Remodeling the Stroma: The Role of ncRNAs in Cancer-Associated Fibroblasts and Therapeutic Implications for Digestive Tumors

Abstract

The crosstalk between cancer-associated fibroblasts (CAFs) and tumor cells presents a formidable barrier to treating digestive malignancies, yet the molecular syntax governing this interaction is often overlooked. Recent studies have highlighted the indispensable function of non-coding RNAs (ncRNAs), which act not merely as bystanders but as architects of stromal remodeling. Here, we dissect how ncRNAs dictate the heterogeneity and function of CAFs, regulating fibroblast activation, metastatic niche formation, drug resistance, and metabolic reprogramming. We further examine how the exosomal transfer of ncRNAs reshapes the immune landscape, engineering a tumor-permissive microenvironment. Current clinical observations indicate, however, that indiscriminate depletion of CAFs is often counterproductive, underscoring the necessity of “stromal normalization.” Targeting specific ncRNA master regulators to reprogram activated CAFs to a quiescent state—rather than destroying them—provides robust mechanistic support for the stromal normalization paradigm previously proposed.

Introduction

Even with recent progress in molecular therapy, treating digestive system tumors—such as gastric, colorectal, pancreatic, and liver cancers—is still a major clinical challenge. The main reason for treatment failure is drug resistance.1,2 Historically, cancer research focused mostly on genetic mutations inside tumor cells. Newer studies show that the limits of current chemotherapy and immunotherapy are actually set by the tumor microenvironment (TME), particularly the dense fibrosis known as the desmoplastic reaction.3 Inside this environment, cancer-associated fibroblasts (CAFs) do more than just provide structural support. They actively block drugs and immune cells from entering the tumor, creating a physical barrier that helps cancer cells survive and grow.4

Thanks to advanced tools like single-cell RNA sequencing and spatial transcriptomics, we now have a better understanding of the stroma. It is clear that CAFs are not a single, uniform group. Instead, they consist of diverse populations with different functions. Key subtypes include inflammatory CAFs (iCAFs), myofibroblastic CAFs (myCAFs), and antigen-presenting CAFs (apCAFs), each found in specific locations within the tumor.5,6 Despite this knowledge, we still do not fully understand the epigenetic mechanisms that control these different phenotypes and their precise interactions with cancer cells. Recent research suggests that non-coding RNAs (ncRNAs)—including microRNAs (miRNAs), long non-coding RNAs (lncRNAs), and circular RNAs (circRNAs)—act as the main messengers in this communication.

Unlike cytokines, which usually spread through passive diffusion, ncRNAs work differently. Current research suggests they rely on a specific sorting mechanism. RNA-binding proteins load these ncRNAs into extracellular vesicles, mostly exosomes, allowing them to travel safely between stromal and tumor cells.7 This interaction creates a cycle that promotes cancer. For example, ncRNAs from the tumor can change normal fibroblasts into activated CAFs. In return, CAFs release ncRNAs that alter tumor metabolism, maintain stem-like properties, and stop immune cells from entering the tissue.8

Although many studies have linked specific ncRNAs to cancer, we still lack a clear picture of how these mechanisms relate to the newly discovered diversity of CAFs. This review aims to bridge that gap for digestive system tumors. From a biological standpoint, these digestive tract malignancies—particularly pancreatic, gastric, and colorectal cancers—are uniquely characterized by an exceptionally prominent desmoplastic reaction and a dense, rigid extracellular matrix. This shared histopathological architecture serves not merely as a passive physical impediment to drug delivery but as an active biochemical signaling hub driven by mechanical tension, providing a compelling biological rationale for synthesizing their stromal regulatory networks under a unified framework. Instead of just listing phenotypes, we organize the current knowledge into five main areas. First, we discuss how ncRNAs turn normal cells into activated CAFs. Next, we look at how they change the physical environment to support metastasis. We also examine their role in drug resistance, for instance by protecting tumor cells from apoptosis. We analyze how these molecules affect the immune system and alter cell metabolism. To conclude, we discuss the clinical potential of these findings. We suggest that identifying key ncRNAs could lead to treatments that “normalize” the stroma, rather than simply trying to destroy it.

By systematically applying a five-domain framework—comprising CAF activation, metastasis, metabolic reprogramming, therapeutic resistance, and immune evasion—this review uncovers critical insights often overlooked in recent pan-cancer syntheses that treat CAFs as a monolithic entity. Specifically, we argue that ncRNAs act as the precise “programmers” of CAF heterogeneity rather than mere biomarkers of activation.

In the realm of lineage commitment, ncRNAs bidirectionally dictate subtype trajectories; for instance, circCUL2 specifically drives the inflammatory iCAF phenotype in pancreatic ductal adenocarcinoma (PDAC), while tumor-derived exosomal miR-4736 can precisely restrain myCAF activation by targeting the ATG7-autophagy axis. This precision extends to metabolic and metastatic remodeling, where myCAF-derived exosomal PWAR6 in colorectal cancer facilitates liver metastasis by intricately altering stromal glutamine availability—a subtype-specific dynamic that traditional “bulk” analyses fail to capture.

Furthermore, ncRNAs redefine our understanding of the stromal barrier. In therapeutic resistance, specific iCAF subpopulations in PDAC actively drive chemoresistance by secreting IL-8, which induces UPK1A-AS1 in cancer cells to mediate DNA damage repair, challenging the notion of CAFs as a purely physical obstacle. Finally, in reshaping the immune landscape, ncRNAs enable “surgical strikes” on immune surveillance, such as PWAR6-mediated compromise of NK cell function or RP11-161H23.5-induced degradation of HLA-A, allowing for exact immune evasion without global immunosuppression.

Collectively, these mechanism-driven insights across five domains provide a robust biological rationale for a paradigm shift. By nominating miR-21, UPK1A-AS1, and circCUL2 as tractable targets, we advocate for transitioning from broad “stromal depletion” toward precision “stromal normalization”—a strategy particularly vital for gastrointestinal malignancies characterized by dense, therapy-resistant desmoplasia.

CAFs: The Orchestrators of the Desmoplastic Niche

Lineage Plasticity and Epigenetic Origins

Cancer-associated fibroblasts (CAFs) are the activated form of the stromal mesenchyme. They behave differently from the quiet, normal fibroblasts that keep tissues healthy. Unlike cancer cells, which are driven by random genetic mutations, CAFs usually have stable genomes. Instead, their main feature is high epigenetic plasticity. This means that becoming a CAF is not a genetic event. Rather, it is a reprogramming process caused by epigenetic changes. This change happens because of the constant inflammation in the tumor, often described as “a wound that never heals”.9 In the past, scientists thought CAFs came mostly from local normal fibroblasts (NFs) stimulated by tumor signals like TGF-β. New tracking studies show that the origin of CAFs is more complex.10 In digestive system tumors, these cells come from several different sources, not just local fibroblasts. For example, in liver and pancreatic cancers, the main source is quiescent stellate cells (HSCs or PSCs). These cells activate quickly to produce collagen and scar tissue. Non-fibroblast cells also contribute. Endothelial cells can change their type through a process called endothelial-to-mesenchymal transition (EndMT) to become fibroblast-like. Also, mesenchymal stem cells (MSCs) from the bone marrow can travel to the tumor and turn into CAFs.

The origin of the cell matters. It is not just history; it determines what the CAF does. CAFs keep an “epigenetic memory” of where they came from. This memory influences where they live in the tumor and how they behave.11,12 For instance, CAFs that come from endothelial cells tend to stay near blood vessels to help with blood supply and metastasis. In contrast, those from stellate cells are better at changing metabolism and stiffening the tissue. Because of these different origins, different ncRNA networks are likely needed to control these changes. We should not see CAFs as one uniform group. Instead, they are a mix of different subsets, and treatments need to be specific to their developmental origins.

