Spatial Fibroblast–Macrophage Niches in Gastric Cancer: Mechanisms of Immune Exclusion, Therapeutic Resistance, and Targeted Intervention

Abstract

Recent advances in immune checkpoint blockade, chemotherapy, anti-angiogenic therapy, and molecularly targeted therapies have expanded therapeutic options for gastric cancer. However, primary and acquired resistance remain major barriers to achieving durable clinical benefit. Tumor cell-intrinsic molecular features, including HER2 status, PD-L1 expression, MSI/MMR status, claudin 18 isoform 2 (CLDN18.2), and fibroblast growth factor receptor 2b (FGFR2b) alterations, do not fully account for the substantial heterogeneity in treatment response, highlighting the need to consider the spatial organization of the tumor microenvironment. Cancer-associated fibroblasts (CAFs) and tumor-associated macrophages (TAMs) are key components of this microenvironment. Evidence from single-cell sequencing, spatial transcriptomics, multiplex imaging, clinicopathological cohorts, and functional experiments indicates that CAFs and TAMs can form spatially organized niches through chemokine-mediated monocyte recruitment, macrophage-state reprogramming, extracellular matrix remodeling, complement signaling, cytokine networks, and immunometabolic regulation. Our synthesis identifies recurrent gastric cancer-associated niche archetypes, including CCL2⁺ fibroblast–STAT3-activated macrophage, GREM1⁺ fibroblast–SPP1⁺ macrophage, FAP⁺ fibroblast–SPP1⁺ macrophage, and THBS2⁺ matrix CAF–SPP1⁺ TAM niches. Among these, the THBS2⁺ matrix CAF–SPP1⁺ TAM niche in peritoneal metastasis has particularly strong spatial and functional support, including experimental evidence that CAF-derived C3 promotes C3AR1-dependent accumulation of SPP1⁺ TAMs and resistance to immune checkpoint blockade. Other proposed interactions are supported to varying degrees by spatial colocalization, ligand–receptor inference, clinicopathological association, or experimental perturbation, but their causal and clinical relevance remains incompletely established. Collectively, the available evidence suggests that selected CAF–TAM niches may promote CD8⁺ T-cell exclusion, vascular dysfunction, impaired therapeutic delivery, and persistence of treatment-resistant tumor cells. Actionable pathways include CCL2–CCR2, CSF1–CSF1R, IL-6–JAK–STAT3, CXCL12–CXCR4, SPP1–integrin, FAP, TGF-β, LOX, and C3–C3AR1. Future strategies should move beyond non-selective depletion of CAFs or macrophages and instead integrate molecular subtype, metastatic site, treatment context, and spatial biomarker patterns to identify and selectively disrupt dominant pathogenic niches.

Introduction

The therapeutic landscape of advanced gastric cancer has expanded substantially beyond conventional cytotoxic chemotherapy. Platinum–fluoropyrimidine doublets remain the backbone of first-line systemic treatment, while PD-1/PD-L1 blockade has improved outcomes in selected patients when combined with chemotherapy.^1–3 In parallel, biomarker-directed strategies have reshaped treatment algorithms, including HER2-targeted therapy, VEGFR2 blockade, CLDN18.2-directed zolbetuximab, and emerging FGFR2b-targeted approaches.^4–10 Nevertheless, the clinical benefit of these therapies is uneven across molecular contexts, anatomical patterns of spread, and metastatic niches. TCGA and Asian Cancer Research Group (ACRG) classifications have clarified the genomic diversity of gastric cancer, but they do not fully explain why tumors sharing nominally targetable alterations may show primary resistance, mixed response, or early relapse.^11,12 Increasing evidence suggests that treatment failure is not determined by tumor-cell genotype alone. Stromal barriers, immune exclusion, aberrant wound-repair programs, and compensatory microenvironmental signals can blunt drug penetration, limit cytotoxic T-cell access, and provide survival cues even when a recognizable therapeutic target is present.¹³

Historically, cancer-associated fibroblasts (CAFs) and tumor-associated macrophages (TAMs) have often been discussed as separate protumor components of the gastric cancer microenvironment. This separation is useful analytically, but biologically incomplete. Single-cell studies have shown that gastric cancer CAFs are heterogeneous, with inflammatory, extracellular matrix-remodeling, and pro-invasive programs that differ across patients and histological contexts.^14,15 TAMs are similarly plastic rather than a single M2-like population. More importantly, CAFs and macrophages frequently operate as a coupled stromal–myeloid unit. CAF-derived chemokines and cytokines can recruit monocytes and promote their differentiation toward immunosuppressive or tissue-repair-like macrophage states, whereas macrophages can reinforce fibroblast activation, extracellular matrix deposition, angiogenesis, and epithelial repair programs. In this sense, the CAF–TAM axis is not merely a coexistence of two unfavorable cell populations, but a self-amplifying circuit that can stabilize a fibrotic, poorly inflamed, and therapy-resistant tumor ecosystem.^16–18

The emergence of single-cell RNA sequencing has provided a high-resolution catalogue of these cellular states, but cell-state annotation alone cannot define how resistance is organized in tissue. Spatial transcriptomics, digital spatial profiling, and multiplex tissue imaging have added a second layer of information: which CAF subsets are physically adjacent to which macrophage subsets, whether these neighborhoods are located in the tumor core, invasive margin, stromal septa, or peritoneal metastatic implants, and whether they coincide with CD8⁺ T-cell exclusion or poor response to immune checkpoint blockade. Recent gastric cancer studies have begun to identify such spatially organized niches. For example, CCL2⁺ fibroblast regions have been linked to STAT3-activated macrophages and an immunosuppressive gastric cancer microenvironment, while peritoneal metastasis studies have described SPP1⁺ TAMs and matrix-producing CAFs as a dominant stromal–myeloid niche associated with immune escape and limited immunotherapy efficacy.^16,17 These observations suggest that CAF–macrophage crosstalk is best understood not only as a ligand–receptor network, but also as a spatially constrained ecological structure. This spatial information extends beyond that provided by conventional molecular biomarkers. Bulk measurements of CAF- or TAM-associated genes can indicate the overall abundance of stromal and myeloid programs but cannot determine whether the corresponding cell populations are physically adjacent, separated into different tissue compartments, or positioned at interfaces that restrict immune-cell access to tumor nests. Spatial analysis can identify the cellular source and anatomical localization of these signals, distinguish tumor-core, invasive-margin, stromal, and metastatic-site-specific niches, and determine whether CAF–TAM neighborhoods coincide with CD8⁺ T-cell exclusion or treatment-resistant tumor regions.^16–18 Thus, tumors with similar overall expression of CAF and macrophage markers may differ substantially in their tissue architecture and therapeutic behavior. Spatial organization should therefore be considered a complementary biomarker dimension that captures cellular proximity and tissue topology. However, spatial colocalization alone does not establish a causal interaction and requires validation through functional perturbation.

In this review, we focus on gastric cancer-specific evidence for CAF–macrophage niches and evaluate how strongly different lines of evidence support their biological and clinical relevance. Rather than treating CAFs and TAMs as broad markers of a “cold” tumor microenvironment, we discuss how defined CAF–macrophage neighborhoods contribute to therapeutic resistance, how they may serve as spatial biomarkers of immune exclusion and metastatic behavior, and how they could be exploited as therapeutic targets. Particular attention is given to the value of this axis as an entry point for rational combination strategies, including chemotherapy, immune checkpoint blockade, anti-angiogenic therapy, matrix modulation, and emerging approaches targeting chemokine, cytokine, and macrophage-reprogramming pathways. By framing CAF–macrophage interactions as spatially organized and therapeutically actionable niches, this review aims to connect descriptive microenvironmental atlases with clinically testable hypotheses in gastric cancer. Figure 1 summarizes the proposed model of CAF–TAM spatial niche formation, illustrating how fibroblast activation, monocyte recruitment, macrophage reprogramming, and bidirectional stromal–myeloid feedback collectively shape immune exclusion and therapeutic resistance in gastric cancer.

Figure 1.

Proposed formation and functional evolution of fibroblast–macrophage spatial niches in gastric cancer. This schematic illustrates a proposed sequence linking CAF activation, monocyte recruitment, macrophage reprogramming, reciprocal CAF–TAM feedback, spatial niche formation, and downstream functional consequences. In the CAF activation stage, quiescent fibroblasts may acquire activated CAF states under the influence of tumor-derived and microenvironmental signals, including TGF-β, IL-6, and matrix-remodeling programs. Activated CAFs may promote the recruitment of circulating monocytes through chemokine-mediated signals, particularly CCL2 acting through CCR2 on monocytes, while CSF1-related signaling contributes primarily to macrophage survival, differentiation, and maintenance. Recruited monocytes may subsequently acquire context-dependent TAM states under the influence of IL-6–STAT3 signaling, hypoxia, and other local stromal cues, including the acquisition of SPP1-associated macrophage programs. CAFs and reprogrammed TAMs may further reinforce each other through reciprocal signaling involving extracellular matrix deposition, collagen remodeling, cytokines, growth factors, and tissue-repair-like programs. These interactions may generate spatially organized stromal–myeloid niches characterized by dense fibrotic stroma, CAF–TAM clustering, compressed vessels, and barrier-like tissue architecture. Such niches may restrict CD8⁺ T-cell entry into tumor nests, reduce drug penetration, support the persistence of residual tumor cells, and contribute to therapeutic resistance. Solid single-headed arrows indicate the proposed direction of cellular transitions, recruitment, signaling, niche formation, or downstream consequences. The paired curved arrows between activated CAFs and reprogrammed TAMs denote reciprocal CAF–TAM feedback. Branched arrows extending from the spatial niche indicate the two proposed downstream consequences: immune exclusion and therapeutic resistance. Crossed-circle symbols indicate restricted CD8⁺ T-cell entry into tumor nests. Black dotted connector lines in the CAF activation panel link the illustrated gastric lesion to the magnified tumor-tissue context and do not represent signaling interactions.

Literature Search and Evidence Assessment

This review was designed as a focused narrative synthesis rather than a systematic review or meta-analysis. PubMed, Web of Science, and Scopus were searched up to July 1, 2026, which was the final search date. Search terms included combinations of “gastric cancer”, “gastric adenocarcinoma”, “cancer-associated fibroblast”, “tumor-associated macrophage”, “fibroblast–macrophage crosstalk”, “spatial niche”, “spatial transcriptomics”, “single-cell sequencing”, “multiplex imaging”, “immune exclusion”, “treatment resistance”, “targeted therapy”, “immune checkpoint blockade”, and relevant ligand–receptor or signaling-pathway terms. The reference lists of key clinical studies, single-cell atlases, spatial profiling studies, mechanistic investigations, and relevant reviews were also examined to identify additional eligible publications.

Studies were included if they provided evidence relevant to at least one of the following areas: CAF or TAM heterogeneity in gastric cancer; spatial organization or functional communication between fibroblasts and macrophages; CAF- or TAM-mediated immune exclusion, metastasis, or therapeutic resistance; clinicopathological associations involving stromal–myeloid features; or experimental evaluation of potentially targetable CAF–TAM pathways. Priority was given to studies using human gastric cancer tissues, clinically annotated cohorts, spatially resolved methods, single-cell analyses, patient-derived models, or functional perturbation experiments. Evidence from other cancer types was included selectively only when it addressed a biologically relevant mechanism that had not been sufficiently investigated in gastric cancer. General reviews and methodological articles were used primarily to provide conceptual or technical context.

Studies were excluded if they lacked relevance to CAFs, TAMs, or their associated microenvironmental mechanisms; addressed an unrelated disease context without a clearly transferable mechanism; provided insufficient methodological or outcome information for interpretation; or duplicated data reported more comprehensively elsewhere. Titles and abstracts were initially assessed for relevance, after which potentially eligible full texts were evaluated according to their disease context, experimental design, spatial resolution, mechanistic depth, and clinical relevance. Because this was a narrative review, study selection was based on relevance and evidentiary contribution rather than on a formal PRISMA-based screening process. The final synthesis prioritized direct gastric cancer evidence and used evidence from other malignancies only to define mechanistic plausibility or unresolved research directions.

Throughout the review, evidence was interpreted according to its proximity to human gastric cancer. Findings derived from gastric cancer patient tissues, clinical cohorts, gastric cancer models, or gastric cancer-specific spatial datasets are identified as gastric cancer-specific evidence. Findings derived exclusively from other malignancies are explicitly introduced as non-gastric cancer evidence and are treated as mechanistic extrapolation rather than direct proof in gastric cancer. This distinction was applied particularly to proposed mechanisms of immune exclusion, matrix remodeling, vascular dysfunction, and treatment resistance. This tiered approach follows general principles for transparent narrative evidence synthesis while avoiding overstatement of mechanisms supported mainly by non-gastric cancer data.^19,20

Defining Fibroblast–Macrophage Spatial Niches in Gastric Cancer

Fibroblast States Relevant to Macrophage Regulation

CAF heterogeneity in gastric cancer should not be reduced to a single “activated fibroblast” label. For the purpose of understanding macrophage regulation and therapeutic resistance, the most relevant fibroblast states are those that either recruit monocytes, impose macrophage polarization, remodel extracellular matrix, or mark immune-excluded stromal territories. These states partly overlap, and their markers are not mutually exclusive; nevertheless, several recurrent CAF programs have emerged from gastric cancer single-cell, spatial, and functional studies.

CCL2⁺ inflammatory fibroblasts represent one of the clearest gastric cancer-specific examples of fibroblast-driven macrophage regulation. These cells are characterized by inflammatory chemokine expression, particularly CCL2, and are enriched in stromal regions where myeloid JAK–STAT3 activity is prominent. Spatial transcriptomic analysis has shown that CCL2⁺ fibroblast-rich regions are closely associated with STAT3-activated macrophages, and tumors with concomitantly high CCL2⁺ fibroblast and STAT3-activated macrophage signatures show worse clinical outcomes in public gastric cancer cohorts.¹⁶ Mechanistically, this niche is plausible because CCL2–CCR2 signaling can recruit circulating monocytes, while downstream IL-6/JAK/STAT3-type programs support immunosuppressive macrophage differentiation. The therapeutic implication is therefore not simply “depleting fibroblasts”, but disrupting fibroblast-to-myeloid inflammatory instruction through the CCL2–CCR2 or JAK–STAT3 axis. However, clinical translation in gastric cancer remains exploratory, and patient selection would likely require spatial or transcriptomic confirmation of this coupled fibroblast–macrophage state.

GREM1⁺ fibroblasts appear to define a more matrix-associated and developmentally patterned stromal niche. GREM1 is a secreted BMP antagonist, and GREM1⁺ fibroblasts in gastric cancer have been linked to extracellular matrix organization, TGF-β-related programs, diffuse-type enrichment, and poor prognosis when co-enriched with SPP1⁺ macrophages. Spatial transcriptomics and multiplex immunohistochemistry further suggest that GREM1⁺ fibroblasts and SPP1⁺ macrophages are not merely correlated at the bulk level but can occupy neighboring tissue sites, with prominent SPP1–integrin and collagen-related communication.¹⁸ This makes the GREM1⁺ fibroblast–SPP1⁺ macrophage niche particularly relevant to fibrotic tumor architecture, peritoneal dissemination, and immune exclusion. Directly targeting GREM1 remains less mature than targeting chemokine or cytokine pathways, but the niche may be therapeutically approachable through broader interruption of TGF-β, SPP1–integrin, or matrix-remodeling pathways.

FAP⁺ CAFs are among the most clinically visible fibroblast states because FAP is a cell-surface protease and has become accessible to imaging and therapeutic targeting strategies. In gastric cancer, FAP⁺ fibroblasts have been associated with tumor progression, drug resistance, impaired antitumor T-cell activity, and resistance to immune checkpoint blockade in experimental models.²¹ More recent clinicopathological evidence also links FAP positivity with poor recurrence-free survival and higher infiltration of FOXP3⁺ regulatory T cells and CD163⁺ M2-like macrophages in resected gastric cancer specimens.²² Thus, FAP is best interpreted as both a stromal activation marker and a practical entry point for targeting immune-tolerant fibroblast-rich regions. Nevertheless, FAP is not a macrophage-polarizing ligand by itself; its importance lies in identifying a CAF compartment capable of producing immunoregulatory cytokines, chemokines, matrix cues, and exclusionary stromal architecture.

THBS2⁺ matrix CAFs reflect a matricellular fibroblast program rather than a purely inflammatory one. THBS2, together with other extracellular matrix-associated molecules such as COL1A1, COL1A2, COL3A1, POSTN, SPARC, and INHBA, marks fibroblast states involved in collagen organization, cell–matrix adhesion, and stromal remodeling. Gastric cancer studies have associated THBS2 expression with advanced stage, poor prognosis, extracellular matrix-receptor interaction, focal adhesion, and immune infiltration, including macrophage-related signals.^23,24 In single-cell gastric cancer data, extracellular matrix CAFs can interact with immune cells and are correlated with M2-like macrophage enrichment, particularly through matrix-associated mediators such as POSTN.¹⁵ THBS2 itself is not yet a validated therapeutic target in gastric cancer, but THBS2⁺ matrix CAFs are useful as a marker of stromal stiffness, immune exclusion, and possible resistance to chemotherapy or immunotherapy. Their therapeutic relevance may therefore lie in selecting patients for matrix-modulating, anti-fibrotic, or stroma-normalizing combinations rather than in direct THBS2 blockade.

LOX⁺ matrix-remodeling fibroblasts provide a more mechanistic bridge between extracellular matrix remodeling and macrophage polarization. LOX participates in collagen crosslinking and stromal stiffening, and a gastric cancer single-cell study identified LOX⁺ fibroblasts as enriched in tumor tissues, spatially close to CD206⁺ M2-like macrophages, and associated with poor prognosis. Ligand–receptor analysis suggested that LOX⁺ fibroblasts may stimulate monocyte-to-M2 macrophage differentiation through IL6–IL6R signaling, particularly within hypoxic stromal regions.²⁵ This state is therefore relevant not only because it remodels matrix, but because it connects hypoxia, collagen remodeling, macrophage polarization, and checkpoint-rich immune suppression. LOX inhibition or collagen-remodeling strategies may have theoretical value, but in gastric cancer these approaches still require careful validation, especially because matrix disruption can have context-dependent effects on invasion and drug delivery.

Finally, IL-6-high and CPT1C-related CAFs represent an immunometabolic fibroblast state with direct functional evidence for macrophage reprogramming. Gastric CAF-derived IL-6 has been shown to promote epithelial–mesenchymal transition, metastasis, and chemotherapy resistance through JAK/STAT signaling.^26,27 Separately, IL-6 can induce STAT3-dependent M2 macrophage differentiation in gastric cancer, linking cytokine abundance to a macrophage phenotype associated with tumor progression.²⁸ More recently, CPT1C⁺ CAFs were identified as a gastric cancer CAF subset enriched in extracellular matrix and immunosuppressive features; these cells promoted M2-like macrophage polarization in vitro, and CPT1C knockdown reduced IL-6 secretion and attenuated macrophage reprogramming. High CPT1C⁺ CAF infiltration was also associated with poor immunotherapy response in a clinical cohort.²⁹ Compared with THBS2⁺ matrix CAFs, CPT1C⁺/IL-6-high CAFs are more directly actionable because the IL-6–JAK–STAT3 axis has existing pharmacological inhibitors. The main unresolved issue is not whether the pathway is biologically plausible, but how to identify the subgroup of gastric cancer patients in whom IL-6-producing CAFs are the dominant upstream driver of macrophage-mediated immune suppression.

Collectively, these CAF states suggest that macrophage regulation in gastric cancer is organized through at least three partially overlapping stromal programs: inflammatory recruitment, matrix remodeling, and cytokine-metabolic instruction. CCL2⁺ and IL-6-high/CPT1C⁺ CAFs primarily explain monocyte recruitment and macrophage polarization; GREM1⁺, THBS2⁺, and LOX⁺ CAFs emphasize matrix remodeling and spatial immune exclusion; FAP⁺ CAFs provide a clinically tractable marker of activated, immunosuppressive stroma. The key issue for the field is to move from marker-based CAF annotation to spatially resolved functional niches that can predict which stromal–myeloid circuit is actually sustaining resistance in an individual tumor.

Macrophage States Beyond the M1/M2 Model

The M1/M2 framework remains useful as a simplified language for describing inflammatory and tissue-repair-like macrophage tendencies, but it cannot adequately represent macrophage biology in gastric cancer tissues. Macrophages in the gastric cancer microenvironment occupy continuous and context-dependent states shaped by tumor cells, fibroblasts, hypoxia, metabolic stress, cytokines, and spatial constraints. Current single-cell and spatial studies show that gastric cancer macrophages include SPP1⁺, C1QC⁺, FCN1⁺, interferon-responsive, lipid-associated, and STAT3-activated states rather than two fixed M1 and M2 populations.^14,30–32 Therefore, M1/M2 terminology should be used only as a descriptive approximation, not as a substitute for tissue-defined macrophage subsets or spatially organized stromal–myeloid programs.

SPP1⁺ macrophages are among the most consistently recognized protumor macrophage states in gastric cancer. They are typically characterized by high expression of SPP1 and are often accompanied by tissue-repair, matrix-interacting, angiogenic, and immunosuppressive features.^31–33 Multi-transcriptomic analysis has shown that SPP1⁺ macrophages are enriched in gastric cancer tissues and can engage malignant epithelial cells through SPP1–CD44 signaling, supporting intratumoral crosstalk and tumor-promoting behavior.³³ In metastatic gastric cancer, SPP1⁺ TAMs have also been linked to exhausted CD8⁺ T cells and an immunosuppressive microenvironment.³⁴ More recently, analyses integrating single-cell transcriptomics, spatial data, and immunofluorescence suggested that SPP1⁺/C1QC⁺ macrophages can participate in a stromal immunosuppressive barrier associated with gastric cancer immunotherapy resistance.³⁵ These findings make SPP1⁺ macrophages particularly relevant to CAF-rich niches, because SPP1 can interact with CD44 and integrins, connect macrophages to matrix-rich regions, and contribute to T-cell exclusion in fibrotic tumor territories.

STAT3-activated macrophages represent a cytokine-instructed signaling state rather than a conventional lineage-defined subset. Spatial dissection of gastric cancer tissues identified an immunosuppressive crosstalk between CCL2⁺ fibroblasts and STAT3-activated macrophages, and patients with high levels of both CCL2⁺ fibroblast and STAT3-activated macrophage signatures showed worse clinical outcomes in TCGA and ACRG cohorts.¹⁶ This provides direct gastric cancer evidence that fibroblasts can organize macrophage function in situ. Mechanistically, the finding is consistent with the broader role of CCL2–CCR2 in monocyte recruitment and IL-6/JAK/STAT3 signaling in macrophage polarization; in gastric cancer, IL-6 has been shown to induce STAT3-dependent M2-like macrophage differentiation and correlate with tumor progression.²⁸ Thus, STAT3-activated macrophages should not be treated simply as “M2 macrophages”; they mark a spatially coupled and potentially targetable cytokine-response state within fibroblast-rich regions.

C1QC⁺ macrophages are context-dependent and should not be automatically classified as either antitumor or protumor. They are generally marked by complement-related genes, including C1QA, C1QB, and C1QC, and often show phagocytosis- or tissue-resident-associated features in single-cell datasets.^31,32 In some gastric cancer contexts, this population may preserve antigen-handling or homeostatic functions; in others, it appears as part of an immune-suppressive barrier. For example, a recent immunotherapy-resistance analysis suggested that C1QC⁺ macrophages, together with SPP1⁺ macrophages and exhausted CD8⁺ T cells, contribute to stromal immune suppression through cell–cell communication pathways involving MIF, LGALS9, CXCL16, and related signals.³⁵ Therefore, the biological interpretation of C1QC⁺ macrophages depends heavily on their spatial neighbors, ligand–receptor partners, and coexistence with exhausted T cells or fibroblast-rich matrix.

TREM2⁺ lipid-associated macrophages add an immunometabolic layer to gastric cancer macrophage heterogeneity. The concept of TREM2-dependent lipid-associated macrophages was first defined in metabolic tissues, where these cells showed conserved programs related to lipid handling, phagocytosis, and tissue stress adaptation.³⁶ In gastric cancer, recent evidence indicates that TREM2⁺ macrophages are enriched in tumor tissues and may cooperate with fibroblasts through the CXCL12–CXCR4 axis to promote tumor growth and immunosuppression.³⁷ Another gastric cancer study reported that TREM2 in TAMs promoted tumor progression and immune evasion by stabilizing STAT1, enhancing CCL8 and PD-L1 expression, and impairing CD8⁺ T-cell function.³⁸ These data suggest that TREM2⁺ macrophages may be relevant to immune checkpoint resistance, although additional clinical validation is still needed to determine whether TREM2 is best used as a therapeutic target, a myeloid biomarker, or both.

