Cancer-associated fibroblasts and tertiary lymphoid structure orchestrate the spatial architecture of tumor immunity Authors: Haijia Tang, Youyang Shi, Qintao Ge, Rui Wang, Jianfeng Yang, Yao Zhu, Sheng Liu, Wenhao Xu Published: 2026 ABSTRACT Tumor immunity is shaped not only by the cellular composition of the tumor microenvironment (TME), but also by how stromal and immune elements are spatially organized within it. Among these regulators, cancer-associated fibroblasts (CAFs) and tertiary lymphoid structures (TLS) have emerged as two major determinants of the spatial niche of anti-tumor immunity. CAFs comprise heterogeneous and plastic stromal populations that can remodel the extracellular matrix, restrict lymphocyte infiltration, sustain chronic inflammation, and promote therapeutic resistance, while selected subsets may also support antigen presentation and lymphoid organization. In contrast, TLS function as ectopic immune hubs that coordinate local T- and B-cell priming, clonal expansion, and B-cell maturation, and are increasingly associated with favorable prognosis and improved responses to immune checkpoint blockade. In this review, we synthesize recent advances from single-cell and spatial multi-omics studies to examine how CAF states and TLS maturation programs arise, interact, and remodel the immune landscape across tumor types. We propose that the balance between CAF-driven stromal restriction and TLS-associated immune organization is a central determinant of immune exclusion versus immune activation. We discuss how this spatial framework may refine tumor classification, improve patient stratification, and guide therapeutic strategies aimed at stromal reprogramming, TLS induction or maturation, and rational combination immunotherapy, while highlighting innate immune-stromal circuits, including IL-33-activated ILC2s, NCR⁺ ILC3s, LTi-like programs, and myeloid organizer cells, as upstream regulators of TLS neogenesis. By integrating CAF biology, TLS dynamics, and spatial context, this review provides a conceptual framework for understanding and targeting the stromal architecture of tumor immunity.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13045-026-01814-6. Introduction The tumor microenvironment (TME) is a complex ecosystem composed of malignant cells, vasculature, extracellular matrix (ECM), and diverse non-malignant host cells, including immune and stromal cells [1, 2]. It is now well established that the TME actively participates in tumor initiation, progression, metastasis, and therapy response. The efficacy of modern treatments, particularly immune checkpoint blockade (ICB), is profoundly dictated by TME composition [3]. It has led to the classification of tumors as immunologically “hot” or “cold”: “hot” tumors are characterized by high densities of infiltrating cytotoxic T lymphocytes (CTLs) and generally respond well to ICB, whereas “cold” tumors lack such infiltrates and exhibit profound immunotherapy resistance [4]. This dichotomy is conserved across cancer types, with TME subtypes serving as a generalized biomarker for immunotherapy response [5]. The stromal compartment is increasingly recognized as a master regulator of this hot-versus-cold dichotomy. Cancer-associated fibroblasts (CAFs), a predominant stromal cell type, are remarkably heterogeneous and can exert both pro- and anti-tumorigenic functions. Certain CAF subtypes promote an immune-excluded “cold” phenotype enriched in immunosuppressive myeloid cells; for instance, ApoE-high CAFs promote cancer stemness, contributing to tumor progression and drug resistance. Conversely, antigen-presenting CAFs (apCAFs) occupy distinct spatial niches and may play more complex immunomodulatory roles [6]. In contrast, tertiary lymphoid structures (TLS) are hallmarks of ‘hot’, immune-active TMEs and are frequently associated with favorable prognosis and enhanced ICB response [7]. They serve as local sites for generating adaptive immune responses. Although TLSs are often interpreted as adaptive immune hubs, their initiation is increasingly recognized as being regulated by innate immune and stromal organizer circuits. Alarmin cytokines, innate lymphoid cells, lymphoid tissue inducer-like programs, myeloid organizer cells, and fibroblast-lineage stromal cells can cooperate to initiate chemokine gradients and stromal scaffolds required for TLS neogenesis [8]. Foundational studies demonstrated that a single chemokine, CXCL13 (BLC), can be sufficient to trigger de novo lymphoid tissue formation via B cell- and lymphotoxin-dependent mechanisms [9]. This lymphoid neogenesis involves fibroblastic reticular cell (FRC)-like fibroblasts expressing CCL19, which are crucial for organizing lymphoid architecture. Notably, certain B cell populations can paradoxically disrupt TLS formation and suppress anti-tumor immunity, highlighting an additional regulatory layer. Advances in single-cell and spatial technologies have revolutionized TME analysis [10]. Single-cell RNA sequencing (scRNA-seq), integrated with spatial transcriptomic and proteomic imaging platforms, enables the construction of detailed cellular atlases of tumors and surrounding stroma [11]. These approaches allow characterization of rare cell subtypes, their functional states, and precise spatial organization, revealing the specific niches where CAFs and TLS components interact to shape immune responses [12]. This review explores the functional heterogeneity and spatial organization of CAFs and TLS, discussing their interplay in dictating the hot-versus-cold immune status of tumors and, consequently, immunotherapy response. We further summarize emerging therapeutic strategies targeting specific CAF subpopulations or promoting TLS formation, with the goal of reprogramming the TME to foster robust and durable anti-tumor immunity. CAFs: a spectrum of phenotypes and functions The origin and heterogeneity of CAF CAFs are a predominant and highly dynamic component of the tumor microenvironment (TME), exhibiting remarkable heterogeneity in their origins, phenotypes, and functions. As summarized in Fig. 1, CAF heterogeneity arises from diverse cellular origins and is continuously reshaped by tumor-derived, stromal, and microenvironmental cues, giving rise to a spectrum of phenotypic states with distinct immunological and functional consequences. Additional schematic overview is provided in Supplementary Fig. 1. The framework highlights two key concepts that underpin the following sections. First, CAFs do not represent a single static lineage, but rather a dynamic and context-dependent stromal continuum. Second, distinct CAF states are not merely phenotypic labels; they are functionally linked to extracellular matrix remodeling, immune regulation, antigen presentation, and therapeutic response. Fig. 1Diverse origins and context-dependent activation cues shape the phenotypic and functional spectrum of CAFs Diverse origins and context-dependent activation cues shape the phenotypic and functional spectrum of CAFs CAFs arise from multiple potential cellular sources, including resident fibroblast-like populations and non-fibroblastic precursors such as mesenchymal stem cells, endothelial cells, pericytes, fibrocytes, adipocytes, and stellate cells. During tumor evolution, these precursor states are continuously reprogrammed by cancer cell-derived and microenvironmental signals, including TGF-β, PDGF, FGF, exosomes, extracellular matrix stiffness, hypoxia, metabolic stress, inflammatory cytokines, and chemokines. These inputs drive the emergence of distinct but interconvertible CAF states, including myofibroblastic CAFs (myCAFs), inflammatory CAFs (iCAFs), antigen-presenting CAFs (apCAFs), and pro-tumorigenic CAF programs, which differentially regulate matrix remodeling, immune modulation, antigen presentation, angiogenesis, metastasis, therapeutic resistance, and overall tumor progression. A large-scale study analyzed over 16,000 stromal cells across multiple tumor types [13], it has consistently demonstrated that CAFs comprise reproducible and functionally distinct subpopulations rather than a uniform stromal compartment [14, 15]. Compelling evidence indicates that CAFs predominantly arise from local tissue-resident fibroblasts, although additional sources, including pericytes, mesothelial cells, stellate cells, and endothelial cells via endothelial-to-mesenchymal transition (EndMT), have also been reported [16, 17]. Lineage-tracing studies in murine tumor models demonstrate that distinct normal fibroblast subsets give rise to iCAFs and myCAFs, supporting a model of activation and diversification of pre-existing fibroblast populations rather than extensive recruitment from distant progenitors [18]. CAF plasticity is driven by continuous crosstalk within the TME. Cancer cell-derived signals, including TGF-β and MAPK pathway activation, can induce myofibroblastic differentiation, whereas inflammatory cues such as IL-1 promote transitions toward iCAF states [19]. Environmental pressures, including hypoxia and therapeutic stress, further reshape CAF phenotypes, highlighting their dynamic adaptability. CAFs form a continuum of conserved programs rather than fixed subtypes As outlined in Fig. 1, these CAF states are better understood as dynamic programs rather than fixed subtypes. The functional diversity of CAFs is often framed as a dichotomy between pro- and anti-tumorigenic subtypes, yet available evidence supports a continuum of conserved, interconvertible programs rather than rigid cell types. Pan-cancer analyses have identified a potential foundational population, termed progenitor CAFs (proCAFs), in early-stage lung, stomach, colon, and breast cancers. Context-specific transcriptomic stratification and translational relevance of CAF states in triple-negative breast cancer were illustrated in Supplementary Fig. 2. Characterized by expression of genes such as IGF1, OGN, and NFIX, these cells may serve as a common precursor for several more specialized CAF subpopulations [20, 21]. Within this architecture, the most extensively studied pro-tumorigenic populations are myCAFs and iCAFs. A foundational single-cell RNA sequencing study in pancreatic cancer corroborated the co-existence of these two distinct subtypes and defined their unique in vivo gene signatures, solidifying this classification [22]. Importantly, this dichotomy is not absolute and can be layered upon other axes of heterogeneity. In pancreatic cancer, both myofibroblastic and inflammatory characteristics are found within two distinct, stable fibroblast lineages defined by the presence or absence of CD105 expression [23]. Moreover, the transition between these states is an active process: a distinct CAF state characterized by high expression of repeat RNAs like LINE-1 marks a transitional phenotype between the myCAF and iCAF lineages [24]. Spatial transcriptomic analysis in PDAC further revealed that myCAFs in close proximity to tumor glands exhibit higher expression of repeat RNAs, correlating with loss of canonical myofibroblastic markers (e.g., ACTA2, TAGLN) and gain of inflammatory genes (IL6, IL11), suggesting a tumor-proximal conversion of myCAFs toward an iCAF-like state. Together, these data support a model in which CAF subtypes represent conserved functional programs and dynamic state transitions shaped by lineage and spatial context. To facilitate interpretation across studies, it is important to recognize that current CAF classification systems are partially overlapping and context-dependent rather than universally standardized. As summarized in Supplementary Fig. 3, CAF nomenclature can be organized along several complementary axes, including cellular origin, functional programs, spatial localization, and marker-based phenotypes [25]. Canonical states such as myofibroblastic CAFs (myCAFs), inflammatory CAFs (iCAFs), and antigen-presenting CAFs (apCAFs) are among the most widely adopted functional classifications, particularly in pancreatic cancer and pan-cancer single-cell studies. However, related terms used in other studies, including matrix CAFs (mCAFs), vascular CAFs (vCAFs), reticular CAFs (rCAFs), and proliferating or progenitor CAFs (pCAFs/proCAFs), may partially overlap with these canonical programs depending on tissue context, spatial location, and analytical platform. Importantly, these labels should not be interpreted as rigid or mutually exclusive lineages, but rather as dynamic and interconvertible stromal states shaped by tumor-derived signaling, inflammatory cues, and spatial organization within the tumor microenvironment. Pro-tumor programs dominate: myCAF and iCAF states drive immune exclusion and therapy resistance Among the CAF programs summarized in Fig. 1, myCAF- and iCAF-dominant states most consistently associate with immune exclusion and therapy resistance. MyCAFs are characterized by expression