Spatial Heterogeneity and Functional Polarization of CAFs

Historically, CAFs were identified by a monolithic set of mesenchymal markers, such as α-smooth muscle actin (α-SMA), fibroblast activation protein (FAP), and Vimentin. The lack of exclusivity of these markers often led to the conflation of CAFs with normal physiological fibroblasts or other mesenchymal lineages. The recent application of high-resolution single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics has dismantled this singular view. While large-scale pan-cancer atlases (Luo et al) have recently mapped conserved clusters—including myofibroblastic, inflammatory, and adipogenic CAFs—across diverse solid tumors,13 in the specific context of digestive malignancies, CAFs are typically classified into two dominant transcriptional states governed by their spatial topography and distinct signaling cues: myofibroblastic CAFs (myCAFs) and inflammatory CAFs (iCAFs).14

Myofibroblastic CAFs (myCAFs) are typically located in the juxtatumoral niche, maintaining direct physical contact with neoplastic cells. Driven by high local concentrations of TGF-β, myCAFs exhibit a contractile phenotype characterized by high α-SMA expression and low inflammatory cytokine production, functioning primarily to remodel the extracellular matrix (ECM) and form the desmoplastic barrier.15 Conversely, inflammatory CAFs (iCAFs) reside in the distal stroma, spatially separated from tumor nests. Induced by soluble factors such as IL-1 and TNF-α propagated through the stroma, iCAFs are characterized by low α-SMA levels but robust secretion of inflammatory cytokines and chemokines (eg., IL-6, LIF, CXCL12), acting as the primary architects of the immunosuppressive microenvironment.15

Beyond this dichotomy, a more specialized subset termed antigen-presenting CAFs (apCAFs) has recently been identified in specific digestive tumors.16 These cells express Major Histocompatibility Complex class II (MHC-II) molecules (eg., HLA-DRA) and theoretically possess the capacity to present antigens to CD4+ T cells. Unlike professional dendritic cells, apCAFs lack essential costimulatory molecules (such as CD80 and CD86), leading to T-cell anergy or differentiation into regulatory T cells (Tregs) rather than effective immune activation.17 This spatial and functional heterogeneity implies that “anti-CAF” therapies must be highly selective; indiscriminate depletion of all CAFs—including tumor-restraining myCAF subsets—has paradoxically accelerated tumor progression in preclinical models. Deciphering the ncRNA networks that act as “molecular switches” between the tumor-promoting iCAF phenotype and the potentially restrictive myCAF phenotype represents a critical frontier for therapeutic intervention.

Functional Duality: From Pro-Tumorigenic Engines to Stromal Restraints

Elucidating the functional spectrum of CAFs reveals a complex landscape where these cells act not merely as supportive bystanders but as active architects of the TME. While historically viewed as strictly pro-tumorigenic, contemporary consensus acknowledges the functional duality of CAFs. They show malignancy through a “two-pronged” mechanism: the secretion of a soluble cargo (paracrine signaling) and direct physical remodeling of the stromal architecture (contact-dependent signaling).

In the context of tumor promotion, CAFs function as the primary drivers of invasion and metabolic reprogramming. Mechanistically, CAFs generate physical forces to remodel the ECM, creating “tracks” for cancer cell dissemination. For instance, Labernadie et al demonstrated that CAFs form a mechanically active heterotypic adhesion with cancer cells via the E-cadherin/N-cadherin axis, literally “pulling” cancer cells to invade surrounding tissues.18 In colorectal cancer, this mechanical remodeling extends to the basement membrane; CAFs have been shown to soften and stretch this barrier in a protease-independent manner, facilitating neoplastic penetration.19 Beyond physical assistance, CAFs engage in metabolic cross-feeding. In the nutrient-deprived milieu of pancreatic ductal adenocarcinoma (PDAC), CAFs undergo autophagy to secrete non-essential amino acids (specifically alanine), which tumor cells uptake to fuel their central carbon metabolism, thereby sustaining growth under metabolic stress.20 CAFs serve as crucial angiogenic hubs. By secreting pro-angiogenic factors such as VEGF and TGF-β, they remodel the vascular architecture to meet the high oxygen demands of the expanding tumor mass.21,22

Parallel to these structural and metabolic roles, CAFs are the master regulators of the immune landscape. They erect an immunosuppressive barrier that blunts the efficacy of immunotherapy. In colorectal malignancies, CAF-derived IL-6 has been implicated in a synergistic loop that recruits M2-like tumor-associated macrophages (TAMs) while simultaneously suppressing the cytotoxic activity of Natural Killer (NK) cells, effectively rendering the tumor “cold” to immune surveillance.23 Other studies confirm that CAF-derived interleukins, such as IL-6 and IL-33, act as potent inducers of the epithelial-mesenchymal transition (EMT), further linking immune evasion with metastatic potential.24,25

The paradigm that “all CAFs are bad” has been challenged by the identification of cancer-restraining CAFs (rCAFs). This discovery highlights a critical pitfall in therapeutic strategies: the total depletion of the stromal compartment often removes these protective brakes, paradoxically accelerating tumor progression. Recent high-profile studies have begun to characterize markers for these restrictive subsets. Mizutani et al identified Meflin (ISLR), a GPI-anchored protein, as a marker of rCAFs in pancreatic cancer; its loss correlates with straightened collagen fibers and aggressive disease, suggesting Meflin preserves “stromal softness” to inhibit malignancy.26 Similarly, the composition of the ECM itself acts as a double-edged sword. While specific collagen architectures drive invasion, Bhattacharjee et al revealed that type I collagen produced by myCAFs can mechanically restrain tumor dissemination.27 This was corroborated by Chen et al, who demonstrated that the specific deletion of type I collagen in α-SMA+ myofibroblasts collapsed the immune barrier, recruiting myeloid-derived suppressor cells (MDSCs) and accelerating PDAC progression.28 These findings include the necessity of stromal normalization—targeting specific protumorigenic signaling (often regulated by ncRNAs) while preserving the protective rCAF populations.

General Principles of CAF-Associated ncRNAs: Origins and Mechanisms

Cellular Origins and Bidirectional Dynamics

When considering the origin of ncRNAs within the tumor microenvironment, it is essential to focus on their functional cellular source over their transcriptional origin. In this bidirectional dialogue, while immune infiltrates, such as TAMs, also contribute to the extracellular ncRNA pool by releasing signals that modulate stromal plasticity and inflammation, this review specifically concentrates on the direct, bidirectional dialogue between CAFs and tumor cells. In this context, ncRNAs primarily originate from two key cellular sources, creating a complex signaling network.9 CAFs can serve as the source, releasing ncRNAs that directly modulate the malignant behavior of tumor cells, while conversely, tumor cells can secrete their own ncRNAs to “educate” and reprogram the surrounding stroma.

This dual-sourcing and the resulting feedback loops are illustrated by several compelling examples. Similarly, in gastric cancer, tumor-derived exosomes transport miR-124-3p to normal fibroblasts, suppressing local gene expression to induce CAF activation.29 Conversely, miR-21 acts as a reciprocal messenger; highly secreted by activated CAFs, it is transferred to esophageal cancer cells to enhance metastasis and chemoresistance, reinforcing the symbiotic nature of this relationship.30 These elegant mechanisms demonstrate that the cellular source of an ncRNA is not a static identity but a dynamic role within a complex interplay. Deciphering whether an ncRNA is CAF-derived, tumor-derived, or sustains a bidirectional feedback loop is crucial for identifying precise therapeutic targets to disrupt this malignant crosstalk.31

Molecular Modes of Action: From Decoys to Scaffolds

While various classes of ncRNAs contribute to the TME crosstalk, lncRNAs are particularly distinguished by their structural complexity. The functional versatility of lncRNAs stems from their ability to operate through several distinct modes of action, which can be broadly classified into four archetypal roles.32 First, they can act as signals, where their expression levels serve as indicators of cellular states or responses to stimuli. Second, they can function as decoys, acting as molecular sponges that bind and sequester other biomolecules, such as microRNAs (in a mechanism known as competing endogenous RNA, or ceRNA) or transcription factors, thereby titrating their availability and function.33 Third, lncRNAs can serve as guides, recruiting regulatory protein complexes (eg., chromatin modifiers) to specific genomic loci to modulate gene expression in cis or trans.34 Finally, they can operate as scaffolds, providing a structural platform to facilitate the assembly of multiple proteins into a stable and functional ribonucleoprotein complex.32 Within the tumor microenvironment, these modes of action are the fundamental strategies by which ncRNAs mediate the intricate crosstalk between CAFs and tumor cells, and a single ncRNA may even employ multiple strategies to exert its influence.

