The changes in collagen in terms of stiffness, spatial structure, and fiber morphology jointly promote tumor progression.

A primary alteration is the dramatic stiffening of the tumor stroma, a process driven by both the aberrant deposition of fibrillar collagens and their extensive cross-linking. Tumors frequently show higher levels and enzymatic activity of LOX enzymes, catalyzing the formation of covalent bonds that dramatically increases tissue tension [14]. Clinically, this is evidenced by the elastic modulus of breast cancer tissue, which can reach 4–12 kPa, starkly contrasting with the 0.4–2 kPa of normal mammary tissue [15]. This heightened matrix stiffness is not a passive feature but is a potent protumorigenic signal.

Concurrently, the matrix undergoes profound architectural reorganization. The physiological, porous, and randomly oriented collagen meshwork transforms into a dense, linearized architecture, resulting in a significant reduction in interstitial pore size from the normal range of 10–50 µm down to 1–10 µm in tumors [16]. This densified structure forms a formidable physical barrier that severely impedes the intratumoral trafficking of immune effector cells, such as CTLs, as well as the perfusion of therapeutic agents. This phenomenon, widely termed “immune exclusion”, is an important mechanism contributing to resistance against both chemotherapy and immunotherapy [5].

On a microscopic scale, the morphology of individual collagen fibers is also remodeled. Tumor-associated collagen fibers become significantly thicker and longer, with diameters increasing from 50–500 nm in healthy tissue to 200–2000 nm [17]. Critically, at the tumor’s invasive front, these fibers are often realigned and bundled into taut, linear tracks oriented perpendicularly to the tumor boundary. These unique topological features are known as TACS [18].

Beyond its role as a physical scaffold, the collagen matrix operates as a dynamic and potent biochemical signaling hub. On the one hand, the TME’s collagen network functions as a crucial reservoir, concentrating and modulating the bioavailability of growth factors, cytokines, and chemokines. The high density of positively charged residues on collagen fibers facilitates the electrostatic sequestration of numerous anionic signaling molecules, including TGF-$\beta$, vascular endothelial growth factor (VEGF), and FGF [8]. This interaction is further stabilized and specified by intermediaries such as heparan sulfate proteoglycans, which act as co-receptors [17]. This creates a localized, high-concentration gradient of protumorigenic signals.

On the other hand, the collagen fibers themselves serve as solid-phase ligands, directly initiating intracellular signaling cascades upon binding to cell surface receptors on both cancer and immune cells. On cancer cells, specific cryptic binding motifs within the collagen triple helix, such as the GFOGER (for Gly-Phe-Hyp-Gly-Glu-Arg) and RGD (Arg-Gly-Asp) sequences, become exposed during matrix remodeling [19]. These motifs are recognized by integrin heterodimers and discoidin domain receptors 1 and 2 (DDR1/2), which are unique receptor tyrosine kinases [20]. This engagement establishes a pernicious feed-forward loop: activated cancer cells increase their secretion of MMPs, which further remodel the collagen matrix, exposing more binding sites and thus amplifying these protumorigenic signals. Additionally, collagen can bind to inhibitory receptors on immune cells (e.g., leukocyte-associated immunoglobulin-like receptor-1 [LAIR-1]) [21].

The primary physical function of collagen is to provide a critical three-dimensional structural support framework for tumor tissues. Tumor cells tend to proliferate preferentially in regions with higher collagen density. These areas not only offer stable anchor points necessary for cells to resist mechanical stress and maintain their shape but also function as migration tracks [22]. Cells can migrate along these collagen fibers through contact guidance, at speeds up to 5 to 10 times faster than free migration in the matrix. Live imaging techniques have confirmed that tumor cells preferentially invade outward along collagen fibers perpendicular to the tumor boundary (TACS-3) [18]. Notably, different types of immune cells have varying abilities to utilize these collagen tracks, which may be related to their deformability: dendritic cells and macrophages can efficiently use these tracks for migration, while T-cell migration is more dependent on the interstitial network’s gaps [23].

High-density collagen deposition can form a fibrous capsule around the tumor, serving a protective function. On the one hand, this dense barrier severely hinders the effective penetration of chemotherapy drugs and antibody drugs, which is a significant cause of treatment resistance. For instance, in pancreatic cancer characterized by dense stroma, the penetration distance of drugs is often limited to 50–100 micrometers around blood vessels, far below the range required to achieve an effective therapeutic concentration [24]. On the other hand, it significantly restricts the infiltration of immune effector cells. Studies have shown that for every doubling of collagen density, the infiltration depth of CD8+ T cells decreases by approximately 30% [11]. Notably, some immune cells also have regulatory roles in tumor fibrosis. Research has found that TAMs, through mannose receptor-mediated mechanisms, upregulate inducible nitric oxide synthase, thereby generating reactive nitrogen species to promote tumor fibrosis [25].

