in the tumor environment [474]. Nanoparticles not only improve the immediate effects of immune cells but also help them survive longer [475, 476]. By attaching to CAR-T/TCR-T cells, nanoparticles reduce cell death and functional loss, prolonging immune cell activity [477]. For example, studies show that liposomes delivering IL-7 and IL-15 enhance CAR-T cell growth, survival, and anti-tumor activity, enabling longer persistence and sustained tumor fighting [454]. In tumor treatment, substances and cells in the TME, such as TGF-β, IL-10, Tregs, and MDSCs, are major contributors to treatment failure [315, 473]. Nanoparticles can target these substances or cells, improving the TME for immune cells. For example, nanoparticles can target CAFs and suppressive cells (such as Tregs, MDSCs), preventing tumor immune escape and enhancing immunotherapy [385, 449]. Studies show that nanoparticles with anti-FAP antibodies targeting CAFs can reduce the number of suppressive substances they produce, enhance T cell activity, and increase the anti-tumor efficacy of immune cells [315, 454]. In summary, nanoparticles enhance immunotherapy by targeting tumors, improving immune cell function, and making the tumor environment more favorable for treatment. The use of nanoparticles in immunotherapy extends beyond enhancing CAR-T and TCR-T cell function. Nanoparticles can also be combined with other treatment methods to create multimodal therapies that enhance treatment outcomes [478]. Nanomaterials can work in conjunction with traditional treatments such as chemotherapy, radiotherapy, and phototherapy to create a more comprehensive treatment plan [479, 480]. In this combined therapy, nanoparticles serve as carriers for various therapeutic molecules, enhancing drug targeting and efficacy [481, 482]. For example, in multimodal therapy, nanoparticles can carry both chemotherapy drugs (like paclitaxel or cisplatin) and immune checkpoint inhibitors at the same time [481]. This combination not only kills tumor cells but also makes tumors more visible to the immune system, which increases the effectiveness of CAR-T cells [481, 482]. Nanoparticles enable more accurate delivery of chemotherapy drugs and immune checkpoint inhibitors to the tumor site, improving drug availability and activating the immune system, creating a synergistic effect [483]. This combined therapy improves immunotherapy by killing tumor cells, regulating the immune environment, and enhancing immune cell activity [480]. This approach can overcome the limitations of single therapies, especially in tumors that are resistant to drugs [478, 480]. Nanoparticles, as drug carriers, improve targeting and enhance immune cell recognition and tumor cell destruction, increasing treatment efficacy. Therefore, combining nanoparticles with various therapies provides a novel and more effective strategy for cancer treatment [484]. Challenges and future perspectives Translating the integration of CAF-targeted nanomedicine and immunotherapy into clinical practice requires addressing critical knowledge gaps and technical challenges. A central priority is to achieve more precise CAF targeting. Given the functional heterogeneity of CAFs and the limited specificity of current markers such as FAP and α-SMA, which may deplete potentially tumor-restraining subsets, future studies should leverage single-cell and spatial transcriptomics to identify definitive subtype-specific biomarkers [60, 485, 486]. These biomarkers may include MHC II/CD74 for antigen-presenting CAFs and distinct perivascular signatures for vascular CAFs [186]. Nanocarriers can then be engineered with tailored targeting ligands to selectively modulate pro-tumorigenic CAF subpopulations while preserving beneficial subsets [401]. Importantly, such precision is also essential for safety during clinical translation because several commonly used CAF markers, particularly FAP, are not absolutely tumor-restricted. They may also be expressed in normal fibroblasts, wound-healing tissues, fibrotic lesions, and bone marrow stromal compartments [487, 488]. This pattern raises the possibility of on-target/off-tumor toxicities, including disruption of physiological tissue remodeling and repair [50, 488]. Therefore, future targeting strategies should not rely solely on single-marker recognition. Instead, they should incorporate combinatorial or logic-gated designs based on dual-marker recognition, microenvironment-triggered activation, or locally activated nanocarriers to improve selectivity. Advancing multifunctional nanosystems is equally essential. Next-generation carriers should evolve beyond single-stimulus responsiveness to incorporate multi-stimulus release mechanisms (e.g., pH/ROS/MMP triple-responsive) [489, 490], aligning with the complex biochemistry of the TME. Incorporating theranostic capabilities, such as MRI/PET/fluorescence imaging, would enable real-time, non-invasive monitoring of nanocarrier biodistribution, drug release kinetics, and CAF modulation, facilitating efficacy assessment and personalized dose optimization [491]. However, the translational feasibility of these platforms is complicated by pronounced inter- and intrapatient heterogeneity in nanoparticle delivery [492]. In particular, reliance on the enhanced permeability and retention (EPR) effect remains problematic, as vascular permeability, stromal density, interstitial fluid pressure, and lymphatic drainage vary considerably across tumor types, anatomical sites, and treatment stages. Such variability may lead to inconsistent tumor accumulation, suboptimal CAF exposure, and unpredictable therapeutic responses [493, 494]. To mitigate this limitation, future systems should combine passive targeting