Elucidating the precise mode of action for a given ncRNA is a critical step in understanding its ultimate impact on cancer progression. For lncRNAs acting as decoys, a primary mechanism is the “miRNA sponge” effect within the ceRNA network. For example, CAF-derived LINC01410 acts as a sponge for miR-122-5p in esophageal cancer, thereby liberating target mRNAs like PKM2 from suppression.35 As guides, the mechanism involves the lncRNA directly binding to chromatin-modifying enzymes, such as Polycomb Repressive Complex 2 (PRC2), to induce epigenetic silencing. A case in point is ANRIL, which guides chromatin modifiers to the promoter of the ABCB1 transporter gene in gastric cancer, upregulating its expression and causing drug resistance.36 When functioning as scaffolds, the mechanism entails the lncRNA serving as a structural backbone to bring multiple proteins into close proximity. For instance, exosomal FAL1 in colorectal cancer acts as a scaffold for the interaction between Beclin1 and the E3 ligase TRIM3, promoting Beclin1 degradation and suppressing autophagy.37 Finally, as signals, the mechanism can involve direct interaction with a protein to regulate its stability. Dissecting these diverse molecular mechanisms is fundamental to understanding how a single ncRNA can show complex biological outcomes in the CAF-tumor dialogue.

Functional Roles of CAF-Associated ncRNAs in Digestive System Neoplasms

Unlike generalized pan-cancer frameworks that treat stromal crosstalk as a universal phenomenon across disparate solid tumors, the regulatory networks of CAF-associated ncRNAs in digestive system neoplasms are inherently governed by a unique microenvironmental architecture. Malignancies of the digestive tract—most notably pancreatic ductal adenocarcinoma (PDAC), gastric cancer (GC), and colorectal cancer (CRC)—are histopathologically distinguished by an exceptionally profound desmoplastic reaction and an ultra-dense extracellular matrix.38 Crucially, this specialized stromal compartment does not merely pose a late-stage obstacle but actively dictates early oncogenesis, as evidenced by pioneering data demonstrating that CAFs directly induce acinar-to-ductal cell transdifferentiation and pancreatic cancer initiation via the LAMA5/ITGA4 axis.39 Furthermore, this rigid physical barrier generates intense mechanical solid stress and high interstitial fluid pressure, which actively compresses local microvessels to exacerbate intratumoral hypoxia and nutrient deprivation. Consequently, the intercellular delivery and decoy dynamics of CAF-associated ncRNAs in these organs are uniquely hardwired to cross-amplify mechanical tension, hypoxic stress, and metabolic starvation. This microenvironmental wiring is intimately coupled with the functional stratification of the stroma, as evidenced by landmark profiling in PDAC demonstrating that distinct CAF subpopulations uniquely orchestrate IL-6-mediated therapy resistance within this dense matrix.40 Therefore, isolating and synthesizing these specific ncRNA circuits under a dedicated digestive system framework is not merely an anatomical classification, but a biological necessity to uncover stromal normalization vectors that remain entirely obscured in broader, pan-cancer syntheses.

Functionally, the diverse molecular modes described in General Principles of CAF-Associated ncRNAs: Origins and Mechanisms—whether acting as decoys, scaffolds, guides, or signals—translate into tangible biological impacts within the tumor microenvironment. Given that the bidirectional interactions between CAFs and cancer cells are critical for disease trajectory, ncRNAs serve as the pivotal “molecular syntax” regulating these processes. In this review, the regulatory roles of CAF-associated ncRNAs in digestive system neoplasms are broadly categorized into five major aspects: (1) CAF activation and reprogramming, where ncRNAs induce the transformation of normal fibroblasts into activated phenotypes; (2) tumor progression and metastasis, driven by molecules like circ-0076535, and LINC01410 that often operate as ceRNAs or signaling activators; (3) therapeutic resistance, mediated by ncRNAs including ANRIL, DNM3OS, and Lnc-HOTAIR via mechanisms such as drug efflux and apoptosis inhibition; (4) immune evasion and regulation, involving modulators such as RP11-161H23.5 and NORAD; and (5) metabolic reprogramming, exemplified by NNT-AS1 and TUG1 which regulate glycolysis and lipid metabolism. These diverse biological impacts are comprehensively summarized in Tables 1, 2, and 3 and detailed in the following sections. The intricate interplay of these five functional categories is schematically summarized in Figure 1.

Table 1.

CAF-Associated LncRNAs and Their Potential Mechanisms in Digestive System Neoplasms