of α-smooth muscle actin (α-SMA) and are primary drivers of desmoplasia. While some studies classify ECM-remodeling and myofibroblastic/contractile functions as distinct CAF states, other work suggests that a single subpopulation is often enriched for both ECM-associated genes (e.g., collagens) and contractile proteins, highlighting the close relationship between these roles [25]. Functionally, myCAFs orchestrate deposition and remodeling of the ECM, creating a dense fibrotic stroma that acts as a physical barrier to immune cell infiltration [26, 27]. In a large NSCLC cohort, a high density of matrix CAFs (mCAFs), defined by expression of MMP11 and POSTN and enrichment for TGF-β signaling, was directly correlated with low immune cell infiltration and worse patient survival [25]. Consistent with a role in immune evasion and immunotherapy resistance, extensive fibrosis correlates with reduced T-cell numbers and impaired immune surveillance [28]. A key pro-tumorigenic myofibroblast subset marked by LRRC15 has been identified as a critical driver of immunosuppression [29]. In addition to ECM-mediated exclusion, CAFs can also directly enforce suppressive lymphocyte programs. A novel mechanism of immunosuppression has been attributed to the broader α-SMA⁺ myCAF population: these CAFs phagocytose tumor antigens and, in an autophagy-dependent process, form immunological synapses with Foxp3⁺ regulatory T cells (Tregs), leading to their antigen-specific activation and proliferation [30]. In contrast, iCAFs are defined by a cytokine- and chemokine-rich secretome. General iCAF markers include PLA2G2A and CFD, with enrichment for IL6-JAK-STAT3 signaling. A key cytokine secreted by iCAFs is IL-6, which has been shown to mediate resistance to gemcitabine in PDAC [31]. Beyond broad inflammatory programs, a specific pro-tumorigenic subtype in colorectal cancer is defined by high expression of the IL-1 receptor IL1R1. These IL1R1⁺ iCAFs, which co-express markers such as FAP and PDPN, are associated with T-cell suppression, an increased Th17 cell signature, and worse overall survival, particularly in the highly stromal CMS4 molecular subtype. The causal relevance of this pathway is supported by in vivo evidence: both genetic ablation of IL1R1 in fibroblasts and pharmacological blockade with the IL-1 receptor antagonist Anakinra reduced tumor growth. Collectively, myCAF-driven fibrosis and iCAF-driven inflammatory signaling provide complementary mechanisms for immune exclusion and therapy resistance. Context-dependent immune-supportive states: lymphoid-organizing and antigen-presenting CAF programs Notably, Fig. 1 also illustrates that CAF heterogeneity includes immune-supportive states, such as antigen-presenting or lymphoid-organizing fibroblasts, which challenge the view of CAFs as uniformly immunosuppressive. Despite the predominance of pro-tumorigenic programs, CAFs can also adopt immune-supportive states linked to lymphoid organization and adaptive immune activation. A study in HNSCC identified a CCL19⁺ FRC-like population enriched in immune-hot tumors. These FRC-like CAFs were found within TLS, associated with CD4⁺ T cells and B cells, and regulated by lymphotoxin-driven non-canonical NF-κB signaling. Pan-cancer analysis further confirmed that while this FRC-like population is rare, its presence is associated with significantly better survival in patients receiving immune checkpoint inhibitors [32]. Challenging the paradigm of CAFs as exclusively pro-tumorigenic, apCAFs were first described in PDAC. In a seminal single-cell RNA sequencing study of human and mouse PDAC, researchers identified a CAF population characterized by expression of MHC class II molecules and CD74, but lacking classical co-stimulatory molecules. Crucially, these apCAFs were functionally active, as they could activate CD4⁺ T cells in an antigen-specific manner in vitro [22]. Their status as a major, ubiquitous CAF component across solid tumors has since been confirmed through large-scale pan-cancer analysis [17], and further characterization in pancreatic cancer linked apCAFs to the tumor-suppressive CD105-negative fibroblast population [23]. Spatially, apCAFs are not randomly distributed: in gastric cancer, spatial transcriptomics and multiplex imaging revealed that apCAFs are preferentially located near TLS, with a high density of apCAFs correlating with increased TLS number within the tumor [33]. Functionally, in vivo co-implantation experiments demonstrated that apCAFs can slow tumor growth and reprogram the immune landscape, with increased infiltration of activated CD4⁺ and cytotoxic CD8⁺ T cells; clinically, higher apCAF infiltration in baseline tumors predicts favorable response to immunotherapy across multiple cancer types. CAF-structured spatial barriers and immune exclusion at the tumor-stroma interface The TME cellular architecture dictates function and patient outcome, and high-resolution mapping reveals that TMEs are built from recurring cellular “circuits” that are often organized around the tumor-stroma interface. A dominant spatial program relevant to the tumor immune niche is immune exclusion, which is frequently constructed by CAF subpopulations. In lung tumors, MYH11+α-SMA+ CAFs form single-layer barriers in early stages, while FAP+αSMA+ CAFs create thicker, multi-layered structures in advanced tumors, each using distinct collagen programs. CAF-secreted collagen XII is critical for organizing collagen I fibers to increase tumor stiffness and promote invasion [34], and this barrier is cooperatively built, as macrophages can enhance CAF-mediated collagen deposition two- to three-fold [35]. Consistent with these spatial principles, computational topic modeling supports CAF concentration at the tumor edge [36], and functionally, depleting FAP+ CAFs disrupts this barrier and promotes T-cell infiltration. Spatially segregated metabolic programs further define CAF function, as glycolytic CAFs (glyCAFs) at sarcoma margins impede T-cell entry [37], and in NSCLC adenocarcinoma, high glucose uptake is a stromal phenomenon linked to matrix remodeling [38]. Together, these findings position CAF-driven architecture at the tumor boundary as a key determinant of immune exclusion and a tractable target for reshaping the tumor immune niche. In summary, CAFs exhibit a profound duality in shaping the tumor immune niche. On the immunosuppressive side, CAFs promote physical exclusion of CD8+ T cells and reinforce immunotherapy resistance [28, 39], exemplified by the LRRC15+ subset whose depletion relieves CD8+ T-cell suppression and augments anti-PD-L1 responses [29]. Functional heterogeneity can be lineage-linked, as CD105-negative CAFs restrict tumor growth in an adaptive immunity-dependent manner whereas CD105-positive CAFs are tumor-permissive. CAFs also directly impose antigen-dependent suppression through autophagy-dependent synapses with Tregs [30], and through suppressive cross-presentation pathways that eliminate or exhaust CD8+ T cells via PD-L2/FASL or cathepsin S-dependent mechanisms [40, 41], while additionally remodeling myeloid and NK compartments via subset-specific programs (e.g., CD146+ CAF-COL4A1-macrophage polarization; IL1R1+ CAFs) [42, 43]. In stark contrast, specific CAF states support immune-active niches by organizing TLS; CCL19+ lymphoid organizer/FRC-like CAFs localize to TLS, are enriched in immune-hot tumors, and predict improved survival under immunotherapy. Mechanistically, the transition of CAFs between immune-suppressive and immune-activating states is unlikely to be dictated by a single marker or lineage alone, but rather by how local tissue-repair cues are interpreted within the inflammatory, spatial, and mechanical context of the TME. Acute epithelial stress or tissue injury can release type 2 alarmins, including TSLP, IL-25, and IL-33, thereby activating type 2 immune circuits, classically characterized by ILC2/TH2 activation and IL-4, IL-5, and IL-13 production, which coordinate eosinophil, macrophage, B-cell, and fibroblast responses. In selected contexts, such alarmin-driven repair programs may cooperate with eosinophils, macrophages, B-cell responses, vascular remodeling, and fibroblast IL-4 responsiveness to favor tumor restraint, immune surveillance, and stromal conditions permissive for lymphoid organization. By contrast, when the same wound-healing logic becomes chronic or is reinforced by dominant TGFβ-, IL-13-, hypoxia-, stiffness-, and tumor-derived inflammatory cues, CAFs are more likely to consolidate myofibroblastic or inflammatory programs characterized by collagen deposition, ECM remodeling, immune exclusion, and myeloid/Treg-skewed suppression [44–46]. Thus, CAF polarization reflects a dynamic balance between transient alarmin-driven immune-repair programs and persistent fibrotic inflammatory remodeling, further shaped by tumor genotype, spatial proximity to tumor nests, stromal stiffness, immune-cell partners, and therapeutic pressure. This dual functionality therefore reflects the contextual and plastic nature of CAFs and a “push-pull” between CAF programs, where successful checkpoint blockade shifts CAF composition toward immune-supportive states, whereas fully activated programs associate with resistance. Collectively, CAF heterogeneity and origin-driven plasticity position CAFs as central architects of tumor immune niches, and as critical, yet challenging, targets for therapy [32]. Tertiary lymphoid structures: ectopic hubs of anti-tumor immunity TLS are ectopic lymphoid aggregates formed in non-lymphoid tissues like tumors during chronic inflammation, acting as local immune hubs that recapitulate features of secondary lymphoid organs (SLOs). Unlike encapsulated SLOs, their unencapsulated nature may permit freer cell entry. TLS presence often signals a potent anti-tumor response, and their maturation is categorized from Stage I (T/B cell aggregates) to Stage II (FDC presence, no germinal centers) and Stage III (mature GCs, FDCs, and high endothelial venules) [47]. As outlined in Fig. 2, TLS emerge in chronically inflamed tissues as unencapsulated immune aggregates that progressively acquire structural and functional features of secondary lymphoid organs, with increasing maturation closely linked to stronger anti-tumor immunity. Fig. 2Chronic inflammation drives the formation and maturation of tertiary lymphoid structures as ectopic hubs of anti-tumor immunity Chronic inflammation drives the formation and maturation of tertiary lymphoid structures as ectopic hubs of anti-tumor immunity Persistent inflammatory signals within the tumor microenvironment, including chemokine production and lymphoid tissue inducer (LTi)-like cellular programs, promote the formation of ectopic lymphoid aggregates. These aggregates develop into tertiary lymphoid structures (TLS), which recapitulate key features of SLOs, including organized T-cell and B-cell zones, dendritic cells, follicular dendritic cells (FDCs), and specialized vasculature such as high endothelial venules (HEVs), while remaining unencapsulated and therefore more permissive to local cellular entry. TLS mature progressively from Stage I lymphocyte aggregates, to Stage II structures containing FDC networks, and ultimately to Stage III mature TLS with germinal centers, increasingly supporting local adaptive immune activation and correlating with stronger anti-tumor responses, tumor regression, and improved prognosis. The overview emphasizes three organizing principles for the TLS field. First, TLS are not passive histological bystanders, but dynamically assembled immune structures that arise in response to chronic inflammatory cues. Second, TLS function depends not merely on their presence, but on the degree to which they acquire organized stromal scaffolds, follicular architecture, and vascular specialization. Third, TLS maturation provides a useful conceptual bridge linking developmental biology, stromal architecture, immune function, and clinical relevance, which are discussed in the following sections. Developmental and innate immune regulation of TLS neogenesis Innate lymphoid cells provide an upstream regulatory layer that links tissue injury, alarmin release, and stromal lymphoid organization. Recent work identified an IL-33-ILC2 axis capable of inducing tertiary lymphoid structures in pancreatic cancer. In this model, IL-33 activates lymphotoxin-expressing lymphoneogenic ILC2s, which engage LTβR⁺ myeloid organizer cells to initiate tertiary lymphoneogenesis. Putative lymphoneogenic ILC2s and IL-33-expressing cells were also detected within human PDAC TLSs and correlated with improved prognosis, suggesting that alarmin-driven ILC2 programs may represent an inducible and potentially targetable upstream pathway for TLS formation [48]. ILC3s provide a complementary link between classical lymphoid tissue-inducer biology and