The Spatiotemporal Heterogeneity of Tumor-Associated Stromal Cells: Reprogramming Plasticity to Unlock Precision Cancer Immunotherapy.

Introduction
The tumor microenvironment (TME) has emerged as an indispensable component of the complex tumor ecosystem, spanning tumor initiation, development, and metastasis [1]. Tumor-associated stromal cells (TASCs) are prominent cells of the TME and include cancer-associated fibroblasts (CAFs), tumor-associated endothelial cells (TECs), cancer-associated pericytes (CAPs), cancer-associated adipocytes (CAAs), and carcinoma-associated mesenchymal stem cells (CA-MSCs; Fig. 1). These stromal components display a wide range of heterogeneity regarding their origins, subtypes, biomarkers, and functions (Table 1). Stromal cells modulate tumor progression through multiple mechanisms, including providing structural support, orchestrating immunosuppressive environments, and influencing tumor cell behaviors through the secretion of growth factors, cytokines, enzymes, and extracellular matrices (ECMs) [2]. In addition, other nonimmune stromal cells, including Schwann cells, astrocytes, and osteoblasts, are present in specialized TMEs [3,4]. These cells are not described in this review, as they are relatively understudied and excluded from mainstream discussions.
Single-cell analysis and spatial histology techniques have revealed the diversity of TASCs. The subtype heterogeneity of CAFs and TECs has been extensively explored, but the heterogeneity of other cells remains unclear. CAFs, as the mainstay of TASCs, can be divided into the following key types: myofibroblastic CAFs (myCAFs), inflammatory CAFs (iCAFs), vascular CAFs (vCAFs), antigen-presenting CAFs (apCAFs), and lipid-laden CAFs (lpCAFs). The heterogeneity of CAFs was initially identified in pancreatic ductal adenocarcinoma (PDAC), in which myCAFs highly expressed α-smooth muscle actin (α-SMA) and fibroblast activation protein-α (FAP). myCAFs promote fibrosis and remodel the ECM [5]. iCAFs are characterized by high expression of inflammatory factors and low expression of myofibroblast markers, thus contributing to the immunosuppressive microenvironment [6]. apCAFs, which express major histocompatibility complex (MHC) class II molecules but not classical costimulatory molecules, were first identified in murine models of PDAC [7]. This unique phenotype was later confirmed in human breast and lung cancer [8,9]. vCAFs, first systematically identified as a separate subpopulation of cells in the mouse mammary carcinoma (MMTV-PyMT) model, exhibit vascular features and angiogenesis-related gene expression profiles [10]. lpCAFs are characterized by lipid storage and transport, which accelerate malignant tumor progression through metabolic interactions [11,12]. Based on single-cell RNA sequencing (scRNA-seq) analyses of cross-cancer datasets, an increasing number of subgroups have emerged, further resolving the complexity and heterogeneity of CAFs [13,14]. Eight CAF subsets, which are widely explored across diverse tumor contexts, are shown in Fig. 2. Analogously, the taxonomy of annotated TEC subpopulations has been refined, mainly encompassing arterial TECs, capillary TECs, venous TECs, tip cells [15], activated postcapillary veins (PCVs), lymphatic TECs [16], and immunomodulatory TECs [17].
Novel computational frameworks leveraging single-cell omics data now enable predictions of intractable biological processes, such as cell-to-cell interaction networks and spatiotemporal gene expression dynamics. CellChat, as a competent tool for cellular interactions, is important for exploring the interaction networks among TASCs, tumor cells, and immune cells [18]. Trajectory analysis, RNA velocity, and dynamic scRNA-seq capture the dynamic changes in TASCs and increase the understanding of complex spatiotemporal heterogeneity [19,20]. From a temporal perspective, as tumors progress, the functions of stromal cells evolve as follows: (a) orchestrating tumor angiogenesis, providing oxygen and nutrition to support tumor growth [2], (b) metabolic reprogramming to supply energy and shape the TME [21], (c) participating in epithelial–mesenchymal transition (EMT) and endowing tumor cells with invasive properties [22], and (d) constructing the ECM and interacting with immune cells, leading to the formation of immunologically inert tumors [3]. Spatially, stromal cells in different tumor regions display distinct characteristics. In the hypoxic tumor core, iCAFs predominantly promote tumor growth, whereas at the boundary, myCAFs shift toward immunosuppression and facilitate tumor cell dissemination [4,23]. This spatiotemporal heterogeneity profoundly affects the functions of TASCs and, in turn, shapes the immune landscape and therapeutic responses.
To elucidate the spatiotemporal plasticity of TASCs, recent studies have increasingly emphasized the spatial architecture of tumors, aiming to map the spatial distribution, neighborhood topologies, and colocalization patterns of stromal constituents [24]. Spatial transcriptomics (ST) and its derivative platforms (e.g., Visium, MERSCOPE, CosMx, and Stereo-seq) provide insights into the cellular ecological niche of the TME while preserving spatial topological information [25]. A meta-analysis integrating 7 spatial-omics platforms across 10 cancer types has resolved the spatial subtypes of CAFs (S1 to S4), which exhibited differentiated spatial localization and diverse cellular interaction networks. S1-CAFs were adjacent to tumor cells, orchestrating collagen synthesis and matrix remodeling; S2-CAFs were localized in the peritumoral stroma and displayed a gene expression profile reminiscent of that of iCAFs; S3-CAFs encircled the vasculature and engaged in crosstalk with macrophages and neutrophils; and S4-CAFs were enriched in lymphoid aggregates and facilitated the recruitment of adaptive immune cells [26]. Despite these advances, most existing reviews focus on the molecular typing or functional heterogeneity of TASCs [27–29]. There is still a gap in the systematic exploration of their spatial heterogeneity. In this paper, we break through the traditional framework and systematically compare the spatial heterogeneity of TASCs from 3 perspectives: regional heterogeneity within the tumor, spatial evolution between the primary foci and metastases, and differences across various cancer types. Overall, this review provides a comprehensive overview of TASCs, focuses on their spatiotemporal heterogeneity, elucidates their dynamic associations with immune responses, and summarizes current therapeutic strategies. By dissecting these complex cellular interactions, we aim to establish a framework for novel targeted therapies and offer avenues to improve patient outcomes in clinical settings.

Temporal Trajectory: Dynamic Evolution of TASCs
As dynamic entities, TASCs undergo marked functional transformations throughout the continuum of tumor development. Distinct subpopulations exert predominant influences at different stages.