Type	LncRNAs	Source	Expression	Mechanism	Impact on Tumor	References
CAF Activation and Reprogramming	LINC00691	Cancer cells	Up	Triggers the JAK2/STAT3 signaling cascade	Reprograms NFs into activated CAFs in gastric cancer	[41]
	MALAT1	Cancer cells	Up	Directly inhibits autophagic flux	Enforces fibroblast activation and α-SMA expression in gastric cancer	[42]
	LNC-FSD2-31:1	Cancer cells	Down	Promotes vesicular packing of miR-4736 to target ATG7-autophagy axis	Activates CAFs to promote PDAC progression/restrains myCAF trajectory	[43]
	LINC01915	Cancer cells/CAFs	Up	Sponges miR-92a-3p to derepress KLF4/CH25H cascade	Limits NF uptake of tumor exosomes and blocks NF-to-CAF transition in CRC	[44]
Tumor Progression and Metastasis	PWAR6	myCAF-derived exosomes	Up	Alters stromal glutamine availability and compromises NK cell surveillance	Transforms local environment into a permissive sanctuary to accelerate CRC liver metastasis	[45]
	LINC01410	CAF-derived exosomes	Up	Sponges miR-122-5p to liberate target mRNAs like PKM2	Fuels glycolytic shifts, drives EMT programs, and promotes ESCC invasion	[35]
	Lnc-HOTAIR	CAF-derived exosomes	Up	Represses E-cadherin and upregulates N-cadherin and glycolytic enzymes	Enhances tumor proliferation, migration, and invasion, promoting EMT in ESCC	[46]
	LINC00659	CAF-derived exosomes	Up	Sponges miR-342-3p to upregulate ANXA2 via LINC00659/miR-342-3p/ANXA2 axis	Promotes CRC cell proliferation, migration, invasion, and EMT	[47]
	SNHG3	CAFs	Up	Binds miR-34b-5p to structurally stabilize HuR protein and enhance HOXC6 transcription	Facilitates carrying vesicles to drive colorectal cancer progression	[48]
	WEE2-AS1	CAF-derived exosomes	Up	Promotes MOB1A degradation to inhibit canonical Hippo pathway	Facilitates downstream cascades to drive colorectal cancer progression	[49]
	UCA1	CAFs	Up	Provokes long noncoding RNA UCA1 to activate the mTOR/KRAS signaling axis	Enhances colorectal cancer cell proliferation and metastatic potential	[50]
	LINC00152	CAFs	Up	Regulates miR-205-5p/CXCL11 signaling pathway	Sustains transcriptional activity to affect tumor cell growth in HCC	[51]
	MAPKAPK5-AS1	CAFs	Up	Binds and inhibits SMURF2 to stabilize TCF12, promoting FOXH1 transcription	Carries into malignant lineages to drive HCC cell proliferation	[52]
Therapeutic Resistance	UPK1A-AS1	CAFs	Up	Acts as molecular scaffold to enhance Ku70/Ku80 interaction and NHEJ	Driven by iCAFs to enhance double-strand break-induced oxaliplatin resistance in PDAC	[53]
	ST7-AS1	Cancer cells	Up	CAF-secreted midkine upregulates ST7-AS1 to activate PI3K/AKT pathway	Promotes EMT programs and acquired cisplatin resistance in gastric cancer	[54]
	LINC00460	Cancer cells	Up	Binds PDAP1 and mediates PDGFA/PDGFR signaling to promote CAF proliferation	Sustains stemness markers to confer gemcitabine chemoresistance in PDAC	[55]
	H19	CAFs	Up	Acts as a ceRNA for miR-141 to activate β-catenin and Notch1 signaling	Reinforces Wnt-driven stemness and reduces chemo-sensitivity in CRC	[56]
	CCAL	CAFs	Up	Interacts with HuR to stabilize β-catenin, activating Wnt pathway	Promotes oxaliplatin and 5-fluorouracil resistance in CRC	[57]
	POU3F3	Cancer cells	Up	Activates CAFs and promotes IL-6 secretion	Enhances cisplatin resistance of ESCC cells via protective niche	[58]
	DACT3-AS1	CAFs	Down	Regulates miR-181a-5p/SIRT1 signaling pathway as ferroptosis promoter	Loss of DACT3-AS1 suppresses ferroptosis to drive oxaliplatin resistance in gastric cancer	[59]
	FAL1	CAF-derived exosomes	Up	Acts as molecular scaffold to promote TRIM3-mediated Beclin1 degradation	Suppresses oxaliplatin-induced autophagy-dependent cell death in CRC	[37]
	ANRIL	CAFs	Up	Guides chromatin modifiers to ABCB1 promoter to induce epigenetic silencing	Epigenetically upregulates ABCB1 efflux pumps to drive cisplatin resistance in gastric cancer	[36]
	AFAP1-AS1	CAFs	Up	Activates internal ATM/CHK2 pathway to promote DNA damage repair	Attenuates radiation-induced double-strand breaks to enhance radioresistance in ESCC	[60]
	DNM3OS	Cancer cells	Up	Regulates PDGFβ/PDGFRβ/FOXO1 signaling network	Modulates DNA Damage Response (DDR) to confer ESCC radioresistance	[61]
Immune Evasion and Regulation	RP11-161H23.5	CAFs	Up	Transferred via exosomes to downregulate HLA-A expression in tumor cells	Diminishes MHC Class I density to impair CD8⁺ T cell antigen recognition in PDAC	[62]
	NORAD	Cancer cells	Up	Functions as a decoy sponge for miRNAs to regulate IL-33 release	Alters the TME cytokine profile to promote immunosuppressive cell infiltration in GC	[63]
Metabolic Reprogramming	NNT-AS1	CAF-derived exosomes	Up	Regulates the downstream target miR-889-3p/HIF-1α signaling axis	Unlocks transcriptional activation of glycolysis under hypoxia in PDAC	[64]
	TUG1	Cancer cells	Up	Sequesters and sponges miR-524-5p to release its target SIX1	Sustains glycolysis-related transcriptional programs to drive liver cancer proliferation	[65]

Table 2.

CAF-Associated MiRNAs and Their Potential Mechanisms in Digestive System Neoplasms

Type	miRNAs	Source	Expression	Mechanism	Impact on Tumor	References
CAF Activation and Reprogramming	miR-4736	Cancer cell EVs	Up	Encapsulated in EVs to target ATG7 and arrest autophagy	Restrains myCAF activation trajectory in PDAC	[43]
	miR-27	CAFs	Up	Promotes NF-to-CAF transition and TGF-β secretion	Establishes activated tumor stroma in esophageal cancer	[66]
	miR-31	CAFs	Up	Inhibits expression of autophagy-related genes (eg., Beclin-1)	Suppresses CAF autophagy to alter interactions and promote CRC	[67]
	miR-124-3p	Cancer cell exosomes	Up	Downregulates SNAI2 expression in recipient fibroblasts	Promotes fibroblast programmatic activation and migration in GC	[29]
	miR-141-3p	Cancer cells	Up	Directly targets STAT4 to inhibit Wnt/β-catenin pathway	Inhibits gastric cancer cell migration, invasion, and CAF formation	[68]
	miR-200	Cancer cells	Up	Targets ZEB1 to inhibit TGF-β–driven differentiation	Decreases CAF accumulation and tumor metastatic potential in CRC	[69]
Tumor Progression and Metastasis	miR-3656	CAF-derived exosomes	Up	Targets ACAP2 to activate PI3K/AKT and β-catenin signaling	Promotes proliferation, migration, and invasion of ESCC cells	[70]
	miR-199a-5p	CAFs	Up	Downregulates FKBP5 to activate the mTORC1 pathway	Promotes gastric cancer cell proliferation and malignancy	[71]
	miR-200b	CAFs	Down	CAFs induce promoter hypermethylation and silencing in tumor cells	Unleashes ZEB1/2-driven GC invasion and peritoneal metastasis	[72]
	miR-20a-5p	CAFs	Up	Targets PTEN and stimulates paracrine IL-6 production	Drives colorectal cancer cell progression	[73]
	miR-345-5p	CAF-derived exosomes	Up	Inhibits CDKN1A to lower the threshold for invasion	Promotes CRC cell proliferation, migration, and metastasis	[74]
	miR-5703	Cancer cells	Up	Reduces APJ suppression in fibroblasts induced by tumor cells	Drives nuclear reprogramming and stromal remodeling in CRC	[75]
	miR-181b-3p	CAF-derived exosomes	Up	Regulates target gene SNX2 and BMP5 expressions	Promotes CRC cell proliferation and metastatic cascades	[76]
	miR-146a-5p	Cancer cells	Up	Targets fibroblasts to activate reactive phenotypes	Promotes CRC xenograft formation and lung metastasis	[77]
	miR-155-5p	Cancer cells	Up	Interacts in synergistic loops to remodel stroma and drive EMT	Collaborates with miR-146a-5p to accelerate CRC aggressiveness	[77]
	miR-100-5p	CAF-derived exosomes	Down	Inhibits IGF1R/PI3K/AKT pathway in endothelial lineages	Facilitates tumor lymphatic endothelial cell (TLEC) tube formation in ESCC	[78]
	miR-522-3p	CAF-derived exosomes	Up	Targets BMP5 to downregulate its expression	Enhances CRC cell migration, invasion, and angiogenesis	[79]
	miR-21	CAFs	Up	Reciprocally targets tumor cell programs via S100A4 feed-forward loop	Enhances tumor cell invasion and reinforces reactive stroma in ESCC	[30]
Therapeutic Resistance	miR-522	CAFs	Up	Targets ALOX15 and blocks lipid-ROS accumulation	Suppresses ferroptosis to induce acquired chemoresistance in GC	[80]
	miR-24-3p	CAFs	Up	Downregulates CDX2/HEPH axis upon exosomal transfer	Accelerates methotrexate (MTX) resistance in CRC cells	[81]
	miR-3173-5p	CAF-derived exosomes	Up	Sponges ACSL4 to suppress lipid peroxidation after uptake	Inhibits ferroptosis and induces gemcitabine resistance in PDAC	[82]
Immune Evasion and Regulation	miR-214	CAF-derived exosomes	Down	Relieves direct PD-L1 suppression, elevating cytotoxic thresholds	Exhausts effector T cells and shields CRC from surveillance	[83]
	miR-889-3p	CAF-derived exosomes	Up	Targets STAT1 to suppress pro-inflammatory M1 phenotype	Represses M1 macrophage polarization to boost ESCC growth	[84]
	miR-4253	CAFs	Up	Interferes with macrophage transcriptional networks via exosomes	Induces anti-inflammatory M2 macrophage polarization in GC	[85]

Table 3.