tumor-associated TLSs. In human NSCLC, NCR⁺ ILC3s are enriched within the lymphoid infiltrate and mainly localize at the edge of tumor-associated TLSs. These cells display lymphoid tissue-inducing properties, produce IL-22, TNF-α, IL-8, and IL-2 upon activation, and activate endothelial cells, supporting their potential contribution to TLS-associated stromal and vascular organization [49]. Adipose-associated lymphoid structures further support the concept that permissive stromal niches can be licensed by innate immune signals. Fat-associated lymphoid clusters are inducible lymphoid aggregates associated with visceral adipose tissues; they express CXCL13, retain B-1 cells, and support B-cell proliferation and germinal-center differentiation during immune challenge. In pleural inflammation, FALC stromal cells produce IL-33, which activates ILC2-dependent IL-5 production and promotes local B1-cell activation and IgM secretion [50]. Although FALCs are not equivalent to tumor-associated TLSs, they provide a useful model in which adipose or perivascular stromal niches, IL-33, ILC2s, and B-cell responses converge to generate inducible lymphoid organization. As illustrated in Fig. 2, TLS neogenesis begins in chronically inflamed stromal niches and proceeds through chemokine-guided assembly of increasingly organized lymphoid aggregates. Lymphoid neogenesis mirrors SLO development but is strongly context-dependent [51]. In tumors, TLS therefore represent a spatially organized immune niche, an ectopic lymphoid microanatomy embedded within the TME, whose presence, density, maturation state, and topographic relationship to tumor nests and stromal barriers provide a tractable spatial readout of the local immune ecosystem. Mechanistically, lymphoid neogenesis is frequently initiated within permissive stromal microenvironments, particularly in adipose-rich tissues, where adipocyte precursor-derived stromal populations can acquire lymphoid tissue organizer-like properties under chronic inflammatory conditions [52, 53]. Through coordinated interactions with lymphoid tissue inducer (LTi)-like immune cells, these activated stromal precursors initiate a developmentally conserved program that recapitulates core principles of secondary lymphoid organogenesis. Central to this process is bidirectional immune-stromal crosstalk mediated by LTα1β2 signaling and the induction of lymphoid-organizing chemokines, including CXCL13, CCL19, and CCL21, which collectively promote immune-cell recruitment, compartmentalization, and scaffold assembly [54]. As illustrated in Fig. 3, TLS neogenesis proceeds through coordinated stromal scaffold formation involving FRCs, FDCs, and mural cell populations, which are thought to arise from a shared embryonic mesenchymal progenitor [55, 56]. These stromal elements establish the structural framework required for the progressive spatial organization of mature TLS, including segregated T-cell zones, B-cell follicles with germinal-center-like features, dendritic cell positioning, and specialized vasculature such as high endothelial venules (HEVs) [57]. Importantly, this process reflects a conserved organizational logic rather than a single invariant cellular composition, allowing TLS to diverge into distinct structural and functional states according to tissue context and disease type. Fig. 3Developmental Reactivation and Stromal-Immune Crosstalk Drive Tumor-Associated TLS Neogenesis Developmental Reactivation and Stromal-Immune Crosstalk Drive Tumor-Associated TLS Neogenesis TLS formation is initiated through interactions between stromal precursor populations and LTi-like immune cells, recapitulating core developmental principles of secondary lymphoid organ formation. In this model, adipose-derived or tissue-resident precursor cells differentiate into stromal organizer cells that engage in reciprocal signaling with LTi-like cells, thereby establishing the chemokine and structural framework required for lymphoid assembly. In both tumors and chronic inflammatory lesions, stromal-immune crosstalk provides the architectural basis for lymphoid assembly; however, the transcriptional programs, chemokine gradients, vascular remodeling, and ultimate functional consequences of TLS may differ substantially across pathological settings. The context dependency is particularly relevant in cancer, where tumor-associated TLS may support anti-tumor immunity while also exhibiting marked heterogeneity in cellular composition, maturation state, spatial topology, and prognostic significance depending on the surrounding stromal niche and inflammatory milieu [58]. Emerging evidence further suggests that tumor-associated TLS can acquire distinct interferon-associated transcriptional programs and CXCL9-enriched chemokine signatures that distinguish them from TLS observed in chronic inflammatory diseases [59]. At the molecular level, TLS initiation can be triggered through multiple convergent inflammatory pathways. TLS neogenesis is orchestrated by a coordinated network of inflammatory mediators, lymphoid-organizing chemokines, and stromal activation signals. Although these pathways recapitulate core elements of secondary lymphoid organ development, their engagement in tumors is highly context-dependent and may differ according to tissue niche, inflammatory intensity, and stromal composition. The principal molecular drivers implicated in this process are summarized in Table 1. Table 1Key chemokines, cytokine axes, and signaling pathways that regulate TLS neogenesisNameContent/FunctionReferencesCXCL9Marks a strong IFN signature in tumor-associated TLS in nasopharyngeal carcinoma (NPC).[51]LTi-stromal cell crosstalkInitiates lymphoid neogenesis in adipose tissue, a program reactivated in chronic inflammation.[52, 53]Stromal Cell ProgenitorsA common embryonic progenitor gives rise to key stromal cells like FRCs, FDCs, and mural cells.[55, 56]TNF-driven cascadeAn initiation pathway for TLS neogenesis that involves the activation of fibroblasts.[47]Melanoma TLS PathwayA multi-step process involving TNFR signaling on CAFs, CD8+ T cell organization, and a B cell/LTα1β2 feedback loop.[60]IL-13/IL-4R & IL-22 SignalingAn alternative pathway that primes and expands immunofibroblasts for TLS formation.[61]LTβR-NF-κB2/RelB axisPromotes lymphoid stromal/FRC-like differentiation.[56]LTβR-independent PathwaysMediate TLS formation in contexts like fat-associated lymphoid clusters (FALCs) and kidney injury.[47, 52, 62]GPR183-oxysterol SignalingA crucial axis for the correct positioning of ILC3s during lymphoid neogenesis.[63]Canonical LTβR SignalingInduces the expression of key chemokines such as CXCL13, CCL19, and CCL21.[9, 47]TNFα/IFNγ-driven PathwayAn LTβR-independent mechanism for CCL19/CCL21 production by pericytes.[64]OPG-RANKL Feedback LoopA negative feedback mechanism that controls M cell density in gut isolated lymphoid follicles (ILFs).[65] Table 1Key chemokines, cytokine axes, and signaling pathways that regulate TLS neogenesisNameContent/FunctionReferencesCXCL9Marks a strong IFN signature in tumor-associated TLS in nasopharyngeal carcinoma (NPC).[51]LTi-stromal cell crosstalkInitiates lymphoid neogenesis in adipose tissue, a program reactivated in chronic inflammation.[52, 53]Stromal Cell ProgenitorsA common embryonic progenitor gives rise to key stromal cells like FRCs, FDCs, and mural cells.[55, 56]TNF-driven cascadeAn initiation pathway for TLS neogenesis that involves the activation of fibroblasts.[47]Melanoma TLS PathwayA multi-step process involving TNFR signaling on CAFs, CD8+ T cell organization, and a B cell/LTα1β2 feedback loop.[60]IL-13/IL-4R & IL-22 SignalingAn alternative pathway that primes and expands immunofibroblasts for TLS formation.[61]LTβR-NF-κB2/RelB axisPromotes lymphoid stromal/FRC-like differentiation.[56]LTβR-independent PathwaysMediate TLS formation in contexts like fat-associated lymphoid clusters (FALCs) and kidney injury.[47, 52, 62]GPR183-oxysterol SignalingA crucial axis for the correct positioning of ILC3s during lymphoid neogenesis.[63]Canonical LTβR SignalingInduces the expression of key chemokines such as CXCL13, CCL19, and CCL21.[9, 47]TNFα/IFNγ-driven PathwayAn LTβR-independent mechanism for CCL19/CCL21 production by pericytes.[64]OPG-RANKL Feedback LoopA negative feedback mechanism that controls M cell density in gut isolated lymphoid follicles (ILFs).[65] Key chemokines, cytokine axes, and signaling pathways that regulate TLS neogenesis Together, these pathways indicate that TLS neogenesis is not governed by a single initiating signal, but rather emerges from convergent inflammatory and stromal-organizing programs that vary across tissue contexts. TNF-driven cascades can activate fibroblasts to enter lymphoid-organizing programs [47], while type 2 cytokine axes (IL-13/IL-4R) and IL-22 can prime and expand immunofibroblasts [61]. Across these inputs, lymphotoxin β receptor (LTβR) signaling functions as a central regulatory axis, inhibiting adipogenesis-associated transcriptional programs (Pparγ, Cebpα) to promote a lymphoid stromal fate [52] and serving as a critical differentiation switch for FRCs [56]. Nevertheless, LTβR-independent routes also exist in specific tissue contexts [47, 52, 62], and additional positioning cues, such as GPR183-oxysterol signaling for ILC3 localization, contribute to spatial assembly of lymphoid structures [63]. Downstream of these pathways, stromal-immune circuits orchestrate chemokine programs that recruit and compartmentalize lymphocytes: LTβR canonically induces CXCL13 as well as CCL19/CCL21 [9, 47], although expression patterns vary with TLS maturation stage [66]. Together, these mechanisms explain how TLS formation is both developmentally grounded and context-tuned, providing a structured immune niche whose spatial state can be interpreted alongside stromal programs to define tumor immune ecosystems. The stromal architecture of TLS The staged maturation model shown in Fig. 2 also underscores why stromal scaffolding and vascular specialization are prerequisites for fully functional TLS rather than optional architectural refinements. The TLS stromal scaffold is built by specialized stromal programs spanning distinct CAFs and perivascular cells, with adipose-rich niches proposed to act as a progenitor reservoir in permissive contexts [52]. As summarized in Fig. 4, TLS architecture depends on a coordinated stromal scaffold in which FRC-like fibroblasts, mesenchymal progenitors, and perivascular support cells jointly organize extracellular matrix structure, chemokine gradients, lymphocyte positioning, and vascular specialization. Fig. 4Stromal scaffolds built by fibroblastic and perivascular programs organize tertiary lymphoid structure architecture Stromal scaffolds built by fibroblastic and perivascular programs organize tertiary lymphoid structure architecture In permissive tissue niches, including adipose-associated and perivascular microenvironments, local precursor cells can differentiate into FRC-like stromal cells that organize extracellular matrix architecture, generate chemokine gradients, and guide lymphocyte recruitment, positioning, and compartmentalization. In parallel, perivascular stromal cells and pericyte-like populations support vascular specialization, including the formation and functional maintenance of HEVs, thereby facilitating immune-cell trafficking into the TLS niche. The stromal view of TLS organization underscores that lymphoid assembly requires more than immune-cell recruitment alone. Functional TLS depend on a structured mesenchymal framework capable of providing mechanical support, topographic compartmentalization, survival cues, and vascular interfaces. Accordingly, the following discussion focuses on how FRC-like fibroblasts and perivascular stromal populations together sustain TLS integrity, maturation, and immune function within tumors. Key among these are FRC-like fibroblasts [32, 67] that differentiate from local mesenchymal cells [47] and provide an architectural framework for lymphocyte positioning and organization. Scaffold construction is tightly controlled: FRCs employ polarized ECM secretion to prevent fibrosis and, during expansion, downregulate ECM synthesis through a CLEC-2/podoplanin mechanism [68]. Consistent with lymphoid tissue organizer-like activity, these specialized fibroblasts express organizing chemokines (e.g., CCL19, CXCL12), survival factors (e.g., IL-15, IL-33, TNFSF13B), and antigen-presentation molecules [67], and exhibit stromal heterogeneity, including PDPN+CD34+ subsets enriched for survival factors (e.g., BAFF, IL-7) versus PDPN+CD34− subsets enriched for organizing chemokines (e.g., CXCL13, CCL19, CCL21) [61]. An immunologically active pericyte population also contributes to the TLS niche [64]. Pericytes share a mesenchymal progenitor relationship with FRC-lineage stromal cells [55], and