Collagen-derived biomarkers reveal distinct fibrotic responses among cancer-associated fibroblasts Authors: Annika Hettich, Morten A. Karsdal, Nicholas Willumsen Published: 2026 ABSTRACT Cancer-associated fibroblasts (CAFs) are key drivers of extracellular matrix (ECM) deposition and remodeling in the tumor microenvironment (TME), processes that facilitate tumor progression and metastasis. However, CAFs exhibit significant heterogeneity and functional complexity, posing challenges for effective therapeutic targeting. In this study, we evaluated the production of three important ECM proteins - type I (PRO-C1), type III (PRO-C3), and type VI (PRO-C6) collagen - by CAFs in vitro. Four distinct CAFs were cultured in Ficoll-containing media supplemented with ascorbic acid for up to 12 days. Cells were stimulated with profibrotic or inflammatory factors (TGF-β1, PDGF-AB, IL-1α, IL-6) and/or treated with antifibrotic compounds (ALK5i, Fresolimumab). Collagen production was quantified in collected cell culture media using competitive ELISA. Our results reveal distinct fibrotic responses among CAFs. Two CAFs displayed high intrinsic fibrotic activity and minimal additional fibrotic responsiveness to profibrotic stimuli, whereas two CAFs exhibited low intrinsic fibrotic activity and significant increases in PRO-C1, PRO-C3, and PRO-C6 upon profibrotic stimulation. Notably, TGF-β1 was the primary driver of PRO-C3, PDGF-AB was the primary driver of PRO-C6, while IL-1α and IL-6 had no effect on PRO-C1, PRO-C3 and PRO-C6 levels. Antifibrotic treatments with ALK5i and Fresolimumab effectively reduced collagen biomarkers elevated by TGF-β1 to baseline levels or below.These results underscore the heterogeneity of CAFs in ECM remodeling, highlighting the need for tailored therapeutic strategies to target tumors exhibiting high fibrotic activity. Introduction Cancer remains one of the most significant global health challenges, with the number of newly reported cases continuing to rise steadily1. It is a complex disease characterized by the uncontrolled proliferation of cells, evasion of normal growth controls, and the capacity to invade surrounding tissues and metastasize to distant organs2. Although much of the early focus in cancer research has been on the genetic and molecular characteristics of tumor cells themselves, there is now growing recognition that the surrounding tumor microenvironment (TME) plays an equally critical role in driving cancer progression and metastasis3. The TME consists of a dynamic and complex network of non-malignant cells, signaling molecules, blood vessels, and extracellular matrix (ECM) components that collectively support tumor growth, angiogenesis, immune evasion, and therapeutic resistance. This intricate ecosystem includes immune cells, endothelial cells, and fibroblasts, each contributing to tumor development in unique ways4–6. Among the stromal cells in the TME, cancer-associated fibroblasts (CAFs) are recognized as key mediators of tumorigenesis and metastasis7. Unlike normal fibroblasts, which usually remain quiescent in healthy tissue, where they maintain tissue architecture and are temporarily activated during wound healing, CAFs adopt a chronically activated state. This chronic activation, often described as part of the “cancer as a wound that never heals” concept, drives a predominantly pro-tumorigenic phenotype that supports tumor growth, modulates immune responses, and promotes pro-tumorigenic ECM remodeling8. CAFs are defined as spindle-shaped cells that lack epithelial, endothelial, and leukocyte markers but express various mesenchymal markers, including α-smooth muscle actin (α-SMA), fibroblast activation protein (FAP), and vimentin (VIM), among others. Despite expressing the same markers, CAFs are highly heterogeneous, varying in their origin, function, and secretory profiles, which can differ both between tumor types and within the same tumor9. CAFs can originate from various sources, including quiescent stellate cells, normal fibroblasts, epithelial cells, and endothelial cells. Additionally, certain