CAF-Associated circRNAs and Their Potential Mechanisms in Digestive System Neoplasms

Type	circRNAs	Source	Expression	Mechanism	Impact on Tumor	References
CAF Activation and Reprogramming	circCUL2	CAFs	Up	Activates the MyD88-dependent NF-κB signaling pathway	Specifically drives the inflammatory iCAF phenotype and lineage trajectory in PDAC	[86]
Tumor Progression and Metastasis	circ_0076535	CAF-derived exosomes	Up	Sponges miR-145-5p to release downstream NF-κB pathways	Induces epithelial-mesenchymal transition (EMT) to promote ESCC proliferation and invasion	[87]
Immune Evasion and Regulation	circ_0067557	CAF-derived exosomes	Up	Upregulates the downstream Lin28A/Lin28B signaling axis	Reinforces the immunosuppressive microenvironment to boost CRC progression and chemoresistance	[88]

Figure 1.

Schematic overview of ncRNA-mediated crosstalk between CAFs and tumor cells in digestive system neoplasms. Bidirectional communication between CAFs and tumor cells is mediated by the transfer of ncRNAs (lncRNAs, circRNAs, miRNAs), often via exosomes. This crosstalk drives five major biological outcomes. (A) CAF Activation and Reprogramming: ncRNAs regulate signaling pathways and autophagy to transform normal fibroblasts into activated CAFs, a process counter-acted by tumor-suppressive ncRNAs. (B) Tumor Progression and Metastasis: ncRNAs promote key malignant phenotypes including proliferation, invasion, angiogenesis, and EMT. (C) Therapeutic Resistance: ncRNAs confer resistance by enhancing drug efflux, inhibiting apoptosis, and promoting DNA damage repair. (D) Immune Evasion: ncRNAs facilitate immune escape by downregulating antigen presentation, enhancing immune checkpoints, and promoting M2 macrophage polarization. (E) Metabolic Reprogramming: ncRNAs mediate a metabolic shift towards aerobic glycolysis (Warburg Effect) and facilitate metabolic support from CAFs to tumor cells (Reverse Warburg Effect).

CAF Activation and Reprogramming

Historically, the process of CAF activation was viewed through a macroscopic lens, where various ncRNAs were identified as general “on-switches” driving the transition from normal fibroblasts (NFs) to an activated pan-CAF phenotype (characterized by generic markers like α-SMA). For instance, lncRNAs such as LINC00691 and MALAT1, as well as miRNAs like miR-27, have been well-documented to reprogram NFs into generic CAFs via classical signaling pathways like JAK2/STAT3 and TGF-β.

However, a critical insight overlooked in these earlier bulk-level studies is the tremendous plasticity and heterogeneity of the resulting stroma. With the advent of advanced transcriptomics, recent high-resolution evidence reveals that this “activation” is not a uniform, monolithic process, but rather a precise lineage reprogramming. Rather than just turning fibroblasts “on”, specific ncRNAs act as molecular rheostats that bidirectionally dictate specific phenotypic trajectories. For example, circCUL2 specifically drives the inflammatory iCAF phenotype in PDAC,86 whereas the tumor-derived LNC-FSD2-31:1/miR-4736 axis precisely restrains myCAF activation.43 Therefore, transitioning from the concept of “generic activation” to “subtype-specific lineage reprogramming” is essential for understanding the true complexity of the initial stromal response.

To execute this context-dependent reprogramming—whether driving generic activation or specific lineage commitment—ncRNAs fundamentally hijack key intracellular signaling axes. A primary strategy involves the constitutive activation of inflammatory and pro-fibrotic pathways. This direct pathway engagement is also prominent in gastric cancer, where LINC00691 acts as a critical switch to reprogram NFs by triggering the JAK2/STAT3 cascade.41

Beyond canonical signaling, the modulation of autophagy represents a distinct mechanistic tier through which ncRNAs enforce the CAF phenotype. By regulating protein stability and autophagic flux, ncRNAs can stabilize pro-tumorigenic factors. MALAT1 in gastric cancer facilitates CAF formation by directly inhibiting autophagic flux, a process that concurrently induces α-SMA expression.42 This theme extends to colorectal cancer, where miR-31 similarly alters the fibroblast phenotype by suppressing autophagy-related genes,67 positioning autophagy modulation as a critical node in stromal reprogramming.

The reprogramming of fibroblasts is further orchestrated by a sophisticated network of exosomal crosstalk and paracrine loops. Tumor cells release ncRNAs packaged in exosomes to “educate” recipient fibroblasts. For instance, the aforementioned myCAF-restraining effect in pancreatic cancer relies heavily on this exosomal network: the lncRNA LNC-FSD2-31:1 promotes the packaging of miR-4736 into extracellular vesicles, which then enter fibroblasts to target the ATG7-autophagy axis.43 Meanwhile, in gastric cancer, tumor-derived exosomes loaded with miR-124-3p are taken up by fibroblasts to downregulate SNAI2, thereby unleashing proliferation and migration.29 NcRNAs amplify these interactions by enhancing the secretion of soluble cytokines; specifically, miR-27 has been shown to promote TGF-β secretion and signaling,66 establishing a feed-forward loop that reinforces the activated state. This communication is regulated by decoy mechanisms; for instance, LINC01915 in colorectal cancer sponges miR-92a-3p to inhibit the uptake of tumor-derived exosomes, effectively blocking the NF-to-CAF transition.44

CAF activation is not unidirectional; it is tightly controlled by counter-regulatory ncRNAs that act as “guardians” of stromal homeostasis. These molecules function as brakes to prevent aberrant activation. For instance, tumor-cell-secreted miR-200 can inhibit TGF-β-driven CAF formation, representing a crucial brake on stromal activation in colorectal cancer.69 Similarly, miR-141-3p in gastric cancer serves as a protective shield; it directly targets STAT4 to inhibit the Wnt/β-catenin pathway, thereby suppressing the transition of normal fibroblasts and bone marrow mesenchymal stem cells into CAFs.68 The ultimate state of the stroma is likely dictated by the dynamic equilibrium between these pro-fibrotic drivers and anti-fibrotic guardians.

Collectively, ncRNA-mediated CAF activation is a highly context-dependent interplay that integrates hijacked signaling pathways (NF-κB, JAK/STAT), autophagy modulation, sophisticated exosomal communication, and a complex balance of activators and inhibitors. While this activation is often reinforced by epigenetic reprogramming, the inherent plasticity of fibroblasts raises a critical question: can this pro-tumorigenic state be reversed? Exploring strategies to therapeutically revert or “normalize” CAFs back to a quiescent, NF-like state represents a promising frontier, as discussed in the section on therapeutic interventions.

Tumor Progression and Metastasis

The interplay between CAFs and tumor cells is not merely a structural interaction but a dynamic exchange of molecular information that drives the dual engines of malignancy: EMT and unchecked proliferation. Crucially, recent single-cell paradigms reveal that the formation of a metastatic niche is not driven by generic stromal stress, but by the highly coordinated activities of specific CAF lineages. A premier example of this subtype-specific orchestration is found in colorectal cancer (CRC), where myofibroblastic CAF (myCAF)-derived exosomal lncRNA PWAR6 acts as a dual metabolic and immunomodulatory architect. By intricately altering stromal glutamine availability and concurrently compromising natural killer (NK) cell surveillance, this myCAF-specific axis transforms the local hepatic environment into a permissive sanctuary for liver metastasis.45 Operating beneath this overarching subtype-specific framework, ncRNAs serve as the primary messengers in this crosstalk through several distinct mechanistic tiers.