CXCL9/CXCL10⁺ inflammatory macrophages represent a more immune-active macrophage state. These cells are usually characterized by interferon-inducible chemokines such as CXCL9 and CXCL10 and may contribute to T-cell recruitment in inflamed tumor regions.^14,32 Compared with SPP1⁺ or STAT3-activated macrophages, CXCL9/CXCL10⁺ macrophages are more likely to reflect an IFN-responsive, T-cell-attracting microenvironment. In gastric cancer peritoneal metastases, CXCL9⁺ macrophages were identified as a distinct macrophage state, whereas THBS2⁺ matrix CAFs were shown to recruit tissue-resident macrophages and promote their transition toward SPP1⁺ TAMs through the C3–C3AR1 axis.¹⁷ This contrast is important: CXCL9/CXCL10⁺ macrophages may mark an inflamed niche, while SPP1⁺ macrophages more often mark a fibrotic, macrophage-dominated, immune-excluding niche. However, inflammatory chemokine expression alone does not guarantee effective antitumor immunity, because these macrophages may still coexist with exhausted T cells or checkpoint-rich inflammation.

Taken together, gastric cancer macrophage biology is better understood as a spectrum of spatially organized states than as a binary M1/M2 system. SPP1⁺ macrophages emphasize matrix interaction, tumor promotion, and T-cell exhaustion; STAT3-activated macrophages mark fibroblast-instructed cytokine signaling; C1QC⁺ macrophages reflect complement-associated and context-dependent phagocytic programs; TREM2⁺ macrophages connect lipid metabolism with immune evasion; and CXCL9/CXCL10⁺ macrophages indicate an IFN-responsive inflammatory state. The key question is therefore not whether TAMs are simply “M1” or “M2”, but which macrophage state is spatially coupled to which CAF state, and whether that coupled niche excludes T cells, maintains fibrosis, supports metastasis, or predicts treatment resistance.

Functional Interpretation of CAF and TAM Markers

CAF and TAM markers should be interpreted as indicators of context-dependent cellular states rather than as lineage-exclusive labels or direct evidence of biological function.^30,39 Among CAF-associated markers, CCL2 indicates a chemokine-producing fibroblast program with the potential to recruit CCR2⁺ monocytes.^16,40 GREM1 identifies a matrix-associated fibroblast state and functions as an antagonist of bone morphogenetic protein signaling.^18,41 FAP is a cell-surface serine protease associated with activated fibroblasts and tissue-remodeling stroma, but its expression alone does not establish an immunosuppressive function.^21,39 THBS2 marks a matricellular fibroblast program associated with extracellular matrix organization and cell–matrix interactions.^17,24 LOX is a collagen-crosslinking enzyme, and LOX⁺ fibroblasts define a matrix-remodeling fibroblast state associated with M2-like macrophage enrichment in gastric cancer.²⁵ CPT1C identifies a CAF state that can promote IL-6-dependent macrophage polarization and immunosuppression.²⁹ These markers therefore capture different aspects of fibroblast biology, including inflammatory recruitment, matrix remodeling, cytokine production, and potential therapeutic accessibility.

TAM-associated markers likewise represent overlapping functional states rather than mutually exclusive macrophage lineages.³⁰ SPP1 encodes osteopontin and commonly identifies macrophage states associated with matrix interaction, tumor-promoting activity, and immunosuppressive communication.^17,18,33,34 C1QC expression is associated with a complement-related and phagocytic macrophage program whose functional consequences depend on the surrounding cellular and spatial context.^31,32,35 TREM2 is associated with lipid sensing, phagocytosis, and adaptation to lipid-rich or stressed tissue environments; although this program was initially characterized outside cancer, TREM2⁺ macrophage states have subsequently been identified in gastric cancer.^36–38 Activated STAT3 reflects a cytokine-response state rather than a macrophage lineage and may indicate local IL-6/JAK/STAT3 signaling.^16,28 CXCL9 and CXCL10 indicate an interferon-responsive chemokine program that may be compatible with T-cell recruitment, although their expression alone does not demonstrate effective antitumor immunity.^14,32 Conventional markers such as CD68, CD163, and CD206 can assist with macrophage identification or phenotypic description but cannot independently establish an immunosuppressive macrophage function.³⁰

Accordingly, the designation of a CAF or TAM state in this review is based, whenever possible, on a combination of marker expression, transcriptional program, anatomical location, neighboring cell populations, inferred communication pathways, and functional evidence. Marker positivity alone is not considered sufficient to demonstrate a functional CAF–TAM niche.^30,39

Evidence Levels for Defining a Niche

Because the term “spatial niche” is increasingly used to describe heterogeneous forms of cell–cell association, we applied a three-level framework to distinguish associative or spatially unresolved evidence, candidate spatial organization, and functionally validated spatial interactions. This framework is an interpretive approach developed for the present review rather than a formally validated evidence-grading instrument. It evaluates the strength of evidence supporting a spatially organized CAF–TAM interaction rather than the overall methodological quality of an individual study. Each interaction was assigned to the highest level for which all required criteria were fulfilled. Evidence from separate studies was not combined to imply a higher level unless those studies examined biologically comparable CAF and TAM states connected by the same or a closely related signaling circuit.

Level I: associative or spatially unresolved interaction. This level includes correlated CAF and TAM signatures identified by bulk transcriptomics, computational deconvolution, or dissociation-based single-cell sequencing. It also includes experimentally supported fibroblast–macrophage mechanisms for which the spatial organization of the two cell populations has not been demonstrated in intact gastric cancer tissue. Level I evidence can establish co-enrichment, clinical association, or mechanistic plausibility, but it cannot establish a spatial niche. Representative examples include LOX⁺ fibroblast–M2-like macrophage communication inferred through IL-6–IL-6R signaling,²⁵ CPT1C⁺ CAF-mediated IL-6-dependent macrophage polarization,²⁹ FAP⁺ fibroblast and SPP1⁺ macrophage co-enrichment in mismatch repair-proficient gastric cancer,⁴² CXCL12–CXCR4-related communication between fibroblasts and TREM2⁺ macrophages,³⁷ and NAMPT–NNMT-dependent metabolic communication between macrophages and fibroblasts.⁴³ These findings support biologically relevant stromal–myeloid interactions, but their organization as reproducible spatial CAF–TAM niches in human gastric cancer tissue remains insufficiently resolved.

Level II: candidate spatial niche. This level requires tissue-architecture-preserving evidence demonstrating that defined CAF and TAM states occupy the same or adjacent tissue regions. Such evidence may be generated by spatial transcriptomics, multiplex immunofluorescence, multiplex immunohistochemistry, digital spatial profiling, or quantitative imaging. Predicted ligand–receptor communication or pathway consistency may strengthen the biological interpretation, but spatial proximity and computational inference alone do not establish functional dependence. Current gastric cancer examples include the CCL2⁺ fibroblast–STAT3-activated macrophage interaction identified through spatial dissection of gastric cancer tissues¹⁶ and the GREM1⁺ fibroblast–SPP1⁺ macrophage interaction supported by spatial transcriptomics and multiplex immunohistochemistry.¹⁸ These interactions meet the criteria for candidate spatial niches, but direct perturbation studies are still required to determine whether disrupting the corresponding CAF–TAM interaction alters macrophage state, immune exclusion, tumor progression, or treatment response.

Level III: functionally validated and treatment-relevant spatial niche. This level requires evidence of spatial organization in human gastric cancer tissue together with mechanistic perturbation demonstrating that interruption of the proposed CAF–TAM pathway alters macrophage recruitment or state, antitumor immunity, tumor behavior, or treatment response. Association with clinical outcome or treatment response further strengthens, but is not by itself sufficient for, Level III classification. The THBS2⁺ matrix CAF–C3/C3AR1–SPP1⁺ TAM circuit in gastric cancer peritoneal metastasis currently provides the clearest Level III example. In this setting, single-cell sequencing, spatial analysis, and multiplex imaging demonstrated the organization of THBS2⁺ matrix CAFs and SPP1⁺ TAMs, while mechanistic experiments showed that CAF-derived C3 promoted C3AR1-dependent macrophage accumulation and transition toward the SPP1⁺ TAM state. Experimental disruption of the C3–C3AR1 pathway reduced this stromal–myeloid program and improved the response to immune checkpoint blockade in preclinical models.¹⁷

This framework does not imply that Level I evidence is unreliable or that a Level III interaction has achieved clinical validation. Rather, it distinguishes hypothesis-generating or spatially unresolved interactions from candidate spatial niches and experimentally supported spatial dependencies. Level I findings support further spatial investigation, Level II findings support spatial biomarker development and mechanistic testing, and Level III findings provide a rationale for translational evaluation. However, even Level III classification does not establish therapeutic efficacy in prospective clinical trials. Figure 2 summarizes representative gastric cancer fibroblast–macrophage niche archetypes according to their current evidence levels, ranging from mechanistically supported interactions to functionally validated treatment-related niches. Representative CAF–TAM interactions, their current evidence levels, proposed mechanisms, candidate biomarkers, potential therapeutic targets, and major evidentiary limitations are summarized in Table 1.

Figure 2.

Gastric cancer-relevant fibroblast–macrophage niche archetypes and their evidence levels. Representative cancer-associated fibroblast (CAF)–tumor-associated macrophage (TAM) interactions are organized according to the strength of evidence supporting their spatial and functional organization in gastric cancer. This three-level framework reflects evidentiary maturity rather than the overall quality of an individual study or the clinical validation of a biomarker or therapeutic target. Level I represents mechanistically supported but spatially unresolved interactions. These modules are supported by co-enrichment, single-cell or ligand–receptor inference, pathway consistency, or functional evidence in gastric cancer, but definitive CAF–TAM spatial organization has not been demonstrated in human gastric cancer tissues. Representative examples include LOX-associated fibroblast–M2-like macrophage interactions, the CPT1C⁺ CAF–IL-6–STAT3 macrophage axis, TGF-β-, CSF1-, or extracellular-vesicle-mediated crosstalk, and CXCL12–CXCR4/TREM2-related interactions. Level II represents candidate spatial niches supported by tissue-preserving evidence of cellular proximity or regional co-organization in gastric cancer, but without sufficient perturbation evidence to establish causal interdependence between the corresponding CAF and TAM states. Representative examples include the CCL2⁺ fibroblast–STAT3-activated macrophage niche, the GREM1⁺ fibroblast–SPP1⁺ macrophage niche, and the FAP⁺ fibroblast–SPP1⁺ macrophage module. The strength and type of spatial evidence may differ among these candidate niches. Level III represents a spatially organized and functionally evaluated niche for which pathway perturbation alters a relevant macrophage, immune, tumor, or treatment-response phenotype. The THBS2⁺ matrix CAF–C3/C3AR1–SPP1⁺ TAM circuit in gastric cancer peritoneal metastasis currently provides the clearest Level III example. Experimental complement-pathway blockade may enhance the response to programmed cell death protein 1/programmed death-ligand 1 (PD-1/PD-L1) blockade; however, this therapeutic effect has not been prospectively validated in patients. Solid single-headed arrows indicate the direction of the interactions depicted in the corresponding studies or mechanistic model. In the Level III panel, CAF-derived C3 signaling through macrophage C3AR1 has direct mechanistic support, whereas SPP1-related feedback toward CAF activation represents a proposed reinforcing component and should not be interpreted as having an equivalent level of causal validation. Double-headed arrows in the Level II panels denote candidate reciprocal crosstalk or spatial association and do not establish causality. Dashed connectors between fibroblasts and macrophages in the Level I panels indicate mechanistically supported but spatially unresolved interactions. Large arrows extending from the evidence-level labels indicate classification within the framework rather than biological signaling, whereas dashed borders delineate individual modules or evidence categories and have no independent signaling meaning. Evidence derived exclusively from other malignancies may provide mechanistic context but was not used alone to classify an interaction as a gastric cancer-specific niche.

Table 1.

Representative CAF–TAM Interactions in Gastric Cancer and Their Current Evidence Levels

CAF–TAM Interaction	Evidence Level	Proposed Mechanism	Candidate Biomarkers	Potential Therapeutic Targets	Major Limitations
LOX⁺ fibroblast–M2-like macrophage interaction25	Level I	Hypoxia-associated LOX⁺ fibroblasts may promote monocyte-to-M2-like macrophage differentiation through IL-6–IL-6R signaling and contribute to matrix remodeling.	LOX, IL-6, IL-6R, CD68, CD163, CD206	LOX inhibition; IL-6/IL-6R or JAK/STAT3 blockade	Evidence is based mainly on single-cell analysis, correlation, pseudotime analysis, and ligand–receptor inference; direct spatial and perturbation evidence in human gastric cancer tissue is lacking.
CPT1C⁺ CAF–IL-6–M2-like macrophage interaction29	Level I	CPT1C⁺ CAFs secrete IL-6 and promote an M2-like macrophage phenotype associated with immunosuppression and poor immunotherapy response.	CPT1C, IL-6, CD163, CD206, ARG1	CPT1C modulation; IL-6/IL-6R or JAK/STAT3 blockade	Functional evidence is predominantly in vitro, and the clinical association was derived from a limited cohort; spatial organization and prospective predictive value remain unvalidated.
FAP⁺ fibroblast–SPP1⁺ macrophage module42	Level I	Co-enrichment of activated fibroblasts and SPP1⁺ macrophages may support stromal activation, tumor progression, and an immune-suppressive microenvironment in pMMR gastric cancer.	FAP, SPP1, CD68, CD163, CD8	FAP-directed approaches; SPP1–CD44 or SPP1–integrin blockade	Current evidence primarily demonstrates co-enrichment and clinical association; tissue-preserving spatial proximity and functional interdependence have not been established.
CXCL12-producing fibroblast–TREM2⁺ macrophage interaction37	Level I	Fibroblast-derived CXCL12 may interact with CXCR4-expressing TREM2⁺ macrophages and contribute to macrophage retention, tumor growth, and immune suppression.	CXCL12, CXCR4, TREM2, CD68	CXCL12–CXCR4 blockade; TREM2-directed macrophage modulation	The proposed interaction is supported mainly by single-cell analysis and inferred cell–cell communication; direct spatial and causal validation remains limited.
CCL2⁺ fibroblast–STAT3-activated macrophage niche16	Level II	CCL2⁺ fibroblasts may recruit CCR2⁺ monocytes and support a local cytokine environment associated with STAT3 activation and immunosuppressive macrophage states.	CCL2, CCR2, CD68, CD163, phosphorylated STAT3	CCL2–CCR2 blockade; IL-6/JAK/STAT3 inhibition	Spatial and clinical associations are available, but direct perturbation of the coupled CAF–TAM circuit and prospective treatment-response validation are lacking.
GREM1⁺ fibroblast–SPP1⁺ macrophage niche18	Level II	Spatially adjacent GREM1⁺ fibroblasts and SPP1⁺ macrophages may communicate through SPP1–integrin and collagen-related pathways in matrix-rich, immune-suppressive regions.	GREM1, SPP1, CD44, ITGAV/ITGB-family integrins, collagen	GREM1 targeting; SPP1–CD44/integrin blockade; matrix normalization	Evidence is based on multi-omics analysis, spatial transcriptomics, multiplex imaging, and predicted communication; direct pathway perturbation and treatment-response validation are still required.
THBS2⁺ matrix CAF–C3/C3AR1–SPP1⁺ TAM niche in peritoneal metastasis17	Level III	THBS2⁺ matrix CAF-derived C3 promotes C3AR1-dependent accumulation and transition of macrophages toward an SPP1⁺ TAM state, contributing to immune suppression and resistance to immune checkpoint blockade.	THBS2, C3, C3AR1, SPP1, CD68, collagen	C3–C3AR1 blockade; SPP1-related targeting; matrix modulation combined with immune checkpoint blockade	Evidence is strongest in gastric cancer peritoneal metastasis and remains predominantly preclinical; generalizability to primary tumors or other metastatic sites and efficacy in prospective clinical trials are unknown.

Gastric Cancer-Specific Fibroblast–Macrophage Niche Archetypes

CCL2⁺ Fibroblast–STAT3-Activated Macrophage Niche

Among the currently described fibroblast–macrophage archetypes in gastric cancer, the CCL2⁺ fibroblast–STAT3-activated macrophage niche provides one of the clearest examples of how a stromal chemokine field can be converted into local myeloid immune suppression. CCL2, also known as MCP-1, is a major chemoattractant for CCR2⁺ classical monocytes. In gastric cancer, high intratumoral CCL2 expression and high CCR2 expression in the tumor microenvironment have both been associated with poor postoperative survival, supporting the clinical relevance of this recruitment axis.^44,45 From a mechanistic perspective, CCL2⁺ CAFs can act as a stromal source of monocyte-attracting signals, drawing circulating CCR2⁺ monocytes into fibroblast-rich tumor regions. These recruited cells may then differentiate under local cytokine pressure into macrophage states that support tumor progression rather than effective antitumor immunity.

The defining feature of this niche is that recruitment and activation are spatially coupled. Spatial transcriptomic analysis of gastric cancer has shown that CCL2⁺ CAF-enriched regions are located close to STAT3-activated macrophage regions, and experimental models confirmed that CCL2⁺ fibroblasts can enhance macrophage accumulation and promote an immunosuppressive tumor microenvironment.¹⁶ This finding is important because it moves the CCL2–CCR2 axis beyond a simple chemotaxis model. In a gastric cancer tissue context, CCL2 does not merely increase macrophage numbers; it helps position macrophages within a fibroblast-dense territory where IL-6, CSF1, IL-8, and other stromal or tumor-derived signals can activate JAK–STAT3 signaling and reinforce suppressive myeloid differentiation.^16,28,46,47 Therefore, STAT3 activation should be interpreted as a local signaling state generated by the tissue niche, rather than as a generic marker of M2 polarization.

Several gastric cancer studies support this cytokine-amplification model. IL-6 can induce STAT3-dependent M2-like macrophage differentiation in gastric cancer and is associated with CD163⁺CD206⁺ macrophage infiltration, tumor progression, and reduced survival.²⁸ Gastric cancer-derived mesenchymal stromal cells, which share functional overlap with activated CAF-like stromal cells, promote macrophage polarization through high secretion of IL-6 and IL-8; neutralizing IL-6/IL-8 or inhibiting JAK2/STAT3 attenuates this macrophage-reprogramming effect in coculture systems.⁴⁶ In addition, high CD68⁺ TAM density in gastric cancer has been linked to poor prognosis in an IL-6-associated context, while high CSF-1 expression in gastric cancer tissues is an independent adverse prognostic factor and may further support macrophage survival and differentiation within the tumor bed.^47,48 Together, these data suggest a self-reinforcing stromal–myeloid loop: CCL2⁺ CAFs recruit monocytes, local IL-6/CSF1-rich signals activate and sustain macrophages, and activated macrophages in turn consolidate immune suppression, angiogenesis, matrix remodeling, and epithelial plasticity.

Therapeutically, this niche offers several possible intervention points, including CCL2–CCR2 blockade to reduce monocyte influx, IL-6–IL-6R or JAK–STAT3 inhibition to interrupt macrophage instruction, and CSF1–CSF1R targeting to limit macrophage survival or reprogram TAM composition. These strategies are conceptually attractive because they do not simply attack one cell type; they attempt to dismantle the recruitment–activation circuit that maintains a macrophage-rich suppressive stroma. However, broad inhibition of monocyte recruitment or macrophage survival requires caution. CCR2-dependent monocytes contribute to antimicrobial defense and inflammatory tissue repair, and macrophages are central to tissue homeostasis, wound healing, and resolution of injury.^40,49 Moreover, preclinical cancer studies have shown that interruption of CCL2 blockade can provoke rebound monocyte mobilization and enhanced metastasis in some settings.⁵⁰ Thus, in gastric cancer, the most rational path is unlikely to be indiscriminate long-term depletion of monocytes or macrophages. A more precise strategy would be spatially or molecularly selected combination therapy: targeting the CCL2⁺ CAF–STAT3⁺ macrophage niche in patients whose tumors show this coupled stromal–myeloid program, ideally together with chemotherapy, immune checkpoint blockade, or anti-angiogenic treatment.

GREM1⁺ Fibroblast–SPP1⁺ Macrophage Niche

The GREM1⁺ fibroblast–SPP1⁺ macrophage niche represents a matrix-centered stromal–myeloid archetype in gastric cancer. Unlike the CCL2⁺ fibroblast–STAT3-activated macrophage niche, which is mainly organized around monocyte recruitment and cytokine activation, this niche appears to be more closely related to extracellular matrix remodeling, fibroblast activation, and macrophage adaptation within collagen-rich stromal regions. Multi-omics analyses have identified GREM1⁺ fibroblasts and SPP1⁺ macrophages as tumor-enriched cell states in gastric cancer, and patients with concomitantly high infiltration of both populations showed enrichment of cancer-related pathways and poorer survival than patients with low infiltration of both populations.¹⁸ These findings suggest that the two cell states may define a clinically unfavorable stromal–myeloid module, although association alone is not sufficient to prove direct functional dependence.

The spatial evidence supporting this niche is stronger than simple bulk co-enrichment. Spatial transcriptomic and multiplex validation analyses indicate that GREM1⁺ fibroblasts and SPP1⁺ macrophages can occupy adjacent regions in gastric cancer tissues. In this setting, GREM1⁺ fibroblasts are likely to mark an activated stromal compartment, whereas SPP1⁺ macrophages mark a macrophage state linked to matrix interaction, tissue repair, angiogenesis, and immunosuppressive signaling. Cell–cell communication analysis further predicted SPP1–integrin and collagen-related interactions between SPP1⁺ macrophages and fibroblast-rich matrix territories.¹⁸ Therefore, SPP1–ITGAV/ITGB-family signaling and collagen–integrin signaling may mediate local anchoring, retention, or activation of macrophages within a desmoplastic microenvironment, but these mechanisms still require direct perturbation experiments in gastric cancer.

The biological meaning of this niche is closely tied to extracellular matrix organization. In patients with high GREM1⁺ fibroblast and SPP1⁺ macrophage infiltration, enrichment analyses highlighted extracellular matrix structure, collagen-containing extracellular matrix, and myeloid leukocyte migration programs.¹⁸ This is consistent with independent pathological evidence showing that collagen content, maturity, morphology, and architecture are substantially reorganized in the gastric cancer microenvironment and that collagen width has prognostic value.⁵¹ Moreover, type I collagen can promote the growth of ITGB1-positive gastric cancer cells through a BCL9L/β-catenin-dependent pathway, supporting the broader concept that collagen–integrin signaling is not a passive scaffold but an active growth- and survival-promoting cue in gastric cancer.⁵² In this context, GREM1⁺ fibroblasts may contribute to a collagen-rich stromal frame, while SPP1⁺ macrophages may help maintain a reparative and matrix-interacting myeloid state.

GREM1 itself provides an additional layer of therapeutic interest. As a secreted antagonist of bone morphogenetic protein signaling, GREM1 can reshape BMP-dependent tissue differentiation and repair programs, and stromal GREM1 expression has been associated with advanced pathological features and worse survival in gastric cancer.^41,53 However, the therapeutic interpretation of GREM1 is more complex than simple ligand blockade. BMP signaling has context-dependent effects in cancer, and GREM1 may also act through BMP-independent mechanisms. Thus, targeting the GREM1–BMP balance could potentially remodel stromal differentiation or reduce tumor-supportive fibroblast programs, but this strategy remains immature in gastric cancer and should be tested with careful attention to tissue context, histological subtype, and stromal dependency.

From a therapeutic perspective, this niche provides several candidate entry points rather than one established target. SPP1 blockade could potentially weaken macrophage–matrix and macrophage–tumor communication. Inhibition of ITGAV/ITGB-family integrins may reduce SPP1- or collagen-dependent cell adhesion and survival signaling. Matrix-modulating strategies could disrupt collagen-mediated exclusion and reduce the stromal anchoring of suppressive myeloid cells. Finally, approaches aimed at restoring the GREM1–BMP balance may alter the fibroblast state that sustains this niche. At present, however, the GREM1⁺ fibroblast–SPP1⁺ macrophage niche should be considered a strong Level II candidate spatial niche rather than a fully validated Level III therapeutic niche. It potentially contributes to ECM deposition, stromal activation, and poor prognosis, but direct evidence that it causes gastric cancer immunotherapy resistance is still insufficient.