cell types, such as adipocytes or macrophages, can undergo transdifferentiation to become CAFs10. Tumor cell-derived growth factors and chemokines, such as transforming growth factor beta 1 (TGF-β1), epidermal growth factor (EGF), platelet-derived growth factor (PDGF), interleukin-6 (IL-6), and interleukin-1 (IL-1), play pivotal roles in shaping CAF phenotypes by activating diverse signaling pathways11. Depending on the specific stimuli, distinct CAF subtypes may emerge, including myofibroblast-like CAFs (myCAFs), inflammatory CAFs (iCAFs), and antigen presenting CAFs (apCAFs), which show differences in their functional roles in tumorigenesis12,13. This heterogeneity can furthermore impact their interactions with cancer cells, as well as their influence on the surrounding ECM14–16. Recent research has highlighted the role of myCAFs in regulating the composition and organization of the ECM, particularly through the production of collagens, fibronectin, among other structural proteins. The remodeling of the ECM by myCAFs leads to a stiffer, fibrotic microenvironment that favors tumor invasion and metastasis11. This process, termed desmoplasia, is commonly observed in solid tumors such as those of the breast, pancreas, colon and lung, and is associated with poor prognosis17–20. Moreover, the stiffened ECM acts as a physical barrier that impedes the delivery of anti-cancer therapies, including chemotherapy, antibody-drug conjugates (ADCs) and immunotherapy, ultimately leading to drug resistance21,22. As the major components of the ECM, collagens provide not only structural support but also influence cell signaling, impacting migration, differentiation, and tumor progression23,24. Fibrillar collagens such as type I and III are among the most extensively studied. As key constituents of the interstitial matrix, they are frequently upregulated in tumors and contribute to tumorigenesis by promoting cell proliferation, migration, invasion, metastasis, inflammation, and drug resistance, ultimately shaping the fibrotic landscape that influences disease progression and therapy response25,26. Type I collagen, the most abundant fibrillar collagen, provides tensile strength to tissues such as skin, bone, and tendons. Type III collagen can be found in skin, vessel walls, lung, liver, spleen, among others and is co-expressed along with type I collagen. The known central function of type III collagen is cell-matrix adhesion, collagen fibril organization, ECM organization and the response to TGF-β1. Type I and type III collagen provide a scaffold for fibroblast attachment during tissue repair and wound healing, thereby improving the mechanical integrity of the healing tissue27. Type VI collagen, a non-fibrillar collagen forming microfibrillar networks, plays a crucial role in anchoring cells to the ECM and maintaining tissue architecture. Importantly, type VI collagen is actively involved in wound healing by supporting fibroblast adhesion, migration, and survival, as well as modulating the inflammatory response - processes that are essential for granulation tissue formation and ECM remodeling27,28. Additionally, elevated levels of type VI collagen, are frequently observed in various cancers and serve as a marker for poor prognosis, tumor progression, and increased metastasis28,29. CAFs overexpress various types of collagens, including type I, III, and VI, which have been linked to increased tumor aggressiveness30. However, collagen production by CAFs is not uniform; different CAFs produce varying levels and types of collagens, reflecting their functional heterogeneity14,30,31. Understanding the mechanisms regulating this heterogeneity is crucial for elucidating ECM-driven cancer progression and identifying therapeutic targets. In this context, ECM-based biomarkers serve as prognostic indicators, predictive markers of therapy response, and potential treatment targets32. PRO-C1, PRO-C3, and PRO-C6 are blood-based biomarkers that measure fragments of collagen (types I, III, and VI, respectively) released into circulation during tissue