In the TIME, the physical functions of collagen and its biochemical functions are not independent of each other but form a continuous spectrum through “mechano-chemical transduction”: extracellular mechanical stress is first integrated into intracellular signals, which then cascade to affect cell proliferation, metabolism, and immune regulation [26].

Mechanical cues of collagen are sensed by cell-surface receptors (integrins $\alpha$2$\beta$1, $\alpha$11$\beta$1 and DDR1/2), which activate mechanosensitive pathways such as FAK–extracellular signal-regulated kinase (ERK), phosphoinositide 3-kinase–Akt, and the Hippo–YAP/TAZ axis [27]. Studies have shown that the stiff collagen matrix can accelerate the transition of the cell cycle from G1 to S phase by 2 to 3 times through activation of the FAK–ERK signaling pathway, thereby promoting tumor cell proliferation [13]. At the same time, collagen signals can upregulate the expression of cyclins D1 and E and downregulate the levels of cyclin-dependent kinase inhibitors p21 and p27 [28]. However, for T cells, excessive collagen signals can lead to functional exhaustion through continuous activation, thereby inhibiting their proliferation [21].

These mechanosignals further orchestrate a profound metabolic reprogramming. In response to matrix stiffness, the Hippo–YAP/TAZ signaling axis is activated, which, in concert with the hypoxia common in poorly vascularized fibrotic regions, stabilizes hypoxia-inducible factor 1 alpha (HIF-1$\alpha$) [29]. This axis collaboratively upregulates key glycolytic enzymes (e.g., facilitative glucose transporter 1, hexokinase 2, and lactate dehydrogenase A), forcing cancer cells into a state of enhanced aerobic glycolysis (the Warburg effect). This metabolic shift not only fuels rapid proliferation but also acidifies the microenvironment [30]. The stiff environment imposes a metabolic penalty on immune cells, leading to dysfunctional mitochondria in T cells that produce less ATP and forcing macrophages to develop a protumor M2-like phenotype characterized by altered oxidative phosphorylation [8].

In parallel, collagen provides precise directional guidance to immune cells by forming stable concentration gradients through its degradation products and chemokines bound to it. For instance, the degradation product PGP (Pro-Gly-Pro) of type I collagen can specifically recruit neutrophils through the C-X-C motif chemokine receptor 2 receptor, while the NC1 domain fragments of type IV collagen (e.g., tumstatin and canstatin) have anti-angiogenic activity [31]. Particularly in the TIME, the chemotactic effect of collagen tends to recruit immunosuppressive cells, such as Tregs and myeloid-derived suppressor cells. For the few effector cells that do infiltrate, the dense, stiff matrix impairs the maturation and antigen-presentation capacity of dendritic cells, evidenced by a marked reduction in MHC-II and co-stimulatory molecule (cluster of differentiation 80 [CD80]/CD86) expression [32]. Furthermore, matrix stiffness induces cytoskeletal tension in T cells, physically disrupting the formation of a stable immunological synapse. Crucially, studies have found that collagen can directly transmit strong inhibitory signals to various immune cells (including T cells and natural killer [NK] cells) by binding to LAIR-1 [8].

First, the acidic pH of the TIME (typically 6.5–6.8) can affect the collagen network by modulating charge changes. Under acidic conditions, the protonation of key amino acid residues on collagen (e.g., lysine and arginine) increases, raising the surface positive charge density by approximately 30%. This not only alters the electrostatic binding ability of growth factors but also influences the interaction between collagen fibers, potentially leading to relaxation of the fiber network. Additionally, acidic pH can directly activate pH-sensitive proteases such as cathepsins B and L, accelerating collagen degradation [33].

Second, the MMP family is the main enzymatic driver of collagen turnover. The TIME is characterized by the aberrant expression and activity of specific MMPs, including interstitial collagenases (MMP-1, MMP-8, MMP-13) that execute the initial cleavage of the native collagen triple helix, and gelatinases (MMP-2, MMP-9) that degrade the denatured products and associated basement membrane structures [34]. Secreted predominantly by CAFs and tumor cells, these MMPs form highly active “hotspot regions” at the tumor invasion front. This localized degradation not only carves out pathways for cellular invasion but also liberates bioactive collagen fragments and releases matrix-sequestered signaling molecules, which in turn regulate angiogenesis and inflammation [35].