Tumor initiation phase: The activation of TASCs
In the TME, cancer cells recruit and educate stromal cells, triggering a transition from indolent to active states. Different cytokines produced by tumor cells shape the heterogeneity of TASCs (Fig. 3A). In particular, the transformation of the CAF phenotypes relies on unique modes of action (Table 2). For CAPs and TECs that acquire invasive phenotypes, pro-angiogenic factors, such as vascular endothelial growth factor (VEGF) and fibroblast growth factor (FGF), and anti-angiogenic factors (e.g., endostatin and angiostatin) released by cancer cells are dominant regulators of angiogenic activity [30–33]. CAA activation is orchestrated by inflammatory factors, including interleukin-6 (IL-6) and interleukin-8 (IL-8) [34–36]. Mesenchymal stem cell (MSC) migration toward the tumor site is guided by a concentration gradient of factors. For example, cervical cancer cells recruited MSCs via the C–X–C motif chemokine ligand 12 (CXCL12)/C–X–C chemokine receptor type 4 (CXCR4) axis [37]. Upon arriving at the tumor site, MSCs further differentiate into CA-MSCs in response to inflammatory or hypoxic signals. A key example is that in breast cancer (BC), hypoxia-inducible factor (HIF) can educate MSCs to promote tumor progression [38].
Notably, epigenetic alterations in tumor cells also reprogram TASCs. In pancreatic cancer, the ectopic enrichment of histone H3 lysine 27 (H3K27) acetylation in tumor cells has been observed during the transition of lpCAFs. This epigenetic mechanism triggered bone morphogenetic protein 2 (BMP2) signaling, reinforcing the protumorigenic phenotype of CAFs [39]. In addition, noncoding RNAs also engage in the regulation of stromal cell behaviors. miR-204-5p from tumor cells up-regulated hypoxia-inducible factor 1A (HIF1A) expression in white adipose tissue, leading to lipolysis and CAA maturation in engineered mice with BC [40].

Tumor development phase: Dynamic progression of TASCs

TASC metabolic reprogramming regulates tumor development
As nurse cells, TASCs provide nutrients and energy for highly proliferative cancer cells. The Warburg effect, a crucial hallmark of cancer, indicates that cancer cells prefer glycolysis for energy production over oxidative phosphorylation, even under aerobic conditions [41]. Similarly, TASCs exhibit preferred metabolic patterns in pathological states, such as glycolysis in CAFs and TECs [42,43] and lipid metabolism in CAAs [44] (Fig. 3B). Cancer cells stimulate glycolysis of CAFs, which is referred to as the reverse Warburg effect. In this state, CAFs increase glucose uptake and lactate production [45]. Moreover, CAFs engage in glutamate/glutamine metabolism, compensating for the Warburg effect-induced suppression of glucose metabolism in tumor cells [46,47]. Similarly, enhanced glycolysis in TECs is advantageous for satisfying the oxygen demand for tumors [48]. Especially during angiogenesis, tip cells transform into a glycolytic phenotype via vascular endothelial growth factor A (VEGFA) to obtain sufficient adenosine triphosphate (ATP) [49]. Fructose-6-phosphate-2-kinase/fructose-2,6-bisphosphatase 3 (PFKFB3), an activator of glycolysis, has been demonstrated to drive metabolic dysregulation in TECs. Pharmacological blockade or inhibition of PFKFB3 could remodel the tumor vascular microenvironment, indicating potential therapeutic promise [50]. In parallel to glycolysis, lipid metabolism plays a central role. CAAs and lpCAFs fulfill malignant energy demands by releasing fatty acids, glycerol, and leptin [39,51]. Under the action of CAAs, lipid metabolism predominates in tumor cells. For example, CAAs up-regulated HIF1A via the IL-6 and Janus kinase (JAK)/signal transducer and activator of transcription 3 (STAT3) signaling pathways, rewiring the glucose metabolism of ovarian cancer cells to phosphatidylcholine synthesis [52].
The metabolic rewiring of the TME exerts far-reaching effects beyond energy provision, critically regulating tumor immune responses. In the context of immunosuppression, accumulated lactate inhibits the activity of cytotoxic T lymphocytes [53]. Indoleamine-2,3-dioxygenase (IDO), an up-regulated metabolic enzyme in CA-MSCs, establishes links between tryptophan metabolism and T cell exhaustion. This leads to tryptophan depletion and kynurenine accumulation, disrupting the normal metabolism of T cells [54].
More importantly, metabolic–epigenetic crosstalk is gaining considerable attention. Metabolites are essential substrates for epigenetic remodeling (e.g., acetylation and methylation). For instance, acetyl-coenzyme A derived from glycolysis drove tumor progression by enhancing histone acetylation [55]. Analogously, CAF-derived acetate mediated SP1 protein acetylation, up-regulating spermidine/spermine N1-acetyltransferase 1 (SAT1) expression to support pancreatic cancer growth [56]. S-Adenosyl methionine (SAM), as a substrate for nicotinamide N-methyltransferase (NNMT), participates in histone methylation. Research combining bladder cancer samples and in vivo experiments revealed that NNMT up-regulation in CAFs depleted SAM within the TME. Low levels of SAM led to restricted methylation and overexpression of serum amyloid A (SAA). SAA, in turn, promoted the infiltration of tumor-associated macrophages (TAMs), driving uroepithelial bladder cancer [57]. These findings underscore that the complex interplay between metabolic rewiring and epigenetic reprogramming drives tumor aggressiveness.

Tumor angiogenesis leads to deterioration
As the tumor enlarges, the tumor core becomes increasingly hypoxic and nutrient-deprived [58]. Angiogenesis is essential for overcoming this constraint. To initiate this process, TASCs release pro-angiogenic factors, such as VEGFA [59,60], platelet-derived growth factor (PDGF) [61], and CXCL12 [55], to recruit endothelial cell precursors and increase vascular permeability (Fig. 3C). Additionally, adipokines secreted by CAAs contribute to neovascularization in BC. These underlying mechanisms are involved in the activation of the interleukin-1 (IL-1) receptor pathway and Wnt/β-catenin signaling in endothelial cells [62,63].
In addition to angiogenesis, vascular co-option is an alternative survival strategy for tumors, whereby cancer cells co-opt the preexisting vasculature to obtain nutrients and support metastasis (Fig. 3C). This phenomenon has been extensively observed, such as in BC liver metastases and glioblastoma [64,65]. The presence of vascular co-option explains the limitations of anti-angiogenic therapy (AAT) and points to novel directions for therapeutic strategies.

Tumor metastasis phase: Functional shift of TASCs

EMT facilitates invasion
During the entire course of carcinogenesis, EMT is vital for metastasis. Tumor cells undergoing EMT adopt a spindle-like morphology and acquire enhanced invasive abilities, which are conducive to extravasation [66] (Fig. 3D). TASC-derived secretions constitute the central mechanism of metastatic EMT. In a coculture experiment involving CAFs and BC cells, carcinoma cells in EMT states were characterized by increased expression of mesenchymal markers (e.g., N-cadherin and vimentin) and decreased expression of epithelial markers (e.g., E-cadherin and ZO-1) [67]. This phenotype shift can be pharmacologically reversed by inhibiting the transforming growth factor-β (TGF-β) pathway [68]. TGF-β, a dominant secretory factor, has been implicated in EMT activation across multiple malignancies. For example, all CAFs in colorectal cancer (CRC) [69], gastric cancer (GC) [70], and bladder cancer [71], TECs in hepatocellular carcinoma (HCC) [72], and CA-MSCs in melanoma [73] secrete TGF-β, thereby promoting EMT. Moreover, other factors, such as IL-6 derived from CAAs [36] and Semaphorin 3C from CAFs [74], trigger EMT during tumor invasion. Recently, more precise molecular mechanisms underlying EMT have been illuminated, revealing several specific functional subsets. Thrombospondin 2 (THBS2+) mCAFs, as a matrix CAF subset, were associated with EMT [75]. THBS2 accelerated EMT in lung adenocarcinoma by binding to syndecan-4 on tumor cells [76]. In addition, they secreted collagen type VIII α1 chain (COL8A1) to activate phosphoinositide 3-kinase (PI3K)–protein kinase B (AKT) signaling in malignant cells and drove EMT of CRC [77]. Complement factor D (CFD+) iCAFs, which were inflammatory CAFs found in metastatic CRC, accelerated EMT via the secreted frizzled-related protein 1 (SFRP1)–fibroblast growth factor receptor 2 (FGFR2)–HIF1 signaling axis [78].