FAP⁺ Fibroblast–SPP1⁺ Macrophage Module in pMMR Gastric Cancer

The FAP⁺ fibroblast–SPP1⁺ macrophage module provides a useful framework for understanding why many proficient mismatch repair or microsatellite-stable gastric cancers remain poorly responsive to immune checkpoint blockade. Compared with dMMR/MSI-H tumors, pMMR/MSS gastric cancers generally have lower neoantigen burden and less robust immune activation, although mismatch repair status alone does not fully explain therapeutic sensitivity.^42,54 Recent single-cell analysis comparing gastric cancers with different MMR statuses showed that pMMR tumors were enriched for FAP⁺ fibroblasts, SPP1⁺ macrophages, GC2 malignant cells, and E2 endothelial cells. These cell populations were associated with hypoxia, angiogenesis, EMT, VEGF signaling, MIF signaling, SPP1 signaling, and ECM remodeling.⁴² This pattern suggests that pMMR gastric cancer may not simply be “immune cold” because of fewer mutations; in at least a subset of tumors, immune low-sensitivity may also reflect a stromal–myeloid program that restricts effective T-cell access and function.

Within this module, FAP⁺ fibroblasts appear to represent an activated stromal compartment with strong angiogenic and matrix-remodeling features. In pMMR gastric cancer, FAP⁺ fibroblasts were reported to show enhanced interactions with malignant epithelial cells and endothelial cells through VEGF, MK, CXCL12–CXCR4, and MIF-related pathways. In parallel, SPP1⁺ macrophages showed enhanced communication with tumor cells, endothelial cells, and fibroblasts through SPP1, MIF, VEGF, and ECM-related signals.⁴² These findings support a model in which FAP⁺ fibroblasts and SPP1⁺ macrophages may jointly shape a hypoxic, pro-angiogenic, matrix-rich microenvironment. Such a microenvironment potentially contributes to a “myeloid-rich, T-cell-restricted” state, in which macrophages and fibroblasts are abundant, but cytotoxic T cells are either quantitatively limited, spatially displaced, or functionally ineffective.

Several signaling axes are particularly relevant. SPP1–integrin signaling may support macrophage adhesion, matrix interaction, and tumor-promoting communication within fibroblast-rich regions. MIF–CD74/CXCR4 signaling may provide an additional immunosuppressive route. In gastric cancer, a stromal immune barrier analysis suggested that macrophage-derived MIF signaling through CD74/CXCR4/CD44 may contribute to immune dysfunction and non-response to immune checkpoint inhibitors.³⁵ Another single-cell and functional study showed that gastric cancer-derived MIF promoted immunosuppressive TAM remodeling through the MIF–CD74 axis, whereas CD74 blockade reversed MIF-induced TAM suppression in vitro.⁵⁵ VEGF signaling is also relevant because it connects hypoxia, endothelial remodeling, and myeloid immune suppression. A recent gastric cancer study identified SPI1⁺CD68⁺ macrophages as a pro-angiogenic TAM subset associated with VEGFA secretion, metastasis, immunosuppressive contexture, and potential sensitivity to combined antiangiogenic and immunotherapeutic strategies.⁵⁶ Although this does not prove that VEGF is the driver of the FAP⁺ fibroblast–SPP1⁺ macrophage module, it supports the broader rationale for combining vascular and myeloid modulation in angiogenic gastric cancer microenvironments.

The therapeutic implications of this module are therefore multidimensional. FAP-directed approaches may be used not only for imaging but also for selective drug, radionuclide, or immune-modulating payload delivery into fibroblast-rich tumors.^57,58 Targeting SPP1–integrin interactions could potentially weaken macrophage–matrix anchoring and reduce stromal myeloid reinforcement. MIF–CD74/CXCR4 blockade may attenuate macrophage-mediated immunosuppression and improve T-cell function, particularly where MIF-high macrophage or tumor-cell communication is present.^35,55 Anti-VEGF or VEGFR-directed therapy, when paired with immunotherapy or macrophage-targeting strategies, may help remodel hypoxic and pro-angiogenic stromal–myeloid territories; early clinical experience with regorafenib plus nivolumab in gastric cancer supports the feasibility of combining anti-angiogenic pathway inhibition with PD-1 blockade, although optimal patient selection remains unresolved.⁵⁹

At present, the FAP⁺ fibroblast–SPP1⁺ macrophage module in pMMR gastric cancer should be classified as a candidate functional module rather than a fully validated spatial niche. The evidence supports cell-state enrichment, pathway-level communication, immunofluorescence validation of FAP⁺ fibroblast abundance, and biologically plausible links to hypoxia, angiogenesis, ECM remodeling, and immune low-sensitivity.⁴² However, direct spatial colocalization between FAP⁺ fibroblasts and SPP1⁺ macrophages in pMMR gastric cancer remains incompletely resolved, and functional perturbation of this exact module in spatially defined patient-derived models is still needed. Therefore, its value at this stage lies in generating testable therapeutic hypotheses: pMMR/MSS tumors with FAP⁺ fibroblast, SPP1⁺ macrophage, VEGF/MIF, and ECM-high features may represent a subgroup in which stromal targeting, macrophage reprogramming, anti-angiogenic therapy, and checkpoint blockade should be rationally combined rather than applied in isolation.

THBS2⁺ Matrix CAF–SPP1⁺ TAM Niche in Peritoneal Metastasis

The THBS2⁺ matrix CAF–SPP1⁺ TAM niche in gastric cancer peritoneal metastasis represents one of the strongest currently available examples of a functional and therapy-relevant fibroblast–macrophage niche. Peritoneal metastasis is not simply a secondary deposit that reproduces the immune ecology of the primary gastric tumor. Earlier pathological studies already suggested that macrophages in the peritoneal cavity are enriched in patients with gastric cancer peritoneal metastasis and may support metastatic colonization through growth and angiogenic factors such as EGF and VEGF.⁶⁰ More recent single-cell and spatial studies further indicate that peritoneal metastases and malignant ascites develop a distinct microenvironment marked by stromal remodeling, macrophage lineage transition, epithelial plasticity, and treatment resistance.^61–63 This organ-specific divergence is important because it explains why therapeutic response in the peritoneal cavity may differ from that predicted by the primary tumor alone.

Within this metastatic setting, THBS2⁺ matrix CAFs and SPP1⁺ TAMs form a spatially organized stromal–myeloid niche. THBS2⁺ matrix CAFs represent a collagen- and extracellular matrix-associated fibroblast state, whereas SPP1⁺ TAMs represent a macrophage state linked to tissue repair, matrix interaction, immunosuppression, and T-cell dysfunction. In gastric cancer peritoneal metastasis, single-cell RNA sequencing, spatial transcriptomics, multiplex immunofluorescence, and functional validation collectively showed that the peritoneal metastatic ecosystem is dominated by a stroma–myeloid compartment composed of THBS2⁺ matrix CAFs and SPP1⁺ TAMs. This finding is stronger than simple co-enrichment, because the two cell states were not only abundant but also spatially organized within metastatic lesions and associated with immune checkpoint blockade resistance.¹⁷

Mechanistically, the most important feature of this niche is the CAF-derived complement signal. THBS2⁺ matrix CAFs were shown to release complement C3, which acts on C3AR1-expressing macrophages and promotes the recruitment and accumulation of peritoneum-specific tissue-resident macrophage-derived SPP1⁺ TAMs.¹⁷ This mechanism provides a direct bridge between matrix fibroblasts and macrophage state conversion. Rather than viewing SPP1⁺ TAMs as passively enriched in peritoneal lesions, this model suggests that the fibroblast-rich peritoneal matrix actively instructs macrophage localization and phenotype through the C3–C3AR1 axis. The macrophage compartment may therefore arise from both tissue-resident and recruited precursors, but its terminal tumor-supportive state is shaped by the metastatic stromal niche.

The functional consequences of this niche are closely tied to extracellular matrix deposition and immune suppression. THBS2⁺ matrix CAFs can create a dense stromal scaffold, while SPP1⁺ TAMs may reinforce tissue repair-like signaling, collagen-associated communication, angiogenesis, and T-cell dysfunction. Independent ascites-based single-cell studies have also shown that macrophages in gastric cancer peritoneal metastasis can undergo immunosuppressive transitions, including CTS^high-to-C1Q^high macrophage differentiation, and interact with disseminated tumor cells in ways that promote immune evasion.⁶² Another treatment-process study of GCPM reported that therapy-failure samples showed spatial and ligand–receptor features linking macrophages, fibroblasts, endothelial cells, and tumor cells, including SPP1–CD44 and LGALS9-related interactions associated with reduced cytotoxic T-cell activity.⁶¹ Together, these findings support the broader idea that peritoneal metastasis is sustained by a tissue-specific stromal–myeloid barrier, not by malignant epithelial cells alone.

Clinically, this niche is especially important because it is linked to the low responsiveness of some gastric cancer peritoneal metastases to immune checkpoint blockade. In ICB-resistant peritoneal tumors, extensive accumulation of THBS2⁺ matrix CAFs and SPP1⁺ TAMs was associated with impaired antitumor immunity. More importantly, blockade of the C3–C3AR1 axis disrupted macrophage accumulation and improved the antitumor efficacy of immune checkpoint blockade in experimental models.¹⁷ This places the THBS2⁺ matrix CAF–SPP1⁺ TAM niche close to Level III evidence: it has human spatial evidence, mechanistic ligand–receptor support, pathway perturbation, experimental validation, and a treatment-response connection. The remaining challenge is to translate this axis into prospective biomarker-selected clinical trials.

Several therapeutic implications follow from this model. First, C3–C3AR1 is the most direct intervention point because it connects THBS2⁺ matrix CAFs to SPP1⁺ TAM recruitment and state conversion. Second, SPP1 itself may be targeted to weaken macrophage–matrix and macrophage–tumor communication, although the optimal approach to SPP1 blockade in gastric cancer remains undefined. Third, the THBS2-associated matrix program may serve as a biomarker of desmoplastic peritoneal immune exclusion rather than as a simple single-molecule target. Finally, complement blockade combined with PD-1/PD-L1 inhibitors is a rational strategy for selected patients with C3–C3AR1-high peritoneal stromal–myeloid niches. This approach is supported not only by the gastric cancer peritoneal metastasis model, but also by broader cancer evidence showing that complement C3 can promote immunosuppressive TAM activity and resistance to PD-L1 blockade through C3a–C3aR signaling.^64,65 However, because complement also participates in host defense, inflammation resolution, and tissue homeostasis, clinical translation will require spatial or molecular selection rather than indiscriminate complement suppression.

Mechanistically Supported but Spatially Unresolved Interactions

Not all biologically plausible fibroblast–macrophage interactions in gastric cancer should be described as established spatial niches. Several axes have convincing mechanistic or functional support, but they still lack the integrated evidence chain required for a fully defined niche: high-resolution spatial organization, cell-type-specific pathway perturbation, and treatment-response validation in the same biological context. These interactions are therefore best regarded as mechanistically supported but spatially unresolved stromal–myeloid programs.

The LOX⁺ fibroblast–M2-like macrophage interaction is one such example. LOX⁺ fibroblasts are enriched in gastric cancer tissues and are associated with extracellular matrix remodeling, hypoxic stromal features, and poor prognosis. Single-cell analysis and immunofluorescence double labeling suggest that LOX⁺ fibroblasts are closely located with CD206⁺ M2-like macrophages, and ligand–receptor analysis predicts that LOX⁺ fibroblasts may induce monocyte-to-M2-like macrophage differentiation through IL6–IL6R signaling.²⁵ This axis is mechanistically attractive because it links collagen crosslinking, stromal stiffness, hypoxia, and macrophage polarization. However, the current evidence does not yet establish a spatially resolved fibroblast–macrophage niche comparable to the THBS2⁺ matrix CAF–SPP1⁺ TAM niche in peritoneal metastasis. The remaining gaps include higher-resolution spatial transcriptomic confirmation, lineage tracing of recruited versus resident macrophages, and direct testing of whether LOX or IL6–IL6R blockade reverses immune exclusion or improves treatment response in gastric cancer models.

The CPT1C⁺ CAF–IL-6–macrophage axis is more functionally developed but still spatially incomplete. CPT1C⁺ CAFs have been identified as an immunosuppressive CAF subset in gastric cancer. These cells are enriched in stroma-high tumors, correlate with M2-like macrophage infiltration, and promote CD163⁺CD206⁺ macrophage polarization in coculture systems. Mechanistically, CPT1C knockdown reduces IL-6 secretion from CAFs, while IL-6 neutralization attenuates macrophage reprogramming, supporting a CPT1C–IL-6–STAT3-related immunosuppressive route. Clinically, high CPT1C⁺ CAF infiltration has been associated with poorer response to immunotherapy and adjuvant treatment.²⁹ Even so, this program should not yet be framed as a fully defined spatial niche, because the evidence mainly connects CAF state, cytokine output, macrophage phenotype, and treatment association rather than demonstrating a reproducible spatial architecture in patient tissues.

TGF-β, CSF1, and extracellular vesicle-related communication add further complexity to gastric cancer stromal–myeloid regulation. TGF-β signaling can contribute to macrophage recruitment and M2-like polarization in gastric cancer; for instance, the SPI1–INHBA axis has been reported to activate TGF-β signaling, increase CCL2 expression, and promote macrophage recruitment and M2 polarization.⁶⁶ CSF1–CSF1R signaling is another macrophage-supporting pathway, but in gastric cancer the most recent spatial and functional evidence highlights a CSF1⁺ tumor cell–SPP1⁺ macrophage axis rather than a purely CAF-derived CSF1 signal.⁶⁷ Exosome-mediated communication also participates in macrophage polarization. Gastric cancer cell-derived exosomal circATP8A1 induces M2 polarization through the miR-1-3p/STAT6 axis, while SERPINE1-mediated exosomal let-7g-5p promotes M2 polarization through SOCS7 suppression and STAT3 activation.^68,69 CAF-derived extracellular vesicles can also promote aggressive gastric cancer phenotypes, but their direct role in fibroblast-to-macrophage spatial instruction remains insufficiently defined.²³ These studies support extracellular vesicles as important carriers of stromal or tumor-derived immunoregulatory information, but they do not yet define a spatial CAF–macrophage niche.

The CXCL12–CXCR4/TREM2 axis is also promising but should be interpreted cautiously. Recent work indicates that CAFs can secrete CXCL12, which interacts with CXCR4 on TREM2⁺ lipid-associated macrophages, promoting immunosuppression, tumor growth, and metastatic behavior in gastric cancer models.⁸ Blocking the CXCL12/CXCR4 axis reduced tumor-promoting effects and partially restored T-cell markers in vivo.³⁷ These data support a functional CAF–TREM2⁺ macrophage interaction and suggest that CXCL12–CXCR4 may be a therapeutic entry point. However, the current evidence still lacks sufficient gastric cancer spatial validation showing where these CAFs and TREM2⁺ macrophages are positioned relative to tumor nests, vessels, invasive margins, or metastatic sites. Therefore, this axis is best described as a mechanistically supported candidate interaction rather than an established spatial niche.

Overall, these interactions should be kept in the review, but with explicit evidence boundaries. LOX⁺ fibroblast–M2-like macrophage communication links matrix remodeling to myeloid polarization; CPT1C⁺ CAFs provide strong functional evidence for IL-6-mediated macrophage reprogramming; TGF-β, CSF1, and exosomal cargoes illustrate additional routes of macrophage instruction; and CXCL12–CXCR4/TREM2 signaling connects fibroblast-derived chemokines with lipid-associated macrophage function. Their therapeutic potential is real, but their classification should remain below fully validated Level III niches until spatial localization, perturbation, and clinical-response evidence are integrated in gastric cancer-specific systems.

Mechanisms Linking Spatial Niches to Immune Exclusion

Preferential Recruitment and Lineage Programming

A central feature of fibroblast–macrophage spatial niches is preferential immune recruitment. These niches do not simply increase immune-cell infiltration; they reshape which immune cells are allowed to enter, survive, and mature within the tumor tissue. In gastric cancer, CAF-rich regions can produce chemokines, cytokines, complement components, and survival factors that preferentially attract circulating monocytes or sustain tissue-resident macrophages, while failing to generate sufficient effector T-cell-attracting signals such as CXCL9/CXCL10–CXCR3. The consequence is a microenvironment enriched in suppressive or reparative myeloid cells but relatively restricted in effective cytotoxic lymphocytes.

The CCL2–CCR2 axis is a prototypical monocyte recruitment pathway. CCL2 produced by stromal or tumor cells attracts CCR2⁺ classical monocytes from the circulation, allowing them to enter tumor tissue and differentiate into TAMs under local instructive signals. In gastric cancer, high CCL2 or CCR2 expression has been associated with poor survival after surgical resection, supporting the clinical relevance of this pathway.^44,45 When CCL2 is produced by CAF-rich stromal territories, recruited monocytes are not entering a neutral tissue space; they are exposed immediately to fibroblast-derived and tumor-derived cytokines, matrix ligands, and hypoxic cues that bias their differentiation toward immunosuppressive macrophage states. Thus, CCL2–CCR2 helps explain why some tumors accumulate abundant myeloid cells without developing a productive antitumor lymphocyte response.

CSF1–CSF1R signaling provides a second layer of selection by supporting macrophage survival, differentiation, and maintenance. In gastric cancer, elevated CSF1 expression has been associated with adverse prognosis, and a recent spatially validated study identified a CSF1⁺ tumor cell–SPP1⁺ macrophage axis linked to gastric cancer progression and immunotherapy resistance.^48,67 Although this particular axis is tumor-cell centered rather than strictly CAF-derived, it illustrates a broader principle relevant to stromal–myeloid niches: once monocytes or macrophages enter the tumor bed, CSF1-rich regions can sustain macrophage persistence and favor TAM programs over antigen-presenting dendritic-cell-like or lymphocyte-supportive immunity. In other solid tumor models, CSF1R blockade can deplete or reprogram TAMs and improve checkpoint inhibitor responsiveness, supporting the therapeutic rationale for targeting macrophage maintenance pathways.^70,71

IL-6–JAK–STAT3 signaling is especially important for lineage programming after recruitment. IL-6 does not primarily act as a chemotactic gate; rather, it converts recruited or resident macrophages into suppressive, tissue-repair-like states. In gastric cancer, IL-6 can induce STAT3-dependent M2-like macrophage differentiation, and gastric cancer-derived stromal cells can promote macrophage polarization through IL-6/IL-8–JAK2/STAT3 signaling.^28,46 STAT3 activation also links macrophage polarization to broader immune exclusion because STAT3-driven myeloid cells can produce IL-10, VEGF, arginase-related suppressive programs, and checkpoint-associated signals that inhibit T-cell activation or reduce T-cell access to tumor nests. Within CAF-rich areas, IL-6 therefore helps transform monocyte recruitment into a durable suppressive macrophage niche.

Complement signaling adds a tissue-specific route to macrophage recruitment and state conversion. In gastric cancer peritoneal metastasis, THBS2⁺ matrix CAFs release complement C3, which acts on C3AR1-expressing macrophages and promotes the recruitment and transformation of peritoneum-specific tissue-resident macrophage-derived SPP1⁺ TAMs. This mechanism is particularly informative because it shows that a fibroblast-derived innate immune signal can organize macrophage fate within a metastatic niche. Rather than attracting effector lymphocytes, the C3–C3AR1 axis enriches a macrophage population associated with extracellular matrix remodeling, impaired antitumor immunity, and poor response to immune checkpoint blockade. Experimental blockade of this pathway improved antitumor immune activity and enhanced checkpoint therapy efficacy in gastric cancer peritoneal metastasis models.¹⁷

The CXCL12–CXCR4 axis represents another mechanism by which CAF-rich regions may favor myeloid retention and T-cell restriction. CXCL12 is frequently produced by activated fibroblasts and can attract or retain CXCR4⁺ immune and stromal cells. In gastric cancer, CAF-derived CXCL12 has been reported to interact with CXCR4 on TREM2⁺ lipid-associated macrophages, enhancing immunosuppressive macrophage features and tumor progression.³⁷ More recent gastric cancer studies further suggest that CXCL12/CXCR4 signaling can enhance CAF–TAM communication, promote M2 macrophage polarization, increase PD-L1 expression, and confer anti-PD-1 resistance.^72,73 This axis may therefore create a dual barrier: it supports suppressive macrophage programming while limiting effective T-cell activation or access.

These pathways also explain why immune exclusion is not merely a matter of low immune-cell density. A tumor may be heavily infiltrated by immune cells and still be immunologically ineffective if the dominant chemokine and cytokine network recruits CCR2⁺ monocytes, sustains CSF1R⁺ macrophages, activates STAT3-dependent suppressive programs, or retains CXCR4⁺ myeloid cells in fibrotic regions. In contrast, effective antitumor immunity usually requires chemokine circuits that support dendritic-cell activation and CXCL9/CXCL10–CXCR3-dependent CD8⁺ T-cell recruitment. In gastric cancer, restoration of CXCL9/CXCL10 expression has been linked to increased CD8⁺ T-cell infiltration in experimental models, and multidimensional immune profiling has shown that spatial organization of T-cell and macrophage subsets predicts response to anti-PD-1/PD-L1 therapy.^74,75 Therefore, the core mechanism of CAF–macrophage immune exclusion is selective recruitment plus selective education: the niche chooses myeloid cells, programs them locally, and leaves cytotoxic lymphocytes either outside the tumor nests or functionally constrained inside a suppressive stromal field.

ECM Remodeling and Mechanical Exclusion

Fibroblast–macrophage niches promote immune exclusion not only through chemokines and cytokines, but also through physical remodeling of the tumor tissue. In gastric cancer, collagen deposition, fibronectin accumulation, lysyl oxidase-mediated collagen crosslinking, and matricellular ligand–receptor signaling can convert the stroma into a dense and mechanically restrictive compartment. This remodeled matrix does not simply surround tumor cells; it creates a structural boundary that influences where immune cells can move, where they pause, and whether they can reach malignant epithelial nests.

Collagen is the central scaffold of this barrier. Gastric cancer tissues show marked reorganization of collagen architecture, including changes in collagen content, width, alignment, and maturity, and these collagen features have been associated with tumor progression and prognosis.⁵¹ More recent functional evidence suggests that collagen-rich ECM can directly promote gastric cancer immune evasion. In a 3D collagen model, gastric cancer cells became more resistant to HER2-specific CAR-T cell cytotoxicity, and collagen-induced integrin αvβ1 signaling increased IL4I1 expression, leading to aryl hydrocarbon receptor (AHR) activation and CAR-T cell exhaustion.⁷⁶ Although CAR-T models are not identical to endogenous CD8⁺ T-cell immunity, this study provides gastric cancer-specific evidence that collagen ECM can actively suppress effector lymphocyte function rather than merely serving as a passive structural scaffold.

LOX-mediated collagen crosslinking further strengthens this barrier. LOX and related lysyl oxidase family enzymes crosslink collagen fibers, increase matrix stiffness, and help generate a mechanically rigid tumor microenvironment. In gastric cancer, LOX⁺ fibroblasts are enriched in tumor tissues, associated with poor prognosis, and spatially close to CD206⁺ M2-like macrophages by immunofluorescence validation.²⁵ This observation links matrix stiffening to myeloid immune suppression: LOX⁺ fibroblasts may simultaneously reorganize collagen and support suppressive macrophage polarization through IL6–IL6R-related signaling. A stiff collagen network can reduce matrix pore size, align fibers along tumor–stroma interfaces, and provide contact-guidance tracks that direct immune cells parallel to tumor nests instead of permitting deep penetration.

Matricellular adhesion signals reinforce this physical exclusion. SPP1, also known as osteopontin, binds integrins and CD44 family receptors, enabling cell–matrix adhesion, migration, and local survival signaling.⁷⁷ In gastric cancer, SPP1/CD44-mediated intratumoral crosstalk has been associated with macrophage infiltration, advanced stage, and poor survival.³³ Within a CAF-rich ECM, SPP1–integrin signaling may help retain SPP1⁺ TAMs in matrix-dense territories, strengthen macrophage–fibroblast communication, and indirectly limit T-cell access. FN1 provides a parallel matrix-associated route. FN1 is highly expressed in gastric cancer tissues, correlates with poor prognosis, and is closely associated with macrophage infiltration, particularly M2-like macrophage markers.⁷⁸ Because fibronectin fibers are major regulators of cell adhesion and tissue migration, FN1-rich stromal tracks may favor macrophage residence and tumor-cell invasion while failing to provide a permissive route for CD8⁺ T cells to enter tumor epithelial nests.

These biochemical and structural changes converge on mechanical exclusion. Dense collagen and fibronectin networks can trap T cells in the stroma or force them to move along the tumor boundary. Live imaging of human lung tumor slices showed that T cells migrated actively in loose collagen- and fibronectin-rich stromal regions but poorly in dense matrix areas; aligned fibers surrounding tumor islets directed T-cell trajectories parallel to the tumor–stroma border and restricted contact with cancer cells, whereas collagenase treatment increased T-cell contact with tumor cells.⁷⁹ Although this evidence comes from lung cancer, it provides a mechanistic explanation that is highly relevant to desmoplastic gastric cancer: when CAFs and TAMs organize dense and aligned ECM at the tumor edge, CD8⁺ T cells may accumulate in the peritumoral stroma without efficiently entering malignant epithelial compartments.