Furthermore, oxidative stress profoundly affects the integrity of collagen. Elevated reactive oxygen species (ROS) levels in the TIME can directly oxidize proline and lysine residues in collagen, destabilizing the triple-helical structure and even causing peptide chain breaks. Additionally, ROS can activate latent MMPs, indirectly accelerating collagen degradation. This oxidative damage is highly spatially specific, mainly occurring in the hypoxia-reoxygenation junction areas within the TIME [34].

Finally, the tumor-associated inflammatory cytokine network regulates collagen synthesis and degradation. For instance, tumor necrosis factor alpha and interleukin 1 beta (IL-1$\beta$) can upregulate MMP expression and downregulate their endogenous inhibitors, TIMPs, promoting degradation [34]. TGF-$\beta$ is the most potent promoter of collagen synthesis and inhibitor of its degradation [36]. Additionally, IL-6 contributes to fibrosis through the STAT3 signaling pathway [37]. The complex spatiotemporal distribution patterns of these factors ultimately determine the overall pattern of collagen remodeling in tumors.

The dynamic remodeling of collagen architecture has been systematically characterized into distinct TACS, which serve as structural indicators of tumor invasiveness and potential biomarkers for evaluating treatment responses [7].

In the early stage of tumor occurrence (precancerous lesions and early in situ carcinoma), TACS-1 is defined by a locally disorganized and wavy collagen configuration. These fibers, showing only a modest increase in density, loosely encircle nascent tumor foci. This signature reflects a period of localized matrix degradation, primarily mediated by tumor cell-secreted MMPs, which creates the necessary space for initial tumor expansion and has relatively immune accessibility [38]. Early, less‑aligned matrices can sequester antigens and chemokines within peritumoral niches and support local immune surveillance, acting as a transient barrier to systemic spread; this immunoprotective compartmentalization diminishes as fibers straighten, pores narrow, and inhibitory collagen–immune receptor pathways intensify [39].

As the tumor progresses and its volume increases, it exerts significant mechanical pressure on the surrounding stroma, leading to the TACS-2 signature. This stage is characterized by collagen fibers that become linearized, bundled, and aligned tangentially (parallel) to the tumor boundary [38]. This architectural shift is dominated by the robust activation of CAFs, which become the primary source of de novo collagen synthesis, dramatically increasing overall matrix density and stiffness [40].

When tumor enters the invasion stage, it is marked by the emergence of TACS-3. These fibers are reorganized into thick, straightened bundles oriented perpendicularly to the tumor margin. This radial alignment creates veritable ‘highways’ that facilitate directed cell migration into the surrounding tissue. At this stage, the matrix reaches its peak density and degree of cross-linking, providing stable, pre-established conduits for invasion [13]. Critically, intravital imaging has confirmed that the vast majority of motile, invasive tumor cells migrate along these TACS-3 fibers [18]. In addition, TACS-3 also features dual physical–ligand barriers and T-cell rejection characteristics, including collagen–LAIR-1 signaling and DDR1/DDR2-directed invasion [41].

Notably, the tumor immune response itself can also have a reverse effect on the evolution of collagen. In the early stage of immune activation, local degradation of collagen occurs by effector T cells and NK cells, forming “hot zones” where immune cells aggregate [42]. However, as chronic inflammation is established, the elevated levels of T helper 2-type cytokines and TGF-$\beta$ conversely promote fibrotic reactions, leading to a dense, restrictive matrix [6]. This dynamic relationship is clinically important, as the study showed a clear link between effective immune checkpoint inhibitor (ICI) therapy correlates and a measurable reduction in collagen density [9].

The collagenous stroma within the TIME is organized into distinct spatial regions, from the peritumoral capsule to the invasive front and perivascular niches. This structural heterogeneity has functional consequences.

The central regions of many tumors are often characterized by a sparse and fragmented collagen landscape. This is typically a consequence of localized hypoxia and necrosis, which limit the metabolic activity of stromal cells and favor a degradative enzymatic environment [14]. However, in some dense tumors (e.g., pancreatic cancer), even the central region can maintain a high collagen content, which is mainly related to the tumor type and distribution pattern of CAFs [24].

Encapsulating the tumor mass is the peritumoral region or invasive front, which typically represents the zone of maximal collagen deposition. Here, a dense, highly cross-linked fibrous capsule (often 50–200 µm thick) forms a strong physical barrier. This structure can achieve a mechanical stiffness 10- to 20-fold greater than that of adjacent normal tissue [14]. Spatially resolved transcriptomic studies have confirmed that CAFs in this area express the highest levels of collagen synthesis genes (e.g., COL1A1) and cross-linking enzyme genes (e.g., LOX). This marginal barrier is a critical determinant of both immune cell exclusion and drug penetration [9].