fate-mapping of Ccl19-expressing progenitors supports a perivascular origin for key organizing populations [56]. Functionally, these perivascular stromal cells, particularly those positioned near high endothelial venules (HEVs), support HEV development and chemokine programs that recruit effector lymphocytes, including a CXCL9-expressing IFN-HEV subset that attracts CXCR3+ T cells [51, 64, 69]. Moreover, FRC-like cells can adopt “disease-adapted” states characterized by CXCL9 upregulation [67] and are implicated in sustaining anti-tumor T-cell activity [69]. These innate lymphoid pathways likely converge on stromal organizer programs. Depending on tissue context, fibroblast-lineage cells, pericytes, CAF subsets, and LTβR⁺ myeloid organizer cells may respond to lymphotoxin, TNF, IL-22, IL-13, and IL-33-associated inflammatory cues by acquiring FRC-like or lymphoid organizer-like functions [48]. Such states are characterized by chemokine production, including CCL19, CCL21, CXCL12, and CXCL13, survival factors such as IL-7 and BAFF, and matrix scaffolding programs that support immune-cell positioning. Thus, selected CAF states may participate in TLS initiation and maturation not as generic fibroblasts, but as specialized lymphoid-organizing stromal programs [60, 70]. Beyond lymphocyte recruitment, TLS formation requires a specialized stromal scaffold capable of providing structural support, compartmentalization, vascular interfaces, and chemokine-guided immune positioning. These functions are executed by a heterogeneous set of stromal populations, including fibroblastic reticular-like cells, mesenchymal organizer cells, and perivascular stromal subsets. The main stromal cell types implicated in TLS architecture are summarized in Table 2. Table 2Stromal cell populations that organize and sustain TLS architectureStromal cell typeRepresentative markers/featuresRole in TLS architecture/functionReferencesGeneral Stromal ScaffoldFibroblastic and perivascular programs; adipose-associated progenitor nichesStructural framework for TLS assembly[52]FRC-like fibroblastsLocal mesenchymal origin; CLEC-2/podoplanin axis; CCL19, CXCL12, IL-15, IL-33, TNFSF13B; inducible CXCL9ECM organization, chemokine production, lymphocyte recruitment and positioning[32, 51], 67– [69]FAP-negative CAF subsetLymphoid tissue organizer-like phenotypeDrives TLS formation[60]PDPN+CD34+ CAF subsetBAFF- and IL-7-enrichedSupports lymphocyte survival and TLS maintenance[61]PDPN+CD34− CAF subsetCXCL13, CCL19, CCL21 enrichedPromotes immune recruitment and TLS organization[61]Pericytes/perivascular stromal cellsShared progenitors with FRCs; CCL19, CCL21Supports HEV development and immune-cell trafficking[55, 64]CCL19+ progenitorsFate-mapped perivascular stromal progenitorsSource of TLS stromal organizer cells[56]CXCL9⁺ IFN-HEV subsetIFN-associated HEV program; recruits CXCR3⁺ T cellsFacilitates effector T-cell entry[51]General Stromal CellsIntrinsic lymphoid organizer activityOrganizes ectopic lymphoid structures[55] Table 2Stromal cell populations that organize and sustain TLS architectureStromal cell typeRepresentative markers/featuresRole in TLS architecture/functionReferencesGeneral Stromal ScaffoldFibroblastic and perivascular programs; adipose-associated progenitor nichesStructural framework for TLS assembly[52]FRC-like fibroblastsLocal mesenchymal origin; CLEC-2/podoplanin axis; CCL19, CXCL12, IL-15, IL-33, TNFSF13B; inducible CXCL9ECM organization, chemokine production, lymphocyte recruitment and positioning[32, 51], 67– [69]FAP-negative CAF subsetLymphoid tissue organizer-like phenotypeDrives TLS formation[60]PDPN+CD34+ CAF subsetBAFF- and IL-7-enrichedSupports lymphocyte survival and TLS maintenance[61]PDPN+CD34− CAF subsetCXCL13, CCL19, CCL21 enrichedPromotes immune recruitment and TLS organization[61]Pericytes/perivascular stromal cellsShared progenitors with FRCs; CCL19, CCL21Supports HEV development and immune-cell trafficking[55, 64]CCL19+ progenitorsFate-mapped perivascular stromal progenitorsSource of TLS stromal organizer cells[56]CXCL9⁺ IFN-HEV subsetIFN-associated HEV program; recruits CXCR3⁺ T cellsFacilitates effector T-cell entry[51]General Stromal CellsIntrinsic lymphoid organizer activityOrganizes ectopic lymphoid structures[55] Stromal cell populations that organize and sustain TLS architecture These observations also provide a conceptual basis for later discussion of CAF-TLS interactions, suggesting that fibroblast-lineage stromal states may diverge toward either immune-excluding or lymphoid-organizing programs depending on tumor context and activation state. Collectively, these data support a model in which TLS formation and maintenance emerge from coordinated fibroblast-perivascular scaffolds that couple stromal architecture with lymphocyte recruitment, survival, and vascular specialization within the tumor immune niche. TLS as functional hubs for anti-tumor immunity Beyond their architectural features, TLS act as active sites of local adaptive immunity by supporting coordinated B-cell and T-cell responses. As summarized in Fig. 5, mature TLS function as local hubs of adaptive immunity by coordinating B-cell maturation and plasma-cell differentiation together with T-cell clonal expansion, memory formation, and tumor-directed cytotoxic responses. Fig. 5Tertiary lymphoid structures function as local hubs of anti-tumor B-cell and T-cell immunity Tertiary lymphoid structures function as local hubs of anti-tumor B-cell and T-cell immunity Within mature TLS, coordinated adaptive immune programs support both humoral and cellular anti-tumor responses. On the B-cell arm, TLS promote clonal diversification, maturation, and differentiation into plasma cells, thereby enabling local antibody-associated immune activity that has been linked to favorable outcomes in several tumor settings. On the T-cell arm, TLS support local priming, clonal expansion, and memory differentiation, including the generation of CD8⁺ memory T-cell populations enriched in mature TLS. Importantly, Fig. 5 supports a shift in perspective from viewing TLS as static prognostic markers to recognizing them as spatially organized sites of coordinated adaptive immunity. On the B-cell side, TLS can support clonal diversification, affinity maturation, and plasma-cell generation; on the T-cell side, they promote local activation, clonal expansion, and memory differentiation. The major functional roles of TLS across different cancer contexts are summarized in Table 3. Together, these observations indicate that TLS function as active sites of coordinated adaptive immunity, although their dominant immune outputs may vary across tumor types and local microenvironmental contexts. Table 3Functional roles of TLS across different cancer contextsCancer type/ContextTLS functionClinical or immune implicationReferencesMultiple tumor typesB-cell maturation and plasma-cell differentiationSupports local humoral immunity and anti-tumor surveillance[71, 72]Renal Cell Carcinoma & Colorectal Cancer Liver MetastasesCoordinated B-cell and T-cell immune activationAssociated with tumor cell apoptosis and immune control[71, 73]Gastric CancerMemory T-cell enrichment in mature TLSLinked to stronger anti-tumor immunity[66]Cervical CancerLocal adaptive immune activationMay support anti-tumor responses in functional TLS states[74]GliomasT-cell priming and clonal expansionSupports local cellular immunity[75]HNSCCGerminal-center activity and B-cell responsesAssociated with favorable prognosis[72]Merkel Cell CarcinomaCoordinated adaptive immune responsesCorrelates with improved outcome[76] Table 3Functional roles of TLS across different cancer contextsCancer type/ContextTLS functionClinical or immune implicationReferencesMultiple tumor typesB-cell maturation and plasma-cell differentiationSupports local humoral immunity and anti-tumor surveillance[71, 72]Renal Cell Carcinoma & Colorectal Cancer Liver MetastasesCoordinated B-cell and T-cell immune activationAssociated with tumor cell apoptosis and immune control[71, 73]Gastric CancerMemory T-cell enrichment in mature TLSLinked to stronger anti-tumor immunity[66]Cervical CancerLocal adaptive immune activationMay support anti-tumor responses in functional TLS states[74]GliomasT-cell priming and clonal expansionSupports local cellular immunity[75]HNSCCGerminal-center activity and B-cell responsesAssociated with favorable prognosis[72]Merkel Cell CarcinomaCoordinated adaptive immune responsesCorrelates with improved outcome[76] Functional roles of TLS across different cancer contexts Functionally, TLS operate as local hubs for adaptive immune activation within the tumor immune niche. A defining feature of TLS is their capacity to support B-cell clonal diversification and affinity maturation, culminating in plasma-cell differentiation. Consistently, in renal cell carcinoma and colorectal cancer liver metastases, mature plasma cells can disseminate from TLS, localize in proximity to apoptotic tumor cells, and associate with favorable clinical outcomes, highlighting a potential antibody-mediated effector arm generated in situ. In parallel, TLS provide a microenvironment for T-cell clonal expansion and differentiation, whereby mature TLS are often enriched in memory-like cytotoxic T-cell states and coincide with tumor-cell apoptosis. Mechanistically, mature TLS can contain stem-like CD8⁺ T cells together with Tfh-like progenitor exhausted CD4⁺ T cells, which coordinate local B-cell activation and promote CD8⁺ T-cell priming, establishing a “priming-to-effector” pipeline. Supporting this model, in Merkel cell carcinoma, identical T-cell clones are detected as memory populations within TLS and as effector populations in the tumor bed, consistent with TLS acting as a source of locally generated anti-tumor effectors. Notably, TLS function is context-dependent and can be subverted by immunosuppressive myeloid infiltration; for example, high PMN-MDSC density within cervical cancer TLS correlates with the paucity of tumor-specific T cells, indicating that TLS architecture alone does not guarantee productive anti-tumor immunity. Clinical relevance and therapeutic modulation of TLS The clinical relevance of TLS is profound but context-dependent, indicating that TLS density alone is not sufficient. As summarized in Fig. 6, the clinical significance of TLS is highly context-dependent, with mature TLS generally associated with favorable prognosis and improved immunotherapy responsiveness, whereas immature or regulatory TLS states may correlate with ineffective immunity and poorer clinical outcomes. Fig. 6The clinical relevance and therapeutic modulation of TLSs are shaped by their maturation state and immune context The clinical relevance and therapeutic modulation of TLSs are shaped by their maturation state and immune context The prognostic and therapeutic significance of TLS is not uniform across tumors, but depends on their degree of maturation, cellular composition, and surrounding immune milieu. Mature TLSs, characterized by organized lymphoid architecture and functional B-cell and T-cell programs, are frequently associated with favorable prognosis, enhanced anti-tumor immunity, and improved responsiveness to immune checkpoint blockade. In contrast, immature, poorly organized, or immunoregulatory TLS-like states may be associated with limited effector function, ineffective immune priming, or adverse clinical outcomes in selected contexts. Importantly, TLS are increasingly recognized as dynamic immune niches that can be therapeutically modulated, with emerging strategies aimed at promoting TLS induction, maturation, or functional reinforcement to improve cancer immunotherapy. The clinical framework reinforces two important points. First, TLS abundance alone is insufficient for interpretation unless maturation state, cellular composition, and local immune context are also considered. Second, TLS should not be viewed solely as prognostic biomarkers, but also as dynamic immune niches that may be therapeutically reinforced or reprogrammed. These concepts provide a foundation for understanding why mature TLS often predict benefit from immune checkpoint blockade and why strategies that promote TLS induction or maturation are increasingly being explored in cancer therapy (Table 4). Table 4Prognostic and immunotherapy-associated TLS features across cancer contextsTLS-associated characteristicsClinical relevanceCancer type/settingReferencesImmature TLSLinked to poor prognosisGastric cancer, especially peritoneal metastases[66]Mature TLS (mTLS) with Germinal CentersPredicts favorable ICB responseMultiple tumor types[72, 77]ICB-induced TLSAssociated with tumor reductionImmunotherapy-treated tumors[60]mTLS-enriched Stem-like CD8+ T cellsAssociated with ICB benefitHNSCC[72]CXCL9-producing IFN-HEV signatureCorrelates with ICB responseNPC[51]Abundant HEVsCorrelates with