with active targeting ligands, microenvironment-responsive penetration strategies, stromal priming approaches, or image-guided patient stratification to identify tumors most likely to support efficient nanoparticle delivery [494, 495]. A key translational priority is to improve safety, biodistribution control, and pharmaceutical developability. In addition to demonstrating short-term antitumor efficacy, studies should systematically evaluate off-target stromal effects in normal tissues, innate immune sensing, complement activation-related pseudoallergy, cytokine perturbation, and adaptive immunogenicity associated with repeated dosing [496]. Long-term biodistribution and clearance also require careful investigation, particularly for inorganic, hybrid, or long-circulating materials that may accumulate in the liver, spleen, lung, or other reticuloendothelial organs [496]. Rational design should therefore emphasize biodegradable or metabolically clearable materials, minimization of persistent residues, and control over size, surface chemistry, and protein corona formation, all of which critically influence circulation behavior and immune compatibility [497, 498]. To overcome clinical translation barriers, research should focus on enhancing the biocompatibility and safety profiles of nanomaterials by optimizing metabolic clearance and minimizing immunogenicity. Developing scalable and reproducible manufacturing processes, as well as employing clinically relevant models, such as patient-derived organoids that recapitulate the human TME, will be crucial for enhancing the predictive value of preclinical studies. Chemistry, manufacturing, and controls (CMC) considerations are equally important, as they often represent a major barrier to regulatory advancement. Multifunctional nanomedicines often suffer from batch-to-batch variability, complex surface modification procedures, unstable drug loading, and challenges in sterilization, storage, and quality control [499, 500]. Thus, future development should prioritize simplified yet robust formulations with well-defined critical quality attributes, scalable synthesis workflows, and reproducible release specifications that are compatible with regulatory expectations [501, 502]. In parallel, preclinical evaluation should move beyond conventional xenograft models to include patient-derived organoids, ex vivo stromal-immune coculture systems, and immunocompetent or humanized models that more faithfully capture CAF heterogeneity, stromal barriers, and treatment-related toxicities in the human TME [503, 504]. Furthermore, the integration of multi-omics technologies and artificial intelligence will profoundly shape future development [491, 505–507]. For example, integrative single-cell spatial multi-omics analysis of over 14 million cells across 10 cancer types recently identified four conserved spatial CAF subtypes with distinct cellular neighborhoods, revealing previously uncharacterized ligand-receptor crosstalk with cancer cells and immune infiltrates (e.g., specific CAF-secreted factors modulating T-cell exclusion and macrophage polarization), which represent promising novel therapeutic targets [60]. Multi-omics can delineate the dynamic crosstalk among CAFs, cancer, and immune cells to reveal novel therapeutic targets, such as uncharacterized CAF-secreted factors. Meanwhile, AI and machine learning can predict ligand-receptor interactions, optimize nanocarrier physicochemical properties for improved TME penetration, and simulate drug release profiles as demonstrated by machine-learning frameworks that integrate nanoparticle size, surface charge, and morphology data to predict tumor accumulation, stromal penetration, and pH/enzyme-triggered release kinetics, thereby enabling rational, patient-stratified design of CAF-targeted nanomedicines[508], thereby accelerating the rational design of nanomedicines[509]. These approaches may also help address translational bottlenecks by identifying biomarker-defined patient subsets, predicting delivery heterogeneity, and establishing exposure–response-safety relationships that support more rational trial design (Fig. 13). Schematic workflow illustrating the integration of multi-omics technologies and artificial intelligence (AI)/machine learning (ML) in advancing CAF-targeted nanomedicine for cancer immunotherapy. (a) Overview of the general machine learning pipeline applied to nanomedicine design. The process begins with Data Preparation (e.g., compiling large-scale experimental datasets on nanoparticle properties and biological interactions), followed by Feature Engineering (extracting and selecting relevant physicochemical and biological features such as size, surface charge, and ligand-receptor affinities), Model Selection (training predictive models, including neural networks or ensemble methods), and culminates in ML Applications for rational optimization of nanocarriers. This workflow enables accelerated, data-driven design of multifunctional nanosystems tailored to the complex TME. (b) Detailed application of ML-enhanced approaches across key stages of nanoparticle (NP) development and performance in CAF-rich solid tumors. In the Synthesis & Formulation of Nanoparticles phase, ML predicts and optimizes critical parameters including shape, size, solvent-accessible surface area (SASA), zeta potential, polydispersity index (PDI), surface chemistry, encapsulation efficiency, drug loading/release kinetics, and aggregation behavior. NP–protein Interactions & Blood Circulation focuses on forecasting protein corona formation, pharmacokinetic profiles, and organ-specific accumulation to improve systemic stability and reduce off-target effects. Extravasation to the