ECM remodeling supports tumor metastasis
With respect to tumor metastasis, TASCs establish a favorable immune microenvironment through ECM remodeling, such as ECM deposition and ECM degradation [79] (Fig. 3E). In terms of deposition, myCAFs are the master architects of the ECM and secrete collagens, fibronectin, proteoglycans, and periostin [80,81]. Notably, heterogeneous CAF subtypes with unique collagen expression profiles have been proposed, which result in diverse ECM compositions and properties (Table 2). In addition to directly generating ECM constituents, CAFs secrete lysyl oxidase (LOX) and lysyl oxidase-like proteins (LOXLs) to mediate the cross-linking and maturation of collagen [82,83]. Dense components and robust structures ultimately catalyze matrix stiffening. Stiff ECMs subsequently affect cellular signaling pathways, ultimately exacerbating tumor progression [84]. For example, mechanical cues from rigid ECMs trigger transduction signals, such as Ras-homologous (Rho)/Rho-associated coiled-coil containing kinase (ROCK), which regulate the cytoskeleton and accelerate metastasis [85]. Moreover, the remodeled stroma acts as a native immune barrier, aiding tumor invasion. In PDAC, disturbed fiber orientations and increased collagen densities failed to mediate T cell migration into tumor islets [86]. Interestingly, periostin (POSTN+) CAFs, which were involved in ECM remodeling, suppressed TAMs via IL-6, establishing their dual role in tumor progression [87–89].
To degrade the ECM, CAFs secrete matrix metalloproteinases (MMPs) and urokinase-type plasminogen activators (uPAs) [90] (Fig. 3E). Similarly, CAAs participate in degradation by producing MMP2, MMP7, MMP9, MMP11, and plasminogen activator inhibitor-1 (PAI-1) [91]. Without intricate 3-dimensional (3D) structures, cancer cells are apt to undergo metastatic dissemination. Given that the ECM serves as a reservoir for cytokines, its degradation releases various growth factors such as TGF-β, VEGF, and FGF, which are instrumental in tumor angiogenesis and metastasis [92]. Moreover, ECM fragments act as chemotactic agents, accelerating tumor development. For instance, the collagen VI α3 chain attracted macrophages, facilitating EMT in BC [93]. Overall, the balance between ECM deposition and degradation is crucial for maintaining tissue homeostasis. Normalizing ECM remodeling or combining it with immunotherapy holds potential value as a new therapeutic direction [94].

Spatial Decoding: Regional Specificity of TASCs
Recently, advances in spatial technologies have attached importance to spatial heterogeneity, including tumor internal ecological niches, cellular neighborhoods, and overall tumor spatial architecture. Spatial profiling of heterogeneity enables a novel biomarker framework for precise prognostic stratification and personalized immunotherapy optimization [95].

Ecological niche in the TME: Distribution and interaction of TASCs
TASCs, epithelial cells, and immune cells collectively shape the cellular landscape and spatial architecture of tumors. Insights into the spatial structure of tumors elucidate the distribution and intercellular communication among these cells, offering novel perspectives on tumor growth, invasion, and metastasis.

Distributions and spatial heterogeneity of TASCs within the TME
From the perspective of the TME, spatial heterogeneity is characterized by 3 functionally distinct compartments: the tumor core, the invasive margin, and the tumor stroma [96]. The characteristics of the cellular composition within each compartment are delineated in Fig. 4A. The tumor core, as an immune-exhausted zone, is dominated by tumor cells, immunosuppressive leukocytes [e.g., TAMs, myeloid-derived suppressor cells (MDSCs), and regulatory T cells (Tregs)], as well as a few CAFs and TECs [96]. The invasive margin serves as a 2-sided battlefield for immune surveillance and tumor escape. The dynamic balance between protumor factors (e.g., CAFs and M2 macrophages) and antitumor factors [e.g., tertiary lymphoid structures (TLSs), CD8+ T cells, and natural killer (NK) cells] in this region makes it critical for metastasis [97]. The outermost stromal region of the tumor constitutes a supportive microenvironment, composed of CAFs, CAAs, the ECM, and blood vessels [98].
The spatial heterogeneity of TASCs affects their functions and interactions with other cells. Here, we delve into the distribution of CAFs and TECs from the perspective of the overall TME. The unique niches of major CAF subsets have been revealed with spatial mapping. myCAFs are in proximity to tumor cells and are concentrated within the stromal compartment or the invasive front [99,100]. vCAFs are enriched in the vascular zone [101], and apCAFs colocalize with immune cells such as macrophages and T cells, resulting in synergistic immunological effects [102]. In contrast to the peritumoral localization of myCAFs, iCAFs are located in fibrogenic areas distant from tumor cells according to the initial definition [6,103]. The spatial resolution of multiple tumors has further confirmed this phenomenon, suggesting that the differences in tumor origins and activated signals are potential drivers of spatial heterogeneity [104]. Interestingly, defined by their enrichment in hypoxic regions, hypoxic CAFs (HCAFs) exhibit a stronger correlation with hypoxia levels than other CAFs [105].
TECs are distributed in the tumor core and margins along with the vascular system. Compared with peritumoral endothelial cells, tumor-core TECs exhibit marked heterogeneity in their abundances, molecular subtypes, and function. First, the overall abundance of TECs is markedly increased along with a rise in pericyte abundance [106]. Second, the proportions of TEC subtypes are dysregulated, with a reduction in capillary endothelial cells and an enrichment of immature TECs and pro-angiogenic tip cell subtypes [107]. These changes reflect profound dysregulation of vascular maturation. Third, the TEC-enriched signaling pathways in the core region are biased toward angiogenesis. According to scRNA-seq analysis of glioblastoma patients, TECs in the tumor core acquired genetic signatures linked to basement membrane remodeling and tip cell formation. This subpopulation was predominantly clustered in the microvascular proliferation (MVP) region, which functionally manifested enhanced angiogenic capacity [108,109]. TECs also exhibited impaired immune defense functions, as evidenced by the suppressed expression of interferon-stimulated genes and the transcriptional program for antigen presentation (pS05) in CRC [110]. Furthermore, TECs actively enforced immune exclusion via FasL-mediated CD8+ T cell exhaustion and secretion of CXCL12/14, inhibiting cytotoxic T cell maturation [111–113].
While TECs in the tumor core possess a highly abnormal and immunosuppressive phenotype, their counterparts at the tumor margin present a distinct, yet equally consequential niche. The tumor margins are more densely packed with relatively normal functioning vessels and a higher percentage of venous endothelium. For example, a venous microenvironment consisting of atypical chemokine receptor 1-positive (ACKR1+) TECs and C–C motif chemokine ligand 21-positive (CCL21+) peripheral pericytes was present in the tumor margin area and metastatic lymph nodes in BC. ACKR1+ TECs were enriched in postcapillary microvessels, whereas CCL21+ pericytes formed concentric laminar structures around these vessels. These cells spatially synergized to form a chemokine-rich microenvironment that promoted the infiltration of C–C motif chemokine receptor 7-positive (CCR7+) lymphocytes, highlighting the critical role of spatial heterogeneity in immunotherapy [114].