ICB responseMerkel cell carcinoma[76]Tfh-like progenitor exhausted CD4+ T cellsCoordinates TLS-mediated ICB responsesMultiple tumor types[72]CXCL13 gene signaturePredictive of clinical outcomeGastric and breast cancer[77, 78]CTRscorePredicts clinical outcomeNPC[51]Immature TLS trajectoriesDevelopmental path linked to immunotherapy responseHCC[79] Table 4Prognostic and immunotherapy-associated TLS features across cancer contextsTLS-associated characteristicsClinical relevanceCancer type/settingReferencesImmature TLSLinked to poor prognosisGastric cancer, especially peritoneal metastases[66]Mature TLS (mTLS) with Germinal CentersPredicts favorable ICB responseMultiple tumor types[72, 77]ICB-induced TLSAssociated with tumor reductionImmunotherapy-treated tumors[60]mTLS-enriched Stem-like CD8+ T cellsAssociated with ICB benefitHNSCC[72]CXCL9-producing IFN-HEV signatureCorrelates with ICB responseNPC[51]Abundant HEVsCorrelates with ICB responseMerkel cell carcinoma[76]Tfh-like progenitor exhausted CD4+ T cellsCoordinates TLS-mediated ICB responsesMultiple tumor types[72]CXCL13 gene signaturePredictive of clinical outcomeGastric and breast cancer[77, 78]CTRscorePredicts clinical outcomeNPC[51]Immature TLS trajectoriesDevelopmental path linked to immunotherapy responseHCC[79] Prognostic and immunotherapy-associated TLS features across cancer contexts It is evident when high TLS density correlates with improved survival in HPV+ head and neck cancer but not HPV+ cervical cancer, where high regulatory cell infiltration predicts poor outcomes. Context dependence is further supported in gastric cancer, in which TLS are linked to poor prognosis, an effect driven by a predominance of immature TLS in peritoneal metastases [66], highlighting the importance of maturation state and disease site. Accordingly, mature TLS (mTLS) with germinal centers strongly predict favorable responses to ICB [72, 77], and ICB can induce TLS formation and maturation, which correlates with tumor reduction [60]. A Tfh-like progenitor exhausted CD4+ T cell subset is key to coordinating these responses [72]. Complementing histology, TLS gene signatures, such as high CXCL13, are predictive in gastric and breast cancers [77]. A similar principle is emerging in HCC, where TLS are not simply favorable or unfavorable according to their presence alone, but appear to differ according to their developmental trajectory and functional maturation state. In HCC, immature TLS can follow “conforming” or “deviating” developmental trajectories that determine immunotherapy response. Conforming immature TLS show features of productive lymphoid maturation and are associated with immunotherapy-responsive immune niches, whereas deviating immature TLS exhibit impaired maturation and reduced response-associated immune organization [60]. This example further supports the view that TLS should be evaluated by maturation trajectory, spatial localization, and functional immune architecture rather than by presence or density alone. Together, these observations indicate that TLS density alone is insufficient for clinical interpretation, whereas maturation state, vascular specialization, and immune-cell composition provide more informative biomarkers of anti-tumor immunity and treatment responsiveness. TME-mediated inhibition of TLS function This context-dependent clinical framework also helps explain why poorly organized or immunoregulatory TLS-like states may fail to confer benefit, as summarized in Fig. 6. The TME can inhibit TLS function by arresting TLS maturation or co-opting TLS into a suppressive niche. As illustrated in Fig. 7, the TME can suppress TLS function through multiple non-mutually exclusive mechanisms, including arrested maturation, metabolic restriction, suppressive niche co-option, and dysfunctional immune reprogramming, thereby limiting effective anti-tumor immunity. Fig. 7TME cues arrest TLS maturation and reprogram TLSs into dysfunctional or suppressive immune niches TME cues arrest TLS maturation and reprogram TLSs into dysfunctional or suppressive immune niches TLS can be rendered ineffective by multiple forms of tumor microenvironmental suppression. In hepatocellular carcinoma, a tryptophan-enriched metabolic milieu may arrest TLS maturation, whereas combined inhibition of tryptophan metabolism and anti-PD-1 therapy has been shown to restore TLS maturation and enhance therapeutic synergy. In metastatic gastric cancer, immature TLS may be co-opted into suppressive niches enriched for regulatory T cells, macrophages, and protumorigenic signaling pathways, thereby fostering T-cell exhaustion rather than productive anti-tumor immunity. In HPV-positive cervical cancer, TLS-like structures may also persist in functionally impaired states associated with poor clinical outcomes. The inhibitory framework is important for interpreting TLS biology beyond simple presence-absence models. In tumors, TLS may fail to confer benefit not because lymphoid structures are absent, but because their maturation is interrupted, their cellular composition is skewed toward suppressive programs, or their local metabolic environment prevents productive immune activation. These observations help explain why immature or dysfunctional TLS-like states can coexist with poor anti-tumor immunity and adverse outcomes, and they provide a rationale for therapeutic strategies aimed at restoring TLS maturation and function rather than merely increasing lymphocyte aggregation. In HCC, a tryptophan-enriched microenvironment arrests TLS maturation; inhibiting tryptophan metabolism restores maturation and synergizes with anti-PD-1 therapy. In gastric cancer peritoneal metastases, immature TLS are associated with high Treg and macrophage infiltration and linked to protumorigenic pathways like T-cell exhaustion. Similarly, in HPV+ cervical cancer, TLS in patients with poor outcomes are heavily infiltrated by Tregs and arginase-1-expressing PMN-MDSCs, which impair T-cell receptor function, and are associated with an absence of tumor-specific T cells and arrested TLS development [80]. Thus, TLS efficacy is context-dependent and can be subverted by metabolic and cellular inhibitory signals, as well as other inhibitory signals such as from immunosuppressive extracellular matrix-producing myofibroblasts [81]. The major microenvironmental factors associated with TLS maturation arrest, suppressive remodeling, or functional impairment are summarized in Table 5. Table 5Tumor microenvironmental factors associated with TLS dysfunction and impaired maturationCancer type/contextRepresentative inhibitory factor or TLS stateConsequence for TLS functionReferenceHepatocellular carcinomaTryptophan-rich metabolic nicheArrested TLS maturation[79]Gastric cancer, especially peritoneal metastasesTreg- and macrophage-enriched suppressive nicheImmature TLS and T-cell exhaustion[66]Tumors with suppressive immune remodelingRegulatory-cell enrichmentReduced TLS immune competence[82]HPV-positive cervical cancerDysfunctional TLS-like stateAssociated with poor survival[80]Immature TLS-rich tumorsLocal metabolic suppressionDysfunctional immune activation[79]Poorly organized TLS-rich tumorsLimited lymphoid compartmentalizationReduced anti-tumor immunity[82] Table 5Tumor microenvironmental factors associated with TLS dysfunction and impaired maturationCancer type/contextRepresentative inhibitory factor or TLS stateConsequence for TLS functionReferenceHepatocellular carcinomaTryptophan-rich metabolic nicheArrested TLS maturation[79]Gastric cancer, especially peritoneal metastasesTreg- and macrophage-enriched suppressive nicheImmature TLS and T-cell exhaustion[66]Tumors with suppressive immune remodelingRegulatory-cell enrichmentReduced TLS immune competence[82]HPV-positive cervical cancerDysfunctional TLS-like stateAssociated with poor survival[80]Immature TLS-rich tumorsLocal metabolic suppressionDysfunctional immune activation[79]Poorly organized TLS-rich tumorsLimited lymphoid compartmentalizationReduced anti-tumor immunity[82] Tumor microenvironmental factors associated with TLS dysfunction and impaired maturation Consistent with the inhibitory framework shown in Fig. 7, the examples summarized in Table 5 indicate that TLS dysfunction may result from maturation arrest, suppressive niche remodeling, or metabolically constrained immune activation. These observations reinforce the concept that TLS presence alone does not guarantee TLS competence. Instead, maturation status, metabolic context, and suppressive cellular composition collectively determine whether TLS function as productive immune hubs or as ineffective lymphoid-like aggregates. Spatial organization and functional heterogeneity of TLS in anti-tumor immunity and therapy response TLS are spatially organized immune cell aggregates within the tumor immune niche and are associated with robust anti-tumor responses. Their clinical impact is context-dependent, varying by anatomical location [82] and proximity to the tumor, with intratumoral TLS (iTLS) being most favorable. Prognostically favorable iTLS are enriched in CXCL13+ T cells, germinal center B cells, LAMP3+ dendritic cells, and specialized HEVs, and spatially-derived TLS gene signatures from TNBC and HGSOC are powerful prognostic and predictive markers [77]. In lung cancer, a defined cellular community composed of stem-like TCF7+ T cells, LAMP3+ DCs, and CCL19+ fibroblasts strongly predicts favorable PD-1 blockade outcome, and the presence of the complete community is a more powerful predictor than the abundance of any single cell type [83]. Importantly, TLS pathways are not universally protective, as TLS-formation pathways can be co-opted by malignant cells for pro-tumorigenic purposes [84]. TLS are most likely to support productive anti-tumor immunity when they are intratumoral or tumor-proximal, structurally mature, supported by HEVs and germinal-center programs, and embedded within coordinated immune-stromal neighborhoods. For example, in renal cell carcinoma, TLS can generate and disseminate anti-tumor antibody-producing plasma cells, whereas in head and neck squamous cell carcinoma, mature TLS harbor coordinated T- and B-cell responses driven in part by progenitor exhausted CD4⁺ T cells. Similarly, in lung cancer, spatially organized stem-immunity hubs containing TCF7⁺ T cells, LAMP3⁺ dendritic cells, and CCL19⁺ fibroblasts predict favorable PD-1 blockade outcomes [72]. By contrast, TLS-like aggregates may be poorly functional, immunosuppressive, or clinically unfavorable when their maturation is arrested or when they are embedded in suppressive anatomical, metabolic, or stromal contexts. In gastric cancer peritoneal metastases, TLS are predominantly immature and associated with increased macrophage and Treg infiltration [66]; in hepatocellular carcinoma, tryptophan metabolism can restrict intratumoral TLS maturation [79]; and in cervical cancer, fatal cases show immature TLS enriched in Tregs and PMN-MDSCs rather than productive tumor-specific T-cell responses [80]. These examples indicate that TLS should not be interpreted simply as beneficial structures; rather, their immune output depends on maturation state, anatomical location, cellular composition, metabolic constraints, and the surrounding stromal context. Integrative spatial model of CAF-TLS interaction Defining CAF-TLS spatial competition We define CAF-TLS spatial competition as a dynamic balance between two opposing stromal-immune organizational programs within the tumor microenvironment: CAF-dominated immune-exclusion niches and TLS-supportive immune-organizing niches. In this framework, “competition” does not necessarily imply direct physical antagonism between CAFs and TLS. Rather, it refers to the spatial allocation and organization of stromal space, extracellular matrix architecture, chemokine gradients, vascular access, and immune-cell positioning toward either exclusionary matrix-rich architectures or lymphoid-organizing immune hubs. When pro-fibrotic CAF programs dominate, dense extracellular matrix deposition, TGF-β-associated stromal activation, CXCL12-mediated chemokine exclusion, and suppressive myeloid recruitment may restrict lymphocyte entry, disperse immune-cell aggregates, and limit the stromal permissiveness required for TLS initiation or maturation [85]. Conversely, when lymphoid-organizing fibroblast programs prevail, CCL19-, CCL21-, and CXCL13-enriched stromal niches, together with HEV-supportive vascular programs, can support immune-cell recruitment, lymphocyte compartmentalization, and mature TLS development [60, 73]. Hence, the immune state of a tumor may be interpreted as the spatial outcome of a transition between CAF-mediated exclusion and TLS-associated immune organization. Importantly, this framework accommodates several non-mutually exclusive causal scenarios: CAF programs may actively suppress TLS