Matrix remodeling also changes tissue pressure and vascular function. Accumulation of collagen, hyaluronan, fibroblasts, and proliferating cancer cells generates solid stress; in parallel, leaky vessels and impaired lymphatic drainage elevate interstitial fluid pressure. These mechanical forces can compress blood and lymphatic vessels, reduce perfusion, worsen hypoxia, and limit delivery of both immune cells and therapeutic agents.^80,81 In gastric cancer, abnormal angiogenesis and VEGF-rich stromal–myeloid programs frequently coexist with CAF and TAM enrichment, suggesting that mechanical and vascular barriers are coupled rather than independent.⁵⁶ Vessel compression and poor perfusion further reduce lymphocyte entry, while hypoxia promotes VEGF, IL-6, LOX, and macrophage-polarizing signals, thereby closing a self-reinforcing loop between matrix stiffness, hypoxia, myeloid enrichment, and T-cell exclusion.

Thus, ECM remodeling provides a spatial mechanism for immune exclusion: collagen deposition builds the scaffold, LOX crosslinking stiffens it, SPP1–integrin and FN1–CD44-related adhesion signals stabilize stromal–myeloid residence, and elevated solid stress/interstitial pressure compresses vessels and restricts tissue access. The result is not necessarily an absence of CD8⁺ T cells, but their mislocalization. T cells may be present at the invasive margin, perivascular regions, or fibroblast-rich stromal zones, yet remain physically separated from tumor cells. For this reason, future biomarker assessment in gastric cancer should not only quantify CD8⁺ T-cell density, but also measure whether CD8⁺ T cells penetrate tumor nests or are retained within collagen-, fibronectin-, and macrophage-rich stromal compartments.

Chemokine-Mediated T-Cell Sequestration

Chemokine dysregulation links fibroblast–macrophage niches to immune exclusion, but different spatial immune phenotypes should be distinguished carefully. A tumor with few CD8⁺ T cells, a tumor with CD8⁺ T cells trapped in stromal regions, and a tumor with intratumoral but dysfunctional CD8⁺ T cells are biologically different. They may all respond poorly to immune checkpoint blockade, but they are not the same form of immune resistance. In this review, “immune exclusion” is used mainly for tumors in which T cells are present in the tissue but remain spatially separated from malignant epithelial nests, usually at the invasive margin, perivascular zones, or fibroblast-rich stromal compartments.

The first pattern is insufficient T-cell recruitment, often described as an immune-desert state. In this setting, tumor tissues lack the chemokine gradients needed to bring effector lymphocytes into the lesion. Effective CD8⁺ T-cell trafficking commonly depends on interferon-inducible CXCL9/CXCL10/CXCL11–CXCR3 signaling, together with antigen presentation and dendritic-cell activation. In advanced gastric cancer, high CXCR3 expression has been associated with increased CD4⁺ and CD8⁺ tumor-infiltrating lymphocytes and favorable overall survival, supporting the relevance of this chemokine axis in gastric cancer immunity.⁸² More recent work in MSS gastric cancer also suggests that T-bet⁺CD8⁺ T cells, which are functionally linked to interferon-related chemokine circuits, are important determinants of anti-PD-1 responsiveness.⁸³ Therefore, when CXCL9/CXCL10–CXCR3-type recruitment is weak, the problem is not that T cells are excluded by a stromal wall after arrival; rather, effective T-cell entry may never be properly initiated.

The second pattern is true chemokine-mediated sequestration. Here, T cells are recruited into the tumor-bearing tissue but are diverted away from cancer-cell-rich regions. CAF-derived CXCL12 provides the best-known example of this mechanism. In pancreatic cancer, FAP⁺ CAFs produce CXCL12, which coats malignant cells and contributes to T-cell exclusion from tumor-cell regions; CXCR4 inhibition rapidly permits T-cell accumulation near cancer cells and sensitizes tumors to PD-L1 blockade.⁸⁴ Although this exact mechanism has not been fully resolved in gastric cancer, it provides a useful model for interpreting CAF-rich gastric tumors in which CD8⁺ T cells accumulate at the tumor–stroma interface but fail to penetrate tumor nests. In gastric cancer, CXCL12–CXCR4 signaling has been linked to CAF–TAM communication, M2-like macrophage polarization, PD-L1 upregulation, and anti-PD-1 resistance, suggesting that this pathway may sequester or misdirect immune cells while simultaneously strengthening myeloid suppression.^72,73

The third pattern is intratumoral dysfunction rather than exclusion. In this case, CD8⁺ T cells have entered tumor-cell-rich areas, but their effector function is constrained by chronic antigen stimulation, checkpoint signaling, suppressive macrophages, Tregs, metabolic stress, or inhibitory cytokines. This distinction matters because intratumoral exhausted T cells can still mark an immunologically engaged tumor and may be reinvigorated by checkpoint blockade if sufficient precursor-like or effector-competent T-cell states are present. Multidimensional immune profiling in gastric cancer has shown that response to anti-PD-1/PD-L1 therapy is better predicted by combinations of immune-cell density, phenotype, and spatial organization than by a single marker alone.⁷⁵ Similarly, spatial RNA sequencing in MSS gastric cancer suggests that T-bet⁺CD8⁺ T-cell states and their interaction with PD-L1⁺ tumor cells help determine immunotherapy responsiveness.⁸³ Thus, the presence of PD-1⁺ or exhausted CD8⁺ T cells inside tumor nests should not automatically be called immune exclusion; it reflects an inflamed but functionally restrained state.

These distinctions are important for therapeutic design. Immune-desert tumors may require strategies that restore antigen presentation, type I/II interferon signaling, or CXCL9/CXCL10-driven T-cell recruitment. Immune-excluded tumors may require stromal, matrix, CXCL12–CXCR4, VEGF, TGF-β, or macrophage-targeting interventions to move T cells from the stromal border into tumor nests. Intratumorally infiltrated but dysfunctional tumors may be more amenable to checkpoint blockade, costimulatory agonists, metabolic modulation, or myeloid/Treg suppression. Therefore, the term “immune exclusion” should be reserved for a spatial phenotype, not used as a general label for all immunosuppressive microenvironments.

In CAF–macrophage niches, chemokines can therefore produce resistance in at least three non-equivalent ways: by failing to recruit effector T cells, by attracting or retaining them in the wrong compartment, or by allowing entry but imposing functional paralysis. A rigorous spatial assessment of gastric cancer should report not only how many CD8⁺ T cells are present, but where they are located relative to tumor nests, CAF-rich matrix, macrophage clusters, and vessels, and whether their transcriptional state indicates recruitment failure, stromal sequestration, or intratumoral dysfunction.

Hypoxia and Vascular Dysfunction

Hypoxia and vascular dysfunction provide a metabolic and anatomical layer to fibroblast–macrophage-mediated immune exclusion. In gastric cancer, hypoxia is not only a consequence of rapid tumor growth; it is also reinforced by dense ECM, compressed vessels, abnormal angiogenesis, and stromal–myeloid inflammation. High HIF-1α expression in gastric cancer has been correlated with CD68⁺/CD163⁺ TAM infiltration, aggressive clinicopathological features, EMT-related changes, and poor survival, suggesting that hypoxia and macrophage enrichment are biologically connected in human tumors.⁸⁵ Within CAF-rich regions, hypoxia can further strengthen fibroblast activation, matrix remodeling, VEGF production, and macrophage polarization, thereby creating a local environment that favors SPP1⁺ or M2-like TAM states rather than productive T-cell immunity.

Hypoxia can shape SPP1⁺ macrophage biology through both direct and indirect routes. Pan-cancer single-cell analysis has shown that SPP1⁺ TAMs are expanded in hypoxia-high macrophage states, and SPP1 expression can be induced in macrophages under hypoxic culture conditions.⁸⁶ In gastric cancer, pMMR tumors enriched for SPP1⁺ macrophages and FAP⁺ fibroblasts display hypoxia, pro-angiogenesis, ECM remodeling, VEGF, and MIF-related communication programs.⁴² Although these findings do not prove that hypoxia alone generates SPP1⁺ TAMs in every gastric cancer, they support a plausible spatial model: poorly perfused fibroblast-rich regions provide hypoxic and matrix-associated cues that stabilize a macrophage state adapted to tissue repair, angiogenesis, and immune suppression.

VEGF-driven vascular abnormality is a major downstream consequence of this hypoxic niche. Rather than producing mature and well-perfused vessels, persistent VEGF signaling often generates tortuous, leaky, poorly covered, and functionally inefficient vessels. This abnormal vasculature worsens regional hypoxia and increases interstitial pressure, creating a vicious cycle in which hypoxia induces VEGF, VEGF produces dysfunctional angiogenesis, and dysfunctional angiogenesis further reduces oxygenation and drug delivery. Gastric cancer studies support this coupling: HIF-1α expression is associated with VEGF expression, excessive angiogenesis, invasion, metastasis, and poor prognosis, while SPI1⁺CD68⁺ macrophages have recently been identified as a pro-angiogenic macrophage subset that promotes VEGFA secretion, metastatic behavior, and an immunosuppressive contexture.^56,87

Low perfusion has two direct therapeutic consequences. First, it limits the delivery of cytotoxic drugs, antibodies, immune cells, and oxygen into tumor nests. Even when a molecular target is present, a poorly perfused and high-pressure stromal region may receive insufficient drug exposure. Second, hypoperfusion reduces the probability that circulating effector lymphocytes will efficiently enter the tumor parenchyma. In this sense, vascular dysfunction and mechanical exclusion are inseparable: ECM remodeling compresses vessels and elevates interstitial pressure, while abnormal vessels fail to support uniform immune-cell and drug trafficking. This is one reason why vascular normalization has been proposed as a rational partner for immunotherapy in gastric cancer, aiming not simply to starve tumors but to improve perfusion, reduce hypoxia, and restore immune-cell access.^88,89

Endothelial cells also actively regulate whether T cells can cross from blood into tumor tissue. Effective lymphocyte transmigration requires endothelial adhesion molecules, chemokines, and permissive vascular structure. However, tumor endothelial cells can acquire an immunosuppressive or “anergic” phenotype under the influence of VEGF, hypoxia, prostaglandins, and myeloid-derived signals. In gastric cancer single-cell analysis, tumor-associated endothelial subtypes showed increased angiogenic and cell motility pathways but reduced antigen-presenting and immune-defense programs, suggesting that immune-cell attraction through endothelial cells is impaired in at least part of the gastric cancer vasculature.³¹ More broadly, tumor endothelial FasL has been shown to selectively limit CD8⁺ T-cell infiltration while permitting immunosuppressive cells, illustrating how the vasculature can function as an immune gate rather than a passive conduit.⁹⁰

Thus, hypoxia and vascular dysfunction link CAF–macrophage niches to immune exclusion through several connected mechanisms: hypoxia reinforces activated fibroblast and SPP1⁺/M2-like macrophage programs; VEGF generates abnormal vessels and strengthens myeloid immunosuppression; low perfusion reduces drug and lymphocyte delivery; and tumor endothelial cells can restrict T-cell adhesion and transmigration. The result is a spatially uneven immune landscape in which CD8⁺ T cells may remain perivascular, stromal, or marginal, while hypoxic tumor nests and macrophage-rich matrix regions remain poorly accessible. Therapeutically, these observations support combining PD-1/PD-L1 blockade with vascular normalization, VEGF/VEGFR inhibition, macrophage reprogramming, or matrix modulation in selected gastric cancers with hypoxic, angiogenic, and stromal–myeloid signatures.

Metabolic and Extracellular-Vesicle Communication

Metabolic and extracellular-vesicle communication adds a less visible but highly dynamic layer to fibroblast–macrophage spatial niches. Unlike chemokine recruitment or ECM remodeling, these signals do not always map neatly onto fixed tissue borders. They are shaped by local hypoxia, nutrient availability, macrophage state, fibroblast activation, treatment pressure, and regional perfusion. Therefore, lactate, lipid metabolites, amino acid flux, nicotinamide metabolism, and extracellular vesicles should be interpreted as context-dependent regulators of the gastric cancer microenvironment rather than uniformly protumor signals.

Lactate is a representative example of this complexity. In gastric cancer, glycolytic tumor cells can release lactate into the microenvironment, where lactate uptake by macrophages through monocarboxylate transporters activates HIF-1α and promotes M2-like macrophage polarization.⁹¹ Lactate may also act through epigenetic mechanisms. Histone lactylation was originally described as a lactate-derived modification capable of regulating gene transcription in macrophages, including late activation of tissue-repair-like programs.⁹² These mechanisms provide a plausible link between hypoxic glycolytic tumor nests and macrophage polarization within CAF-rich niches. However, lactate should not be described as uniformly immunosuppressive. In preclinical models, lactate separated from acidic stress can increase stem-like CD8⁺ T-cell states and augment antitumor immunity.⁹³ Thus, the biological effect of lactate depends on concentration, pH, cellular transporter expression, oxygen tension, and the responding cell type.

Lipid metabolism provides another axis of macrophage adaptation. TREM2⁺ lipid-associated macrophages in gastric cancer display lipid-handling features, including molecules such as FABP4, FABP5, and FASN, and can acquire immunosuppressive properties in cooperation with CAF-derived CXCL12–CXCR4 signaling.³⁷ This suggests that lipid-rich or hypoxic stromal regions may select for macrophages capable of lipid uptake, oxidative stress adaptation, and suppressive cytokine production. Nevertheless, lipid metabolism is not intrinsically protumor. Some lipid-handling programs may support efferocytosis, tissue repair, or inflammatory resolution, whereas excessive lipid accumulation, oxidized lipids, or ferroptosis-prone states may impair CD8⁺ T cells and favor myeloid suppression. The key issue is therefore not whether lipid metabolism is “good” or “bad”, but whether a given lipid-associated macrophage state is coupled to immune exclusion, matrix remodeling, or restoration of tissue homeostasis.

Amino acid metabolism also influences the balance between antitumor immunity and suppressive myeloid programming. Tryptophan catabolism through IDO1/TDO2 can deplete local tryptophan and generate kynurenine, which suppresses effector T-cell activity and supports regulatory or tolerogenic immune states. Arginine metabolism can similarly shape immunity: ARG1⁺ myeloid cells consume extracellular arginine, weakening T-cell receptor signaling and favoring macrophage programs associated with repair or suppression. Glutamine and methionine metabolism further affect tumor growth, epigenetic regulation, and immune-cell function in gastric cancer.⁹⁴ These pathways are relevant to fibroblast–macrophage niches because dense stromal regions often combine hypoxia, poor perfusion, amino acid competition, and myeloid accumulation. However, amino acid targeting must be approached cautiously because effector T cells also require amino acids for proliferation, cytokine production, and memory formation. A therapy that starves tumor cells may also starve immune cells if not properly timed or targeted.

Nicotinamide metabolism is particularly important because it provides direct evidence of metabolic competition between macrophages and fibroblasts in gastric cancer. Jiang et al identified a “metabolic face-off” in which NAMPT-expressing macrophages and NNMT-expressing fibroblasts regulate the nicotinamide/1-methylnicotinamide ratio and thereby influence CD8⁺ T-cell function. Mechanistically, NNMT in CAFs is transcriptionally activated by NOTCH pathway signaling through RBP-J, whereas macrophage-derived extracellular vesicles containing NAMPT suppress NNMT expression in CAFs through the SIRT1/NICD axis. Importantly, manipulating this pathway with NAMPT-containing extracellular vesicles restored CD8⁺ T-cell cytotoxicity and enhanced anti-PD-1 response in gastric cancer models.⁴³ This finding is conceptually valuable because it shows that macrophage–fibroblast crosstalk is not always suppressive; depending on the metabolic direction of the signal, macrophage-derived EVs can restrain CAF-mediated immune inhibition.

Extracellular vesicles should therefore be considered bidirectional carriers of niche information rather than simple tumor-promoting particles. Gastric cancer-derived exosomes can induce PD1⁺ macrophages with M2-like and CD8⁺ T-cell-suppressive features.⁹⁵ Conversely, macrophage-derived vesicles may have opposite effects depending on their source state and cargo. M2 TAM-derived exosomal MALAT1 promotes glycolysis, metastasis, and chemoresistance in gastric cancer cells, creating a feedback loop in which macrophages reinforce tumor metabolic adaptation.⁹⁶ In contrast, M1 macrophage-derived exosomal miR-16-5p downregulates PD-L1 in gastric cancer cells and enhances T-cell-dependent antitumor immunity.⁹⁷ These examples illustrate why EV-mediated communication cannot be assigned one biological direction. The same delivery format—an extracellular vesicle—can either strengthen immune suppression or restore antitumor activity depending on its cellular origin, cargo, and recipient cell.

Taken together, metabolic and EV-mediated communication transforms CAF–macrophage niches into adaptive ecosystems. Lactate can promote macrophage polarization but may also support selected CD8⁺ T-cell states under different conditions. Lipid metabolism can sustain TREM2⁺ suppressive macrophages but may also reflect tissue-repair adaptation. Amino acid metabolism can suppress T cells through tryptophan or arginine depletion, yet broad metabolic deprivation may harm antitumor lymphocytes. NAMPT–NNMT signaling demonstrates a rare but important antitumor direction of macrophage–fibroblast metabolic competition. EVs, likewise, may carry either suppressive or immune-restorative cargo. For this reason, these mechanisms should be framed as context-dependent modulators of spatial niches, not as uniformly protumor pathways. Their therapeutic value will depend on identifying which metabolic or vesicular signal dominates in a particular gastric cancer tissue region and whether intervention can selectively weaken suppressive stromal–myeloid loops without disabling effector T-cell metabolism.

Fibroblast–Macrophage Niches and Therapeutic Resistance

Resistance to Immune Checkpoint Blockade

Immune checkpoint blockade has improved the treatment landscape of advanced gastric cancer, but its benefit remains heterogeneous and often transient. From the perspective of fibroblast–macrophage niches, resistance to PD-1/PD-L1 blockade should not be viewed only as failure of checkpoint inhibition at the T-cell surface. In many tumors, the more fundamental problem is that the tissue architecture lacks a sufficient pool of tumor-reactive CD8⁺ T cells that can be reactivated, or that these cells are physically and functionally separated from malignant epithelial nests. Fibroblast–macrophage niches contribute to this process by combining stromal exclusion, myeloid suppression, vascular dysfunction, and compensatory innate immune signaling.

Primary resistance occurs when the pretreatment tumor microenvironment is already poorly configured for checkpoint response. In this situation, the tumor core may contain few CD8⁺ T cells with preserved effector potential, even when immune cells are present elsewhere in the tissue. Multidimensional immune profiling of gastric cancer has shown that response to anti-PD-1/PD-L1 therapy depends on the density, phenotype, and spatial organization of tumor-infiltrating immune cells, including CD8⁺ T-cell subsets and CD68⁺ myeloid cells, rather than on a single immune marker.⁷⁵ This observation is important because PD-1 blockade requires pre-existing or newly recruited tumor-reactive T cells. If CD8⁺ T cells are absent from the tumor core, trapped at the stromal edge, or replaced by macrophage-rich immune barriers, there may be little effector immunity for checkpoint blockade to restore.

ECM and vascular barriers further strengthen primary resistance. CAF-derived collagen, fibronectin, hyaluronan, and matricellular proteins can restrict T-cell penetration, while abnormal VEGF-driven vessels reduce perfusion and limit lymphocyte trafficking. In gastric cancer peritoneal metastasis, hyaluronic acid–CD44 signaling has been linked to therapeutic resistance and an immunosuppressive microenvironment, and combined CD44/PD-1 blockade enhanced therapeutic efficacy in experimental models.⁹⁸ Although this pathway is not identical to the THBS2⁺ CAF–SPP1⁺ TAM niche, it reinforces the broader concept that matrix-rich metastatic lesions can limit immune access and reduce checkpoint efficacy. In stromal tumors, therefore, resistance may arise before PD-1 signaling is even therapeutically relevant: the T cell has not reached the malignant cell, or the local vascular and matrix architecture does not permit effective immune-cell delivery.

Myeloid dominance is another major form of primary resistance. Several gastric cancer studies now point to suppressive macrophages, MDSC-recruiting myeloid programs, and macrophage–T-cell inhibitory communication as determinants of poor ICB response. A stromal immune barrier analysis identified SPP1⁺/C1QC⁺ macrophages and exhausted CD8⁺ T cells as a resistance-associated cellular network, with macrophage-derived MIF signaling through CD74/CXCR4/CD44 proposed as an immunosuppressive route.³⁵ Separately, USP14⁺ TAMs were shown to promote anti-PD-1 resistance in gastric cancer by stabilizing IMP2 and inducing CXCL2-mediated recruitment of myeloid-derived suppressor cells, thereby reducing CD8⁺ T-cell infiltration.⁹⁹ These findings suggest that some gastric cancers are not simply immune-cold; they are actively myeloid-conditioned. In such tumors, PD-1 blockade alone may be insufficient because suppressive macrophages and related myeloid populations define the dominant immune logic of the tissue.

Adaptive resistance reflects a different process. Here, checkpoint blockade may initially activate antitumor inflammation, but the tumor microenvironment responds by rebuilding suppressive circuits. IFN-γ and inflammatory cytokines generated after ICB can be double-edged: they can enhance antigen presentation and T-cell activity, yet chronic or spatially restricted IFN-γ signaling may also induce counter-regulatory pathways, checkpoint ligands, stromal activation, and adaptive immune escape.¹⁰⁰ In non-gastric cancer models, CAF-rich or TGF-β-activated stromal programs have been shown to exclude CD8⁺ T cells and confer resistance to checkpoint blockade, supporting the broader plausibility that inflammatory pressure can reinforce fibroblast-dependent immune barriers.^13,101 In gastric cancer, this possibility should be stated cautiously: direct longitudinal proof that ICB itself activates a specific CAF state remains limited, but treatment-pressure studies increasingly show that stromal–myeloid programs can dominate in therapy-failure lesions.

Myeloid recruitment and complement signaling may also compensate after immune activation. In gastric cancer peritoneal metastasis, treatment-failure samples showed higher C1Q⁺ macrophage infiltration and lower GZMA⁺ T-lymphocyte infiltration, with ligand–receptor analyses implicating LGALS9–CD45 and SPP1–CD44 interactions.⁶¹ These data suggest that after treatment pressure, the peritoneal microenvironment may restore immune suppression by enriching macrophage states that reduce cytotoxic T-cell function. Complement may participate in this compensation. Tumor-intrinsic complement C3 has been reported to promote PD-L1 antibody resistance by modulating TAM function in cancer models, while in gastric cancer peritoneal metastasis CAF-derived C3 drives C3AR1-dependent accumulation of SPP1⁺ TAMs and impaired response to ICB.^17,64 Therefore, adaptive resistance may involve not only upregulation of alternative checkpoints, but also reconstitution of stromal–myeloid niches that re-sequester T cells and reestablish macrophage-dominated suppression.

Site-specific resistance is especially important in gastric cancer because metastatic organs impose distinct stromal and immune rules. Peritoneal metastasis provides the most persuasive example. Compared with primary gastric tumors, gastric cancer peritoneal metastases develop a unique immunosuppressive ecosystem dominated by THBS2⁺ matrix CAFs and SPP1⁺ TAMs. Mechanistically, THBS2⁺ matrix CAFs release complement C3, which acts on C3AR1⁺ macrophages and promotes recruitment and transition of peritoneum-specific tissue-resident macrophage-derived cells into SPP1⁺ TAMs. These SPP1⁺ TAMs are associated with immune checkpoint molecules, matrix remodeling, and impaired antitumor immunity. Importantly, blockade of the C3–C3AR1 axis disrupted this stromal–myeloid circuit and enhanced immune checkpoint blockade efficacy in experimental models.¹⁷ This evidence makes the THBS2⁺ matrix CAF–SPP1⁺ TAM niche one of the strongest currently available Level III examples in gastric cancer.

The therapeutic implication is that ICB resistance should be stratified spatially and anatomically. A primary gastric tumor with low CD8⁺ T-cell entry may require T-cell recruitment or vascular normalization strategies; a myeloid-rich tumor may require macrophage reprogramming, CSF1R/CCR2/MIF/SPP1-targeted approaches, or MDSC suppression; and a peritoneal metastasis with a THBS2⁺ CAF–SPP1⁺ TAM–C3/C3AR1 program may require complement or matrix–myeloid targeting combined with PD-1/PD-L1 blockade. In this framework, fibroblast–macrophage niches are not peripheral modifiers of ICB response. They determine whether checkpoint blockade has access to the correct cellular substrate: intratumoral, tumor-reactive, functionally recoverable CD8⁺ T cells.