A predominant mode of regulation involves the exosomal transfer of ncRNAs that function as molecular decoys (ceRNAs) to unlock metastatic programs. In esophageal squamous cell carcinoma (ESCC), CAF-derived exosomes are enriched with LINC01410 and circ-0076535, which act as molecular sponges for miR-122-5p and miR-145-5p, respectively. This sequestration releases the inhibition on key drivers like PKM2 and NF-κB, thereby fueling glycolytic shifts and EMT.35,87 Similarly, exosomal HOTAIR promotes metastasis by directly repressing E-cadherin while upregulating N-cadherin and glycolytic enzymes.46 This “decoy” mechanism is mirrored in colorectal cancer (CRC), where exosomal LINC00659 sponges miR-342-3p to upregulate ANXA2,47 and SNHG3 binds miR-34b-5p to stabilize the RNA-binding protein HuR, creating a synergistic node that amplifies HOXC6 transcription and drives tumor growth.48

Beyond acting as sponges, ncRNAs fundamentally rewire intracellular signaling architectures, effectively dismantling growth checkpoints and activating proliferation cascades. This is particularly evident in the disruption of the Hippo pathway, a central controller of organ size. In CRC, WEE2-AS1 promotes the degradation of MOB1A, thereby releasing the brake on Hippo signaling and enabling unrestrained growth.49 In parallel, kinase signaling networks are frequently hijacked to support survival and invasion. For instance, UCA1 in CRC activates the mTOR/KRAS axis, while miR-3656 in ESCC targets ACAP2 to activate the PI3K/AKT pathway.50,70 This strategy is consistent across tumor types: in gastric cancer, miR-199a-5p downregulates FKBP5 to activate mTORC1 signaling,71 and in hepatocellular carcinoma, LINC00152 modulates the miR-205-5p/CXCL11 axis to sustain transcriptional activity and tumor growth.51

At the nuclear level, specific ncRNAs stabilize aggressive phenotypes by modulating transcription factors and epigenetic states. In hepatocellular carcinoma, MAPKAPK5-AS1 inhibits SMURF2 to stabilize TCF12, which in turn amplifies FOXH1 activity, driving a sustained proliferative program.52 In gastric cancer, the regulation becomes epigenetic: CAFs induce promoter methylation of miR-200b, silencing this EMT-suppressor to unleash ZEB1/2-driven dissemination.72 Furthermore, tumor-derived exosomal miR-5703 acts as a critical pathological messenger in colorectal cancer (CRC); when enriched in exosomes secreted by p53-suppressed cancer cells, miR-5703 directly downregulates apelin receptor (APJ) expression in resident fibroblasts, subsequently triggering Smad2/3 phosphorylation to induce a pro-tumorigenic CAF phenotype that accelerates tumor growth.75

NcRNAs remodel the stromal niche by specifically suppressing tumor-restraining factors to support continuous expansion. In CRC, miR-181b-3p downregulates SNX2 and BMP5 to promote proliferation,76 while miR-20a-5p and miR-345-5p target PTEN and CDKN1A, respectively, to lower the threshold for invasion.73,74 These molecules effectively lower the biological barriers that typically constrain tumorigenesis.

Intercellular signaling is amplified through inflammatory and lymphangiogenic pathways to physically restructure the tumor microenvironment. miR-146a-5p and miR-155-5p induce inflammatory cytokine secretion in CRC CAFs,77 while the downregulation of miR-100-5p promotes lymph node metastasis in ESCC by facilitating lymphangiogenesis.78 This pro-angiogenic role is further enhanced by exosomal miR-522-3p, which has been shown to drive angiogenesis and invasion in CRC.79

This regulatory network is often cyclical. For instance, miR-21, highly secreted by CAFs, forms a feed-forward loop: it not only enhances tumor cell invasion but also reinforces the activated CAF phenotype via S100A4, cementing the pro-tumorigenic stroma.30 These findings reveal a multilayered regulatory landscape where CAF-associated ncRNAs synchronize the acquisition of motility with the capacity for colonization.

Therapeutic Resistance

In Digestive system neoplasms, CAFs function as a sanctuary for tumor cells, orchestrating a drug-resistant microenvironment by secreting a specific repertoire of ncRNAs. These ncRNAs confer therapeutic insensitivity not through a single route, but by systematically activating multidimensional protective mechanisms, primarily by enhancing the intrinsic resilience of tumor cells and blocking regulated cell death pathways. Crucially, recent evidence demonstrates that this therapeutic insensitivity is actively driven by specific CAF subpopulations. A striking example is found in pancreatic ductal adenocarcinoma (PDAC), where the inflammatory iCAF subset secretes high levels of IL-8. This paracrine signal directly induces the expression of lncRNA UPK1A-AS1 in tumor cells, which acts as a molecular scaffold to enhance Ku70/Ku80 interaction, driving efficient double-strand break (DSB) repair and oxaliplatin resistance.53

Operating alongside these programmed subset activities, a major strategy is to enhance tumor cell resilience by stabilizing cancer stem-like properties. This allows a subpopulation of cells to survive cytotoxic insults and initiate recurrence. In gastric cancer, multiple ncRNAs converge on this theme. CAF-derived signals upregulate ST7-AS1, which in turn promotes EMT and cisplatin resistance via activation of the PI3K/AKT pathway.54 This mechanism of maintaining stemness is also observed in other digestive malignancies; for example, in PDAC, LINC00460 sustains stemness markers to confer gemcitabine resistance.55 Similarly, in CRC, H19 activates both β-catenin and Notch1 signaling,56 while CCAL interacts with HuR to stabilize β-catenin,57 thereby reinforcing Wnt-driven stemness and reducing sensitivity to oxaliplatin and 5-fluorouracil. This principle of stromal-driven resistance also extends to ESCC, where tumor-secreted exosomal POU3F3 promotes cisplatin resistance by inducing the differentiation of normal fibroblasts into CAFs, reinforcing the protective stem-cell niche.58

Recent studies highlight the critical role of CAFs in blocking regulated cell death pathways. In gastric cancer, miR-522 targets ALOX15 to suppress lipid peroxidation, effectively inhibiting ferroptosis and reducing the efficacy of cisplatin and paclitaxel.80 Conversely, the downregulation of DACT3-AS1, a promoter of SIRT1-mediated ferroptosis, is strongly associated with resistant phenotypes.59 In CRC, exosomal FAL1 functions as a molecular scaffold to promote Beclin1 ubiquitination and degradation, thereby suppressing autophagy-dependent cell death induced by oxaliplatin.37 Beyond these novel forms of death, classical apoptosis is also thwarted; for instance, miR-24-3p activates the PI3K/AKT pathway in colorectal cancer, establishing a high threshold for drug-induced apoptosis.81

Finally, CAFs directly enhance cellular defense mechanisms against therapy. In gastric cancer, ANRIL upregulates the ATP-binding cassette transporter ABCB1, acting as a molecular pump to accelerate cisplatin efflux,36 while miR-3173-5p suppresses lipid peroxidation and inhibits ferroptosis by targeting ACSL4.82 Resistance to radiotherapy is a key feature, primarily mediated by enhancing the DNA Damage Response (DDR). In esophageal squamous cell carcinoma, exosomal AFAP1-AS1 activates DNA damage repair mechanisms and attenuates radiation-induced double-strand breaks.60 Another study in ESCC demonstrated that CAF-promoted lncRNA DNM3OS confers significant radioresistance by regulating the DDR pathway, highlighting it as an attractive therapeutic target to reverse radioresistance.61

In summary, these ncRNAs delineate a sophisticated resistance network. By orchestrating subtype-specific DNA repair (UPK1A-AS1), stabilizing stemness (H19, CCAL), blocking diverse death programs (miR-522, FAL1), and enhancing cellular defense (ANRIL, AFAP1-AS1), CAFs ensure tumor survival against a broad spectrum of therapeutic modalities.