formation; TLS may arise independently in stromal regions permissive for lymphoid organization; or both CAF and TLS states may be co-regulated by upstream inflammatory, vascular, metabolic, or therapeutic signals. Therefore, CAF-TLS spatial competition is proposed here as a testable spatial-regulatory model rather than a purely correlative observation. Spatial transition between CAF-mediated exclusion and TLS-associated immune organization CAF-dominated exclusion and TLS-associated immune organization should not be viewed as two isolated histological patterns, but as alternative spatial states that can coexist, compete, or transition within the same tumor ecosystem. In CAF-rich regions, stromal architecture is often organized around matrix deposition, tissue stiffening, abnormal vascular access, and immune-cell sequestration. In TLS-supportive regions, by contrast, stromal and vascular programs are reorganized to enable lymphocyte recruitment, compartmentalization, antigen presentation, and local adaptive immune activation. The spatial balance between these programs determines whether immune cells remain trapped or dispersed in the stroma, or instead assemble into organized lymphoid niches capable of supporting productive anti-tumor immunity. The CAF-mediated exclusion state is maintained by both structural and signaling barriers. Matrix-producing CAFs can generate dense collagen-rich compartments that restrict lymphocyte migration and physically separate immune cells from tumor nests. Spatial analysis of human lung tumors identified CAF populations associated with T-cell exclusion, including MYH11⁺αSMA⁺ CAFs in early-stage tumors and FAP⁺αSMA⁺ CAFs in more advanced tumors, highlighting the importance of combining molecular CAF identity with spatial positioning [26]. In pancreatic cancer, FAP-expressing carcinoma-associated fibroblasts produce CXCL12, which coats cancer cells and contributes to T-cell exclusion; pharmacological disruption of the CXCL12-CXCR4 axis synergized with anti-PD-L1 therapy, supporting a causal role for CAF-derived chemokine barriers in immune exclusion. Similarly, stromal TGF-β signaling is strongly linked to immune-excluded tumors and resistance to PD-L1 blockade, and combined TGF-β and PD-L1 blockade can facilitate T-cell penetration into tumor parenchyma [86]. These exclusionary programs may indirectly suppress TLS formation by preventing the prerequisites for lymphoid organization. TLS development requires not only immune-cell infiltration, but also local immune-cell clustering, stromal scaffolding, chemokine gradients, and vascular specialization. When CAF-dominated niches impose excessive matrix constraint, CXCL12-mediated mislocalization, TGF-β-driven stromal activation, or suppressive myeloid recruitment, lymphocytes may accumulate in peritumoral stroma but fail to form stable B/T-cell compartments or mature TLS [87]. In this sense, CAF-rich exclusion niches may not need to directly destroy TLS; they may suppress TLS initiation or maturation by maintaining a spatial state that is incompatible with lymphoid organization. Conversely, TLS-associated immune organization reflects a stromal state in which fibroblast-lineage and vascular programs support lymphoid assembly rather than immune exclusion. Tumor-associated TLS development in melanoma has been shown to be orchestrated by CAFs with lymphoid tissue organizer-like features, involving tumor necrosis factor receptor signaling, CD8⁺ T-cell organization, and B-cell lymphotoxin feedback. In colorectal cancer liver metastasis, CCL19-producing fibroblasts promote TLS formation and enhance anti-tumor IgG responses, supporting the concept that specific fibroblast programs can actively drive lymphoid organization rather than merely coexist with TLS. In nasopharyngeal carcinoma, single-cell and spatial transcriptomic analyses identified TLS-associated cell populations, including CXCL13⁺ CAFs, stem-like CXCL13⁺CD8⁺ T cells, and B-cell/plasma-cell programs, linking stromal organization to TLS maturation and immunotherapy response [70]. Therefore, the CAF-TLS relationship is best understood as a spatial transition model. At one end, pro-fibrotic CAF programs create immune-excluded niches characterized by stromal barriers, lymphocyte mispositioning, and impaired immune organization. At the other end, lymphoid-organizing fibroblast and vascular programs generate TLS-supportive niches characterized by CCL19/CCL21/CXCL13 chemokine networks, HEV-associated immune-cell entry, and B/T-cell compartmentalization [72]. Between these poles, mixed or transitional states may occur, in which CAF programs partially restrict immune access while selected fibroblast populations or inflammatory cues promote local lymphoid aggregation [88]. This model explains why the presence of CAFs or TLS alone is insufficient for interpretation; the key determinant is whether the local stromal program favors exclusionary architecture or lymphoid organization. CAF subsets and regulatory nodes modulate TLS formation The impact of CAFs on TLS formation is subset-, mediator-, and context-dependent. Rather than asking whether CAFs are uniformly suppressive or supportive, a more useful framework is to define which CAF programs act on which regulatory nodes, immune-cell targets, and TLS developmental stages. These effects can be broadly grouped into three categories: negative regulators that impose immune exclusion, positive regulators that promote lymphoid organization, and context-dependent programs that may support or restrain TLS depending on spatial and inflammatory conditions. Pro-fibrotic myofibroblastic and matrix-remodeling CAF programs are the most plausible negative regulators of TLS initiation and maturation. myCAFs, matrix CAFs, FAP⁺αSMA⁺ CAFs, and LRRC15⁺ CAFs can promote ECM deposition, collagen remodeling, tissue stiffness, and TGF-β-associated stromal activation, thereby restricting lymphocyte motility and preventing the formation of organized lymphoid aggregates. A TGF-β-driven LRRC15⁺ CAF lineage has been associated with poor outcomes in immunotherapy-treated solid tumors, supporting the link between specific myofibroblast-like CAF programs and immune-resistant stromal states [89]. In human lung cancer, spatially positioned matrix-producing CAF populations are associated with T-cell exclusion, suggesting that CAF location and matrix program, rather than CAF abundance alone, determine immune access. NOX4 has also been identified as a regulator of CAF-mediated CD8⁺ T-cell exclusion, and NOX4 inhibition can potentiate immunotherapy by relieving CAF-dependent exclusion, indicating that pro-fibrotic CAF states are therapeutically modifiable [39]. Chemokine-mediated exclusion represents a second negative regulatory node. FAP⁺ CAF-derived CXCL12 can exclude T cells from tumor nests and maintain a spatially segregated immune architecture. Although this mechanism has been most directly linked to T-cell exclusion rather than TLS suppression, it is highly relevant to CAF-TLS spatial competition because TLS initiation requires coordinated immune-cell recruitment and positioning. If CXCL12-rich CAF niches misdirect or immobilize lymphocytes in stromal compartments, they may prevent the transition from diffuse immune infiltration to organized lymphoid neogenesis. In this setting, the expected TLS effect is negative or maturation-limiting, especially when CXCL12 signaling coexists with dense ECM, TGF-β activation, and suppressive myeloid recruitment. Inflammatory CAF programs may also limit TLS competence, although their effects are more context-dependent. iCAF-like states enriched for IL-6, IL-1-related signaling, or inflammatory chemokines can sustain chronic inflammation, but chronic inflammatory signaling does not necessarily generate productive lymphoid organization. Depending on the local immune context, these signals may recruit suppressive myeloid cells, reinforce Th17-skewed or exhausted immune programs, or generate dysfunctional lymphoid-like aggregates rather than mature TLS. Therefore, iCAF programs should not be classified as uniformly negative or positive; their effect on TLS likely depends on whether inflammation is coupled to lymphoid-organizing chemokines, stromal scaffolding, and HEV-associated vascular access. In contrast, lymphoid-organizing CAF and fibroblast-like programs can positively regulate TLS formation. The clearest examples are CCL19⁺ fibroblasts, FRC-like CAFs, and TLS-associated lymphoid organizer-like CAFs (Lto-like CAF). Here, “lymphoid tissue organizer-like CAFs” denotes a functional CAF state resembling developmental lymphoid tissue organizer cells, rather than a universally standardized CAF subtype. In melanoma, CAFs with such organizer-like features orchestrate tumor-associated TLS development through TNFR-dependent stromal activation, reticular CAF network formation, CXCL13-mediated recruitment of LTα1β2⁺ B cells, and lymphotoxin-dependent immune-stromal feedback [60]. In colorectal cancer liver metastasis, CCL19-producing fibroblasts promote TLS formation and enhance anti-tumor IgG responses, providing direct evidence that a fibroblast subset can actively support TLS-associated humoral immunity [73]. Earlier work also showed that CCL19-producing fibroblastic stromal cells can foster an immune-stimulating intratumoral niche in lung carcinoma, supporting the broader concept that CCL19⁺ stromal programs help organize anti-tumor immune microenvironments [90]. CXCL13-associated CAF and stromal programs represent another TLS-supportive axis. CXCL13 is central to B-cell recruitment and follicular organization, and TLS-associated CXCL13⁺ CAFs have been identified by single-cell and spatial transcriptomic analyses in nasopharyngeal carcinoma [70]. However, CXCL13 regulation also illustrates why CAF-TLS interactions are not simply positive or negative. In NSCLC, CAFs can drive CXCL13 production in activated T cells via TGF-β while limiting IL-2 availability and expanding regulatory T-cell programs, suggesting that the same upstream pathway may generate lymphoid-recruiting signals while also imposing immunoregulatory constraints [91]. Thus, CXCL13-linked fibroblast programs may support TLS-associated immune organization only when accompanied by productive B-cell recruitment, T-cell help, and maturation rather than suppressive inflammation. Antigen-presenting CAFs may provide a third supportive or context-dependent regulatory node. apCAFs expressing MHC class II and CD74 have been linked to local CD4⁺ T-cell activation, and recent spatial analyses in gastric cancer showed that apCAFs are preferentially located near TLS and can enhance anti-tumor immunity [33]. This does not prove that apCAFs initiate TLS formation in all tumor types, but it supports the idea that selected CAF states may cooperate with TLS niches by contributing antigen-presentation capacity, CD4⁺ T-cell activation, or local immune organization. Therefore, apCAFs should be interpreted as potentially TLS-supportive when spatially associated with lymphoid aggregates, but their function may depend on co-stimulatory context, tumor type, and TLS maturation stage. Taken together, as summarized in Table 6, CAF regulation of TLS formation can be summarized as a set of spatially organized regulatory nodes rather than a single linear pathway. myCAF/matrix CAF/LRRC15⁺/FAP⁺αSMA⁺ programs tend to favor exclusion through ECM, TGF-β, CXCL12, and myeloid-associated barriers. CCL19⁺ FRC-like, CXCL13-associated, lymphoid tissue organizer-like, and selected apCAF programs tend to favor TLS-supportive immune organization. iCAF and TGF-β-linked inflammatory programs occupy an intermediate position because they may either sustain lymphoid-recruiting chemokines or reinforce suppressive inflammation. Thus, CAF identity alone is insufficient to predict TLS outcome; the decisive variables are spatial localization, mediator profile, target immune cells, vascular access, and TLS maturation state. Table 6CAF programs and regulatory nodes shaping TLS formation in CAF-TLS spatial competitionCAF program/subsetMain regulatory mediators/nodesTargeted immune or stromal processEffect on TLSReferencemyCAF/Matrix CAFECM deposition; collagen remodeling; matrix stiffness; TGF-β-associated stromal activationRestricts lymphocyte migration; maintains stromal barriers; prevents immune-cell clusteringNegative [26, 86]FAP⁺CXCL12+CAFCXCL12-CXCR4 axis; FAP-associated stromal barrierExcludes or mislocalizes T cells; maintains immune segregation within stromaNegative [85]LRRC15⁺ CAFTGF-β-LRRC15 axis; pro-fibrotic stromal activationPromotes immune-resistant stromal states and fibrotic remodelingNegative [89]NOX4+ CAFNOX4-maintained CAF activationMaintains immunosuppressive CAF phenotype and CD8⁺ T-cell exclusionNegative [39]IL1R1⁺ iCAFIL-1 signaling; IL-6/JAK/STAT3-related inflammatory programsPromotes T-cell and