Emerging Evidence in Chemotherapy Resistance

Compared with immune checkpoint blockade, the role of fibroblast–macrophage niches in gastric cancer chemotherapy resistance is less spatially resolved. Current evidence is mainly assembled from three sources: studies showing that CAFs independently promote resistance to fluoropyrimidine- or platinum-based chemotherapy; studies showing that TAMs independently enhance chemoresistance through cytokines, exosomes, and survival pathways; and studies from other cancers demonstrating that CAF–TAM cooperation can protect tumor cells from cytotoxic injury. Therefore, chemotherapy resistance should be discussed as an emerging and biologically plausible function of fibroblast–macrophage niches, rather than as a fully validated spatial mechanism in gastric cancer.

The first mechanism is restricted drug access. Dense collagen, fibronectin, hyaluronan, activated fibroblasts, and abnormal vasculature can reduce the effective distribution of cytotoxic agents within tumor tissue. In gastric cancer, activated gastric cancer-associated fibroblasts have been shown to promote malignant phenotypes and decrease 5-FU sensitivity through paracrine mechanisms.¹⁰² CAF-derived IL-8 also mediates cisplatin resistance in gastric cancer, and high pretreatment serum IL-8 was associated with poor response to platinum-based therapy in patients receiving neoadjuvant chemotherapy.¹⁰³ These findings indicate that fibroblast-rich stroma does not merely dilute tumor-cell content in a specimen; it can function as a drug-protective compartment through both physical and soluble-factor-mediated mechanisms. However, direct imaging evidence showing chemotherapy concentration gradients across CAF–macrophage niches in gastric cancer remains limited.

A second mechanism involves survival signaling after drug exposure. CAF-derived IL-6 activates the JAK1–STAT3 pathway in gastric cancer cells and allows tumor cells to resist apoptosis induced by chemotherapy; blockade of IL-6 signaling restored chemosensitivity in experimental models.²⁷ Similar survival logic is seen on the macrophage side. In gastric cancer, M2-polarized macrophages enhance 5-FU resistance through CXCL5-mediated activation of the PI3K/AKT/mTOR pathway, while M2 macrophage-derived exosomal miR-21, miR-588, circTEX2, and circ_0008253 have each been reported to increase resistance to cisplatin or oxaliplatin through distinct downstream targets.^104–108 These studies strongly support TAM-mediated chemoresistance, but most do not resolve whether the responsible macrophages are physically organized with CAF subsets in patient tissues. The likely scenario is that CAFs and TAMs converge on common anti-apoptotic routes—STAT3, PI3K/AKT, NF-κB, ABC transporters, autophagy, and EMT—but the spatial architecture of this convergence remains insufficiently mapped.

An additional possibility is that CAF-derived IL-6 may promote chemoresistance by sustaining stem-like, drug-tolerant tumor-cell states. Gastric cancer stem cells, identified using markers such as CD44 and related stemness signatures, exhibit enhanced tumorigenicity, self-renewal, and resistance to chemotherapy.^109,110 Separately, CAF-derived IL-6 has been shown to activate JAK1–STAT3 signaling in gastric cancer cells, inhibit chemotherapy-induced apoptosis, and reduce responsiveness to 5-fluorouracil and cisplatin.²⁷ These parallel findings raise the possibility that IL-6-rich CAF compartments may support a stemness-associated reservoir of residual tumor cells during chemotherapy. Within fibroblast–macrophage spatial niches, this effect could theoretically be reinforced by TAM-derived cytokines, growth factors, and matrix-remodeling signals that converge on STAT3-related survival programs. However, direct evidence that CAF-derived IL-6 induces chemoresistance specifically through gastric cancer stemness, or that macrophages cooperate in this process within a spatially defined niche, remains limited. This proposed mechanism should therefore be distinguished from the directly demonstrated anti-apoptotic effect of CAF-derived IL-6 and validated using spatial colocalization of CAFs, TAMs, and stemness markers, together with functional IL-6/STAT3 blockade, tumorsphere assays, lineage tracing, and rescue experiments.

TGF-β signaling, hypoxia, EMT, and stemness provide a third layer of resistance. In gastric cancer, hyaluronan-mediated motility receptor promotes chemotherapy resistance through TGF-β/Smad2-induced EMT and cancer stem-like properties.¹¹¹ Hypoxic conditions also induce EMT, increase stem-like features, and enhance invasive behavior in gastric cancer cells.¹¹² These effects are highly compatible with fibroblast–macrophage niches: CAFs are major sources and responders of TGF-β, while TAMs are enriched in hypoxic and tissue-repair-like regions. A dense stromal–myeloid niche may therefore select for tumor cells that are less proliferative, more mesenchymal, more stem-like, and more tolerant of cytotoxic stress. Yet here again, the evidence remains inferential. Gastric cancer studies have documented hypoxia-associated EMT and TGF-β-associated chemoresistance, but few have simultaneously mapped CAF subsets, macrophage states, chemotherapy penetration, and resistant tumor-cell clones in the same spatial framework.

Chemotherapy may also remodel the microenvironment after initial tumor injury. Cytotoxic treatment kills sensitive tumor cells, damages stromal tissue, releases inflammatory signals, and activates wound-repair programs. In gastric cancer, longitudinal single-cell sequencing before and after neoadjuvant chemotherapy showed broad reconstruction of the tumor microenvironment, including altered malignant, stromal, and immune cell states.¹¹³ A separate histological study of paired pre- and post-neoadjuvant chemotherapy gastric cancer samples reported that post-treatment CD8⁺⁺ and CD86⁺M1⁺M2-like macrophage ratios were linked to pathological response, indicating that immune and macrophage remodeling after chemotherapy is clinically meaningful.¹¹⁴ These findings suggest that chemotherapy does not simply reduce tumor burden; it can reshape the ecological balance of fibroblasts, macrophages, lymphocytes, and residual malignant cells. In resistant lesions, damage repair may favor re-enrichment of CAFs and macrophages, secretion of SASP-like cytokines, and reconstruction of matrix-rich survival niches.

Evidence from other cancers strengthens this interpretation but should be used as mechanistic support rather than direct gastric cancer proof. CAFs and macrophages are well-recognized mediators of microenvironment-driven chemoresistance, acting through soluble factors, cell adhesion, ECM remodeling, metabolic reprogramming, exosomes, and therapy-induced repair responses.¹¹⁵ In breast cancer models, blockade of macrophage recruitment pathways combined with chemotherapy improved outcome through CD8⁺ T-cell-dependent mechanisms, showing that macrophage-rich repair programs can limit cytotoxic therapy and that reprogramming the myeloid compartment can improve chemotherapy efficacy.¹¹⁶ More recently, CSF1R-targeted strategies combined with chemotherapy or immunochemotherapy have further supported the principle that macrophage-dominant tumors may require myeloid modulation to achieve durable cytotoxic response.¹¹⁷ These non-gastric findings are relevant because gastric cancer contains similar CAF and TAM programs, but they do not substitute for gastric cancer-specific spatial validation.

Overall, chemotherapy resistance in gastric cancer is likely shaped by overlapping stromal and myeloid mechanisms: CAF-rich matrix limits drug access; IL-6–STAT3, IL-8, TGF-β, and CXCL5–PI3K/AKT/mTOR signaling support tumor-cell survival; hypoxia drives EMT and stemness; and treatment-induced injury may rebuild CAF- and macrophage-rich repair niches. The field now needs spatial pharmacology and longitudinal tissue studies that map cytotoxic drug distribution, CAF states, macrophage states, and residual tumor-cell phenotypes before and after chemotherapy. Until such data are available, fibroblast–macrophage niches should be considered plausible contributors to chemotherapy resistance in gastric cancer, but not yet fully established spatial determinants.

Implications for Targeted and Anti-Angiogenic Therapy

Fibroblast–macrophage niches may also influence targeted and anti-angiogenic therapy, although the evidence is less mature than that for immune checkpoint blockade. In gastric cancer, anti-VEGFR2 therapy with ramucirumab has established clinical benefit in the second-line setting, particularly in combination with paclitaxel.⁷ Biologically, however, inhibition of angiogenic signaling is not a simple process of “cutting off blood supply”. Depending on dose, timing, and baseline vascular structure, anti-angiogenic therapy may transiently normalize vessels and improve immune-cell or drug delivery, but excessive or sustained vascular pruning may worsen hypoxia, increase HIF-dependent escape signals, and promote recruitment of pro-angiogenic myeloid cells.^118,119 In CAF- and TAM-rich gastric tumors, this creates a plausible resistance scenario: VEGF pathway inhibition may reduce some vascular support, while hypoxia and stromal cytokines compensate by reinforcing macrophage recruitment, angiogenic TAM states, and alternative endothelial programs. Direct spatial evidence for this sequence after ramucirumab in gastric cancer remains limited, so it should be regarded as a testable microenvironmental hypothesis rather than an established resistance mechanism.

The stromal barrier is also relevant to antibody-based agents, antibody–drug conjugates (ADCs), and other large-molecule therapeutics. Macromolecular drugs must extravasate from vessels, diffuse or convect through interstitial space, bind antigen-bearing cells, and, in the case of ADCs, undergo appropriate internalization, trafficking, and payload release. Dense collagen, hyaluronan, fibronectin, high interstitial pressure, abnormal vessels, and long diffusion distances can reduce the uniformity of antibody distribution in solid tumors.^120,121 This principle is especially important for gastric cancer because several clinically relevant targets—HER2, CLDN18.2, and FGFR2b—are expressed heterogeneously across tumor regions, histological patterns, and metastatic sites. Therefore, even when a tumor is biomarker-positive, a fibrotic or poorly perfused stromal–myeloid compartment may still limit effective target engagement in some regions. At present, however, this remains better supported by general solid-tumor pharmacology than by gastric cancer-specific spatial pharmacokinetic studies.

CAF-derived or stromal-associated bypass signals provide a second possible route of resistance. In HER2-positive gastric cancer, acquired trastuzumab resistance has been linked to upregulation of HER-family ligands, which can maintain downstream MAPK/ERK and PI3K/mTOR signaling despite HER2 blockade.¹²² A separate gastric cancer study showed that tumor-derived CCL2 reshaped TAMs and promoted trastuzumab resistance; CD40×HER2 bispecific antibody treatment restored macrophage activation and overcame resistance in experimental systems.¹²³ Another study reported that PFKFB3-driven CXCL8 secretion was associated with vascular abnormality and trastuzumab resistance, and PFKFB3 inhibition improved trastuzumab sensitivity partly through vessel normalization.¹²⁴ These findings do not prove that CAF–macrophage niches universally drive HER2 resistance, but they show that growth-factor compensation, chemokine-driven macrophage remodeling, and vascular dysfunction can all reduce the effectiveness of HER2-directed therapy.

For CLDN18.2-directed therapy, the microenvironmental question is different. Zolbetuximab depends in part on immune effector mechanisms such as antibody-dependent cellular cytotoxicity, involving NK cells and monocytes/macrophages. Recent gastric cancer data indicate that CLDN18.2 expression is associated with selected molecular and immune features, but tumor-infiltrating NK-cell and macrophage frequencies are not uniformly increased in all CLDN18.2-positive tumors.^125,126 This suggests that CLDN18.2 positivity alone may not fully predict the quality of the effector-cell compartment required for antibody activity. It is reasonable to hypothesize that CAF-rich, macrophage-dominated, poorly perfused, or T-cell-restricted niches could modify responses to CLDN18.2-targeted therapy, especially in peritoneal or diffuse-type disease. However, direct evidence that a defined CAF–macrophage spatial niche causes zolbetuximab resistance is not yet available; antigen heterogeneity, target loss, tumor burden, effector-cell competence, and tissue delivery should be evaluated together.

FGFR2b therapy raises another microenvironmental issue: ligand and receptor redundancy. Bemarituzumab and other FGFR2/FGFR2b-targeted strategies are built around epithelial tumor dependence on FGFR2b signaling, but FGFR signaling is embedded within a broader network of growth factors and receptor tyrosine kinases.¹⁰ In FGFR2-amplified gastric cancer models, activation of EGFR, HER3, and MET attenuated sensitivity to FGFR2 inhibition by restoring downstream MAPK and AKT signaling.¹²⁷ Because CAFs are major producers of growth factors, including FGF-family ligands, HGF, IGF, and cytokines in many tumor contexts, it is plausible that CAF-rich gastric tumors may support bypass signaling under FGFR2b blockade. Still, this should be framed as a hypothesis: direct evidence connecting a specific CAF–macrophage niche to bemarituzumab resistance in gastric cancer remains insufficient.

Taken together, targeted and anti-angiogenic therapy resistance should be interpreted through three connected but not yet fully proven microenvironmental principles. First, anti-VEGF/VEGFR therapy may create a balance between vascular normalization and hypoxia-driven myeloid compensation. Second, dense ECM and abnormal vessels may limit delivery of antibodies, ADCs, and other large molecules, even when the epithelial target is present. Third, CAF- and macrophage-derived signals may provide bypass survival cues that allow tumor cells to maintain downstream signaling after HER2, CLDN18.2, or FGFR2b targeting. The practical implication is that future biomarker strategies should not stop at epithelial target expression. They should also assess stromal density, vascular function, macrophage state, hypoxia, and regional target heterogeneity to determine whether a targeted agent can actually reach and act on its intended tumor-cell compartment. The major actionable signaling nodes within fibroblast–macrophage niches and their potential therapeutic consequences are summarized in Figure 3.

Figure 3.

Proposed actionable targets within fibroblast–macrophage niches in gastric cancer. This hypothesis-generating schematic summarizes major candidate therapeutic nodes within gastric cancer cancer-associated fibroblast (CAF)–tumor-associated macrophage (TAM) niches and their proposed downstream effects. The targets are grouped into three intervention categories. Recruitment and maintenance-related targets include: CCL2–CCR2 blockade to reduce monocyte recruitment; CSF1–CSF1R inhibition to reduce TAM survival and maintenance; and disruption of C3–C3AR1 signaling to interfere with complement-driven TAM accumulation, particularly within peritoneal metastatic niches. Macrophage signaling and reprogramming-related targets include: IL-6–STAT3 inhibition to attenuate suppressive macrophage states; and SPP1–integrin pathway inhibition to weaken matrix-associated TAM signaling. CAF- and extracellular matrix (ECM)-directed targets include: FAP-directed strategies targeting activated fibroblasts; LOX inhibition to reduce collagen crosslinking; and modulation of TGF-β signaling to normalize, rather than indiscriminately eliminate, activated stroma. The proposed downstream effects comprise: reduced myeloid recruitment; macrophage reprogramming; attenuation of the ECM barrier; improved CD8⁺ T-cell entry into tumor regions; and enhanced drug delivery and immune checkpoint blockade (ICB) responsiveness. These outcomes represent anticipated therapeutic consequences rather than uniformly demonstrated clinical effects. Among the illustrated pathways, the THBS2⁺ matrix CAF–C3/C3AR1–SPP1⁺ TAM circuit in gastric cancer peritoneal metastasis currently has the strongest spatial and functional evidence, whereas the evidentiary support for the other targets varies from gastric cancer-specific experimental findings to mechanistic inference or evidence extrapolated from other tumor types. Accordingly, these interventions require biomarker-guided experimental evaluation and prospective clinical validation. Potential safety concerns associated with systemic fibroblast depletion, macrophage depletion, chemokine inhibition, complement blockade, and broad TGF-β pathway inhibition should also be considered.

The Niche as a Therapeutic Sanctuary

A useful way to integrate the above mechanisms is to view fibroblast–macrophage niches as therapeutic sanctuaries. In this model, the niche is not defined by one ligand, one CAF marker, or one macrophage subset. Rather, it represents a localized tissue state in which several protective conditions converge: low perfusion, dense extracellular matrix, restricted immune pressure, and abundant survival signals. Within such regions, residual cancer cells may be exposed to lower concentrations of cytotoxic drugs or large-molecule therapeutics, encounter fewer functional CD8⁺ T cells, and receive compensatory signals from CAFs, macrophages, endothelial cells, and matrix ligands.

This model helps explain why spatial organization matters. A resistant cell is not always intrinsically resistant at baseline. It may survive because it is positioned in a protected microenvironment. Dense collagen and fibronectin restrict drug and T-cell access; abnormal vessels and elevated interstitial pressure reduce perfusion; hypoxia promotes EMT, stemness, and metabolic adaptation; and CAF- or macrophage-derived IL-6, TGF-β, CSF1, SPP1, VEGF, CXCL12, and complement signals create anti-apoptotic and immune-suppressive support. In this sense, the CAF–macrophage niche functions like a local ecological shelter, allowing tumor cells that would otherwise be vulnerable to persist through therapy.

The concept is supported by broader cancer biology. Microenvironment-mediated drug resistance has long been recognized as a contributor to minimal residual disease and acquired resistance, particularly when stromal cells, extracellular matrix proteins, and soluble factors protect tumor cells from therapeutic injury.¹²⁸ Physical barriers in solid tumors also impair drug penetration, especially for macromolecules and antibody-based agents.¹²¹ Hypoxia adds another layer by promoting therapy resistance, immune evasion, angiogenic escape, and metabolic plasticity.¹²⁹ These principles are not gastric cancer-specific, but they provide the mechanistic foundation for interpreting stromal–myeloid compartments as survival niches rather than passive histological background.

Gastric cancer-specific evidence is beginning to fit this model. Longitudinal single-cell sequencing after neoadjuvant chemotherapy showed that the gastric cancer microenvironment is reconstructed after treatment, with changes in malignant, stromal, and immune cell states.¹¹³ Multiplex immunofluorescence and T-cell receptor repertoire analysis further demonstrated that residual tumors after neoadjuvant chemotherapy have altered immune-cell composition and immune repertoire features.¹³⁰ In peritoneal metastasis, the THBS2⁺ matrix CAF–SPP1⁺ TAM niche provides a particularly strong example: it combines matrix remodeling, complement-mediated macrophage recruitment, immune suppression, and poor response to immune checkpoint blockade; interruption of C3–C3AR1 signaling improved checkpoint response in experimental models.¹⁷ Together, these findings suggest that residual disease in gastric cancer may be maintained not only by resistant tumor clones, but also by treatment-selected tissue niches that favor survival and immune escape.

This sanctuary model also clarifies the origin of recurrence and acquired resistance. After treatment, most sensitive tumor cells may be eliminated, but cells protected in low-perfusion, high-matrix, low-immune-pressure regions can remain viable. Over time, these cells may undergo clonal selection, acquire additional genetic or epigenetic resistance programs, and re-expand within a repaired stromal bed. The niche itself may also evolve: chemotherapy-induced injury, ICB-induced inflammation, or anti-angiogenic pressure can stimulate wound-repair programs, macrophage recruitment, complement activation, and CAF remodeling. Thus, recurrence may arise from the interaction between residual cancer cells and a reconstituted protective microenvironment, rather than from tumor-cell-intrinsic resistance alone.

Clinically, this model argues for a shift in biomarker design. Measuring epithelial targets, PD-L1, MSI status, or tumor mutation burden is not sufficient if the tissue contains sanctuary regions where drugs and effector lymphocytes cannot act effectively. Future assessment should integrate spatial markers of perfusion, ECM density, CAF state, macrophage state, hypoxia, and CD8⁺ T-cell localization. Therapeutically, the goal should not be indiscriminate stromal depletion, which may damage normal repair and immune homeostasis, but selective dismantling of the protective features of the niche: improving perfusion, normalizing matrix, reprogramming suppressive macrophages, blocking compensatory survival pathways, and moving functional T cells into tumor nests. In this framework, CAF–macrophage niches are not merely correlates of poor prognosis; they are candidate tissue-level reservoirs from which minimal residual disease, relapse, and acquired therapeutic resistance may emerge.

Actionable Targets and Therapeutic Strategies

Blocking Monocyte and Macrophage Recruitment

Blocking monocyte and macrophage recruitment is one of the most direct strategies for dismantling fibroblast–macrophage niches. The rationale is straightforward: many suppressive TAM populations are continuously replenished by circulating monocytes or maintained by local macrophage survival signals. If monocyte entry, macrophage retention, or macrophage state conversion can be interrupted, a stromal–myeloid niche may become less suppressive and more permissive to CD8⁺ T-cell infiltration and immune checkpoint blockade. In gastric cancer, the most relevant recruitment or maintenance axes are CCL2–CCR2, CSF1–CSF1R, and C3–C3AR1. These pathways should not be considered interchangeable: CCL2–CCR2 mainly controls monocyte trafficking, CSF1–CSF1R controls macrophage survival and differentiation, and C3–C3AR1 appears particularly important for macrophage accumulation and state conversion in peritoneal metastatic niches.

The CCL2–CCR2 axis is the prototypical route for recruiting CCR2⁺ inflammatory monocytes into tumor tissue. In CAF-rich gastric cancer regions, CCL2 may help draw circulating monocytes into a stromal field where IL-6, CSF1, hypoxia, and matrix cues subsequently program them toward suppressive macrophage states. Therapeutically, this axis can be targeted either by neutralizing CCL2 itself or by blocking CCR2 on monocytes. Carlumab, a human anti-CCL2 monoclonal antibody, was evaluated in early-phase solid tumor studies but showed limited antitumor activity, partly because durable suppression of free CCL2 was difficult to maintain.¹³¹ Small-molecule CCR2 inhibitors have shown more encouraging pharmacodynamic logic in pancreatic cancer. PF-04136309 combined with FOLFIRINOX reduced circulating inflammatory monocytes and showed clinical activity in borderline resectable or locally advanced pancreatic cancer, although this has not established a gastric cancer indication.¹³² More recently, BMS-813160, a dual CCR2/CCR5 antagonist, was tested with chemotherapy with or without nivolumab in advanced pancreatic and colorectal cancers; the combination was tolerable, but clinical benefit was inconsistent across cohorts.¹³³ These experiences suggest that CCR2 blockade is feasible, but its activity likely depends on tumor type, baseline monocyte dependence, treatment partner, and biomarker selection.

For gastric cancer, CCL2–CCR2 blockade is most rational in tumors with a demonstrable CCL2⁺ fibroblast–macrophage program rather than in unselected patients. Potential combinations include CCR2 inhibition plus PD-1/PD-L1 blockade, CCR2 inhibition plus chemotherapy, or CCR2 inhibition plus anti-angiogenic therapy. The ICB rationale is that fewer recruited suppressive monocytes may reduce TAM-mediated T-cell inhibition and allow checkpoint blockade to act on a more favorable immune substrate. However, the risk of rebound recruitment must be taken seriously. In mouse breast cancer models, cessation of CCL2 inhibition accelerated metastasis through rebound monocyte mobilization, angiogenesis, and IL-6/VEGF-dependent effects.⁵⁰ This does not mean CCL2–CCR2 blockade should be abandoned, but it argues for careful scheduling, tapering or continuous blockade strategies, and combination designs that prevent compensatory myeloid or angiogenic escape.

CSF1–CSF1R targeting addresses a different stage of the same problem. CSF1R is expressed on monocytes and macrophages and supports TAM survival, differentiation, and tissue retention. In gastric cancer, high CSF1 expression has been associated with poor prognosis, and CSF1-rich tumor regions may help sustain macrophage-dominant immune suppression.⁴⁸ Drug classes targeting this pathway include anti-CSF1R monoclonal antibodies, such as emactuzumab, cabiralizumab, and AMG 820, and small-molecule tyrosine kinase inhibitors, such as pexidartinib, PLX7486, and BLZ945-like agents. Preclinically, CSF1R blockade can deplete or reprogram TAMs and improve T-cell infiltration or checkpoint responsiveness in several tumor models.^70,71 Clinically, however, the results have been more modest. Emactuzumab depleted immunosuppressive M2-like macrophages in advanced solid tumors but produced limited objective antitumor activity as monotherapy or in combination with paclitaxel.¹³⁴ Pexidartinib and other CSF1R inhibitors also demonstrate that macrophage targeting is biologically active, yet antitumor efficacy in common epithelial cancers has been inconsistent.

The translational lesson is that CSF1R blockade should not be used as nonspecific macrophage depletion. Macrophages are heterogeneous, and some macrophage states may support antigen presentation, tissue repair, or response to therapy. Excessive or poorly timed CSF1R inhibition may also impair dendritic-cell biology or normal host-defense functions. Therefore, in gastric cancer, the most plausible setting for CSF1R-targeted therapy is a macrophage-rich, T-cell-restricted tumor with evidence of CSF1–CSF1R dependence. Combination with PD-1/PD-L1 blockade is rational because CSF1R inhibition may reduce suppressive macrophage survival and create space for effector T-cell activity. Combination with chemotherapy may also be reasonable because chemotherapy-induced tissue injury can recruit macrophages into repair programs. Nevertheless, future trials should include on-treatment biopsies or spatial biomarkers to determine whether CSF1R blockade actually reduces suppressive TAM niches, improves CD8⁺ T-cell localization, or merely alters macrophage markers without changing tissue immunity.