TME and Tumor Penetration & NP Distribution modules leverage ML to quantify TME biological structures (e.g., dense CAF-derived ECM, abnormal vasculature, and high interstitial fluid pressure), predict NP extravasation mechanisms, stromal penetration, accumulation, and delivery efficiency—addressing the well-known heterogeneity and barriers posed by CAFs. Cellular Internalization models forecast uptake by target cells (including pro-tumorigenic CAF subpopulations), identify key biomarkers, and simulate mRNA or drug transfection outcomes. Finally, Bioactivity Assessment & Therapeutic Efficacy integrates predictions of cancer cell cytotoxicity, immune response modulation (e.g., reversing CAF-mediated immunosuppression and enhancing T-cell infiltration), and optimization of treatment parameters to support personalized CAF-targeted combination immunotherapies. Adapted from [508], Copyright © 2025 by the authors. Under the terms of the Creative Commons Attribution License (CC BY) Finally, exploring sophisticated combinatorial strategies will be key. This includes moving beyond simple co-administration with immune checkpoint inhibitors or CAR-T cells, toward leveraging radiotherapy or photodynamic therapy to induce immunogenic cell death, and designing nanosystems that enhance CAR-T cell infiltration and persistence in solid tumors [510]. Nevertheless, such combinations must be evaluated cautiously, because stromal depletion, immune activation, and nanomaterial-related inflammation may interact in nonlinear ways and potentially amplify toxicity [511, 512]. Accordingly, future studies should incorporate translationally relevant safety endpoints, including systemic inflammatory markers, tissue repair capacity, hematologic effects, and organ-specific toxicity, in addition to antitumor efficacy. Together, these approaches will pave the way for precise, effective, and clinically viable stromal-targeted combination therapies for patients with stroma-rich malignancies. Conclusion CAFs are pivotal in shaping an immunosuppressive and treatment-resistant TME, which significantly limits the effectiveness of current cancer immunotherapies. Their functional heterogeneity, dynamic interactions with immune and cancer cells, and contribution to physical and biochemical barriers highlight the need for precise stromal targeting strategies. Nanomedicine offers a powerful and versatile platform to tackle these challenges. When designed effectively, with surface functionalization of CAF-specific ligands, controlled drug release, and multimodal integration, nanocarriers can selectively target CAFs, modulate their tumor-promoting functions, and disrupt the matrix barrier. This not only improves drug penetration and immune cell infiltration but also reverses immune suppression, making the TME more responsive to immunotherapy. Therefore, the synergy between CAF-targeted nanomedicine and immunotherapy represents a promising frontier in oncology. By depleting or reprogramming immunosuppressive CAF subpopulations, nanoparticles can enhance the efficacy of immune checkpoint inhibitors, adoptive cell therapies, and other immune modulators. Future progress will rely on identifying more specific CAF biomarkers, developing multifunctional and intelligent nanosystems, and incorporating artificial intelligence into personalized treatment strategies. Ultimately, the strategic combination of nanomaterial-based CAF targeting and immunotherapy holds great potential to overcome matrix-mediated drug resistance and achieve durable anti-tumor responses in matrix-rich cancer patients. Abbreviations Adoptive cell therapy Antigen-presenting CAFs Bone marrow-derived mesenchymal stem cells Chimeric antigen receptor T Cell Cancer-associated fibroblasts Dendritic cells Extracellular matrix ECM-producing CAFs Epithelial-mesenchymal transition Endothelial-mesenchymal transition Fibroblast activation protein Fibroblast-specific protein 1 Inflammatory CAFs Immune checkpoint inhibitors Indoleamine 2,3-dioxygenase Myeloid-derived suppressor cells Macrophage-to-myofibroblast transition Myofibroblastic CAFs Natural killer cells Platelet-derived growth factor receptor Prostaglandin E2 Reactive oxygen species Small interfering RNA Tumor-associated macrophages T Cell receptor T cell Tumor microenvironment Regulatory T cells Vascular CAFs α-Smooth muscle actin Author contributions Zhongsong Zhang: Conceptualization (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (lead). Rong Wang: Conceptualization (supporting); Supervision (equal); Writing – review & editing (supporting). Long Chen: Supervision (equal); Writing – review & editing (equal). The final manuscript was approved by all the authors above. Funding This study was sponsored by National Natural Science Foundation of China (No. 82504308), University-Level Natural Science Foundation General Project of Chengdu Medical College (2024CDYXY-01), Clinical Science Research Foundation of Chengdu Medical College & the First Affiliated Hospital of Chengdu Medical College (24LHLNYX1-08), Clinical Science Research Foundation of Chengdu Medical College & Nanbu People’s Hospital (2024LHFBM1-04) and Clinical Science Research Foundation of Chengdu Medical College & Chengdu Pidu People’s Hospital (2024LHFYSZ1-41). Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval Consent to participate Consent for publication all the authors were consent for publication. Competing interests The authors declare no competing interests. Footnotes Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Zhongsong Zhang and Rong Wang contributed equally to this work. References Associated Data Data Availability Statement No datasets were generated or analysed during the current study.