formation or remodeling, providing a readout of ECM turnover. Since these biomarkers are assessed from a simple blood sample rather than requiring a tissue biopsy, they provide a non-invasive means to monitor ECM remodeling and fibrotic activity in patients and show promise across various cancer types33,34. The ability to track fibrotic processes through circulating collagen-based biomarkers offers a clinically relevant tool for patient management - enabling early identification of high-risk patients, personalized therapeutic strategies, and real-time monitoring of antifibrotic or stroma-targeting therapies35–37. For instance, elevated PRO-C3 levels in serum have been associated with aggressive desmoplastic tumors and poor prognosis in several cancer types34,35,38,39. Similarly, PRO-C6, a marker of type VI collagen formation, reflects pericellular fibrosis and is emerging as a potential indicator of therapy resistance and stromal remodeling38,40. PRO-C1 reflects fibroblast and myofibroblast activity and indicates active deposition, however, its prognostic value in cancer is less consistent and generally weaker than that of PRO-C3 and PRO-C641. Integrating such biomarkers into clinical decision-making could ultimately improve treatment outcomes by addressing the fibrotic barriers that limit therapeutic efficacy in solid tumors. Given the role of CAFs in ECM remodeling and their association with poor prognosis, therapeutic strategies targeting CAFs and their signaling pathways are actively being explored. While these approaches hold promise, challenges remain in optimizing their clinical efficacy11,42. Preclinical models are essential for evaluating drug candidates and translating therapies into clinical applications. In addition to their clinical utility as blood-based biomarkers, measuring PRO-C1, PRO-C3, and PRO-C6 in vitro allows for the characterization of collagen-producing cells, such as and their contribution to fibrotic ECM remodeling38,43,44. In vitro models provide a controlled environment to test how specific compounds, or therapeutic agents affect collagen production and ECM dynamics, supporting preclinical evaluation of stroma-targeting therapies. Correlating in vitro collagen biomarker levels with in vivo or patient-derived data enhances the translational relevance of experimental findings and helps to identify cellular drivers of fibrosis38. Ultimately, using these biomarkers in vitro may support the development of mechanism-based therapies and the refinement of biomarker-guided treatment strategies. A tool to study fibroblasts is the Scar-in-a-Jar (SiaJ) cell-culture model, that was introduced by Chen et al.45. Fibroblasts are cultured under “crowded” conditions in Ficoll-based media, which mimics the in vivo environment and promotes increased collagen production. The model was further refined to investigate ECM profiles under both non-pathological and pathological conditions, including the evaluation of fibrotic activity in fibroblasts isolated from cancer patients38,43,44,46. Using translational biomarkers to measure collagen fragments in patient serum or collected cell culture media derived from the SiaJ model, Nissen et al.38 demonstrated that CAFs isolated from pancreatic tumors produce higher levels of type III and type VI collagen compared to healthy pancreatic fibroblasts and exhibit distinct responses to pro- and antifibrotic compounds. Treatment with an antifibrotic agent inhibited the production of both collagen types in pancreatic CAFs and healthy fibroblasts. Building on these findings, we aimed to investigate in this study the heterogeneity in collagen production and growth factor responsiveness across CAFs derived from multiple cancer types (pancreas, breast, lung, and colon) and healthy tissue counterparts. Using the SiaJ model, we quantified the production of type I (PRO-C1), type III (PRO-C3), and type VI (PRO-C6) collagen and assessed how CAFs respond to key profibrotic and inflammatory signals, including TGF-β1, PDGF-AB, IL-1α, and IL-6, which are known to shape CAF phenotypes. We