Spatial architectures composed of TASCs
Specific aggregations of TASCs with surrounding cells constitute unique ecological niches. These niches are not only landmarks of tumors’ spatial architectures but also functional units that shape the functional heterogeneity of the TME. Several niches containing the immune-related TLS, tumor-invasive front (TIF), and vascular zone have been extensively characterized (Fig. 4B).
TLSs, as ectopic lymphoid organs, are predominantly found at the invasive margin or stromal regions of tumors [115]. The main components include immune protective cells [e.g., B cells, T cells, and dendritic cells (DCs)] and tumor-associated high endothelial venules (TA-HEVs) [116]. To recruit constituents and orchestrate spatial architectures, TASCs function as organizers. CAFs maintain lymphocyte aggregates by expressing lymphotoxin β receptor (LTβR) and tumor necrosis factor receptor (TNFR) [117]. Specifically, CCL19+ iCAFs were confirmed to recruit B cells through the CCL19/CCR7 axis in a humanized CRC model [118]. TA-HEVs, a specialized subset of TECs, provide convenient entries for immune cell infiltration [119]. In vitro experiments of nasopharyngeal carcinoma have demonstrated that TA-HEVs were induced by type I and type II interferons. These TA-HEVs, in turn, mediated immune responses through colocalization with CD4+ T cells via CXCL9 [120]. Similarly, TA-HEVs enhanced the attraction, adhesion, and infiltration of CD8+ T cells by expressing sialomucins and E/P-selectins [121]. Following insights into the components and forming mechanisms, TLSs have been established as an important marker of good prognoses, which was first revealed in non-small cell lung carcinoma (NSCLC) [116]. Building on this foundation, deepening the knowledge of TASCs in TLSs, especially TA-HEVs, has paved the way for new therapeutic avenues in modulating the tumor immune microenvironment (TIME).
In contrast to the immune protection of TLSs, TIF, a spatially and dynamically heterogeneous region, orchestrates local invasion and distant metastasis by counteracting immune activities [122]. With the advancement of ST, the landscape of TIF has become clear. Zhou’s team [123] achieved subcellular-level localization of TIF in HCC using Stereo-seq, a new ultrahigh-resolution technology. This invasive zone was defined as the special band spanning 500 μm on the transition area of carcinoma and paracarcinoma tissue. Technologies have further characterized the profound immunosuppression within the TIF, shedding light on the critical crosstalk between CAFs and macrophages [124]. On the basis of CellPhoneDB and ST analyses, the spatial proximity of A-kinase anchoring protein 12 (AKAP12+) CAFs and M2 macrophages in triple-negative breast cancer (TNBC) suggested that CAFs induced macrophage polarization via interleukin-34 (IL-34). IL-34 signaling blockade with programmed cell death-1 (PD-1) therapy promoted immunotherapeutic efficacy in vivo experiments [125]. Moreover, TAMs and CAFs cooperate to construct immune barriers around tumor boundaries. Secreted phosphoprotein 1 (SPP1) from TAMs, as a potent mediator of fibrosis, enhances the profibrotic function of CAFs through combination with a cluster of differentiation 44 (CD44) [126]. The fibrotic barrier, formed by CAFs, hinders T cell infiltration into the tumor core [127]. In contrast, in BC, myCAFs promoted CD8+ T cell aggregation through the expression of elastin microfibril interfacer 1 (EMILIN1), which inactivated TGF-β signaling. This finding contradicted the previously held perception of myCAFs as uniformly immunosuppressive [128].
In addition to the above 2 niches, the vascular zone establishes a pathway for tumor cell dissemination. TECs line the inner side of blood vessels [129], whereas CAPs reside on the exterior and are embedded in the basement membrane [130]. In addition, sparse vCAFs encircle the vasculature, acting as critical mediators of angiogenesis [101]. Close spatial apposition and interaction between TECs and CAPs are essential for maintaining vascular homeostasis. Cytokines secreted by TECs, such as PDGF-BB and TGF-β, are crucial mediators of CAP recruitment [131]. In turn, CAPs maintain the quiescent state of TECs through angiopoietin-2 (Ang2)/Tie and PDGF/platelet-derived growth factor receptor (PDGFR) [132,133]. However, in response to the TME, this cooperative relationship is disrupted. CAPs detach from endothelial cells, leading to anomalous vascular structures that facilitate cancer cell metastasis [132]. For example, elevated prostaglandin E2 (PGE2) levels in the TME impaired TEC-CAP connections by down-regulating N-cadherin expression in CAPs [134]. Furthermore, dysregulated pericyte contractility contributed to poor vascular perfusion and vessel leakiness. This characteristic was mechanistically linked to the expression of regulator of G protein signaling 5 (RGS5) [135].
TECs also organize the perivascular microenvironment through interactions with immune cells and cancer stem cells (CSCs). For example, TECs induced the accumulation of M2 macrophages and the depletion of cytotoxic T cells, thereby establishing an immunosuppressive milieu [136,137]. Concurrently, TECs cultivated a fertile ground for CSCs, especially in brain tumors, where their co-occurrence was critical for tumor stemness [138,139]. Another study demonstrated that TECs promoted the malignant transition of glioma stem cell-like cells through the MMP–nuclear factor κB (NF-κB) pathway [140].

Heterogeneous landscape of the tumor core: The spatial heterogeneity of TASCs between primary and metastatic tumors
The malignant evolution of tumors is a highly coordinated spatial process. From primary sites to distal organs, stromal components of the TME display spatial heterogeneity. In primary sites of GC, stromal cells, including CAFs, TECs, and protective immune cells (e.g., DCs and cytotoxic T cells), constituted a relatively well-balanced, yet perturbed microenvironment. In contrast, peritoneal metastases were marked by exhausted CD8+ T cells [141]. Single-cell analyses of metastatic samples from other malignancies further confirm a trend toward reduced stromal infiltration and enhanced immunosuppression of metastatic foci. For instance, the proportion of CAFs was reduced in pancreatic cancer liver metastases [142], and a marked increase in M2 macrophages was detected in lung adenocarcinoma brain metastases [143]. Moreover, CAF-derived insulin-like growth factor-binding protein 2 (IGFBP2) in peritoneal metastatic sites of CRC inhibited macrophage activation and T cell proliferation [144].
The pronounced functional reprogramming of TASCs in metastatic lesions further shapes spatial heterogeneity. Before the arrival of tumor cells, the primary tumor remotely modifies TASCs in distant organs to establish a premetastatic microenvironment (PMN) [145]. CXCL14 from osteosarcoma stimulated CAFs to secrete TGF-β, thereby attracting tumor cells to the lung niche [146]. Following dissemination, TASCs facilitate tumor cell lodging. Melanoma cell-derived extracellular vesicles (EVs) in murine models promoted the adhesion of tumor cells to the endothelium by up-regulating intercellular adhesion molecule 1 (ICAM1) expression on TECs [147]. Lymphatic TECs supported metastasis through stromal cell-derived factor-1 (SDF-1) binding to the CXCR4 on tumor cells [148]. Lastly, unique cell groups are enriched in metastases. Single-cell analysis of 4 resected esophageal squamous cell carcinoma (ESCC) specimens revealed increased numbers of pericytes with angiogenic features in metastatic lymph nodes, whereas in the primary foci, their counterparts predominantly regulated cell migration. A distinct pericyte subset (C6_pericyte), marked by high expression of Thy-1 cell surface antigen (THY1), PDGFRβ, RGS5, and NDUFA4 mitochondrial complex-associated like 2 (NDUFA4L2), was enriched in metastatic lymph nodes. This subset was proposed to orchestrate the pericyte–fibroblast transition (PFT) to accelerate tumor progression [149].