macrophage suppression; sustains pro-tumor inflammatory nichesContext-dependent [43]CCL19⁺ FRC-like CAFCCL19-CCR7 axis; lymphotoxin-related stromal feedback; non-canonical NF-κBRecruits and positions T cells, dendritic cells, and B cells; supports stromal scaffoldingPositive [73]Lto-like CAFTNFR signaling; B-cell lymphotoxin feedback; CXCL13-related B-cell recruitmentOrganizes CD8⁺ T cells and B cells; supports tumor-associated TLS developmentPositive [60]CXCL13+ TLS-associated CAFCXCL13-CXCR5 axis; CAF-associated induction of CXCL13-producing T-cell programs in selected contextsSupports B-cell recruitment, follicle formation, and TLS-associated immune organizationContext-dependent [70, 91]TLS-adjacent apCAFMHC-II/CD74-associated antigen presentation; CD4⁺ T-cell activationSupports local antigen presentation and TLS-associated immune activationContext-dependent [33] Table 6CAF programs and regulatory nodes shaping TLS formation in CAF-TLS spatial competitionCAF program/subsetMain regulatory mediators/nodesTargeted immune or stromal processEffect on TLSReferencemyCAF/Matrix CAFECM deposition; collagen remodeling; matrix stiffness; TGF-β-associated stromal activationRestricts lymphocyte migration; maintains stromal barriers; prevents immune-cell clusteringNegative [26, 86]FAP⁺CXCL12+CAFCXCL12-CXCR4 axis; FAP-associated stromal barrierExcludes or mislocalizes T cells; maintains immune segregation within stromaNegative [85]LRRC15⁺ CAFTGF-β-LRRC15 axis; pro-fibrotic stromal activationPromotes immune-resistant stromal states and fibrotic remodelingNegative [89]NOX4+ CAFNOX4-maintained CAF activationMaintains immunosuppressive CAF phenotype and CD8⁺ T-cell exclusionNegative [39]IL1R1⁺ iCAFIL-1 signaling; IL-6/JAK/STAT3-related inflammatory programsPromotes T-cell and macrophage suppression; sustains pro-tumor inflammatory nichesContext-dependent [43]CCL19⁺ FRC-like CAFCCL19-CCR7 axis; lymphotoxin-related stromal feedback; non-canonical NF-κBRecruits and positions T cells, dendritic cells, and B cells; supports stromal scaffoldingPositive [73]Lto-like CAFTNFR signaling; B-cell lymphotoxin feedback; CXCL13-related B-cell recruitmentOrganizes CD8⁺ T cells and B cells; supports tumor-associated TLS developmentPositive [60]CXCL13+ TLS-associated CAFCXCL13-CXCR5 axis; CAF-associated induction of CXCL13-producing T-cell programs in selected contextsSupports B-cell recruitment, follicle formation, and TLS-associated immune organizationContext-dependent [70, 91]TLS-adjacent apCAFMHC-II/CD74-associated antigen presentation; CD4⁺ T-cell activationSupports local antigen presentation and TLS-associated immune activationContext-dependent [33] CAF programs and regulatory nodes shaping TLS formation in CAF-TLS spatial competition Therapeutic targeting of the stromal microenvironment TLS are not only biomarkers of ongoing anti-tumor immunity, but also immune niches that may be reinforced by restoring TLS competence through the targeting of metabolic suppression, inhibitory myeloid networks, or stromal barriers. In contrast, activated CAFs can remodel the extracellular matrix, impair vascular access, promote hypoxia, and exclude immune cells, thereby establishing an immune-excluded and therapy-resistant tumor state. Therapeutic stromal reprogramming therefore aims not simply to deplete fibroblasts, but to convert hostile stromal architecture into an immune-permissive state characterized by improved vascular function, reduced matrix constraint, enhanced immune infiltration, and conditions more favorable for TLS assembly or maturation. Figure 8 collectively support a therapeutic framework in which TLS and CAF-dominated stromal niches are viewed as dynamic and modifiable components of the tumor microenvironment. The transition from stromal exclusion to immune organization provides a rationale for combining stromal-targeted strategies with immunotherapy. Fig. 8CAF-driven stromal architecture shapes immune exclusion, therapeutic resistance, and opportunities for immune reprogramming in tumors CAF-driven stromal architecture shapes immune exclusion, therapeutic resistance, and opportunities for immune reprogramming in tumors Activated CAFs generate a pro-tumorigenic stromal microenvironment characterized by dense extracellular matrix deposition, vascular collapse, hypoxia, and exclusion of immune effector cells, thereby promoting an immunologically “cold” and therapy-resistant tumor state. Beyond serving as passive structural barriers, CAFs actively support tumor progression through reciprocal signaling with cancer cells, induction of epithelial-mesenchymal transition, facilitation of angiogenesis and invasion, and metabolic or matrix-mediated restriction of drug delivery and immune access. Therapeutic strategies aimed at stromal depletion or, more importantly, stromal reprogramming may convert this hostile architecture into an immune-permissive microenvironment through extracellular matrix normalization, vascular normalization, and enhanced immune-cell infiltration. Translating stromal biology into clinical applications represents a critical frontier in oncology. The stromal microenvironment is now recognized as an active driver of tumor progression, metastasis, and therapy resistance, particularly in immunologically “cold” tumors such as PDAC, where stromal components can constitute up to 90% of tumor volume [92]. In this context, reciprocal interactions between CAFs and tumor-associated macrophages help sustain an immunosuppressive and desmoplastic microenvironment. ECM remodeling and matrix-derived proteolytic products further regulate multiple steps of the cancer-immunity cycle, providing potential therapeutic targets and predictive biomarkers [93]. Consistently, high-resolution spatial transcriptomics in colorectal cancer has shown that cellular organization at the tumor-stroma boundary critically shapes local immune status and response to immunotherapy [94]. Representative strategies for stromal reprogramming, immune-permissive niche restoration, and TLS-supportive remodeling are summarized in Table 7. Table 7Therapeutic approaches for stromal reprogramming and enhancement of TLS-associated anti-tumor immunityTherapeutic strategyMain target/mechanismExpected effect on tumor immunity or TLSReferenceSelective CAF reprogrammingShifts exclusionary stromal statesFavors immune-permissive architecture [92]ECM remodeling/normalizationRelieves physical stromal barriersImproves immune-cell access [93]Vascular normalizationRestores vascular function and perfusionSupports immune trafficking and HEV activity [95]TLS-inductive chemokine programsReinforces lymphoid-organizing signalsPromotes TLS assembly or maturation [96]HEV-supportive vascular remodelingEnhances specialized immune entry sitesFacilitates lymphocyte recruitment [97]Metabolic inhibitionRelieves local suppressive metabolismRestores TLS maturation potential [98]Stromal targeting plus ICBCombines niche remodeling with checkpoint releaseEnhances anti-tumor immunity [99]Multimodal stromal-immune therapyIntegrates matrix, vascular, and immune reprogrammingPromotes durable TLS-associated immune control [100] Table 7Therapeutic approaches for stromal reprogramming and enhancement of TLS-associated anti-tumor immunityTherapeutic strategyMain target/mechanismExpected effect on tumor immunity or TLSReferenceSelective CAF reprogrammingShifts exclusionary stromal statesFavors immune-permissive architecture [92]ECM remodeling/normalizationRelieves physical stromal barriersImproves immune-cell access [93]Vascular normalizationRestores vascular function and perfusionSupports immune trafficking and HEV activity [95]TLS-inductive chemokine programsReinforces lymphoid-organizing signalsPromotes TLS assembly or maturation [96]HEV-supportive vascular remodelingEnhances specialized immune entry sitesFacilitates lymphocyte recruitment [97]Metabolic inhibitionRelieves local suppressive metabolismRestores TLS maturation potential [98]Stromal targeting plus ICBCombines niche remodeling with checkpoint releaseEnhances anti-tumor immunity [99]Multimodal stromal-immune therapyIntegrates matrix, vascular, and immune reprogrammingPromotes durable TLS-associated immune control [100] Therapeutic approaches for stromal reprogramming and enhancement of TLS-associated anti-tumor immunity Therapeutic strategies are increasingly focused on selective stromal modulation, with two major goals: suppressing pro-tumorigenic CAF functions and enhancing TLS-associated anti-tumor immunity. One emerging approach targets master regulator checkpoint (MRC) modules, small transcription factor networks that integrate upstream signals and control immunosuppressive stromal states [100]. Another strategy uses engineered agents that anchor to the stromal extracellular matrix, thereby concentrating therapeutic activity within tumors while limiting systemic exposure. The rationale for CAF-directed intervention is further supported by pan-cancer analyses showing that CAF-derived cancer-associated ECM signatures are strongly associated with poor prognosis across multiple tumor types [101]. Consistent with the framework shown in Fig. 8, the representative strategies summarized in Table 7 reflect a shift from indiscriminate stromal depletion toward selective stromal normalization and immune reorganization, aiming to convert CAF-dominated, immune-excluded tumors into immune-permissive and potentially TLS-supportive states. Targeting pro-tumorigenic CAFs Pro-tumorigenic CAF programs promote therapy resistance by coupling ECM remodeling with immunosuppressive signaling at the tumor-stroma interface. In immune-excluded tumors, CD8⁺ T cells can be sequestered within fibroblast- and collagen-rich stromal compartments, and this exclusion is strongly associated with stromal TGF-β signaling and reduced response to immune checkpoint blockade [86]. Spatial multi-omics further highlights the importance of the tumor-stroma boundary, where aggressive CAF programs have been linked to EMT, immune exclusion, and poor prognosis [102]. However, CAF targeting must be selective, because CAF states can have context-dependent and sometimes opposing effects on tumor immunity and therapy response. In PDAC, functionally distinct CAF populations have been associated with different effects on tumor progression and therapeutic resistance [31]. Moreover, CAF vulnerabilities and therapeutic relevance may vary across experimental systems and tumor contexts, as illustrated by drug-screening studies identifying the DDR1/2-MAPK12-GLI1 axis as a targetable CAF-associated pathway [103]. Accordingly, therapeutic strategies are shifting from broad CAF depletion toward functional stromal reprogramming. FAP-directed approaches may be used to localize immune agonists within the tumor microenvironment and enhance intratumoral T-cell effector functions [104]. TGF-β blockade can remodel the CAF landscape by reducing pro-fibrotic myCAF programs and promoting fibroblast states that are more compatible with anti-tumor immunity, thereby improving responsiveness to PD-L1 blockade [105]. Emerging approaches that target master regulator checkpoint modules may further enable more durable conversion of immunosuppressive stromal states into immune-permissive programs [100]. Collectively, these strategies aim to dismantle CAF-dependent exclusion niches and restore immune access to the tumor parenchyma. Therapeutically harnessing tertiary lymphoid structures In contrast to immune-exclusion niches, “immunity hubs” such as TLS are associated with robust anti-tumor responses, yet their functional impact is clearly context-dependent. Their potency varies by anatomical location [82] and by proximity to the tumor, with intratumoral TLS (iTLS) being most favorable. Conceptually, TLS formation reflects a conserved response to tissue disruption and can also occur outside cancer, including in injured kidneys [106, 107], underscoring that lymphoid-like organization can emerge whenever local inflammatory cues and stromal scaffolding permit. Within tumors, prognostically favorable iTLS are characterized by coordinated cellular and vascular specialization, including enrichment of CXCL13+ T cells, germinal center B cells, LAMP3+ dendritic cells, and specialized HEVs. This organization highlights an important principle: protective immunity is not solely determined by the presence of immune cells, but by whether they assemble into spatially structured communities that support antigen presentation, lymphocyte recruitment, and productive crosstalk. Across multiple tumor types, spatially organized immune-supportive communities, collectively framed as TLS-like “immunity hubs”, share several recurring features: stromal scaffolding by selected fibroblast programs, production of lymphoid-organizing chemokines, and co-localization of antigen-presenting cells with B- and T-cell populations capable of sustaining