further evaluated the effects of TGF-β pathway inhibition using the small molecule ALK5i and the monoclonal antibody Fresolimumab on collagen deposition. We hypothesized that CAF populations exhibit intrinsic variability in collagen production and differential growth factor responses, with TGF-β signaling serving as a central regulator of these phenotypes. The model is illustrated in Fig. 1. Materials & methods Cells Native healthy normal fibroblasts (NFs) and primary CAFs from different donors and tissues - pancreatic (CAF1 and NF1), breast (CAF2 and NF2), lung (CAF3 and NF3) and colorectal (CAF4 and NF4) - were selected to investigate a broad range of fibroblasts, whereas CAFs and NFs were not isolated from the same donor (all CAFs and NF1 used in this study were obtained from Vitro Biopharma (Golden, CO, US) with the following catalog numbers: CAF1 (cat# CAF08), CAF2 (cat# CAF06), CAF3 (cat# CAF07-AD), CAF4 (cat# CAF05), and NF1 (cat# SC00A5)). All CAFs were selected by Vitro Biopharma based on different CAF-specific markers including FAP, GFAP, FSP-1, Vimentin, α-SMA, CD90, CD105. NF2 was acquired from Innoprot (Derio, Spain, cat# P10893). NF3 was sourced from Lonza (Morrisville, NC, US, cat# CC-2512), and NF4 was obtained from ATCC (Manassas, VA, US, cat# CRL-1459). All cells were kept on a low passage number, not higher than passage 7. Cell culture of NFs and CAFs The cells were cultured following the “Scar-in-a-Jar” protocol as described by Nissen et al.38. Briefly, CAF1 were cultured in pancreatic stellate CAF maintenance medium (Vitro Biopharma). CAF2, CAF3, CAF4, and NF1 were cultured in VitroPlus III medium (Vitro Biopharma). NF3 and NF4 were cultured in high serum medium (Dulbecco’s Modified Eagle Medium (DMEM) + GlutaMax (Gibco) containing 10% fetal bovine serum (FBS) (Sigma-Aldrich) and 1% Penicillin/Streptomycin (P/S) (Sigma-Aldrich, cat#P4333)), and NF2 in fibroblast medium containing 2% FBS (Innoprot), until they reached 80–90% confluency in culture flasks that were coated with 5 µg/cm² of type I collagen purified from rat tail tendon (Innoprot) following the manufacturer’s instructions. On day − 1, fibroblasts were detached using trypsin (Sigma-Aldrich), and 30.000 cells/well were seeded into a 48-well plate in high serum medium. The following day (day 0), the media was replaced with Ficoll medium (low serum medium (0.4% FBS in DMEM + GlutaMax, 1% P/S) with 56.25 mg/mL Ficoll 70 and 37.5 mg/mL Ficoll 400 (Sigma-Aldrich) supplemented with 0.05 mg/mL L-ascorbic acid (Wako)) to promote “crowded” conditions. A total of 200 µL of Ficoll medium was added to each well. 100 µL of treatments diluted in low serum medium were added as appropriate, including: 1 ng/mL TGF-β1 (R&D Systems, cat#7754-BH), 100 ng/ml PDGF-AB (R&D Systems, cat#222-AB), 10 ng/ml IL-1α (R&D Systems, cat#200-LA), 62.7 mg/ml IL-6 (R&D Systems, cat#206-IL) in combination of 3.42 mg/ml IL-6Rα (R&D Systems, cat#227-SR), 1.9 µM TGF-β type I receptor inhibitor (ALK5i) (cat# SB-525334) and 10 µg/ml Fresolimumab (MedChem Express, cat#HY-P99020). ALK5i was diluted in dimethyl sulfoxide (DMSO), and an appropriate vehicle control with DMSO was used as control. The final DMSO concentration was < 1%. The concentrations of growth factors and cytokines were selected based on previously established in vitro models38,43. The concentration of ALK5i was selected based on a study by Nissen et al., who reported a reduction in collagen formation without seeing an impact on cell viability38. Cells were incubated at 37 °C with 5% CO₂ and cultured for 12 days. The medium containing freshly prepared treatments was exchanged on day 0, 3, 6 and 9 (Fig. 1). Cell culture media were collected and stored at -20 °C until biomarker assessment by competitive ELISA. Fig. 1Experimental workflow of the Scar-in-a-Jar model. CAFs and NFs were cultured in a 48-well plate with Ficoll-enriched media over 12 days. CAFs and NFs with different numbers originate from different organs, whereas those with the same number are derived from the same organ. All cells were obtained from different