A comparative view of the distribution of TASCs across tumor types
Different tumor types display a preference for specific target organs, shaping the spatial heterogeneity of TASCs at the organ level. CAFs demonstrate conserved and context-dependent heterogeneity across solid tumors (Table 3). While myCAFs are ubiquitous in desmoplastic tumors (e.g., PDAC and BC), the abundance of iCAFs increases in aggressive and immunosuppressive niches [150]. Crucially, iCAF subpopulations drive divergent immune landscapes. In “hot” tumors, iCAFs interact with B cells and T cells via the CXCL12–CXCR4 axis, whereas in “cold” tumors, which are infiltrated by MDSCs and Tregs, such as prostate cancer and liver cancer, iCAFs are deficient [151].
Single-cell mapping of TECs across cancer types, such as lung, liver, and brain malignancies, has revealed intertumor heterogeneity and dynamic plasticity [16,106]. Human scRNA-seq data have indicated that TECs in glioblastoma up-regulate plasma lemma vesicle-associated protein (PLVAP) for transcellular transport but down-regulate transporter protein genes [e.g., ATP binding cassette subfamily B member 1 (ABCB1) and ATP binding cassette subfamily G member 2 (ABCG2)], suggesting compromised blood–brain barrier integrity [108]. In HCC, TECs lack the expression of canonical liver-sinusoidal endothelial markers, C-type lectin domain family 4 member G (CLEC4G), and express macrovascular endothelial signatures [e.g., platelet and endothelial cell adhesion molecule 1 (PECAM1), aquaporin 1 (AQP1), and CD34] as well as PLVAP. PLVAP induces a decrease in endothelial permeability, which in turn mediates immune escape, shaping the immunosuppressive microenvironment [152]. In fat-infiltrated cancers, particularly BC and ovarian cancer, adipocytes constitute a substantial portion of the tumor stroma, promoting inflammation, metabolic rewiring, and cancer cell modulation [153–155]. The differentiation potential of CA-MSCs into CAFs, TECs, and adipocytes may vary across different tumors, leading to diverse functional mechanisms [156,157]. This variability necessitates further research to elucidate the specific roles and mechanisms of CA-MSCs in various tumor contexts.

Immune Escape Resolution: Dynamic Association of TASCs with Immune Checkpoints
Immune checkpoints are molecules that prevent immune overactivation in the physiological state, such as PD-1, programmed death-ligand 1 (PD-L1), and cytotoxic T lymphocyte-associated antigen-4 (CTLA-4). However, tumors exploit these checkpoints to evade immune detection, and inhibitors are used to enhance antitumor immunity [158]. In this context, understanding the regulatory relationships between TASCs and immune checkpoints is necessary.
TASCs are involved in the regulation of immune checkpoints through an array of mechanisms. We systematically summarized the 2 main ways: expressing immune checkpoint ligands and orchestrating the immune checkpoints on other cells through cytokines (Fig. 5). First, TASCs themselves express a range of ligands that interact with immune checkpoints. PD-L1/PD-L2 [159,160] and B7 homolog 3 (B7-H3) [161] are present on the surface of CAFs. TECs express PD-L1, which is induced by interferon-γ (IFN-γ). In vivo experiments demonstrated that PD-L1+ TECs blunted the activation and proliferation of CD8+ T cells [162]. PD-L1 has also been detected in murine brown adipose tissue, suggesting a potential link between CAAs from this tissue and PD-L1 expression [163]. Next, TASCs reprogram checkpoints on immune cells, disrupting immune defense. Analyses of human samples have unveiled that the abundance of CAFs is correlated with an enhanced CTLA-4 gene signature in CD8+ T cells, which accounts for the T cell exclusion [164]. CAFs connect with PD-1/PD-L1 signaling on T cells through the secretion of IL-6 [165] and IL-8 [166,167] to trigger the STAT3 pathway. In parallel, TECs up-regulate the expression of PD-L1, B7-H3, and B7 homolog 4 (B7-H4) on T cells via factors such as VEGF, PGE2, and IDO1 [168,169].

Reprogramming of TASCs: An Emerging Paradigm in Cancer Immunotherapy
The spatiotemporal heterogeneity of TASCs not only drives tumor progression but also presents a major obstacle to successful therapy. A key challenge is that TASCs dynamically remodel the TME during therapeutic interventions, thereby driving the emergence of therapeutic resistance (Table 4). For example, CAF-secreted hepatocyte growth factor (HGF) is responsible for resistance in BRAF-mutated melanoma [170] and thyroid anaplastic carcinoma [171]. Therefore, targeting the context-dependent functions of TASCs is crucial for overcoming therapeutic failure. To address this, Fig. 6 outlines promising therapeutic avenues that exploit this spatiotemporal heterogeneity. Furthermore, multiple clinical trials evaluating TASC-targeted therapeutic strategies are ongoing (Table 5).

Target TASCs in the context of temporal heterogeneity

Targeting TASCs sculpts antitumor niches
Employing molecular markers of TASCs to guide immunotherapy is pivotal for tumor treatment. Accordingly, a number of clinical trials have been designed to target these specific molecular markers, as summarized in Table 5. However, the functional heterogeneity of TASCs necessitates caution, as monolithic inhibition or depletion strategies may compromise therapeutic efficacy and potentially promote tumor progression. Here, we use CAFs as a prominent example to address this challenge. In genetically engineered mouse models of PDAC, eliminating FAP+ CAFs resulted in improved survival, whereas removing α-SMA+ CAFs led to reduced survival [172]. This complexity was further illustrated by the expression of yes-associated protein 1 (YAP1), a key regulator in CAF phenotypic transformation. YAP1 activation drove the differentiation of extracellular matrix-associated CAFs (ECM-CAFs), whereas YAP1 silencing conversely reprogrammed CAFs toward a lymphatic-like phenotype. In Yap1flox/flox mice, selective YAP1 knockdown in ECM-CAFs resulted in damaged ECM immune barriers and stabilized lymphatic vessels. This synergistic effect boosted CD8+ T cell trafficking, thereby augmenting the efficacy of anti-PD-1 therapy [173]. Unfortunately, manipulating the phenotypic transformation of CAFs does not always yield beneficial outcomes. Blocking the Hedgehog pathway inhibited myCAFs but increased the proportion of iCAFs, resulting in a stronger immunosuppressive microenvironment [174].
The intricate and even contradictory outcomes of pathway-specific interventions have motivated the exploration of broader regulatory strategies, such as epigenetic reprogramming. DNA methylation and histone deacetylation constitute pivotal regulatory mechanisms within TASCs. Clinically available agents include approved DNA methyltransferase (DNMT) inhibitors (e.g., 5-azacitidine, 5-aza-2′-deoxycytidine, and GSK3685032) and histone deacetylase (HDAC) inhibitors (e.g., vorinostat, panobinostat, and givinostat) [175]. Nevertheless, the efficacy of these agents in remodeling TASCs remains inadequately characterized.
Recently, mechanistic insights for novel epigenetic therapeutic strategies have been gleaned, typically in CAFs and TECs. CAFs display heterogeneous DNA methylation patterns, but harbor evolutionarily conserved hypomethylated CpG sites. These sites represent actionable therapeutic targets [176]. The ectopic expression of DNMT1 has been identified [177]. Pharmacological inhibition using 5-aza-2′-deoxycytidine attenuated CAF malignancy and showed therapeutic benefits in PDAC murine models [178]. Notably, epigenetic alterations within CAFs are also involved in angiogenesis [179], metabolic reprogramming [180], and tumor invasion [181]. Despite their therapeutic potential, rigorous preclinical validation and clinical trial data are currently limited.
TECs exhibit distinct DNA methylomes, with promoter hypomethylation linked to pro-angiogenic activation and tumor invasion [182]. Furthermore, HDAC-dependent histone deacetylation has been shown to silence adhesion molecules (e.g., ICAM1 and VCAM1) [183], rationalizing combinatorial AAT and epigenetic targeting. Ongoing related combined strategies include 5-azacitidine and anti-VEGF monoclonal antibody (bevacizumab) for renal cell carcinoma (NCT00934440) and vorinostat and epidermal growth factor receptor (EGFR) inhibitor (gefitinib) for NSCLC (NCT02151721). In summary, targeting TASCs through their molecular markers and epigenetic machinery has profound translational potential. Combining epigenetic therapies with existing modalities such as immunotherapy and anti-angiogenic treatment represents a promising frontier for improving cancer treatment outcomes.