productive anti-tumor immunity. Consistent with this, immuno-protective “EcoCellTypes” in breast cancer emphasize that immune-supportive architecture may arise from co-localized stromal-myeloid programs rather than from any single-cell abundance metric. EcoCellTypes are recurrent FAP+ CAF-related cellular niches identified by spatial analysis. They are not single cell types. Instead, each EcoCellType represents a local cell community [81]. This community contains specific CAF states together with neighboring immune or stromal cells. At the level of biomarkers, spatially derived TLS gene signatures in TNBC and HGSOC serve as powerful prognostic and predictive markers, suggesting that TLS-like organization can be captured by spatially informed transcriptional readouts. A particularly potent instantiation is the lung cancer “stem-immunity hub”, defined by a community of stem-like TCF7+ T cells, LAMP3+ DCs, and CCL19+ fibroblasts, which strongly predicts favorable PD-1 blockade outcome [83]. Notably, the complete hub is more predictive than the abundance of any single component. Key effector populations such as CXCL13+ T cells are themselves predictive of immunotherapy response and can also appear dispersed throughout the parenchyma in “hot” MMRd tumors [108], indicating that TLS-like chemokine programs may support immunity both within structured aggregates and in more diffuse architectures. Importantly, the central role of B cells is non-negotiable in this framework: the prognostic value of a TLS chemokine score is lost without a corresponding B-cell signature. Finally, these pathways are not uniformly beneficial, TLS-formation programs can be co-opted by malignant cells for pro-tumorigenic purposes [84], reinforcing that “immunity hub” signals must be interpreted in anatomical and disease-specific context. Combination strategies: remodeling the stroma to overcome therapy resistance Stromal-targeting therapies are most likely to achieve clinical value when integrated with established treatment modalities, including immune checkpoint blockade, chemotherapy, radiotherapy, and targeted therapy. In many solid tumors, therapeutic resistance is not determined solely by tumor-intrinsic alterations, but also by stromal barriers that limit immune-cell trafficking, drug penetration, and effector-cell function. CAF-rich extracellular matrix, including collagen- and hyaluronan-enriched desmoplasia, can physically restrict cytotoxic T-cell infiltration and reduce access of therapeutic agents to tumor nests [92, 93]. In parallel, CAF-derived chemokines such as CXCL12, abnormal tumor vasculature, VEGF-associated dysfunction, and CAF-myeloid feedback loops can further reinforce immune exclusion and suppressive remodeling [109, 110]. These observations suggest that combination strategies should be designed around defined barrier biology, including matrix- or chemokine-mediated exclusion, vascular dysfunction, suppressive myeloid-fibroblast interactions, and therapy-induced stromal rebound. Clinical and translational studies support the idea that stromal state is a determinant of response to immunotherapy-containing regimens. In the GeparNuevo neoadjuvant TNBC trial, higher baseline expression of stromal-related gene sets was associated with reduced pathological complete response and poorer long-term outcomes in patients treated with durvalumab plus chemotherapy [111]. This finding suggests that stroma-high tumors may require concurrent stromal modulation rather than immune checkpoint blockade alone. Importantly, the same study also indicated that durvalumab monotherapy can induce stromal gene expression, implying that stromal adaptation may occur early during treatment and should be considered when optimizing the timing of combination therapy. Evidence from cellular-resolution studies further emphasizes that CAF phenotypes can shape immunotherapy outcome in a treatment-context-dependent manner. In metastatic melanoma, pretreatment CAF subtype densities were associated with survival after anti-PD-1 therapy independently of CD8⁺ T-cell infiltration or PD-L1 expression [112]. This supports a shift away from viewing stromal abundance as a simple negative feature and toward assessing specific stromal configurations that either restrict or permit immune activity. Mechanistic studies also identify several resistance-associated stromal circuits that may be therapeutically actionable. In recurrent renal cell carcinoma, increased CAF infiltration was accompanied by reduced CD8⁺ T-cell infiltration, and CAF-derived Galectin-1 promoted CD8⁺ T-cell apoptosis, thereby limiting anti-PD-1 efficacy [113]. In metastatic urothelial carcinoma, FGFR3 mutation was associated with an immune-cold microenvironment characterized by reduced immune infiltration, impaired T-cell cytotoxicity, TREM2⁺ macrophage enrichment, and altered CAF programs [114]. In colorectal cancer, CSF1R blockade depleted macrophages but induced a compensatory CAF-mediated fibrotic response through tumor-cell PI3K/AKT activation and CTGF secretion; co-targeting CTGF mitigated this fibrotic rebound and restored sensitivity to anti-PD-1 therapy in preclinical models [115]. Together, these findings support a practical framework for stromal-immune combination therapy. First, patients should be stratified according to stromal resistance features, such as high ECM burden, CAF-rich immune exclusion, abnormal vascular access, or suppressive CAF-myeloid programs. Second, stromal intervention should be timed early enough to prevent consolidation of exclusionary barriers, particularly when immunotherapy itself may induce adaptive stromal remodeling. Third, treatment escalation should be guided by on-treatment evidence of stromal reprogramming rather than by tumor shrinkage alone. Spatial profiling, molecular stromal signatures, and imaging-based measurements of stromal normalization may help determine whether matrix stiffness, vascular access, and immune infiltration are improving during therapy [116]. The overall goal is not to eliminate the stroma indiscriminately, but to convert resistant stromal architecture into an immune-permissive state that facilitates effector-cell infiltration, limits compensatory suppression, and improves the durability of anti-tumor responses [110]. Discussion TME is shaped by continuous interactions among CAFs, immune cells, and tumor cells. CAFs are heterogeneous and can have opposing roles. For example, tumor-restraining CD105-negative fibroblasts have been shown to support anti-tumor immunity [23]. Several CAF states have been described, including immunoregulatory, matrix-remodeling, and therapy-resistant populations, associated with markers such as LRRC15, IL1R1, and galectin-9 [117]. These states are not fixed lineages. Instead, they are influenced by tumor-derived signals, including SHH- and EZH2-related pathways, as well as by local cytokine conditions [118]. As a result, treatment response often depends less on the number of stromal cells than on their functional state and interactions, particularly CAF-macrophage signaling that can either permit immune infiltration or maintain immune exclusion [119]. Spatial organization within the TME further affects therapeutic response. Certain CAF configurations limit CD8+ T-cell infiltration and are associated with resistance to immunotherapy and other treatments [120]. In contrast, TLS provide sites for local immune activation. TLS develop through lymphoid neogenesis initiated by lymphoid tissue inducer (LTi) cells and maintained by stromal cells such as Grem1+ FRCs, which support antigen presentation and lymphocyte interactions [121, 122]. Imaging studies suggest that TLS can form interconnected three-dimensional structures rather than isolated clusters. Their position within tumor tissue therefore likely influences their ability to support immune activity. In this sense, TLS can be viewed as an indicator of a stromal environment that permits immune activation. A major unresolved issue is whether current CAF nomenclature is sufficient for pan-cancer interpretation. The myCAF/iCAF/apCAF framework remains useful as a functional shorthand, but it should not be viewed as a fixed lineage-based taxonomy. CAF states are plastic, spatially conditioned, and shaped by tumor-derived signals, cytokines, and neighboring immune cells. Thus, future classification should move beyond single-marker or single-tumor labels toward a pan-cancer consensus that integrates conserved transcriptional programs with spatial localization, cellular origin, ligand-receptor signaling, matrix organization, and immune interactions [123]. Progress in this area will require spatial multi-omics approaches that capture tissue structure and stromal niche interactions, rather than relying only on dissociated cell data. Computational tools will also be needed to reconstruct three-dimensional maps of the TME and to standardize how spatial cell neighborhoods are defined and compared across datasets [124]. Despite the transformative impact of single-cell and spatial multi-omics technologies, several important limitations remain. Current spatial transcriptomic and proteomic platforms often involve trade-offs between spatial resolution, transcriptomic depth, and tissue coverage, with some approaches lacking true single-cell resolution or capturing only predefined molecular panels [125, 126]. In addition, sampling bias and tissue heterogeneity may complicate interpretation, as spatial analyses are frequently derived from limited tissue regions that may not fully represent the broader tumor ecosystem. Integrating datasets across platforms, cohorts, and modalities also remains computationally challenging because of batch effects, variable resolution, and differences in analytical pipelines. Furthermore, many current computational approaches infer spatial interactions from static tissue snapshots, which may incompletely capture the dynamic and temporal nature of stromal-immune organization. These limitations should be considered when interpreting spatial architectures within the tumor microenvironment and highlight the need for standardized analytical frameworks, longitudinal sampling strategies, and improved multimodal integration algorithms. From a therapeutic perspective, the aim is not simply to remove stromal cells but to shift stromal programs toward states that support immune activity. Several approaches illustrate this principle. TGFβ blockade can alter myofibroblastic programs and promote fibroblast states that are more compatible with immune infiltration. Local delivery strategies, including nanoparticle systems, may help concentrate stromal-modifying agents within tumors and improve the effects of immunotherapy. Other strategies exploit vulnerabilities within stromal populations, such as inducing ZBP1-dependent necroptosis in selected CAF states or targeting metabolic pathways that sustain immunosuppressive niches [127]. Some therapies also combine stromal modulation with direct anti-tumor effects. For example, Pin1 inhibition can influence both CAF programs and tumor-cell signaling, providing a rationale for integrated combination strategies. Conclusion and future perspectives In summary, the tumor microenvironment is not merely defined by the abundance of stromal components. It is also shaped by their functional states and spatial organization. CAFs exhibit substantial plasticity. Distinct CAF states can either constrain or support anti-tumor immunity, in part through their interactions with immune cells and their role in shaping tissue architecture. The presence of TLS further highlights a stromal context that is permissive to local immune activation. Together, these observations underscore the need for a unified, pan-cancer framework that integrates cellular phenotypes with spatial context. Several key questions should guide future work. First, are TLS, immuno-protective EcoCellTypes, and other immune-supportive stromal niches functional organizers of anti-tumor immunity, or mainly markers of an inflamed tumor state? Second, how do these spatial architectures change during tumor progression and after therapy? Third, which spatial features can be converted into reproducible clinical readouts? Addressing these questions will require spatially resolved multi-omics, longitudinal sampling, perturbation-based models, and computational approaches that reconstruct three-dimensional stromal niches and signaling networks. Therapeutically, strategies aimed at reprogramming stromal states, rather than depleting them, hold promise for enhancing immunotherapy efficacy. Targeted modulation of key pathways and localized delivery systems may enable precise remodeling of the tumor microenvironment. Such approaches may help support durable anti-tumor immune responses. Supplementary Information Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3