donors. Fibroblasts were either left unstimulated or treated with various profibrotic or inflammatory (TGF-β1, PDGF-AB, IL-1α, IL-6) and/or antifibrotic (ALK5i, Fresolimumab) compounds. Culture media was collected every three days and replaced with fresh media. ELISA-based biomarker levels were quantified in the collected media. On day 12, matrices were decellularized, and collagen deposition was assessed via Picro Sirius Red staining. Created in https://BioRender.com. Experimental workflow of the Scar-in-a-Jar model. CAFs and NFs were cultured in a 48-well plate with Ficoll-enriched media over 12 days. CAFs and NFs with different numbers originate from different organs, whereas those with the same number are derived from the same organ. All cells were obtained from different donors. Fibroblasts were either left unstimulated or treated with various profibrotic or inflammatory (TGF-β1, PDGF-AB, IL-1α, IL-6) and/or antifibrotic (ALK5i, Fresolimumab) compounds. Culture media was collected every three days and replaced with fresh media. ELISA-based biomarker levels were quantified in the collected media. On day 12, matrices were decellularized, and collagen deposition was assessed via Picro Sirius Red staining. Created in https://BioRender.com. Assessment of metabolic activity by Alamar Blue To assess the effects of treatments and stimulations on cell viability and health, 200 µL of 10% Alamar blue (Invitrogen) diluted in low serum medium was added to each well after media removal on day 0 and on day 12. Cells were incubated at 37 °C, 5% CO2 for 1–2 h. Alamar Blue contains resazurin, a blue, non-fluorescent dye. Viable, metabolically active cells reduce resazurin into resorufin, which is pink and fluorescent. The amount of resorufin produced correlates with cell viability and proliferation. 160 µL of Alamar blue solution was transferred to a black 96-well plate and the fluorescence was measured with 540 nm excitation wavelength on a SpectraMac ELISA reader. Decellularization and staining with Picro Sirius Red to assess net-collagen content On day 12, the cells were washed with phosphate buffered saline (PBS). 300 µL of extraction buffer (PBS + 0.5% Triton X-100 + 20 mM NH2OH) was added to each well and incubated at 37 °C for 10–15 min, or until no intact cells were visible under the light-microscope. After, 300 µL of PBS was added to each well to dilute cell debris. Wells were incubated in the diluted extraction buffer overnight at 4 °C. The next day, the cell debris was carefully removed without disturbing the matrices. Wells were once washed with PBS and then with PBS+ (PBS containing 1 mM CaCl2 and 1 mM MgSO4). Wells were stored in PBS+ containing 1% P/S to preserve ECM structure and avoid contamination until further analysis. To visualize the net-collagen content, the decellularized matrices were stained with Picro Sirius Red. For that, PBS+ was removed and 100 µL of Picro Sirius Red solution, containing 0.5 g Sirius Red F3B in 500 mL saturated aqueous solution of picric acid, was added to each well and incubated at room temperature (RT) for 1 h. After incubation, wells were washed first with 70%, then 96% and then absolute ethanol each for 2 min. Matrices were allowed to completely dry under the fume hood. The stained matrices were visualized under a bright field microscope at 20x magnification. Competitive enzyme-linked immunosorbent assay (ELISA) to assess collagen formation in cell cultures The collagen synthesis of type I collagen (PRO-C1, nordicPRO-C1™, cat.no. 2800), type III collagen (PRO-C3, nordicPRO-C3™, cat.no. 1700) and type VI collagen (PRO-C6, nordicPRO-C6™, cat.no. 4000) was assessed in the collected cell culture media on day 3, 6, 9 and 12 using specific competitive ELISAs according to the manufacturer’s instructions (Nordic Bioscience A/S). The PRO-C1 assay measures the N-terminal pro-peptide of type I collagen, describing the formation of type I collagen47. The PRO-C3 assay measures the neo-epitope of the N-terminal pro-peptide of type III collagen after cleavage from the pro-collagen molecule and describes