Turning enemies into friends: Reprogramming TASCs
Beyond eradicating TASCs, reprogramming them from protumorigenic to a quiescent or antitumorigenic state represents an advanced therapeutic strategy. This approach aims to normalize the TME, thereby attenuating stromal support for malignancy and enhancing the efficacy of concomitant therapies.

Reprogramming CAFs
Reprogramming CAFs as an adjuvant strategy has moved from the bench to the bedside. In PDAC, vitamin A deficiency and vitamin D receptor (VDR) expression are key triggers for the activation of pancreatic stellate cells (PSCs). Preclinical experiments have confirmed that normalizing murine PSCs with the external VDR ligand calcipotriol markedly improved survival by 57% when combined with chemotherapy [184,185]. Clinically established all-trans retinoic acid and lenvatinib bind to the retinoic acid receptor β (RAR-β) and FGFR, respectively. These interactions regulate CAFs and hamper tumor progression [186]. Novel therapeutic approaches in preclinical studies have emerged. An NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) oxidase 4 (NOX4) inhibitor was shown to normalize CAFs in vivo, improving CD8+ T cell infiltration [164]. To precisely reprogram CAFs, EVs modified with integrin α5-targeting peptides have been developed. These EVs deliver cargo such as miR-138-5p and pirfenidone to inhibit CAF activation by blocking the TGF-β signaling pathway [187].

Reprogramming TECs
Approaches to reprogramming TECs mostly focus on AAT, which are described in Targeting the tumor vasculature starves tumors. Beyond AAT, an emerging strategy involves reprogramming TECs into HEVs. Evidence indicates that during cancer immunotherapy, infiltrating CD8+ T cells and NK cells activate LTβR, which facilitates post-capillary venule transformation into HEVs. HEVs, in turn, attract T cells to hamper tumor progression [17]. Therefore, actively reprogramming TECs into HEVs represents a promising future therapeutic direction.

Reprogramming CAAs
Although the basic research on CAAs is still weak, breakthroughs in translational research have laid the key technological foundations for CAA reprogramming strategies. A study developed engineered adipocytes with clustered regularly interspaced short palindromic repeat activation (CRISPRa) that exhibited high metabolism, dramatically inhibiting the proliferation of cancer cells. This antitumor effect was mediated by the restriction of glucose and lipid metabolism in tumor cells, as demonstrated in murine models of PDAC and BC [188].

Reprogramming CA-MSCs
Accumulating evidence demonstrates that the carcinogenic effects of CA-MSCs can be rewired by exogenous signals. In mouse melanoma, studies have shown that CA-MSCs recruit macrophages and MDSCs under the orchestration of NF-κB signaling. Retinoic acid blocks the NF-κB pathway, thereby preventing immune cells and reversing tumor progression [189]. Treatment with IL-12 in TNBC mice counteracted the tumor-promoting effects of CA-MSCs by up-regulating IL-12 and IFN-γ expression [190]. What is more, the inhibition of activated signaling is beneficial for tumor immunotherapy. YAP signaling was activated during the CA-MSC phenotypic transition in patients with lymph node metastasis from GC. Verteporfin successfully inhibited YAP signaling and reprogrammed CA-MSCs in mice, disrupting tumor metastasis and invasion [191].
The reprogramming of diverse TASCs has matured from conceptual ideas to tangible strategies. These approaches seek to normalize the entire tumor ecosystem, offering a powerful means to overcome therapeutic resistance.

Targeting the tumor vasculature starves tumors
Pathological blood vessels, as key drivers of cancer progression, are primarily addressed through anti-angiogenesis and tumor vascular normalization strategies. Anti-angiogenic drugs restrain tumor angiogenesis by blocking critical signaling pathways. Targeted therapies against VEGF and PDGF are well established, with agents such as bevacizumab, ranibizumab, and olaratumab already on the market [192]. However, resistance to these AATs has emerged as a major clinical challenge, limiting their long-term efficacy. To address the challenge, targeting CAPs may yield beneficial outcomes. Paracrine signaling from CAPs supports tumor cell proliferation, survival, and migration. Studies in multicancer mouse models have discovered that CAPs stimulate tumor cell proliferation through paracrine secretion of high-mobility group protein 1 (HMGB1) [193]. The limited efficacy of imatinib and sunitinib in treating tumors may arise from this mechanism [194,195]. Therefore, targeting CAPs for anti-angiogenic activity requires further refinement to minimize adverse outcomes.
In response to the limitations of AAT, tumor vascular normalization has evolved into a complementary strategy. Instead of destroying vessels, this approach aims to improve vascular permeability, increase pericyte coverage, and promote vascular maturation [196]. Directly modifying TECs is a feasible strategy. In tumor-bearing mice, salvianic acid A promoted tight junctions between TECs, reversing aberrant structures and functions of tumor vessels through the pyruvate kinase muscle isozyme 2 (PKM2)/β-catenin/claudin-5 signaling axis [197]. Collectively, these multifaceted strategies targeting tumor vasculature are pivotal for overcoming resistance and improving cancer treatment outcomes.

TASC metabolic reprogramming halts tumor progression
Targeting metabolic pathways within TASCs holds profound therapeutic potential for improving tumor treatment and immune responses, but relevant studies remain preclinical.