The C3–C3AR1 axis is particularly important in gastric cancer peritoneal metastasis. Unlike CCL2–CCR2 or CSF1–CSF1R, which are broadly relevant across tumors, C3–C3AR1 currently has its strongest gastric cancer evidence in the peritoneal metastatic setting. THBS2⁺ matrix CAFs release complement C3, which acts on C3AR1⁺ macrophages and promotes the accumulation and transition of peritoneum-specific macrophages into SPP1⁺ TAMs.¹⁷ This axis is highly actionable because it links a defined CAF state, a macrophage receptor, a metastatic site, and resistance to immune checkpoint blockade. In experimental models, blockade of C3–C3AR1 disrupted the stromal–myeloid circuit and improved antitumor response to checkpoint therapy.⁹ Broader cancer studies also support complement C3 as a mediator of TAM-driven immunosuppression and resistance to PD-L1 blockade through C3a–C3aR signaling.⁶⁴

Available complement-targeting approaches include upstream C3 inhibition, C3a/C3aR blockade, and downstream C5a/C5aR inhibition. Clinically approved C3 inhibitors, such as pegcetacoplan, are established mainly in complement-mediated non-malignant diseases rather than gastric cancer, while selective C3aR-targeted strategies remain largely preclinical in oncology.¹³⁵ C5aR-targeting antibodies, such as avdoralimab, illustrate the feasibility of targeting complement receptors on myeloid cells, but C5aR blockade should not be assumed to substitute for C3AR1 blockade because C3aR and C5aR may regulate different myeloid programs.¹³⁶ For gastric cancer, the most compelling therapeutic hypothesis is therefore not generalized complement suppression, but biomarker-selected C3–C3AR1 targeting in peritoneal metastases with THBS2⁺ matrix CAF–SPP1⁺ TAM enrichment.

Combination with immune checkpoint blockade is the most logical development path for C3–C3AR1 inhibition. In peritoneal metastasis, the problem is not only macrophage accumulation but also formation of a stromal–myeloid barrier that weakens T-cell function and reduces ICB responsiveness. Interrupting C3–C3AR1 may reduce SPP1⁺ TAM accumulation, loosen the matrix-associated immune barrier, and allow PD-1/PD-L1 blockade to act on reinvigorated T cells. However, complement is essential for antimicrobial defense, immune complex clearance, and tissue homeostasis. Systemic complement inhibition can increase infection risk, particularly with encapsulated bacteria, and may have different effects depending on whether complement activation is tumor-derived, CAF-derived, hepatic, local, or systemic. Thus, clinical translation should prioritize local pathway activity, metastatic site, macrophage state, and infection-risk management.

Overall, recruitment blockade should be developed as precision myeloid therapy rather than broad macrophage eradication. CCL2–CCR2 blockade is best suited to CCL2⁺ CAF-rich tumors with active monocyte influx, but rebound recruitment and chemokine redundancy are major concerns. CSF1–CSF1R blockade is best suited to macrophage-survival-dependent tumors, but depletion alone has shown limited efficacy and may impair beneficial myeloid functions. C3–C3AR1 blockade is most compelling in gastric cancer peritoneal metastasis, where the THBS2⁺ matrix CAF–SPP1⁺ TAM niche provides a mechanistic and treatment-response rationale. The key translational requirement for all three approaches is the same: patient selection must be based on spatial or molecular evidence that the targeted recruitment pathway is actually sustaining the suppressive fibroblast–macrophage niche.

Reprogramming Macrophage States

Macrophage reprogramming is conceptually different from macrophage depletion. Depletion aims to remove macrophages from the tumor, whereas reprogramming seeks to convert suppressive, repair-like, or immune-excluding TAM states into macrophages that support phagocytosis, antigen presentation, inflammatory cytokine production, and T-cell activation. This distinction is important in gastric cancer because macrophages are heterogeneous and context-dependent. Some TAM states promote matrix remodeling, angiogenesis, and immune suppression, but others may participate in tumor-cell clearance, antigen presentation, and response to therapy. Therefore, indiscriminate macrophage elimination is unlikely to be the optimal long-term strategy. A more precise approach is to identify the dominant suppressive macrophage program in a given niche and redirect it toward a therapeutically useful state.

JAK–STAT3 inhibition is one of the most mechanistically grounded reprogramming strategies. In gastric cancer, IL-6 induces macrophages to acquire CD163⁺CD206⁺ M2-like features through STAT3 activation, and STAT3 knockdown reduces IL-10 and TGF-β expression in these macrophages.²⁸ Gastric cancer-derived mesenchymal stromal cells can also drive M2-like macrophage polarization through IL-6/IL-8–JAK2/STAT3 signaling, while pathway blockade attenuates this effect.⁴⁶ These findings make the IL-6–JAK–STAT3 axis attractive for CAF-rich or stromal-cell-rich tumors in which macrophage suppression is cytokine-driven. Therapeutically, this pathway can be targeted at several levels, including IL-6/IL-6R antibodies, JAK inhibitors, and direct or indirect STAT3 inhibitors. However, clinical translation in gastric cancer remains early. Because JAK–STAT signaling also supports lymphocyte function, hematopoiesis, and host defense, broad systemic inhibition may suppress the same T cells that therapy aims to rescue. The best application may be short-course, biomarker-selected, or locally delivered STAT3 pathway modulation combined with PD-1/PD-L1 blockade, chemotherapy, or vascular normalization.

TREM2-related intervention is an emerging strategy for lipid-associated and immunosuppressive TAM states. TREM2⁺ macrophages are enriched in multiple solid tumors and often mark a myeloid program associated with lipid handling, tissue repair, T-cell suppression, and poor response to immunotherapy. In gastric cancer, recent studies suggest that TREM2⁺ TAMs contribute to tumor progression and immune evasion, and that depletion or blockade of TREM2⁺ TAMs can enhance anti-PD-L1 or anti-PD-1 activity in experimental models.^38,137 Anti-TREM2 monoclonal antibodies such as PY314 have entered early-phase clinical testing in advanced solid tumors, including evaluation as monotherapy or in combination with pembrolizumab.¹³⁸ The gastric cancer implication is clear but still preliminary: TREM2 targeting may be useful in tumors with TREM2⁺ lipid-associated macrophage enrichment, especially when these macrophages coexist with CAF-derived CXCL12–CXCR4 signals or hypoxic stromal regions. Yet TREM2 biology is tissue-specific, and TREM2 can influence macrophage survival, phagocytosis, lipid processing, and inflammation. Thus, TREM2 therapy should be guided by spatial and functional biomarkers rather than by total macrophage abundance alone.

SPP1–integrin or SPP1–CD44 blockade addresses a matrix-associated macrophage program rather than a single receptor checkpoint. SPP1⁺ macrophages are repeatedly identified in gastric cancer single-cell and spatial analyses, particularly in stromal–myeloid niches and peritoneal metastasis. SPP1 can interact with integrins and CD44, promoting adhesion, migration, matrix communication, and immunoregulatory signaling. In gastric cancer, SPP1/CD44-mediated crosstalk between SPP1⁺ macrophages and epithelial tumor cells has been associated with macrophage infiltration, advanced disease, and poor survival.³³ In gastric cancer peritoneal metastasis, SPP1⁺ TAMs are central components of the THBS2⁺ matrix CAF–dominated niche that governs response to immune checkpoint blockade.¹⁷ Therapeutically, SPP1-axis targeting could involve anti-SPP1 antibodies, CD44-directed strategies, integrin inhibition, or disruption of upstream pathways that generate SPP1⁺ TAMs, such as C3–C3AR1 in peritoneal metastasis. However, this area remains less clinically mature than CD47–SIRPα or CSF1R targeting. Direct SPP1 blockade is still largely preclinical in oncology, and integrin inhibitors have historically shown limited success when used without precise patient selection. Therefore, SPP1–integrin/CD44 targeting should be framed as a promising niche-directed strategy that requires spatial biomarker development.

CD47–SIRPα blockade reprograms macrophage function by removing a phagocytic checkpoint. Tumor-cell CD47 binds SIRPα on macrophages and transmits a “do not eat me” signal that suppresses engulfment. In gastric cancer, CD47 has been reported as an adverse prognostic factor, and CD47-blocking antibodies can enhance macrophage phagocytosis of gastric cancer cells in experimental systems.¹³⁹ CD47 expression also correlates with macrophage infiltration and immune evasion features in gastric cancer tissues.¹⁴⁰ In peritoneal metastasis, galectin-3 cooperates with CD47 to suppress phagocytosis and T-cell immunity, further supporting the relevance of this axis in a site-specific stromal–myeloid context.¹⁴¹ Therapeutic formats include anti-CD47 antibodies, SIRPα-Fc fusion proteins or ligand traps, anti-SIRPα antibodies, and bispecific antibodies that combine CD47/SIRPα blockade with tumor-antigen targeting. For gastric cancer, bispecific approaches are particularly attractive because they may increase tumor selectivity. PT886, for example, is a CLDN18.2/CD47 bispecific antibody under clinical development for gastric, gastroesophageal junction, and pancreatic cancers.¹⁴² The main challenge is safety: because CD47 is broadly expressed on normal cells, especially erythrocytes and platelets, anemia, thrombocytopenia, antigen sink, and macrophage overactivation must be carefully managed. Thus, CD47–SIRPα blockade is best viewed as phagocytosis restoration, not nonspecific immune activation.

Innate immune agonists provide another way to re-educate TAMs. Toll-like receptor agonists, especially TLR7/8 agonists such as R848, can promote inflammatory macrophage polarization and enhance antigen presentation, but systemic administration is limited by cytokine toxicity. Nanoparticle-loaded TLR7/8 agonists have been engineered to deliver R848 preferentially to TAMs and convert suppressive macrophages toward inflammatory phenotypes in preclinical cancer models.^143,144 STING agonism follows a related principle by activating cytosolic DNA sensing, type I interferon signaling, and dendritic-cell/macrophage-mediated antitumor immunity. In gastric cancer, targeting the STING pathway in TAMs regulates innate immune sensing of gastric cancer cells, alters macrophage inflammatory programs, and affects tumor-cell apoptosis and T-cell differentiation.¹⁴⁵ These findings provide gastric cancer-specific support for innate reprogramming. However, TLR and STING activation can be double-edged. Excessive activation may trigger systemic inflammation, tissue toxicity, compensatory immunosuppression, or myeloid exhaustion. Therefore, local delivery, intermittent dosing, and combination with checkpoint blockade or antigen-releasing therapies may be more realistic than continuous systemic innate stimulation.

Nanodelivery and macrophage-targeted carriers may help solve the precision problem. The major limitation of macrophage reprogramming is not the lack of immunologically active molecules, but the difficulty of delivering them to the correct macrophage state in the correct tissue region. Nanoparticles, liposomes, polymeric carriers, exosomes, macrophage-membrane-coated particles, mannose receptor/CD206-targeted systems, and M2pep-based platforms are being developed to enrich drugs in TAMs while reducing systemic exposure. Gastric cancer already has relevant proof-of-concept evidence. M2pep-modified chitosan nanoparticles carrying plerixafor were reported to inhibit CXCL12–CXCR4 signaling, reprogram TAMs, and suppress gastric carcinoma in preclinical models.¹⁴⁶ This example is valuable because it integrates a known CAF–macrophage communication axis with macrophage-targeted nanotherapy. Similar platforms could potentially deliver STAT3 inhibitors, TLR/STING agonists, siRNA against TREM2 or SPP1-related programs, or CD47/SIRPα-modulating agents into defined stromal–myeloid niches.

The central translational message is that macrophage reprogramming should be designed around niche biology rather than generic M1/M2 conversion. In a STAT3-dominant niche, IL-6/JAK/STAT3 blockade may reduce suppressive cytokine programs. In a lipid-associated niche, TREM2 targeting may weaken immunosuppressive macrophage adaptation. In an SPP1⁺ matrix niche, SPP1–integrin/CD44 or upstream C3–C3AR1 targeting may disrupt macrophage retention and immune exclusion. In a phagocytosis-checkpoint-dominant niche, CD47–SIRPα blockade may restore tumor-cell engulfment. In a poorly inflamed niche, TLR or STING agonists may ignite innate sensing and antigen presentation. These approaches are not mutually exclusive, but they should not be combined blindly. Rational combination requires spatial biomarkers, on-treatment tissue sampling, and functional readouts such as CD8⁺ T-cell penetration, macrophage antigen-presentation signatures, phagocytosis markers, and reduction of SPP1⁺/TREM2⁺/STAT3-activated suppressive states.

Compared with nonselective depletion, reprogramming better fits the principle of precision therapy. It preserves the possibility that macrophages can become therapeutic allies rather than simply removing an entire myeloid compartment. The future of TAM-directed therapy in gastric cancer will likely depend on matching the intervention to the dominant macrophage state, the anatomical site, the accompanying CAF program, and the intended partner treatment. For OncoTargets and Therapy, this point is particularly important: the actionable target is not merely “macrophages”, but the druggable state transitions that determine whether macrophages sustain immune exclusion or support antitumor immunity.

Targeting Fibroblast Activation and ECM

Targeting fibroblast activation and ECM remodeling is a central strategy for weakening fibroblast–macrophage niches, but the therapeutic goal should not be complete fibroblast eradication. Fibroblasts are required for tissue repair, vascular stability, epithelial homeostasis, and immune regulation. Moreover, CAFs are heterogeneous: some subsets promote matrix deposition, immune exclusion, macrophage recruitment, and therapy resistance, whereas others may restrain tumor growth or preserve tissue structure. Therefore, the most rational strategy is stromal normalization: reducing pathogenic CAF states, loosening excessive ECM barriers, improving perfusion and immune-cell access, and interrupting stromal signals that maintain suppressive macrophages.

FAP is the most visible CAF-associated therapeutic target. In gastric cancer, FAP⁺ fibroblasts have been linked to tumor progression, poor prognosis, drug resistance, and resistance to immune checkpoint blockade.²¹ Therapeutically, FAP can be approached in several ways, including small-molecule FAP inhibitors, FAP-targeted radiopharmaceuticals, antibody–drug conjugates, bispecific immune engagers, CAR-T cells, and FAP-directed delivery systems. FAPI-PET imaging has already demonstrated the clinical feasibility of visualizing FAP-rich tumor stroma, and FAP-targeted radioligand therapy is being actively explored across solid tumors.^57,58 However, FAP targeting in gastric cancer should be interpreted carefully. FAP is a marker of activated fibroblast-rich stroma rather than a universal driver of tumor biology. Depleting all FAP⁺ fibroblasts may disrupt tissue repair or alter stromal restraint. A more precise approach is to use FAP as a theranostic entry point: identifying FAP-high, matrix-rich, immune-excluded tumors and delivering localized radiation, immune modulators, or stromal-normalizing agents to regions where FAP⁺ CAFs sustain macrophage-rich immune barriers.

TGF-β is another major stromal target because it links fibroblast activation, ECM production, EMT, immune exclusion, and macrophage polarization. In solid tumors, TGF-β-driven fibroblast and ECM programs have been associated with poor response to PD-1/PD-L1 blockade, and TGF-β inhibition can enhance T-cell penetration in preclinical models.^13,101 In gastric cancer, bintrafusp alfa, a bifunctional TGF-β trap/anti-PD-L1 fusion protein, showed manageable safety and signs of clinical activity in a Phase I Asian gastric/gastroesophageal junction cancer cohort.¹⁴⁷ This supports the feasibility of simultaneously targeting PD-L1 and TGF-β biology in gastric cancer, but it does not yet establish TGF-β blockade as a standard stromal therapy. TGF-β has pleiotropic and stage-dependent effects, including roles in epithelial homeostasis, immune tolerance, and tissue repair. Thus, TGF-β-targeted therapy should be reserved for tumors with a defined TGF-β/CAF/ECM exclusion signature and ideally combined with spatial biomarkers showing that T cells are being held at the stromal margin.

LOX and LOX-like enzymes represent a more mechanical approach to stromal therapy. By crosslinking collagen, LOX-family enzymes increase matrix stiffness, raise interstitial pressure, compress vessels, and support immune exclusion. In gastric cancer, LOX⁺ fibroblasts have been identified by single-cell analysis and are spatially associated with CD206⁺ M2-like macrophages, suggesting that collagen crosslinking and myeloid suppression may coexist in the same stromal niche.²⁵ In principle, LOX inhibition could soften the matrix, reduce mechanical exclusion, improve perfusion, and enhance delivery of chemotherapy, antibodies, or immune cells. However, clinical experience with LOXL2 inhibition has been disappointing in other desmoplastic tumors, with simtuzumab failing to improve outcomes in pancreatic and colorectal cancer trials.^148,149 This cautions against assuming that ECM enzyme inhibition alone is sufficient. In gastric cancer, LOX-targeted strategies may be more plausible as combination approaches, particularly with immunotherapy, anti-angiogenic therapy, or macrophage reprogramming, and should be guided by collagen organization, stiffness, perfusion, and LOX⁺ CAF spatial signatures rather than by LOX expression alone.

IL-6 is a soluble stromal target that connects activated fibroblasts to tumor-cell survival and macrophage programming. CAF-derived IL-6 activates JAK1–STAT3 signaling in gastric cancer cells and contributes to stroma-induced chemotherapy resistance; IL-6 blockade restored chemosensitivity in experimental models.²⁷ IL-6 can also promote STAT3-dependent M2-like macrophage differentiation in gastric cancer.²⁸ Therefore, IL-6 targeting may weaken both epithelial survival signaling and suppressive macrophage programming within CAF-rich niches. Potential therapeutic approaches include IL-6 or IL-6R antibodies, JAK inhibitors, and STAT3-directed strategies. Yet systemic IL-6/JAK/STAT inhibition may impair immune-cell function, hematopoiesis, and host defense. For this reason, IL-6-targeted therapy should not be used as a generic anti-inflammatory intervention. It is most rational in tumors with high stromal IL-6, STAT3-activated macrophages, chemotherapy resistance, or immune exclusion, preferably with pharmacodynamic monitoring of pSTAT3 in tumor cells and myeloid cells.

CXCL12 is a particularly attractive target because it links CAFs to both immune exclusion and macrophage polarization. In the classical CAF model from pancreatic cancer, FAP⁺ fibroblast-derived CXCL12 excluded T cells from tumor-cell regions, and CXCR4 inhibition restored T-cell access and sensitized tumors to PD-L1 blockade.⁸⁴ In gastric cancer, the CXCL12–CXCR4 axis has gained direct support. CAF-derived CXCL12 can interact with CXCR4 on TREM2⁺ lipid-associated macrophages, and CXCL12/CXCR4 inhibition using macrophage-targeted plerixafor-loaded nanoparticles reprogrammed TAMs and suppressed gastric carcinoma in preclinical models.^37,146 More recent gastric cancer evidence further indicates that CXCL12 can promote PD-L1 expression, M2-TAM polarization, and anti-PD-1 resistance through CXCR4/PI3K/AKT/HIF-1α signaling.⁷³ These data make CXCL12–CXCR4 one of the most actionable CAF–macrophage axes in gastric cancer. Clinically relevant agents include CXCR4 antagonists such as plerixafor and newer CXCR4 inhibitors, but systemic blockade may mobilize hematopoietic and immune cells broadly. Targeted delivery or biomarker-selected combination with PD-1/PD-L1 blockade may therefore be more appropriate than nonspecific CXCR4 inhibition.

Collagen and integrin signaling represent the structural and adhesive side of CAF targeting. Collagen is not merely a scaffold; it can transmit survival and immune-evasion signals through integrins, discoidin domain receptors, and mechanotransduction pathways. In gastric cancer, collagen ECM was shown to promote immune evasion by activating integrin αvβ1/AKT/SP1/IL4I1 signaling, which increased AHR activity and impaired CAR-T cell function in experimental models.⁷⁶ Type I collagen can also support gastric cancer growth through ITGB1/BCL9L/β-catenin signaling.⁵² These findings suggest that collagen–integrin communication may reduce the efficacy of both cellular immunotherapy and conventional antitumor immunity. Therapeutic options include integrin inhibitors, collagen-modifying enzymes, matrix-degrading or matrix-normalizing agents, DDR inhibitors, and collagen-targeted delivery systems. However, broad degradation of ECM can be risky because matrix also restrains invasion in some contexts and maintains vascular and tissue integrity. The aim should be to reduce pathological stiffness and immune exclusion, not to erase the matrix.

Taken together, CAF and ECM targeting should be organized around normalization rather than destruction. FAP targeting may identify and deliver therapy to activated stromal regions; TGF-β blockade may reduce fibroblast-driven immune exclusion; LOX inhibition may loosen collagen crosslinking and improve perfusion; IL-6 blockade may interrupt fibroblast-driven STAT3 survival and macrophage polarization; CXCL12–CXCR4 inhibition may release T cells and reprogram TAMs; and collagen–integrin targeting may weaken adhesive and mechanical immune barriers. These strategies are most likely to succeed when combined with immune checkpoint blockade, chemotherapy, anti-angiogenic therapy, or macrophage reprogramming, but only in tumors where the relevant stromal pathway is active.

For gastric cancer, the next step is not simply to add anti-stromal drugs to existing regimens. It is to define which stromal abnormality dominates in each patient: FAP-high CAF accumulation, TGF-β-driven exclusion, LOX-dependent collagen stiffness, IL-6–STAT3 survival signaling, CXCL12–CXCR4 myeloid retention, or collagen–integrin immune evasion. Spatial transcriptomics, multiplex imaging, FAPI-PET, collagen imaging, perfusion imaging, and on-treatment biopsies could help match patients to the correct stromal-normalizing strategy. This precision approach is especially important because CAFs and ECM are not universally harmful. The therapeutic objective is to dismantle the pathological fibroblast–macrophage niche while preserving stromal functions needed for tissue integrity and effective antitumor immunity.

Disrupting Niche-Specific Reciprocal Signaling

A niche-directed strategy should move beyond single-cell-type targeting and focus on reciprocal signaling circuits that maintain fibroblast–macrophage cooperation. This approach is important because CAFs and TAMs often reinforce each other: fibroblasts recruit or retain macrophages, macrophages sustain stromal remodeling and immunosuppression, and both cell types generate survival signals for malignant epithelial cells. The most actionable strategy is therefore not to block every stromal or myeloid pathway, but to identify which reciprocal circuit dominates in a given tissue compartment and select the least toxic intervention capable of disrupting that circuit.

The CCL2⁺ CAF–STAT3-activated macrophage niche is one of the most druggable stromal–myeloid circuits in primary gastric cancer. Mechanistically, CCL2⁺ fibroblasts recruit myeloid cells and stimulate STAT3 activation in infiltrating macrophages, resulting in an immunosuppressive microenvironment associated with poor prognosis.¹⁶ This axis can be interrupted at several points. CCL2 neutralization, CCR2 antagonism, IL-6/IL-6R blockade, JAK inhibition, and STAT3 inhibition are all pharmacologically feasible. Carlumab, an anti-CCL2 antibody, and small-molecule CCR2 inhibitors have already been tested in solid tumors, although not as established therapies for gastric cancer.^131,132 IL-6/JAK/STAT3 targeting is also biologically supported by gastric cancer studies showing IL-6-dependent M2-like macrophage differentiation and GC-MSC-driven macrophage polarization through IL-6/IL-8–JAK2/STAT3 signaling.^28,46 This niche is suitable for combination with ICB because disrupting monocyte recruitment and STAT3-driven macrophage programming may increase the probability that CD8⁺ T cells can enter and function within tumor tissue. However, the major risks are chemokine redundancy, rebound monocyte recruitment after drug withdrawal, impaired host defense, and excessive suppression of JAK–STAT pathways needed for lymphocyte function. Thus, this strategy should be biomarker-selected and ideally monitored by decreased CCL2⁺ CAF abundance, reduced pSTAT3⁺ macrophages, and improved CD8⁺ T-cell localization.

The GREM1⁺ fibroblast–SPP1⁺ macrophage niche is therapeutically attractive but less clinically mature. Gastric cancer multi-omics and spatial transcriptomic analyses indicate that GREM1⁺ fibroblasts and SPP1⁺ macrophages can be co-localized, are associated with poor prognosis, and communicate through SPP1–integrin and collagen-related pathways.¹⁸ The druggability of this axis is moderate. GREM1 itself is targetable: ginisortamab, also known as UCB6114, is a first-in-class anti-GREM1 antibody in clinical development for cancer.¹⁵⁰ SPP1–integrin/CD44 signaling is also theoretically targetable through anti-SPP1 strategies, integrin inhibitors, CD44-directed approaches, or upstream modulation of macrophage state. However, direct SPP1 blockade remains largely preclinical in oncology, and integrin inhibitors have historically required better biomarker selection to show meaningful clinical benefit. Gastric cancer evidence for this niche is currently strong at the level of spatial association and ligand–receptor prediction, but direct perturbation experiments proving therapeutic reversibility are still insufficient. Combination with ICB is plausible, particularly in GREM1⁺ fibroblast/SPP1⁺ macrophage-high tumors with matrix-rich immune exclusion, but it should be described as a testable hypothesis rather than a validated strategy. Safety concerns include interference with BMP biology, tissue repair, wound healing, macrophage adhesion, and physiological osteopontin functions in bone and immune regulation.