the formation of type III collagen48 and the PRO-C6 assay measures the C-terminal of the released C5 domain of the type VI collagen α3 chain, also referred to as endotrophin, and describes the formation of type VI collagen49. Ethical statement All methods were carried out in accordance with relevant guidelines and regulations. Primary human healthy fibroblasts and CAFs were purchased from commercial suppliers (Vitro Biopharma, Lonza, and Innoprot). The cells were obtained from human donors who had provided informed consent and were collected under the suppliers’ respective ethical and regulatory approvals; documentation is available from the suppliers. As the cells were commercially available and de-identified, no additional institutional ethics approval was required for their use in this study. Cells were provided for research use only. Statistical analysis Statistical significance was determined using one-way ANOVA with Tukey’s or Dunnett’s multiple comparisons test or two-way ANOVA with Tukey’s multiple comparison tests. Biomarker measurements and Alamar blue values are plotted as raw values. Data are presented as mean ± SD of 1–8 separate experiments, each with 2–4 technical replicates. Values are plotted as bar plots comparing biomarker data from day 3, 6, 9 and 12 or as bar plots comparing accumulated data of all four treatment days. All statistical tests were performed in GraphPad Prism software v.6 (GraphPad Software, San Diego, CA). p-values lower than 0.05 were considered significant. Results CAFs and NFs show differential collagen production activity in response to TGF-β1 stimulation The formation of type I (PRO-C1), type III (PRO-C3), and type VI (PRO-C6) collagen was assessed in four different CAFs (designated as CAF1, CAF2, CAF3, and CAF4) and four organ-matched normal fibroblasts (designated as NF1, NF2, NF3 and NF4), isolated from the same organs but from different donors, under both unstimulated conditions and after stimulation with TGF-β1. Fig. 2Differential collagen formation of CAFs and NFs in response to TGF-β1 stimulation. CAF1-4 and NF1-4 were cultured with or without TGF-β1 stimulation. Biomarkers of ECM synthesis were assessed at days 3, 6, 9, and 12 (a–c) or as the accumulated deposition over all four time points (d–f). Synthesis of type I (a, d), type III (b, e), and type VI (c, f) collagen were quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. On day 0 (g) and on day 12 (h) metabolic activity was assessed using Alamar Blue. All data is shown as mean ± SD of 2–8 separate experiments each with 4 technical replicates. Statistical significance was assessed using an ordinary two-way ANOVA followed by Tukey’s multiple comparisons test, comparing the means of all groups within each timepoint (a–c) or CAF/NF type (d–f, h) with each other. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Differential collagen formation of CAFs and NFs in response to TGF-β1 stimulation. CAF1-4 and NF1-4 were cultured with or without TGF-β1 stimulation. Biomarkers of ECM synthesis were assessed at days 3, 6, 9, and 12 (a–c) or as the accumulated deposition over all four time points (d–f). Synthesis of type I (a, d), type III (b, e), and type VI (c, f) collagen were quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. On day 0 (g) and on day 12 (h) metabolic activity was assessed using Alamar Blue. All data is shown as mean ± SD of 2–8 separate experiments each with 4 technical replicates. Statistical significance was assessed using an ordinary two-way ANOVA followed by Tukey’s multiple comparisons test, comparing the means of all groups within each timepoint (a–c) or CAF/NF type (d–f, h) with each other. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Figure 2a-c illustrates the release of PRO-C1, PRO-C3, and PRO-C6 by fibroblasts into the cell culture media collected at multiple time points over a 12-day period. CAF1/NF1 and CAF2/NF2 exhibited similar overall responses to TGF-β1 stimulation. Under unstimulated conditions, NF1 and NF2 released approximately 2-fold lower PRO-C1 levels than their corresponding CAF1 and CAF2 counterparts on days 6, 9 and 12 (Fig. 2a). TGF-β1 stimulation induced a 1.5-fold increase in PRO-C1 in NF1 and a 2-fold increase in NF2. In contrast, CAF1 and CAF2 maintained stable PRO-C1 levels following stimulation, with no significant change relative to unstimulated controls, resulting in comparable PRO-C1 levels between stimulated CAF1 and NF1, and between CAF2 and NF2. Unstimulated CAF3 exhibited ~2-fold higher PRO-C1 levels compared to NF3 on days 3, 6 and 9, whereas CAF4 and NF4 showed similar levels under unstimulated conditions. Both CAF3 and CAF4, as well as NF3 and NF4, responded to TGF-β1 stimulation with a 3-4-fold increase in PRO-C1 levels. Unstimulated NF1 and NF2 produced 3-5-fold lower PRO-C3 levels than their corresponding CAF1 and CAF2 counterparts (Fig. 2b). TGF-β1 stimulation significantly increased PRO-C3 levels in both CAF1/NF1 and CAF2/NF2. Under stimulated conditions, CAF1 maintained 2-3-fold higher PRO-C3 levels than NF1, whereas stimulated CAF2 and NF2 exhibited comparable levels. Neither CAF3/NF3 nor CAF4/NF4 released considerable PRO-C3 levels under unstimulated conditions throughout the 12-day period. Following TGF-β1 stimulation, PRO-C3 levels increased significantly in all fibroblasts. Notably, NF3 released 1.5-2-fold higher PRO-C3 levels than CAF3 at days 9 and 12 under stimulated conditions, while NF4 and CAF4 reached similar levels. For PRO-C6 (Fig. 2c), CAF1 displayed a 2-fold decrease following TGF-β1 stimulation at days 9 and 12, whereas NF1 exhibited no significant change. Similarly, CAF2 exhibited a 2-fold reduction in PRO-C6 release upon stimulation, whereas NF2 showed a 2-3-fold increase. In contrast, CAF3 and CAF4 displayed 2-3-fold increases in PRO-C6 levels following stimulation, and modest increases were also observed in NF3 and NF4. Overall, PRO-C1 levels remained stable over time across all conditions, whereas PRO-C3 increased and reached a plateau between days 9 and 12. PRO-C6 similarly exhibited a consistent release profile over the study period (Fig. 2a-c). Figure 2d-f summarizes the accumulated deposition of PRO-C1, PRO-C3, and PRO-C6 over the 12-day period. While TGF-β1 stimulation did not increase PRO-C1 levels in CAF1 and CAF2, it caused a 2-fold increase in NF1 and NF2. Additionally, both CAF3 and CAF4, as well as NF3 and NF4, exhibited significantly higher PRO-C1 levels after stimulation (P < 0.0001). Under unstimulated conditions, PRO-C1 levels were similar between CAF1 and CAF2 and were 2-3-fold higher than those in CAF3 and CAF4. However, upon TGF-β1 stimulation, PRO-C1 levels became similar across CAF1-3 and NF1-3, while CAF4 and NF4 demonstrated modestly lower levels (Fig. 2d). Baseline PRO-C3 levels were over 30-fold higher in CAF1 and CAF2 compared to CAF3 and CAF4 (CAF1 mean: 996 ng/ml, CAF2 mean: 703 ng/ml, CAF3 mean: 23 ng/ml, CAF4 mean: 38 ng/ml). The same was true for NF1 and NF2 compared to NF3 and NF4. TGF-β1 stimulation significantly increased PRO-C3 levels in all CAFs and NFs, although CAF1 and CAF2 as well as NF1 and NF2 maintained significantly higher levels, compared to CAF3 and CAF4 and NF3 and NF4, respectively (CAF1 mean: 1274 ng/ml, CAF2 mean: 985 ng/ml, CAF3 mean: 161 ng/ml, CAF4 mean: 226 ng/ml) (Fig. 2e). In contrast, TGF-β1 stimulation resulted in a significant decrease in PRO-C6 levels in CAF1 and CAF2, while NF1 remained unresponsive and NF2 exhibited a 3-fold increase. Both CAF3 and CAF4, as well as NF3 and NF4, demonstrated significant increases in PRO-C6 levels after stimulation. CAF1 and CAF4 exhibited similar baseline PRO-C6 levels, as did CAF2 and CAF3, although the latter were slightly lower. Upon stimulation, CAF3 showed 2-fold and CAF4 1.5-fold higher levels compared to CAF1 and CAF2 (Fig. 2f). To determine whether changes in biomarker levels were attributable to alterations in metabolic activity, we conducted the Alamar Blue assay at day 0 (Fig. 2g) and day 12 (Fig. 2h). At the time of seeding (day 0, prior to stimulation initiation), all CAFs and NFs exhibited comparable metabolic activity levels (Fig. 