Targeting CAF-associated metabolic pathways
Evidence from in vivo research has indicated that interrupting the normal metabolism of CAF affects tumor outcomes. In PDAC, owing to the reduction in lactate from glycolysis, nerve invasion is prevented [198]. Consistent with this, pharmacological inhibition of the Src SH3 domain of the facilitative glucose transporter (GLUT1) has been shown to suppress glycolysis and deactivate CAFs [199]. In addition to glucose metabolism, lipid biosynthesis in CAFs represents a key metabolic vulnerability. CAFs up-regulate ATP citrate lyase (ACLY) to produce fatty acids for energy. Therefore, the combination of ACLY inhibitor NDI-091143 and antitumor drugs (e.g., Adriamycin or paclitaxel) has been demonstrated to disrupt normal metabolism [200].

Targeting TEC-associated metabolic pathways
Inhibiting specific metabolic enzymes in TECs could enhance the efficacy of immunotherapy by improving tumor vascular function. For instance, osimertinib represses glyceraldehyde-3-phosphate dehydrogenase (GAPDH), thereby reducing lactate secretion from TECs and reversing the acidic TME [201]. Phosphoglycerate dehydrogenase (PHGDH) is crucial for nucleotide synthesis, serine metabolism, and glycolysis. Inhibitors such as WQ2201 normalize tumor vessels and enhance T cell infiltration, supporting chimeric antigen receptor T cell (CAR-T) therapies in glioblastoma [202].

Targeting CAA-related metabolic pathways
Distinct from traditional inhibition approaches, enhancing metabolism in CAAs seems to be effective. The previously mentioned engineered CAAs with high metabolism are robust evidence. Moreover, a fructose diet induced adipocytes to unleash leptin via a mechanistic target of rapamycin complex 1 (mTORC1)-dependent pathway. High levels of leptin boosted the antitumor immune response of CD8+ T cells [203].
Despite unclear clinical applications, these findings highlight the therapeutic potential of targeting TASC metabolic pathways, which reshape the TME and enhance immunotherapy. Bridging compelling preclinical insights into clinical translation represents a next step in cancer treatment.

Targeting ECM curbs spread and supports immunity
Restraining ECM deposition signals and fostering degradation are crucial strategies for ameliorating cancer immunotherapy. In the context of ECM deposition, LOX is a crucial factor that promotes matrix rigidity. Preclinical studies have shown that specific inhibitors targeting LOX play a role in overcoming immunotherapy resistance [204,205]. β-Aminopropionitrile (BAPN), the first LOX inhibitor, is widely used in animal experiments to prohibit collagen cross-linking. However, its high toxicity has precluded clinical application [206]. PXS-5505 is a potent pan-LOX inhibitor that perturbs cancer cell invasion and enhances the effect of chemotherapy. The therapeutic feasibility of PXS-5505 in combination with gemcitabine in pancreatic cancer has been approved [207]. However, there are currently no effective blockers on the market.
Degrading key ECM components represents a promising strategy to inhibit tumor progression and improve drug delivery. This can be achieved by leveraging bioactive enzymes, such as hyaluronidases, collagenases, and MMPs [208]. A notable advancement in this area involves engineering CAR-T cells with Notch receptors, enabling them to eradicate the ECM by inducing the secretion of heparinase and MMPs [209]. Dual strategies focusing on ECM deposition and ECM degradation hold profound promise for normalizing the tumor stroma and augmenting the response to immunotherapy.

Spatial heterogeneity and clinical relevance
Advances in spatial encoding have accelerated the exploration of the internal architecture of tumors, and the exploration of spatial heterogeneity has deepened the understanding of cellular ecological niches, which are instructive for diagnosis and treatment.

Tumor niches unlock patient prognostic stratification
Recent advances in single-cell resolution and spatial localization technologies have unveiled novel dimensions for prognostic evaluations within the TME. These approaches allow for sophisticated analyses, from the spatial distribution of individual cell subtypes to the architecture of the TME, providing critical insights for patient stratification and clinical decision-making. Here, we synthesize key findings across 3 main areas: the prognostic value of specific cell subtypes, the utility of spatial biomarkers, and the comprehensive landscape of the TME.
The presence and localization of specific stromal cell subpopulations enable the prediction of prognoses. In the context of CAFs, an analysis of 1,070 NSCLC patients has delineated distinct prognostic values for different CAF subpopulations based on their spatial localizations. Those at the tumor–stroma interface, consisting of iCAFs, interferon-responsive CAFs (ifnCAFs) (a subset of CAFs highly responsive to interferon signaling), and α-SMA+ CAFs, were associated with longer patient survival. In contrast, prometastatic subpopulations within the tumor core or deep stroma, comprising tumor-like CAFs (tCAFs) (a type of CAF expressing tumor stem cell-related markers), hypoxic tCAFs, and matrix CAFs, were linked to poor prognosis [210]. The balance of TEC subpopulations also influences clinical outcomes. Pan-cancer studies have identified 2 key subpopulations: pro-angiogenic E02-tip-CXCR4 and immune-favorable E06-vein-selectin E (SELE). An increased ratio of E02 to E06 was positively correlated with improved survival rates. Owing to its role in leukocyte adhesion, the E06 subpopulation also predicted immune checkpoint inhibitor (ICI) response [106]. More importantly, in clinical practice, Jiang et al. [211] established the TEC score based on differentiation genes, which has been validated as an independent poor prognostic factor for intrahepatic cholangiocarcinoma (ICC).
Beyond single cells, spatial structures within tumor tissues serve as powerful spatial biomarkers and facilitate patient stratification. TLSs have emerged as unique spatial biomarkers due to their prominent role in predicting antitumor immune responses. The densities, cellular compositions, and spatial distributions of TLSs correlate strongly with the sensitivity to PD-1/PD-L1 blockade [212,213]. Clinical evidence from a trial (NCT02534649) has confirmed that TLS-positive patients exhibited prolonged median overall survival (24.8 versus 13.3 months) and progression-free survival (6.1 versus 2.1 months) following anti-PD-1/L1 therapy. Moreover, the spatial location of TLSs is crucial, as it represents distinct functional states. In patients with ICC, the presence of intratumoral TLSs extends overall survival, whereas the presence of peritumoral TLSs shortens it. Immunoclassification based on TLS spatial distributions has outperformed the traditional tumor node metastasis classification (TNM) classification in predicting overall survival [214]. The CAF-Epi-Immune score in ESCC further exemplifies the concept of composite spatial scoring systems. This score is biologically grounded in the observation that spatial niches formed by CAFs and epithelial cells drive tumor invasion through ECM remodeling and immunosuppression [215,216].
Finally, a comprehensive assessment of the TME further optimized the hierarchical model. The stromal score and immune score, which reflect the infiltration of stromal and immune cells, respectively, are regarded as valuable tools for survival predictions. High stromal scores are associated with poor prognoses and mediated resistance to immunotherapy; immune scores show the opposite trend [217–219]. Additionally, the combination of these scores allows for more refined patient stratification [220]. Recent pan-cancer analyses have advanced this concept by identifying conserved TME subtypes. For instance, patients with immune-enriched, fibrotic types showed the best response to ICIs; those with immune-enriched, nonfibrotic types had the longest survival; and those with fibrotic and desert types exhibited ICI resistance and poor prognosis, respectively. Dynamic monitoring of these subtypes can inform treatment decisions and precision personalized medicine [221]. Another compelling example from cervical squamous cell carcinoma has illustrated how spatial atlases reveal therapeutic vulnerabilities. The MP6 subtype (epithelial–keratinocyte) drove immune suppression via CAF activation, whereas the MP7 subtype (epithelial–immune) promoted protective immune cell infiltration. Targeting a key driver (fatty acid-binding protein 5) in the MP6 subtype facilitated the transition between tumor states, highlighting a novel therapeutic avenue [222]. These dynamic models are progressively moving from descriptive snapshots toward clinical tools for precision medicine.
As spatial technologies become more accessible and standardized, their integration into clinical trials and routine diagnostics will be crucial for realizing personalized cancer therapy. Future efforts should focus on validating spatial biomarkers prospectively and developing targeted interventions. Ultimately, the gap between spatial biology and clinical practice should be bridged to improve patient outcomes.