At the junction between tumors and normal tissues, there is a dynamic transition zone with a width of approximately 100–500 micrometers. Real-time imaging shows that the collagen in this area is in a highly dynamic remodeling state, with both synthesis and degradation actively occurring simultaneously [17]. This interface represents a critical arena for tumor–immune interactions. Thus, the collagen density and structure in this area directly affect the efficiency of immune surveillance [14].

Finally, a special pattern of collagen tissue is present around the tumor neovasculature. This collagen is frequently arranged in concentric circles, forming a vascular sheath [43]. This structure, on the one hand, provides structural support for the fragile tumor blood vessels, and on the other hand, forms another barrier for the extravasation of immune cells from the blood vessels. A study has shown that a higher density of collagen density surrounding blood vessels is statistically linked to lower levels of immune cell infiltration [44].

Direct ablation or functional modulation of CAFs represents a foundational strategy in antifibrotic tumor therapy. Talabostat (PT-100) [47], a small molecule inhibitor of fibroblast activation protein (FAP), exemplifies this approach. By targeting FAP—a serine protease highly expressed on activated CAFs that facilitates ECM remodeling—Talabostat disrupts the proteolytic activity essential for collagen turnover and CAF maintenance [64]. However, clinical translation has revealed modest single-agent efficacy, likely due to the remarkable plasticity of the stromal compartment and compensatory activation of alternative fibrogenic pathways. This limitation has catalyzed the development of combination therapies, such as combining ICIs with FAP-directed chimeric antigen receptor (CAR) T cells or antibody-drug conjugates (ADCs) [5, 48]. Early-phase trials have shown encouraging tumor regression and stromal depletion, suggesting that cellular immunotherapy may overcome the barriers that hampered small molecule approaches [65].

A complementary strategy is to inhibit the signaling pathways that drive the profibrotic phenotype in CAFs. The TGF-$\beta$ signaling axis orchestrates CAF activation, collagen synthesis, and matrix cross-linking through both canonical (SMAD-dependent) and non-canonical pathways [36]. Galunisertib (LY2157299) [49], a selective TGF-$\beta$ receptor I (ALK5) kinase inhibitor, disrupts SMAD2/3 phosphorylation, thereby attenuating downstream fibrogenic gene expression [50]. Clinical trials have demonstrated not only reduced collagen deposition but also reversal of the immunosuppressive microenvironment, with enhanced responsiveness to ICIs. In addition, fresolimumab [51], a pan-isoform TGF-$\beta$-neutralizing antibody, sequesters all three TGF-$\beta$ isoforms in the extracellular space [66]. While both agents show promise, their clinical development must carefully balance antifibrotic benefits against potential safety concerns, as TGF-$\beta$ also functions as a tumor suppressor in early-stage disease [67].

This concept of pathway inhibition extends to leveraging clinically approved antifibrotic agents. Drugs such as pirfenidone [52] and nintedanib [53], standard-of-care for idiopathic pulmonary fibrosis, exert pleiotropic effects by simultaneously inhibiting multiple fibrogenic pathways including TGF-$\beta$, VEGF, FGF, and PDGF [68]. Their repurposing for oncology is compelling, and preclinical studies combining these agents with ICIs have demonstrated synergistic antitumor activity by remodeling the stromal landscape. Similarly, direct inhibitors of PDGF/FGF receptors, such as the multikinase inhibitor imatinib [54], are also being evaluated for their potential to reduce stromal density and enhance drug delivery, particularly when combined with chemotherapy [69].

Therapeutic strategies for matrix degradation have undergone a significant paradigm shift, moving away from early, crude approaches towards highly specific and targeted interventions.

Initial forays with broad-spectrum MMPs inhibitors were largely unsuccessful in clinical trials. This failure stemmed from significant dose-limiting toxicities, and more critically, a flawed underlying premise, namely, these inhibitors failed to discriminate between protumorigenic MMPs (e.g., MMP-2, MMP-9) and those with antitumorigenic or homeostatic functions (e.g., MMP-8) [70]. Small molecule agonists and antibody fragments that enhance MMP-8 activity have been shown to accelerate collagen turnover, curb metastatic spread, and boost intratumoral CD8+ T-cell density in breast- and pancreatic-cancer models [55, 71]. Moreover, studies have exploited tumor-specific proteases (e.g., FAP or MMP-14) to activate otherwise quiescent pro-MMP-8 or pro-MMP-9 zymogens, thereby confining collagenolysis to the malignant compartment [72].