Immune Evasion and Regulation

In digestive system neoplasms, CAFs act as the architects of an immunosuppressive barrier, orchestrating the paralysis of immune surveillance through non-coding RNAs. Consistent with the subtype-specific paradigm discussed earlier, the construction of this barrier is increasingly recognized as the task of specialized immunomodulatory subsets (such as inflammatory iCAFs). Delivered via exosomes or paracrine signaling, these subset-derived ncRNAs systematically utilize strategies of “antigen cloaking, checkpoint reinforcement, and inflammatory reprogramming” to facilitate tumor immune escape.

A fundamental strategy for evasion is the reduction of tumor visibility to cytotoxic T lymphocytes (CTLs). In pancreatic cancer, CAF-derived exosomal RP11-161H23.5 functions as a critical signal to silence immunogenicity by downregulating HLA-A expression in tumor cells. By diminishing the surface density of MHC Class I molecules, this lncRNA impairs the antigen recognition efficiency of CD8⁺ T cells, effectively establishing a “stealth” phenotype that serves as the foundation for immune evasion.62

Beyond passive concealment, CAFs actively suppress effector cell function. This is often achieved through decoy-like mechanisms or by reprogramming immune cells. In CRC, CAF-derived exosomal miR-214 targets the PD-1/PD-L1 axis and modulates downstream AKT and STAT signaling networks. This intervention elevates the cytotoxic threshold of effector T cells, effectively inducing an exhausted state and shielding the tumor from immune-mediated elimination.83 CAFs reshape the myeloid compartment to be more tumor-permissive. In ESCC, for example, exosomal miR-889-3p secreted by CAFs can repress M1 macrophage polarization by targeting STAT1, acting as a direct molecular rheostat to foster an immunosuppressive microenvironment rich in M2-like macrophages.84

Finally, CAFs secrete a specific repertoire of ncRNAs to reshape the cytokine milieu, creating a permissive “soil” that dampens cytotoxic responses while recruiting suppressive populations. In gastric cancer, NORAD, likely functioning as a decoy for miRNAs, drives IL-33-related signaling to alter the cytokine profile and promote the infiltration of immunosuppressive cells.63 Similarly, exosomal miR-4253 in gastric cancer functions by interfering with inflammatory transcriptional networks. These interactions actively diminish cytotoxic signatures and foster an immunosuppressive phenotype that enhances tumor adaptability.85 In CRC, circ-0067557 reinforces this suppressive microenvironment by promoting malignant phenotypes via the Lin28A/Lin28B signaling axis.88

These ncRNAs delineate distinct yet complementary pathways of immune suppression. Targeting these molecular nodes—specifically those acting as signals (RP11-161H23.5) or decoys (NORAD, miR-214)—holds significant potential to convert “cold,” immune-desert tumors into “hot,” inflamed lesions, thereby sensitizing digestive system neoplasms to immunotherapy.

Metabolic Reprogramming

In digestive system neoplasms, CAFs act as “metabolic feeders,” remodeling tumor cell energy utilization and carbon flux distribution through exosomal delivery or paracrine signaling. While recent single-cell evidence—such as the myCAF-driven glutamine remodeling by PWAR6 discussed earlier—indicates that metabolic cross-feeding is subtype-dependent, the majority of current ncRNA studies in this domain still rely on “pan-CAF” models. These foundational studies reveal that metabolic reprogramming is primarily manifested as the establishment of a glycolytic advantage (the Warburg effect) and the accumulation of lactate, which creates an acidified microenvironment that reinforces stromal interactions and fuels invasive behaviors.

A central function of CAF-derived ncRNAs is to accelerate aerobic glycolysis, frequently operating through a decoy mechanism to regulate rate-limiting enzymes. This glycolytic shift is frequently coupled with hypoxia adaptation, again often mediated by ncRNA decoys. In pancreatic adenocarcinoma, NNT-AS1 regulates the miR-889-3p/HIF-1α axis, releasing the transcriptional activation of glycolytic pathways. This facilitates glucose metabolism and survival under hypoxia, converging on a metabolic state characterized by high glycolytic load.64 Similarly, in hepatocellular carcinoma, exosomal TUG1 sustains the activation of glycolysis-related transcriptional programs by sequestering miR-524-5p to release its target, SIX1. This drives a rapid energy supply which synergizes with stemness and invasive phenotypes to promote tumor progression.65

These specific ncRNA axes establish distinct glycolytic fingerprints and exemplify the sophisticated integration of stromal signaling with tumor metabolism. These findings indicate that CAF-associated metabolic reprogramming is not a monolithic pathway but a context-dependent network, controlled by ncRNAs acting predominantly as decoys to fuel glycolysis and as scaffolds to integrate broader stress responses. Consequently, translational strategies should adopt stratified interventions targeting specific molecular axes (eg., H19–PFKFB3) combined with the blockade of exosome release, aiming to reverse these malignant phenotypes.

Diagnostic and Therapeutic Implications of CAF-Associated ncRNAs in Digestive System Neoplasms

The elucidation of ncRNA-mediated crosstalk offers a dual clinical opportunity: utilizing these molecules as non-invasive biomarkers for precision diagnosis and exploiting them as actionable targets for stromal reprogramming.

Decoding the Circulating Stromal Signature: Non-Invasive Diagnostics

The elucidation of the CAF-ncRNA secretome offers a paradigm shift in liquid biopsy: moving beyond the detection of circulating tumor cells (CTCs) to capturing the “stromal response signature”. A major advantage of CAF-associated ncRNAs is their encapsulation within exosomes, which provides resilience against RNase degradation and ensures high stability in biofluids including plasma, saliva, and urine. Unlike conventional protein markers which often reflect tumor burden only at advanced stages, stromal ncRNAs reflect the early-stage activation of the tumor microenvironment. This positions them as ideal candidates for non-invasive monitoring.

In the context of diagnostic precision, specific lncRNAs have demonstrated the capacity to distinguish malignant lesions from benign inflammation with high specificity. For instance, in gastric cancer, exosomal LINC00691 has proven effective in differentiating early-stage carcinoma from benign gastric diseases, demonstrating diagnostic fidelity that significantly outperforms traditional serological markers, achieving an area under the curve (AUC) of 0.86, with a sensitivity of 0.85 and a specificity of 0.71.41 In ESCC, plasma levels of POU3F3 serve as a sensitive indicator of malignant transformation, effectively discriminating patients from healthy controls.58

Beyond detection, the “ncRNA cargo” mirrors the metastatic potential and prognostic trajectory of the stroma. In hepatocellular carcinoma, high circulating levels of TUG1 have been strongly correlated with advanced AJCC stages and lung metastasis, acting as surrogates for “stromal aggressiveness”.65 This prognostic value extends to CRC, where plasma levels of WEE2-AS1 closely mirror disease burden and TNM staging.49 In PDAC, exosomal enrichment of Lnc-FSD2-31:1 is inversely correlated with differentiation status and overall survival, highlighting its potential as a stratification biomarker.43 In gastric cancer, TGILR expression has been identified as an independent prognostic factor, closely associated with lymph node metastasis and poor progression-free survival outcomes; clinically, its expression directly correlates with pathological stage (p = 0.04), T stage (p < 0.01), and lymph node status (p < 0.01), while high levels predict significantly shorter overall survival and disease-free survival as an independent risk factor.89 These findings establish a solid foundation for translating CAF-related ncRNAs from benchside discoveries into versatile clinical biomarkers.