Cellular interactions within the TIF reveal new therapeutic targets
Therapeutic strategies targeting FAP+ CAFs have attracted increasing attention due to their central role in shaping the TIF and promoting immunosuppression [100,223]. A range of modalities is currently under investigation, from immunotherapies to signaling regulation. Multiple FAP-directed strategies have been developed, ranging from biologics such as monoclonal antibodies and DNA vaccines to advanced cell therapies such as CAR-T cells. Typically, engineering CAR-T cells to secrete anti-FAP molecules has successfully eradicated the PDAC stroma in preclinical models [224]. Beyond direct targeting, modulating the signaling activity of FAP+ CAFs also shows promise. Blocking CXCL12 signaling from these cells has been demonstrated to synergize with anti-PD-L1 therapy and improve treatment outcomes [225]. In contrast, activating specific signaling combined with anti-FAP has potent antitumor efficacy. FAP-IL2v, comprising an interleukin-2 variant (IL-2v) fused to an antibody targeting FAP, triggers JAK-STAT signaling and activates CD8+ T cells [226]. The representative drug, simlukafusp slfa, is currently undergoing phase I/II clinical trials and is poised to offer a transformative strategy for combating advanced solid tumors [227–229]. To conclude, the spatial delineation of FAP+ CAF niches and their molecular interactions provides a roadmap for developing spatially targeted strategies.

Therapies targeting TASCs in combination with immunotherapies
Owing to their pivotal role in immune regulation, targeting TASCs provides a powerful adjunct to cancer immunotherapy. Here, we introduce several synergistic strategies that increase immunotherapeutic efficiency, including specific CAF targeting, AAT, and ECM remodeling. According to clinical trial datasets for several solid tumors, CAFs with leucine-rich repeat-containing 15 (LRRC15) signatures drive resistance to anti-PD-L1 therapy. In response, the antitumor activity of the specific antibody drug ABBV-085 has been evaluated in several clinical trials (e.g., NCT02565758 and NCT02565758) [230].
AAT reprograms the immunosuppressive TME to enhance the efficacy of ICIs. Therapy combining AAT with ICI has demonstrated new synergistic potential in clinical trials. The Impower150 study (i.e., NCT02366143) evaluated the clinical efficacy of the combination of atezolizumab (an anti-PD-L1 antibody) and bevacizumab (a VEGF inhibitor) in NSCLC. This combination therapy remarkably improved both progression-free survival and overall survival in patients. Based on these positive findings, the U.S. Food and Drug Administration (FDA) has approved this regimen as a first-line treatment option [231]. Another combination of carrilizumab (an anti-PD-L1 antibody) and apatinib (a VEGFR inhibitor) has exhibited therapeutic potential for several tumor types, covering NSCLC, HCC, and oral squamous cell carcinoma (OSCC), in preclinical models and clinical trials (e.g., NCT04379739, NCT04297202, and NCT04393506) [232–234].
Furthermore, accelerating ECM degradation disrupts the immune barrier in the TME. In murine models of BC, the nanomedicine dasatinib reduced ECM matrix secretion and improved matrix permeability, providing a novel method for PD-1/L1 combination therapy [235]. Overall, integrating TASC-targeting approaches into current immunotherapeutic regimens is poised to redefine treatment standards and improve survival outcomes for a broader population of cancer patients.

Conclusion and Perspectives
The intricate spatiotemporal heterogeneity of TASCs within the TME has been investigated in this review. We have provided insights into the evolution and functional characteristics of TASCs, with a focus on their spatial distribution and immunosuppression.

Key findings and challenges
TASCs are the main components of the TME and exhibit spatiotemporal heterogeneity. The dynamic evolution and spatial organization of TASCs are fundamental drivers of tumor initiation, progression, and metastasis. Through intricate intercellular crosstalk, TASCs construct diverse functional niches that profoundly shape the antitumor immune response. Consequently, dissecting the heterogeneity of TASCs has emerged as a pivotal breakthrough for advancing cancer immunotherapy.
Nevertheless, many outstanding challenges still exist. First, the lack of a unified classification system for TASC subpopulations poses great difficulties for research and clinical translation. Further clarification of the molecular markers, functions, and origins of TASCs across diverse tumor types and therapeutic scenarios is essential [17]. Employing the classic approach that combines marker genes with functional annotation may improve the accuracy and completeness of the annotations. Second, the scarcity of specific marker genes within stromal cells leads to ambiguous boundaries in delineating distinct cell populations. In particular, the gene expression profiles overlap in cell types, such as vCAFs and CAPs, myCAFs, and smooth muscle cells. This challenge not only complicates subpopulation clustering in research studies but also hampers the precision of targeted stromal cell therapies. In future targeted therapeutic studies, it is important to consider functional diversity. Finally, phenotypic shifts in TASCs may occur during treatment, contributing to resistance. Therefore, accelerating stromal cell heterogeneity research, especially in the translation of animal models to preclinical experiments, will be key. The resolution of the following outstanding questions will help decipher the heterogeneity of TASCs: How can the various classification criteria be harmonized? How can TASCs be precisely targeted in clinical applications without damaging stromal cells in normal tissues? How does the heterogeneity of TASCs change dynamically during tumor therapy?

Technological advances and future directions
With the continuous development of scRNA-seq and ST technologies, the spatial architecture within tumors has gradually been resolved more precisely. These advances have led to revolutionary breakthroughs in the precise stratification of tumor patients, tumor grading, and treatment predictions. For example, Sorger and colleagues [236] have provided insights into tumor progression mechanisms and potential treatment strategies by mapping large-scale spatial profiles of CRC. However, effectively translating spatial information into clinical applications still faces many challenges, such as integrating data from different platforms and eliminating data bias. To address these issues, the combination of machine learning and spatial genomics has shown potential. By integrating multiple data types, the Tumoroscope platform identified colocalization and exclusion phenomena within tumor tissues [237]. Meanwhile, single-cell 3D imaging technology (mLSR-3D) combined with spatial data analysis revealed the spatial phenotypes of tumors at the cellular level, promoting the construction of tumor maps and accurate diagnosis [238]. To realize this potential, the systematic acquisition of high-quality spatial data across a wide spectrum of cancers is essential. A substantial number of clinical trials remain imperative to amass adequate spatial datasets. Multicenter validation and machine learning-driven identification of robust spatial signatures collectively enhance the clinical reliability of spatial information.
In the future, deciphering the spatial atlas of the TME is a promising direction. Continued progress in spatial profiling will critically illuminate the temporal progression of tumors and the spatiotemporal heterogeneity of TASCs. These advancements will provide new opportunities for clinical treatment, expanding the repertoire of precision medicine and targeted therapy.