Another more direct and precise but technically challenging strategy involves the exogenous delivery of matrix-degrading enzymes. Local or intratumoral delivery of recombinant collagenase (Clostridial class I/II) formulated in thermosensitive hydrogels or lipid nanoparticles can liquefy dense type I collagen. In mouse models of pancreatic cancer, a single intratumoral dose followed by gemcitabine doubled intratumoral drug concentrations and extended median survival by 40%. Similar gains have been reproduced with doxorubicin-loaded liposomes and ICIs, indicating that mechanical decompression, rather than enzyme-drug synergy per se, underlies the therapeutic benefit [73, 74]. The primary obstacle remains the safe and effective translation of this powerful local effect into a viable clinical therapy.

Beyond direct collagenolysis, an alternative strategy involves degrading other critical structural components of the ECM. Hyaluronic acid (HA) is a large glycosaminoglycan that creates immense interstitial fluid pressure and organizes the stroma. The enzymatic agent PEGPH20 (pegvorhyaluronidase) [56] works by systemically degrading HA. It melts the gel-like component of the matrix, which not only improves vascular perfusion but also relieves the mechanical constraints on the collagen fibers. While this approach showed significant promise in clinical trials for patient subgroups with “HA-high” tumors, its development has unfortunately faced setbacks [75]. Besides, inhibition of LOX-mediated cross-linking (e.g., $\beta$-aminopropionitrile or simtuzumab) renders collagen fibers more susceptible to endogenous proteases and can be combined with collagenase or selective MMP activation for additive effects [25, 57]. Likewise, photodynamic or focused ultrasound-mediated matrix debulking is being explored to create transient “windows” for drug entry [76].

Beyond physically degrading the matrix, a more nuanced therapeutic strategy aims to sever the protumorigenic communication lines between malignant cells and the collagen scaffold.

DDR1/2 have emerged as a particularly compelling target class. As the only receptor tyrosine kinases activated directly by collagen, they initiate potent and sustained downstream signaling cascades. Strong preclinical data have shown that the selective small molecule inhibitor, nilotinib [58], effectively suppresses the activation of CAFS, as well as collagen-induced invasion and metastasis [58]. The validation of this concept has propelled several DDR inhibitor candidates into Phase I clinical trials [77].

Integrins, fundamental receptors governing cell–ECM adhesion, have been explored as therapeutic targets for several decades. Early enthusiasm was tempered by the high-profile clinical failure of broad-spectrum agents like cilengitide (an RGD-mimetic targeting $\alpha$v$\beta$3/$\alpha$v$\beta$5) [59]. Consequently, a new generation of more refined integrin-targeted therapies is being developed [78]. This includes developing inhibitors that target specific collagen-binding integrin pairs (e.g., $\alpha$2$\beta$1 or $\alpha$11$\beta$1), which are highly upregulated on cancer cells or CAFs, or that selectively block the intracellular signaling adaptors downstream of integrin activation [60]. By doing so, these next-generation agents aim to not only inhibit tumor cell adhesion and migration but also disrupt integrin-mediated angiogenesis and resistance to conventional therapies [79].

A significant frontier in stroma-modulating therapy involves the strategic use of agents whose primary mechanisms of action are not directly antifibrotic, but that exert potent, indirect effects on collagen-producing CAFs and the broader TIME.

The ADC YL201 targets B7-H3 (CD276) [61], a molecule co-expressed on tumor cells, immune cells, and CAFs. This approach offers a dual benefit: directly blocking T-cell immune checkpoint while also weakening the CAF-mediated collagen network to remodel the immunosuppressive stroma [80, 81].

Angiotensin II receptor blockers such as Losartan, originally developed as antihypertensives, have been repurposed to inhibit the profibrotic RAS axis within tumors [82]. Blocking the angiotensin II type 1 receptor on CAFs directly suppresses their activation and collagen synthesis. Critically, this also alleviates solid stress, which decompresses intratumoral blood vessels, thereby improving perfusion and enhancing the drug delivery [83].

The impact of anti-angiogenic drugs on fibrosis is highly dependent on their specific targets. Multikinase inhibitors targeting PDGF and FGF receptors (e.g., sunitinib [62]) can directly suppress CAF activation. By contrast, agents that exclusively target the VEGF pathway (e.g., bevacizumab) can induce tumor hypoxia. This, in turn, can paradoxically promote a more aggressive, profibrotic CAF phenotype via HIF-1$\alpha$ stabilization [44].

The antidiabetic drug Metformin has demonstrated significant antifibrotic effects by activating the central metabolic sensor, AMP-activated protein kinase (AMPK). AMPK activation inhibits the anabolic processes required for CAF activation and ECM protein synthesis. This metabolic intervention effectively suppresses the fibrogenic function of CAFs [37, 63].