Therapeutic Frontiers: From Response Monitoring to Stromal Normalization

The therapeutic landscape targeting the CAF-ncRNA axis is evolving from broad, non-specific inhibition to sophisticated stromal reprogramming. This paradigm shift focuses on three strategic pillars: dynamic response monitoring, chemosensitization, and the “holy grail” of therapy—stromal normalization.

As real-time barometers for therapeutic efficacy, the expression levels of CAF-associated lncRNAs serve as critical companion diagnostics to guide personalized treatment. For instance, in a clinical cohort of ESCC patients undergoing chemoradiotherapy, baseline exosomal POU3F3 levels in patients achieving a complete response (CR) were significantly lower than those in non-CR patients (p < 0.001). Dynamically, its levels profoundly declined at 4 weeks post-treatment but exhibited a prominent re-elevation upon disease progression (p < 0.001).58 This dynamic fluctuation includes the utility of stromal biomarkers in tracking residual disease and anticipating recurrence before clinical manifestation.

Targeting resistance-associated ncRNAs offers a direct strategy to dismantle the protective niche and re-sensitize tumors to standard regimens. Preclinical evidence indicates that silencing AFAP1-AS1 can effectively abrogate DNA damage repair mechanisms, thereby restoring radiosensitivity in esophageal cancer.60 Similarly, targeting key chemoresistance drivers—such as ANRIL, ST7-AS1, or CCAL—has proven effective in reversing recalcitrance to first-line agents like cisplatin and oxaliplatin in gastric and colorectal cancers, effectively lowering the apoptotic threshold of malignant cells.36,54,57

Perhaps the most promising therapeutic avenue is stromal normalization, which aims to target the plasticity of fibroblast phenotypes to induce lineage reversion—effectively turning activated CAFs back into quiescent, tumor-restrictive NFs.

First, signaling blockade offers a viable route to arrest CAF activation. Since lncRNAs like MALAT1 and LINC00691 drive the NF-to-CAF transition via the IL-6/STAT3 and JAK2/STAT3 cascades, respectively, pharmacological intervention with selective inhibitors (eg., the STAT3 inhibitor WP1066 or the JAK1/2 inhibitor ruxolitinib) has successfully demonstrated the ability to inhibit CAF activation and suppress tumor growth in gastric cancer models.41,42

Second, disruption of specific molecular hubs can collapse the supportive stromal network. For example, targeting the Lnc-FSD2-31:1/miR-4736/ATG7 axis has proven effective in blocking the autophagy-dependent support necessary for PDAC progression in vivo.43

Finally, given that exosomes are the primary vectors for this malignant crosstalk, exosomal interception represents a broad-spectrum strategy. These approaches highlight the potential of targeting the CAF-ncRNA axis to not only eliminate tumor cells directly but also to normalize the supportive microenvironment.

However, translating these benchside discoveries into clinical realities faces formidable biological barriers, most notably systemic toxicity and unpredictable off-target profiles. A sobering case in point is the landmark Phase I clinical trial evaluating MRX34, a liposomal miR-34a mimic, in 85 patients with advanced solid tumors. Although MRX34 was designed to replenish tumor-suppressive pathways globally rather than targeting the stroma specifically, its clinical outcomes serve as a cautionary tale: while partial response or stable disease was documented in 19 cases, the trial was prematurely terminated due to severe immune-mediated adverse events that culminated in four patient fatalities.90 This severe clinical regression underscores that abandoning blunt stromal destruction in favor of precision stromal normalization remains heavily contingent upon conquering the universal safety profiles of non-viral nano-delivery architectures.

Conclusion and Future Perspectives

This review highlights how ncRNAs orchestrate the bidirectional communication between CAFs and tumor cells across gastrointestinal malignancies. Operating far beyond passive structural elements, CAFs actively drive malignancy by utilizing ncRNAs to reprogram the stroma, support metabolic shifts, orchestrate immune evasion, and establish therapeutic resistance. Consequently, conquering these highly desmoplastic tumors requires shifting our therapeutic focus from targeting tumor cells in isolation to disrupting this supportive stromal network.However, realizing the clinical potential of the CAF-ncRNA axis is currently hindered by a critical limitation: the historical reliance on “bulk” tumor analyses. Traditional studies have largely treated CAFs as a homogenous, monolithic population characterized by generic markers (eg., $\alpha$-SMA). Recent single-cell transcriptomic paradigms have decisively shattered this view, proving that the stroma comprises highly heterogeneous subpopulations (such as iCAFs, myCAFs, and apCAFs) with distinct, and sometimes diametrically opposing, biological functions.This functional diversity introduces a profound clinical dilemma: while many reactive CAFs promote tumor progression, specific stromal subsets actively restrain it. For instance, while typical $\alpha$-SMA-high myCAFs are often pro-tumorigenic, emerging evidence suggests that subsets characterized by markers like Meflin possess tumor-suppressive properties—actively maintaining the basement membrane and supporting immune surveillance. Therefore, broad “stromal depletion” strategies that indiscriminately eradicate all CAFs are inherently flawed and can paradoxically accelerate tumor growth by eliminating these protective fibroblasts.The pioneering studies highlighted in our review—such as the myCAF-specific restraint by miR-4736 and the precise metabolic remodeling by myCAF-derived PWAR6—represent the critical first steps in addressing this gap. Future research must urgently utilize high-resolution spatial and single-cell technologies to map the specific ncRNA landscapes of these distinct functional lineages. We must answer critical questions: Which specific ncRNA circuits act as the master switches dictating the polarization between pro-tumorigenic and tumor-restraining states? Can we block the “bad” ncRNAs without harming the beneficial ones?Answering these questions provides the foundation for a paradigm shift: from blunt stromal destruction to precision “stromal normalization”. By identifying and targeting the specific ncRNAs that maintain the reactive state, or by delivering therapeutic miRNA mimics to restore tumor-suppressive phenotypes, we can effectively “re-educate” the stroma. Ultimately, leveraging subtype-specific ncRNAs to revert the reactive microenvironment back to a quiescent, normalized state represents a highly promising frontier to overcome therapeutic resistance and improve long-term survival in digestive system tumors.

Funding Statement

Jiangsu Provincial Health Commission Medical Research General Program, Grant/Award Numbers: M2022100, M2022117; Jiangsu Provincial Natural Science Foundation General Program, Grant/Award Number: BK20231250; Zhenjiang Science and Technology Plan (Basic Research Program), Grant/Award Number: JC2025016; Zhenjiang Science and Technology Plan (Social Development Program), Grant/Award Number: SH2025056.

Abbreviations

CAFs, cancer-associated fibroblasts; ncRNAs, non-coding RNAs; lncRNAs, long non-coding RNAs; miRNAs, microRNAs; circRNAs, circular RNAs; ceRNA, competing endogenous RNA; TME, tumor microenvironment; ECM, extracellular matrix; NFs, normal fibroblasts; MSCs, mesenchymal stem cells; rCAFs, cancer-restraining CAFs; pCAFs, tumor-promoting CAFs; myCAF, myofibroblastic CAFs; apCAFs,antigen-presenting CAFs; iCAFs, inflammatory CAFs; EMT, epithelial–mesenchymal transition; EndMT, endothelial-to-mesenchymal transition; TAMs, tumor-associated macrophages; ESCC, esophageal squamous cell carcinoma; GC, gastric cancer; CRC, colorectal cancer; HCC, hepatocellular carcinoma; PDAC, pancreatic ductal adenocarcinoma; AUC, area under the curve; OS, overall survival; DFS, disease-free survival; DDR, DNA damage response; PFS, progression-free survival.

Disclosure

The authors declare no conflicts of interest in this work.