The THBS2⁺ CAF–C3–C3AR1–SPP1⁺ TAM axis is the strongest niche-specific therapeutic candidate in gastric cancer peritoneal metastasis. This circuit is mechanistically coherent: THBS2⁺ matrix CAFs release complement C3, C3 acts on C3AR1-expressing macrophages, and this signaling promotes accumulation and transition of peritoneum-specific macrophages into SPP1⁺ TAMs. Unlike many candidate CAF–TAM axes, this pathway has gastric cancer-specific spatial evidence, functional validation, and treatment-response relevance. Blocking C3–C3AR1 disrupted the stromal–myeloid circuit and enhanced response to immune checkpoint blockade in experimental models of gastric cancer peritoneal metastasis.¹⁷ The druggability is high in principle because complement-targeted drugs already exist, including C3 inhibitors and complement-receptor-directed agents. However, oncology translation remains early, and C3aR-specific approaches are not yet established gastric cancer therapeutics. C5aR inhibitors should not be assumed to substitute for C3AR1 blockade, because C3aR- and C5aR-mediated myeloid biology can differ. This axis is particularly suitable for combination with PD-1/PD-L1 blockade in biomarker-positive peritoneal metastasis, where the therapeutic objective would be to reduce SPP1⁺ TAM accumulation and reopen an immune-excluded metastatic niche. The key safety risk is infection, especially with systemic complement inhibition, as complement pathways are essential for antibacterial defense, immune-complex handling, and tissue homeostasis.⁶⁴ Therefore, clinical development should prioritize metastatic-site selection, local pathway activity, vaccination and infection-risk management, and pharmacodynamic evidence of reduced C3AR1⁺/SPP1⁺ macrophage signaling.

The LOX/IL-6-related CAF–TAM interaction represents a mechanistically supported but less spatially complete target class. In gastric cancer, LOX⁺ fibroblasts have been identified by single-cell analysis and are associated with M2 macrophage infiltration, hypoxia, IL6–JAK–STAT3 pathway enrichment, and unfavorable prognosis.²⁵ This axis is druggable at several levels. LOX or LOXL2 inhibition could theoretically reduce collagen crosslinking, matrix stiffness, and mechanical immune exclusion. IL-6/IL-6R, JAK, or STAT3 inhibition could reduce macrophage polarization and tumor-cell survival signaling. However, clinical experience with LOXL2 inhibition in other cancers has been disappointing, as simtuzumab failed to improve outcomes in pancreatic and colorectal cancer trials.^148,149 This suggests that collagen crosslinking cannot be targeted effectively without selecting patients whose tumors are truly LOX-dependent and without combining matrix modulation with immune or vascular strategies. Gastric cancer experiments support LOX⁺ fibroblast–M2 macrophage association, but direct evidence that LOX blockade reverses CAF–TAM immune exclusion in gastric cancer remains insufficient. Combination with ICB is rational in LOX-high, hypoxic, matrix-stiff tumors, but safety risks include impaired wound healing, vascular or connective-tissue toxicity, and disruption of normal collagen homeostasis. IL-6/JAK/STAT targeting also carries risks of infection, cytopenia, liver toxicity, and suppression of beneficial antitumor lymphocyte signaling.

Across these four axes, the level of therapeutic readiness differs substantially. The CCL2⁺ CAF–STAT3 macrophage axis has strong gastric cancer spatial and functional evidence and several druggable nodes, but clinical translation must manage rebound myeloid recruitment and systemic immune suppression. The GREM1⁺ fibroblast–SPP1⁺ macrophage axis has compelling spatial association and emerging GREM1 druggability, but still lacks gastric cancer-specific perturbation evidence. The THBS2⁺ CAF–C3–C3AR1–SPP1⁺ TAM axis has the clearest site-specific rationale for ICB combination, especially in peritoneal metastasis, but complement safety and pathway specificity are major concerns. The LOX/IL-6 CAF–TAM axis is biologically plausible and partially supported in gastric cancer, but matrix-targeted therapy requires better patient selection and combination design.

A practical development framework would therefore be to match each reciprocal signaling axis with a specific biomarker and partner therapy. CCL2⁺ CAF/pSTAT3⁺ macrophage-high tumors may be candidates for CCR2 or IL-6/JAK/STAT3 modulation plus ICB. GREM1⁺ fibroblast/SPP1⁺ macrophage-high tumors may be candidates for GREM1, SPP1–integrin/CD44, or matrix-normalizing strategies once functional validation improves. THBS2⁺ matrix CAF/C3AR1⁺SPP1⁺ TAM-positive peritoneal metastases are the strongest candidates for complement-axis blockade plus PD-1/PD-L1 inhibition. LOX-high, stiff, hypoxic tumors may require LOX or IL-6 pathway modulation combined with vascular normalization, macrophage reprogramming, or ICB. In all cases, niche disruption should be guided by spatial transcriptomics, multiplex imaging, and on-treatment biopsies rather than by single-marker expression alone.

Rational Combination and Treatment Sequencing

The clinical value of fibroblast–macrophage niche targeting will depend less on the number of agents combined than on how they are sequenced. Many current combination strategies in oncology are built by adding an immune checkpoint inhibitor to chemotherapy, anti-angiogenic therapy, targeted therapy, or another immune modulator. For stromal–myeloid niches, however, simple concurrent escalation may be insufficient or even counterproductive. CAFs, macrophages, vessels, matrix, and T cells change dynamically after each therapeutic pressure. Therefore, rational combination should be designed around temporal logic: first modify the tissue barrier or myeloid context, then apply the therapy that requires improved access, antigen presentation, or T-cell function.

One strategy is to reduce the stromal barrier before immune checkpoint blockade. In immune-excluded tumors, PD-1/PD-L1 blockade may fail because CD8⁺ T cells are retained at the stromal margin or cannot penetrate tumor nests. In this setting, the first therapeutic step should be to loosen or normalize the CAF/ECM barrier, rather than immediately intensifying checkpoint inhibition. Potential priming interventions include TGF-β pathway modulation, CXCL12–CXCR4 blockade, FAP-directed stromal delivery, LOX or collagen-normalizing approaches, and integrin or matrix-adhesion targeting. The rationale is supported by the classical observation that CXCL12 derived from FAP⁺ CAFs can exclude T cells from tumor-cell regions and that CXCR4 inhibition can sensitize tumors to PD-L1 blockade in preclinical models.⁸⁴ In gastric cancer, CXCL12–CXCR4 inhibition has been linked to enhanced T-cell activity, reduced CAF–TAM-mediated immune evasion, and reversal of anti-PD-1 resistance in experimental studies.^72,73 Thus, a rational sequence would be stromal or CXCL12/CXCR4 priming first, followed by PD-1/PD-L1 blockade during the window in which T-cell access is improved.

A second strategy is to block myeloid recruitment before releasing T-cell checkpoints. In tumors dominated by CCR2⁺ monocyte influx, CSF1R-dependent macrophage maintenance, or complement-driven SPP1⁺ TAM accumulation, checkpoint blockade may have little effect unless suppressive myeloid cells are first reduced or reprogrammed. This does not necessarily require complete macrophage depletion. Instead, short-course CCR2/CCR5 blockade, CSF1R pathway modulation, C3–C3AR1 inhibition, or macrophage-targeted nanotherapy could be used as a preparatory step to reduce the density or suppressive tone of TAM-rich niches. After myeloid pressure decreases, PD-1/PD-L1 blockade may act on a more favorable immune substrate. The logic is supported by clinical and preclinical experience with CCR2 inhibition in pancreatic cancer and by gastric cancer-specific evidence that CAF–macrophage crosstalk through C3–C3AR1 governs response to checkpoint blockade in peritoneal metastasis.^17,132 However, this sequence must be designed carefully because abrupt withdrawal of CCL2/CCR2 blockade may trigger rebound monocyte recruitment, and excessive CSF1R or complement suppression may impair host defense.

Vascular normalization should be considered another form of immune priming rather than only an anti-angiogenic effect. Abnormal VEGF-driven vessels restrict perfusion, increase hypoxia, and limit T-cell and drug delivery. Properly timed anti-angiogenic therapy may transiently improve vessel function, reduce hypoxia, and permit greater immune-cell trafficking. This creates a window for combining vascular normalization with macrophage reprogramming or checkpoint blockade. In gastric cancer, tumor vessel normalization has been proposed as a strategy to enhance immunotherapy, and a spatially resolved Phase II study of ramucirumab plus pembrolizumab in refractory PD-L1-positive gastric cancer showed that responders had greater on-treatment vascular normalization coupled with cytotoxic T-cell infiltration, whereas nonresponders had higher baseline hypoxia, TGF-β activity, and greater distances between immune cells, vessels, and tumor cells.^88,151 These findings support a sequencing model in which anti-VEGF/VEGFR therapy is not simply added to ICB, but used to generate a vascular and immune window in which T cells and macrophage-directed interventions can function more effectively.

Complement blockade combined with ICB is most compelling in gastric cancer peritoneal metastasis. The THBS2⁺ matrix CAF–C3–C3AR1–SPP1⁺ TAM axis provides a site-specific example in which stromal and myeloid targeting can be linked directly to checkpoint response. In this setting, C3–C3AR1 signaling promotes accumulation of SPP1⁺ TAMs and contributes to an immunosuppressive peritoneal niche; blocking this pathway improves checkpoint efficacy in experimental models.¹⁷ A rational sequence would therefore be complement-axis inhibition to disrupt macrophage accumulation and stromal–myeloid reinforcement, followed by or combined with PD-1/PD-L1 blockade once suppressive TAM pressure is reduced. This approach should be restricted to biomarker-positive peritoneal metastases rather than applied broadly to all gastric cancers. Because complement pathways are central to antimicrobial defense and immune homeostasis, clinical protocols would need infection-risk management, careful dosing, and pharmacodynamic confirmation that C3AR1⁺/SPP1⁺ macrophage signaling is actually reduced.

Spatial niche targeting may also improve antibody and ADC therapy. Antibodies and ADCs require vascular access, extravasation, tissue penetration, target binding, and, for ADCs, internalization and payload release. Dense ECM, high interstitial pressure, abnormal vasculature, and long diffusion distances can create heterogeneous exposure even when the epithelial target is present.¹²¹ This is relevant for HER2-, CLDN18.2-, and FGFR2b-directed therapies, as well as HER2-directed ADCs such as trastuzumab deruxtecan.⁵ A niche-aware strategy would not simply combine ADCs with immunotherapy, but first assess whether the target-positive tumor region is accessible. In matrix-rich or poorly perfused lesions, vascular normalization, ECM modulation, or FAP-directed delivery could theoretically improve antibody distribution. In macrophage-rich lesions, CD47–SIRPα blockade, TAM reprogramming, or SPP1/CXCL12 pathway targeting could potentially improve antibody-dependent phagocytosis or relieve immune suppression. These concepts are promising but remain incompletely validated in gastric cancer, and should be tested with paired biopsies, imaging, and spatial pharmacodynamic readouts.

The sequence of combination therapy should therefore be adapted to the dominant resistance architecture. In a high-ECM, immune-excluded tumor, the logical sequence is matrix or CAF normalization followed by ICB. In a monocyte-influx or TAM-dominant tumor, the sequence is myeloid recruitment blockade or macrophage reprogramming followed by checkpoint release. In a hypoxic, poorly perfused tumor, vascular normalization should precede or accompany immunotherapy and macrophage modulation. In peritoneal metastasis with THBS2⁺ CAF–SPP1⁺ TAM enrichment, complement-axis targeting may be paired with ICB. In target-positive but stromal-restricted tumors, niche modulation may be needed before antibodies or ADCs can achieve uniform exposure.

This sequencing logic also helps avoid excessive toxicity. Combining multiple immune suppressor-targeting agents at the same time may impair normal macrophage repair, complement defense, cytokine signaling, or lymphocyte function. For example, simultaneous broad blockade of CCR2, CSF1R, IL-6/JAK, complement, and PD-1 could reduce myeloid suppression, but it could also increase infection risk, cytopenia, tissue-repair defects, and immune-related toxicity. A better strategy is modular therapy: identify the dominant barrier, apply a short or spatially targeted priming intervention, confirm pharmacodynamic remodeling, and then introduce the partner therapy that depends on that remodeling. In this framework, the most important clinical question is not “which drugs can be combined?”, but “which microenvironmental barrier must be removed first?”

For future gastric cancer trials, rational sequencing should be built into the study design. Baseline biopsies should define the niche class: CCL2⁺ CAF–STAT3 macrophage, CXCL12⁺ CAF–TREM2⁺ TAM, THBS2⁺ CAF–SPP1⁺ TAM, LOX-high matrix niche, VEGF-high hypoxic vascular niche, or target-positive but ECM-restricted tumor. Early on-treatment biopsies should determine whether the first intervention achieved its intended effect, such as reduced pSTAT3⁺ macrophages, lower SPP1⁺ TAM density, increased vessel normalization, reduced collagen stiffness, or improved CD8⁺ T-cell penetration. Only then should the second component—PD-1/PD-L1 blockade, chemotherapy, anti-angiogenic therapy, antibody therapy, or ADC—be judged fairly. This type of adaptive, niche-guided sequencing is more aligned with precision oncology than fixed simultaneous multi-drug escalation.

Biomarkers and Patient Selection

Candidate Tissue Biomarkers

Patient selection for fibroblast–macrophage niche-targeted therapy should not rely on single-marker positivity alone. High FAP, SPP1, CD68, or CD163 expression may indicate stromal or myeloid enrichment, but it does not prove that a therapeutically actionable niche is present. More useful biomarkers should integrate cell-state abundance, spatial proximity, matrix architecture, and CD8⁺ T-cell localization. This is particularly relevant in gastric cancer, where spatial and multiplex studies have shown that the organization of immune, stromal, and myeloid cells can influence prognosis and response to immunotherapy.^{16–18,75,152}

The THBS2⁺ CAF–SPP1⁺ TAM combined score is currently one of the most actionable candidate biomarkers, especially for gastric cancer peritoneal metastasis. This score could combine the abundance of THBS2⁺ matrix CAFs, the abundance of SPP1⁺ TAMs, and their spatial colocalization within metastatic lesions. A refined version should also include CAF-derived C3 and macrophage C3AR1 expression, because THBS2⁺ matrix CAFs have been shown to promote C3–C3AR1-dependent accumulation of SPP1⁺ TAMs and reduce response to immune checkpoint blockade in gastric cancer peritoneal metastasis.¹⁷ Therefore, this biomarker is most suitable for selecting patients for complement-axis blockade plus PD-1/PD-L1 inhibition, rather than for general gastric cancer stratification.

The spatial distance between GREM1⁺ fibroblasts and SPP1⁺ macrophages is another candidate marker of matrix-centered stromal–myeloid interaction. Multi-omics and spatial analyses have identified interactions between GREM1⁺ fibroblasts and SPP1⁺ macrophages in gastric cancer, with predicted SPP1–integrin and collagen-related communication. A practical assay could measure the nearest-neighbor distance between these two populations, the proportion of SPP1⁺ macrophages located within a defined radius of GREM1⁺ fibroblasts, and the overlap of both populations with collagen-rich regions. At present, this marker should be considered a candidate biological and prognostic stratifier, not yet a validated predictive biomarker for a specific therapy.¹⁸

CCL2⁺ fibroblast–p-STAT3⁺ macrophage proximity may serve as a marker of inflammatory CAF-driven myeloid programming. Spatial dissection of gastric cancer has shown immunosuppressive crosstalk between CCL2⁺ fibroblasts and STAT3-activated macrophages, with clinical relevance. The most informative readouts would include CCL2⁺ fibroblast density, p-STAT3⁺CD68⁺ or p-STAT3⁺CD163⁺ macrophage density, and their spatial proximity. This marker is therapeutically meaningful because it links a recruitment signal to a downstream macrophage activation state, supporting potential selection for CCR2 blockade, IL-6/JAK/STAT3 inhibition, or macrophage-reprogramming strategies combined with ICB.¹⁶

A combined FAP–SPP1–CD8 spatial panel could provide a practical tissue-level readout of immune exclusion. FAP marks activated stromal regions, SPP1 marks a matrix-associated macrophage program, and CD8 indicates cytotoxic T-cell localization. In pMMR gastric cancer, SPP1⁺ macrophages and FAP⁺ fibroblasts have been associated with stromal activation and tumor progression.⁴² For patient selection, the key feature is not whether these markers are simply positive, but whether FAP⁺ stromal bands and SPP1⁺ macrophages surround tumor nests while CD8⁺ T cells remain confined to stromal or marginal regions. Such a pattern would support an immune-excluded phenotype and may identify tumors requiring stromal or myeloid priming before ICB.^42,75

Collagen density and fiber orientation should be included as structural biomarkers. Gastric cancer studies have shown that collagen reorganization in the tumor microenvironment is associated with prognosis.⁵¹ Tumor-associated collagen signatures can also improve prognostic prediction and may help identify patients likely to benefit from adjuvant chemotherapy.¹⁵³ Therefore, collagen assessment should go beyond the presence or absence of fibrosis. Quantitative variables such as collagen area, fiber width, alignment, straightness, and orientation relative to tumor nests may help define matrix barriers that restrict immune-cell entry and drug penetration.

The distance between macrophages and tumor epithelial regions is another useful spatial metric. Multiplex immunohistochemistry has shown that macrophage subsets in gastric cancer are spatially heterogeneous across tumor core, edge, margin, and non-tumor regions. For niche-based selection, relevant measures include macrophage density within tumor nests, macrophage density at the tumor–stroma interface, nearest-neighbor distance from macrophages to pan-cytokeratin-positive epithelial cells, and the ratio of macrophages to CD8⁺ T cells in tumor-proximal compartments.¹⁵² Tumor-proximal SPP1⁺, p-STAT3⁺, CD163⁺, CD206⁺, or TREM2⁺ macrophages may indicate a stronger immunosuppressive barrier than total macrophage abundance alone.^16,152

In practice, these biomarkers should be developed as spatial composite scores. A clinically realistic panel could include pan-cytokeratin, CD8, FAP, THBS2, GREM1, SPP1, CD68, CD163, p-STAT3, C3, C3AR1, and collagen imaging. Spatial transcriptomics is valuable for discovery, but multiplex immunohistochemistry, multiplex immunofluorescence, picrosirius red, Masson staining, second-harmonic generation imaging, and AI-assisted histopathology are more practical for trial stratification.^51,152,153 These markers should complement, not replace, established biomarkers such as MSI/MMR status, EBV status, PD-L1 CPS, HER2, CLDN18.2, and FGFR2b. The key clinical question is whether the target-positive tumor compartment is spatially accessible, immunologically permissive, and free from dominant fibroblast–macrophage sanctuary niches.

Gastric Cancer Populations Most Likely to Benefit

The patients most likely to benefit from fibroblast–macrophage niche-targeted therapy are not defined by one marker, but by a tissue context in which stromal and myeloid programs dominate resistance. These strategies are therefore most relevant for gastric cancers with low intrinsic immunogenicity, abundant matrix, macrophage-rich immune suppression, or treatment-induced stromal repair.

pMMR/MSS gastric cancer is a priority population. Compared with MSI-H/dMMR tumors, pMMR/MSS tumors are generally less dependent on high neoantigen load and may rely more heavily on stromal and myeloid immune regulation. Recent single-cell analysis showed that SPP1⁺ macrophages and FAP⁺ fibroblasts are enriched in pMMR gastric cancer and interact through hypoxia, ECM-remodeling, VEGF, MIF, and SPP1-related pathways.⁴² This suggests that pMMR/MSS patients with FAP⁺/SPP1⁺ stromal–myeloid enrichment may be rational candidates for CAF normalization, TAM reprogramming, CXCL12/MIF/SPP1-axis targeting, or combination strategies that prime the microenvironment before ICB.

Diffuse-type or stroma-rich gastric cancer may also be suitable for niche-targeted approaches. TCGA showed that genomically stable gastric cancers are enriched for diffuse histology, while the ACRG mesenchymal-like subtype includes diffuse-subtype tumors with poor prognosis and high recurrence frequency.^11,12 These tumors often show strong stromal involvement, invasive growth, and extracellular matrix remodeling, making them biologically compatible with CAF/ECM-directed therapy. However, histology alone is insufficient; patient selection should require tissue evidence of FAP⁺, GREM1⁺, THBS2⁺, LOX⁺, or collagen-rich CAF programs together with suppressive macrophage enrichment.

Patients with peritoneal metastasis represent the clearest site-specific population. Gastric cancer peritoneal metastasis has a distinct immunosuppressive ecosystem dominated by THBS2⁺ matrix CAFs and SPP1⁺ TAMs. CAF-derived C3 promotes C3AR1-dependent accumulation and transition of peritoneal macrophages into SPP1⁺ TAMs, and blockade of this axis improves ICB efficacy in experimental models.¹⁷ Therefore, peritoneal metastasis patients with high THBS2⁺ CAF–SPP1⁺ TAM–C3/C3AR1 activity may be the most suitable group for complement-axis blockade, SPP1/TAM targeting, or matrix–myeloid modulation combined with PD-1/PD-L1 inhibition.

Immune-excluded tumors are another rational population. In these tumors, CD8⁺ T cells are present but remain confined to stromal or marginal regions rather than entering tumor epithelial nests. Multidimensional immune profiling in gastric cancer has shown that immunotherapy response is linked not only to immune-cell density but also to the spatial organization of T-cell and myeloid compartments.⁷⁵ Patients with FAP⁺ stromal bands, SPP1⁺ or p-STAT3⁺ macrophage barriers, dense collagen, and poor CD8⁺ tumor-nest penetration may therefore benefit from stromal or myeloid priming before ICB.

ICB primary-resistant patients should be considered for niche reassessment rather than immediate abandonment of immunotherapy. Some primary-resistant tumors may lack reactivatable intratumoral CD8⁺ T cells, whereas others may contain T cells that are spatially excluded by CAF–macrophage niches. Spatial analysis identifying CCL2⁺ fibroblast–STAT3 macrophage proximity or GREM1⁺ fibroblast–SPP1⁺ macrophage interaction could help distinguish these scenarios.^16,18 The former may require T-cell recruitment or antigen-presentation strategies, while the latter may be more suitable for CCR2, IL-6/JAK/STAT3, SPP1–integrin, or matrix-directed combinations with ICB.

Finally, patients showing stromal and myeloid rebound after treatment may represent a dynamic target population. Longitudinal single-cell sequencing has shown that neoadjuvant chemotherapy reconstructs the gastric cancer microenvironment, altering malignant, stromal, immune, fibroblast, monocyte, and macrophage compartments.¹¹³ If post-treatment biopsies show increased CAF activation, collagen remodeling, SPP1⁺/CD163⁺/p-STAT3⁺ macrophage accumulation, or reduced CD8⁺ tumor penetration, niche-targeted therapy may be most useful as a consolidation or resistance-prevention strategy. This population is especially important because acquired resistance may arise from residual cancer cells protected within a repaired stromal–myeloid sanctuary.

Overall, fibroblast–macrophage niche-targeted therapy should be prioritized for biologically selected patients: pMMR/MSS tumors with FAP⁺/SPP1⁺ enrichment, diffuse or stroma-rich cancers with dominant CAF/ECM programs, peritoneal metastases with THBS2⁺CAF–SPP1⁺TAM activity, immune-excluded tumors, ICB primary-resistant tumors with preserved but excluded T cells, and post-treatment tumors showing stromal–myeloid rebound. These groups should be viewed as trial-enrichment populations, not established indications, and should be validated using baseline and on-treatment spatial biomarkers.

Detection Technologies

Detection technologies for fibroblast–macrophage niches should be evaluated by a practical question: can they help select patients, monitor treatment-induced remodeling, or guide combination therapy? Conventional IHC remains the most clinically accessible method. It can measure individual markers such as FAP, CD8, CD68, CD163, SPP1, THBS2, or p-STAT3 on formalin-fixed tissue, and is suitable for initial screening. However, single-marker IHC cannot reliably define a niche, because it provides limited information about cell–cell proximity and tumor–stroma compartmentalization.

Multiplex immunofluorescence or multiplex IHC is more suitable for niche assessment. It can simultaneously define epithelial cells, CAFs, macrophage states, and CD8⁺ T-cell localization in the same tissue section. In gastric cancer, multiplex immunohistochemistry has already demonstrated marked spatial heterogeneity of macrophage subsets across tumor regions.¹⁵² Multidimensional immune profiling has also shown that spatial immune organization is relevant to immunotherapy response.⁷⁵ Therefore, multiplex imaging is currently the most practical bridge between discovery biology and trial-level patient selection.

Digital pathology can convert tissue architecture into quantitative biomarkers. It is particularly useful for collagen density, fiber orientation, tumor–stroma ratio, immune-cell distance to tumor epithelium, and CD8⁺ T-cell penetration. In gastric cancer, tumor-associated collagen signatures derived from pathological image analysis were associated with prognosis and adjuvant chemotherapy benefit.¹⁵³ Digital pathology is therefore well suited for scalable patient stratification and serial comparison of baseline and post-treatment biopsies, provided that algorithms are standardized and externally validated.

Spatial transcriptomics provides higher biological resolution but is less ready for routine clinical screening. Its major value is discovery and validation of spatial programs, such as CCL2⁺ fibroblast–STAT3 macrophage proximity, GREM1⁺ fibroblast–SPP1⁺ macrophage interaction, and THBS2⁺ CAF–SPP1⁺ TAM organization in gastric cancer.^16–18 For clinical trials, spatial transcriptomics can identify which niche axis is active and define a protein panel for downstream testing. At present, cost, tissue requirements, platform variability, and analysis complexity limit its use as a routine diagnostic assay.

Single-cell RNA sequencing is most useful for defining cell states and treatment-induced remodeling. It can distinguish SPP1⁺, C1QC⁺, TREM2⁺, pSTAT3-associated, or inflammatory macrophage programs and classify CAF subtypes beyond conventional IHC markers. In gastric cancer, single-cell studies have mapped stromal and immune heterogeneity and shown that neoadjuvant chemotherapy can reconstruct malignant, stromal, and immune compartments.^14,113 However, scRNA-seq loses spatial information during tissue dissociation, so it should be paired with spatial or multiplex imaging before being used for patient selection.

Ascites cells and circulating biomarkers may be valuable for monitoring, especially in peritoneal metastasis. Ascites single-cell analysis has identified immunosuppressive macrophage transitions in gastric cancer peritoneal metastasis, suggesting that repeated ascites sampling could track TAM states during therapy.⁶² Circulating CCL2, IL-6, CSF1, SPP1, complement fragments, extracellular vesicles, or monocyte subsets may also reflect stromal–myeloid activity, but they cannot localize the niche or prove tumor-compartment exclusion. Thus, liquid biomarkers are best used for dynamic monitoring, while tissue-based spatial assays remain necessary for initial niche classification.

In practice, a tiered workflow is most realistic. Baseline IHC or multiplex imaging can screen for FAP/SPP1/CD8/macrophage patterns; digital pathology can quantify collagen and tumor–immune distances; spatial transcriptomics or scRNA-seq can be used in selected trial cohorts to refine the active pathway; and ascites or blood markers can monitor temporal changes. This combined approach avoids overreliance on expensive spatial technologies while preserving the biological information needed for niche-guided therapy.

Evidence Limitations, Safety Considerations, and Future Directions

The current evidence base has several overarching limitations. Most proposed CAF–TAM interactions have been derived from cross-sectional single-cell datasets, bulk transcriptomic correlations, computational ligand–receptor inference, or spatial colocalization. These approaches can identify candidate cell states and opportunities for communication but cannot independently establish signaling direction, functional dependence, treatment responsiveness, or temporal stability. Direct gastric cancer evidence is strongest for a limited number of spatially resolved interactions, particularly the THBS2⁺ matrix CAF–C3/C3AR1–SPP1⁺ TAM circuit in peritoneal metastasis. Many other pathways discussed in this review remain associative, spatially unresolved, supported by separate CAF- and TAM-focused experiments, or extrapolated from other malignancies. In addition, differences in tissue source, disease stage, treatment exposure, analytical platform, marker definition, and metastatic site limit direct comparison across studies. These limitations should be considered when interpreting candidate niches as biomarkers or therapeutic targets.

Lack of Standardized Nomenclature

A major barrier to translating fibroblast–macrophage niches into clinical practice is the lack of standardized nomenclature. Across studies, CAFs may be named according to marker genes, presumed functions, anatomical location, transcriptional clusters, or disease context. Terms such as inflammatory CAFs, myofibroblastic CAFs, matrix CAFs, FAP⁺ CAFs, THBS2⁺ CAFs, CCL2⁺ fibroblasts, and GREM1⁺ fibroblasts may overlap but are not always equivalent.³⁹ Similarly, TAMs are often described as M1-like, M2-like, SPP1⁺, C1QC⁺, TREM2⁺, STAT3-activated, or lipid-associated macrophages, yet these labels do not necessarily refer to stable or mutually exclusive cell states.³⁰

This problem is especially important in gastric cancer because single-cell and spatial studies increasingly identify niche-specific fibroblast and macrophage states, but different datasets may use different clustering thresholds, marker panels, and naming rules.¹⁴ A single marker gene is rarely sufficient to define a functional subgroup. For example, FAP can identify activated fibroblast-rich stroma, but it does not specify whether the CAF state is inflammatory, matrix-remodeling, immunosuppressive, or therapy-resistant. Likewise, CD163 or CD206 cannot fully define a suppressive macrophage state, and SPP1 expression may reflect matrix interaction, tissue repair, angiogenesis, or immune suppression depending on spatial context.

Future studies should therefore describe CAF and TAM states using a minimum information framework: marker combination, inferred function, anatomical location, neighboring cell types, ligand–receptor programs, and treatment association. Instead of relying on broad labels such as “M2 macrophages” or “activated CAFs”, gastric cancer studies should report more precise terms such as “CCL2⁺ fibroblast-proximal p-STAT3⁺ macrophages” or “THBS2⁺ matrix CAF-associated SPP1⁺ TAMs”. Such descriptive naming is longer, but it preserves biological meaning and improves cross-study comparability.

Standardization does not require forcing all studies into one rigid taxonomy. Rather, it requires separating three concepts: identity, state, and function. Identity defines the broad lineage; state describes the transcriptional or protein phenotype; function requires spatial, experimental, or clinical evidence. This distinction will be essential for biomarker development and therapeutic trials. Without standardized terminology, the same therapeutic target may be tested in biologically different populations, and apparently conflicting results may simply reflect inconsistent definitions of the niche.

Spatial Association versus Causality

A second translational challenge is the frequent confusion between spatial association and causality. Spatial proximity between CAFs and macrophages is biologically informative, but colocalization alone does not prove direct interaction. Two cell populations may occupy the same region because they respond to the same hypoxic, fibrotic, vascular, or inflammatory condition, rather than because one directly instructs the other. Therefore, spatial enrichment should be interpreted as evidence of a candidate niche, not as proof of a functional signaling circuit.^16–18

Ligand–receptor prediction has similar limitations. Tools such as CellPhoneDB can infer potential cell–cell communication from the paired expression of ligands and receptors, but they do not prove protein secretion, receptor engagement, downstream pathway activation, or biological consequence.¹⁵⁴ In gastric cancer, predicted SPP1–integrin, collagen–integrin, CCL2–CCR2, or C3–C3AR1 interactions should therefore be treated as hypotheses unless supported by phospho-signaling, ligand blockade, receptor inhibition, or genetic perturbation. The THBS2⁺ CAF–C3–C3AR1–SPP1⁺ TAM axis in peritoneal metastasis is a stronger example because it combines spatial evidence with pathway intervention and immunotherapy-response testing.¹⁷

Bulk transcriptomic correlation is even less specific. A correlation between THBS2, SPP1, CCL2, CSF1, IL6, or collagen genes in bulk RNA-seq may reflect mixed cell composition, tumor purity, stromal content, or shared inflammatory programs, rather than direct CAF–macrophage communication. Deconvolution methods can estimate cell-type abundance from bulk tissue, but their accuracy depends on reference profiles, tissue composition, and closely related cell states.¹⁵⁵ Thus, bulk data are useful for screening large cohorts and generating hypotheses, but they cannot establish cell source, spatial proximity, or causal direction.

Future studies should move from association to causality through layered validation. First, spatial methods should confirm that the relevant CAF and macrophage states are physically organized in the same compartment. Second, functional assays should test whether blocking the predicted ligand, receptor, or downstream pathway changes macrophage state, CD8⁺ T-cell localization, tumor-cell survival, or drug response. Third, in vivo or ex vivo models should preserve tissue architecture, such as organoid–stroma–immune co-cultures, tumor slices, patient-derived xenografts with stromal reconstruction, or spatially sampled clinical biopsies before and after treatment. Only niches supported by these steps should be advanced as therapeutic targets rather than descriptive spatial patterns.

Primary versus Metastatic Lesions

Another key challenge is that the primary gastric tumor cannot fully represent the microenvironment of metastatic lesions. Single-cell analyses of gastric cancer have shown substantial heterogeneity between primary tumors and organ-specific metastases, including liver, peritoneal, ovarian, and lymph node lesions.¹⁵⁶ Therefore, biomarkers derived only from archival primary tumor tissue may misclassify the active resistance niche in patients with metastatic disease.

This issue is particularly important in peritoneal metastasis. Gastric cancer peritoneal metastasis has a distinct stromal–myeloid ecosystem, characterized by THBS2⁺ matrix CAFs, SPP1⁺ TAMs, and C3–C3AR1-dependent macrophage accumulation, which differs from the immune architecture of primary gastric cancer.¹⁷ Ascites-based single-cell analysis also shows that macrophage lineage transitions contribute to immune evasion in peritoneal metastatic disease.⁶² These findings indicate that peritoneal lesions and malignant ascites should be prioritized for biomarker assessment when complement blockade, SPP1/TAM targeting, or ICB combinations are being considered.

Liver metastasis may also require separate evaluation. Single-cell profiling comparing gastric primary tumors and liver metastases has revealed differences in immune-cell composition, myeloid-cell gene expression, NK-cell features, and cell–cell communication pathways.¹⁵⁷ Thus, a CAF–macrophage signature detected in the primary tumor may not predict the dominant immune or stromal barrier in liver metastasis. For liver-dominant disease, patient selection should ideally use liver metastatic tissue, or at least circulating and imaging markers interpreted in relation to hepatic tumor biology.

In clinical trials, the most relevant specimen should be the lesion that treatment is intended to control. If the dominant measurable disease is peritoneal, liver, or nodal metastasis, baseline and on-treatment biopsies should preferentially sample that site whenever feasible. Primary tumor tissue remains useful when metastatic sampling is unsafe or unavailable, but it should be treated as a surrogate rather than definitive evidence. Future niche-guided trials should therefore report the source of biomarker tissue, distinguish primary and metastatic samples, and avoid assuming that a stromal–myeloid program observed in the primary tumor is therapeutically relevant in all metastatic compartments.

Pathway-Specific Limitations and Safety Concerns

Safety is a central consideration in CAF–TAM niche-directed therapy because fibroblasts, macrophages, chemokines, complement proteins, and related signaling pathways are not restricted to tumors. These systems contribute to antimicrobial defense, apoptotic-cell clearance, wound healing, vascular integrity, epithelial homeostasis, immune regulation, and tissue repair. Consequently, therapeutic benefit cannot be inferred simply from the protumor association of a marker or pathway. The principal translational challenge is to modulate a pathogenic CAF–TAM interaction without eliminating protective stromal or myeloid functions. The preferred objective should therefore be biomarker-selected, reversible, and, where feasible, spatially restricted niche modulation rather than prolonged systemic depletion or pathway ablation.

Direct macrophage targeting may impair host defense, apoptotic-cell clearance, tissue repair, and organ homeostasis. TAM depletion or reprogramming may reduce immune suppression, but total macrophage abundance does not distinguish pathogenic TAM states from tissue-resident or reparative macrophages.^30,158,159 CSF1R blockade is particularly relevant because CSF1R signaling supports macrophage survival and differentiation; however, prolonged or poorly selected blockade may also deplete protective tissue-resident populations.^159,160 Macrophage-directed therapies should therefore be guided by state- and site-specific biomarkers, accompanied by pharmacodynamic assessment, and designed to avoid sustained non-selective depletion.

Chemokine targeting presents both safety and efficacy challenges because multiple chemokine pathways can contribute to myeloid-cell recruitment, while physiological chemokine signaling is required for leukocyte trafficking and host defense.⁴⁰ Inhibition of a single recruitment pathway may be bypassed by alternative chemokine signals or myeloid populations, while systemic blockade could interfere with infection control and tissue repair. CCL2–CCR2 inhibition additionally raises a treatment-sequencing concern: in preclinical breast cancer models, cessation of CCL2 inhibition triggered rebound monocyte recruitment, angiogenesis, and accelerated metastatic outgrowth.⁵⁰ Although this observation has not been demonstrated in gastric cancer, it cautions against abrupt withdrawal and supports longitudinal monitoring for compensatory myeloid recruitment. These limitations reinforce the need for biomarker selection, rational combination therapy, and predefined discontinuation strategies.

Complement targeting may be particularly relevant to gastric cancer peritoneal metastasis, in which C3–C3AR1 signaling promotes the accumulation and transition of macrophages toward an SPP1⁺ TAM state.¹⁷ Nevertheless, complement has context-dependent antitumor and protumor functions and contributes to opsonization, bacterial killing, immune-complex clearance, and tissue homeostasis.⁶⁵ Systemic complement inhibition may therefore have different biological and safety consequences from selective interruption of locally produced CAF-derived complement. Systemic complement blockade can increase susceptibility to infection; this risk is particularly well established for terminal complement inhibition, with invasive meningococcal disease reported in patients receiving eculizumab despite meningococcal vaccination.¹⁶¹ Clinical development should consequently distinguish local from systemic complement activity and incorporate vaccination planning, infection-risk assessment, prophylactic measures where appropriate, and close clinical monitoring. Complement-axis intervention should remain restricted to biomarker-selected patients until its safety and efficacy are established prospectively in gastric cancer.

Stromal targeting carries the risk of disrupting protective tissue architecture and repair. CAFs are heterogeneous, and some fibroblast states may restrain tumor progression, preserve tissue integrity, or limit tissue injury.³⁹ Non-selective depletion of activated fibroblasts could therefore impair wound healing, destabilize tissue structure, or remove tumor-restraining stromal functions. FAP targeting requires particular caution because FAP-expressing stromal cells are not completely tumor-specific. Experimental immune targeting of FAP caused recognition of multipotent bone marrow stromal cells and cachexia in preclinical models.¹⁶² These findings argue against treating FAP positivity as sufficient justification for systemic fibroblast depletion. Safer approaches may include FAP-guided imaging, tumor-localized delivery, transient pathway modulation, or stromal normalization rather than eradication.

TGF-β blockade has a broad therapeutic rationale because TGF-β regulates CAF activation, ECM deposition, immune exclusion, EMT, and macrophage polarization. However, TGF-β also maintains immune tolerance, epithelial integrity, cardiovascular homeostasis, and wound repair. TGF-β pathway inhibitors have been associated with concerns such as skin toxicity and cardiovascular effects in some settings, reflecting the pleiotropic biology of this pathway.¹⁶³ Therefore, TGF-β targeting should be guided by a clear TGF-β/CAF/immune-exclusion signature and may be safer when delivered intermittently or in combination regimens designed to avoid continuous systemic suppression.

Overall, the safety of CAF–TAM niche-directed therapy should be addressed prospectively through trial design rather than treated as a secondary consideration. Future gastric cancer studies should incorporate baseline infection risk, wound-healing status, nutritional reserve, recent surgery, liver function, autoimmune history, and the physiological roles of the targeted pathway. Pharmacodynamic endpoints should demonstrate modulation of the intended pathogenic niche without excessive depletion of protective macrophages, systemic complement activity, physiological leukocyte recruitment, or reparative fibroblast functions. Spatially restricted delivery, intermittent dosing, reversible inhibition, and biomarker-defined treatment duration may reduce on-target toxicity. These safeguards are essential before niche-directed approaches can progress from mechanistic hypotheses to clinically useful strategies.

Required Clinical Trial Designs

Future clinical trials of fibroblast–macrophage niche-targeted therapy should be designed to prove microenvironmental remodeling, not only tumor shrinkage. Because these approaches aim to modify stromal barriers, macrophage states, vascular access, and CD8⁺ T-cell localization, conventional response endpoints alone are insufficient. Trials should include baseline and on-treatment paired biopsies whenever feasible, allowing investigators to determine whether the intended niche was actually disrupted before interpreting clinical efficacy.^113,151

Spatial biomarker stratification should be built into trial design. Patients should not be enrolled only by broad disease category or single-marker positivity. Instead, trials should prospectively define strata such as THBS2⁺ CAF–SPP1⁺ TAM-high peritoneal metastasis, CCL2⁺ fibroblast–p-STAT3⁺ macrophage-rich tumors, GREM1⁺ fibroblast–SPP1⁺ macrophage spatial proximity, FAP⁺/SPP1⁺/CD8 immune-excluded tumors, or collagen-high matrix-restricted tumors.^16–18 These biomarkers may initially be integrated rather than mandatory, but later-phase studies should test whether they predict benefit from specific niche-directed combinations.

Primary and metastatic lesions should be analyzed separately. Trials should record whether biomarker tissue comes from the primary tumor, peritoneal lesion, liver metastasis, lymph node, or ascites. When the therapeutic goal is to control metastatic disease, the treated lesion should be prioritized for biomarker assessment. This is particularly important because gastric cancer metastases can show organ-specific immune and stromal programs, and peritoneal metastasis has a distinct CAF–macrophage ecosystem.^17,156

Pharmacodynamic endpoints should be prespecified. Examples include reduced SPP1⁺ TAM density, reduced p-STAT3⁺ macrophages, decreased C3–C3AR1 signaling, lower collagen density or fiber alignment, improved vessel normalization, reduced hypoxia, shorter CD8⁺ T-cell distance to tumor epithelium, and increased CD8⁺ tumor-nest penetration. These endpoints should be measured early, before radiographic response is expected. Phase 0 or window-of-opportunity designs may be particularly useful when the main question is whether a drug reaches and remodels the target niche in human tumors.^164,165

Trials should also monitor niche regression and reconstruction over time. A treatment may initially reduce macrophage recruitment or stromal density, but chemotherapy, ICB-induced inflammation, or anti-angiogenic pressure may later drive compensatory CAF activation, monocyte influx, or complement signaling. Longitudinal single-cell and spatial studies in gastric cancer show that the tumor microenvironment can be substantially reconstructed after therapy.¹¹³ Therefore, post-progression biopsies, ascites profiling, and circulating stromal–myeloid markers should be incorporated to identify adaptive resistance.

Finally, combination trials should evaluate sequence, not only combination. For example, matrix or vascular normalization may need to precede ICB; CCR2, CSF1R, or C3–C3AR1 blockade may need to reduce suppressive myeloid pressure before PD-1/PD-L1 inhibition; and stromal modulation may need to improve tissue penetration before antibody or ADC therapy. Randomized arms comparing concurrent versus sequential administration would be more informative than adding all agents simultaneously. The optimal schedule should be defined by pharmacodynamic remodeling, safety, and clinical outcome together.

Overall, the next generation of gastric cancer trials should combine spatially informed eligibility, paired tissue sampling, lesion-specific biomarker assessment, pharmacodynamic endpoints, longitudinal monitoring, and sequence-testing arms. This design framework is necessary to determine whether fibroblast–macrophage niches are merely prognostic tissue features or truly actionable therapeutic dependencies. A future clinical framework integrating lesion-specific sampling, spatial niche profiling, biomarker-guided patient selection, rational combination therapy, and longitudinal monitoring is proposed in Figure 4.

Figure 4.

Proposed framework for CAF–TAM niche-informed patient stratification and therapeutic evaluation in gastric cancer. Clinically relevant samples may be obtained from primary gastric tumors, representative metastatic lesions, or malignant ascites when applicable. Multiplex imaging, spatial transcriptomics, and digital pathology may be used to assess CAF–TAM proximity, CD8⁺ T-cell localization, and matrix architecture. These spatial features may support the assignment of candidate inflammatory recruitment, matrix-associated myeloid, or complement-driven peritoneal niches. The inflammatory recruitment niche encompasses CCL2–CCR2-mediated monocyte recruitment and IL-6–STAT3-associated macrophage activation; these are related but distinct signaling processes rather than a single linear pathway. The matrix-associated myeloid niche is characterized by GREM1⁺ fibroblast–SPP1⁺ macrophage proximity and candidate collagen–integrin communication. The THBS2⁺ matrix CAF–C3/C3AR1–SPP1⁺ TAM circuit in gastric cancer peritoneal metastasis currently has the strongest spatial and functional support. Candidate niche-directed interventions may be investigated in combination with established systemic therapies, followed by on-treatment reassessment of niche regression, persistence, or reconstruction. CAF–TAM proximity, CD8⁺ T-cell localization, and matrix architecture are presented as exploratory pharmacodynamic readouts. Arrows indicate the proposed progression of the research and evaluation workflow and do not imply established biological causality or clinical efficacy. This framework is hypothesis-generating and requires prospective experimental and clinical validation.

Conclusions

Fibroblast–macrophage niches provide a spatially organized framework for understanding how immune escape and therapeutic resistance may arise in gastric cancer. Rather than functioning as isolated stromal or myeloid populations, CAFs and TAMs may interact through chemokine networks, complement signaling, cytokine programs, extracellular matrix remodeling, and immunometabolic pathways. Depending on the tissue context and strength of supporting evidence, these interactions may recruit or reprogram macrophage states, reshape collagen-rich stromal architecture, restrict CD8⁺ T-cell infiltration, contribute to vascular dysfunction, and impair therapeutic delivery. Gastric cancer peritoneal metastasis currently provides the strongest example of a spatially organized and functionally supported CAF–TAM niche, whereas analogous interactions in pMMR/MSS, diffuse-type, stroma-rich, immune-excluded, and treatment-resistant gastric cancers remain less completely validated.

Selective targeting of disease-promoting CAF–TAM niches is therefore a promising therapeutic hypothesis rather than an established clinical strategy. Its successful translation will require spatial validation of the relevant cell states in human gastric cancer tissues, direct perturbation of the proposed signaling circuits, confirmation in physiologically relevant experimental models, and prospective evaluation of predictive biomarkers, safety, and therapeutic efficacy in clinical trials. Future studies should also determine whether niche-directed interventions provide greater benefit than non-selective CAF or macrophage depletion and identify which molecular subtype, metastatic site, and treatment context is most likely to respond. Ultimately, the clinical value of fibroblast–macrophage targeting will depend on demonstrating that a specific niche constitutes a causal and therapeutically actionable dependency in a clearly defined patient population.

Funding Statement

No funding was received for this work.

Abbreviations

ACRG, Asian Cancer Research Group; ADC, antibody–drug conjugate; AHR, aryl hydrocarbon receptor; CAF, cancer-associated fibroblast; C1Q, complement component 1q; C3, complement component 3; C3AR1, complement C3a receptor 1; CCL2, C-C motif chemokine ligand 2; CCR2, C-C motif chemokine receptor 2; CLDN18.2, claudin 18 isoform 2; CPS, combined positive score; CPT1C, carnitine palmitoyltransferase 1C; CSF1, colony-stimulating factor 1; CSF1R, colony-stimulating factor 1 receptor; CXCL, C-X-C motif chemokine ligand; CXCR, C-X-C motif chemokine receptor; ECM, extracellular matrix; EMT, epithelial–mesenchymal transition; FAP, fibroblast activation protein; FAPI-PET, fibroblast activation protein inhibitor positron emission tomography; FCN1, ficolin 1; FGFR2b, fibroblast growth factor receptor 2b; GCPM, gastric cancer peritoneal metastasis; GREM1, gremlin 1; HER2, human epidermal growth factor receptor 2; HIF-1α, hypoxia-inducible factor 1α; ICB, immune checkpoint blockade; IFN-γ, interferon-γ; IHC, immunohistochemistry; IL, interleukin; JAK, Janus kinase; LOX, lysyl oxidase; MDSC, myeloid-derived suppressor cell; MIF, macrophage migration inhibitory factor; MMR, mismatch repair; MSI, microsatellite instability; MSI-H, microsatellite instability-high; MSS, microsatellite stable; NK, natural killer; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; pMMR, proficient mismatch repair; scRNA-seq, single-cell RNA sequencing; SIRPα, signal regulatory protein α; SPP1, secreted phosphoprotein 1; STAT3, signal transducer and activator of transcription 3; TAM, tumor-associated macrophage; TCGA, The Cancer Genome Atlas; TGF-β, transforming growth factor-β; THBS2, thrombospondin 2; TLR, Toll-like receptor; TREM2, triggering receptor expressed on myeloid cells 2; VEGF, vascular endothelial growth factor; VEGFR2, vascular endothelial growth factor receptor 2.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests in this work.