2g). By day 12, distinct differences emerged between CAFs and NFs, as well as between unstimulated and TGF-β1-stimulated fibroblasts (Fig. 2h). For CAF1 and NF1, metabolic activity rates were comparable and remained unchanged following TGF-β1 stimulation. Similar levels were observed for unstimulated and stimulated CAF2 as well as unstimulated NF2; however, NF2 exhibited a significant increase in metabolic activity upon stimulation. In contrast, unstimulated CAF3 displayed 2-fold lower metabolic activity compared to unstimulated NF3. Nevertheless, TGF-β1 stimulation led to a highly significant increase in CAF3 metabolic activity, reaching levels comparable to those of NF3, who demonstrated a slight significant increase in metabolic activity following TGF-β1 stimulation. Similarly, CAF4 showed a tendency towards lower metabolic activity than NF4 under unstimulated conditions. Upon stimulation, metabolic activity increased significantly in CAF4, reaching levels comparable to those of NF4, which also showed an increase upon stimulation. In summary, TGF-β1 stimulation revealed distinct collagen production profiles across CAFs and their matched NFs. Unstimulated CAFs generally showed higher levels of PRO-C1 and PRO-C6 than NFs, with CAF1 and CAF2 also exhibiting markedly elevated PRO-C3 levels. In contrast, CAF3 and CAF4 had lower baseline collagen production but responded strongly to TGF-β1, particularly in PRO-C1 and PRO-C6, although their PRO-C3 levels remained significantly lower than those of CAF1 and CAF2. NFs produced lower levels of type I and type III collagen at baseline but showed robust increases in response to TGF-β1, indicating a high responsiveness to profibrotic stimuli. Interestingly, TGF-β1 decreased PRO-C6 in CAF1 and CAF2 but increased it in CAF3 and CAF4 and NFs, except NF3. Overall, these data highlight the distinct fibrotic profiles of CAFs and their healthy counterparts from different cancer indications in both unstimulated and TGF-β1-stimulated conditions. Growth factors and cytokines differentially regulate collagen production across distinct CAFs Next, we aimed to investigate how CAFs responded to stimulation with PDGF-AB and the cytokines IL-1α and IL-6. Figure 3 shows the accumulated deposition of PRO-C1, PRO-C3 and PRO-C6 over a 12-day period. Fig. 3Differential impact of growth factors and cytokines on collagen production across distinct CAFs. CAF1-4 were cultured under unstimulated conditions or stimulated with TGF-β1, PDGF-AB, IL-1α, or IL-6 in combination with IL-6Rα. Biomarkers of ECM synthesis were measured at days 3, 6, 9, and 12, while the accumulated deposition over all four time points is shown. Collagen production of type I (a), type III (b), and type VI (c) was quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. On day 12 metabolic activity was assessed using Alamar Blue (d). All data is shown as mean ± SD of 2–8 separate experiments each with 4 technical replicates. To assess statistical significance, an ordinary one-way ANOVA with Dunnett’s multiple comparisons test was performed, comparing the stimulated groups to unstimulated. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Differential impact of growth factors and cytokines on collagen production across distinct CAFs. CAF1-4 were cultured under unstimulated conditions or stimulated with TGF-β1, PDGF-AB, IL-1α, or IL-6 in combination with IL-6Rα. Biomarkers of ECM synthesis were measured at days 3, 6, 9, and 12, while the accumulated deposition over all four time points is shown. Collagen production of type I (a), type III (b), and type VI (c) was quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. On day 12 metabolic activity was assessed using Alamar Blue (d). All data is shown as mean ± SD of 2–8 separate experiments each with 4 technical replicates. To assess statistical significance, an ordinary one-way ANOVA with Dunnett’s multiple comparisons test was performed, comparing the stimulated groups to unstimulated. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. For CAF1 and CAF2, none of the stimuli led to a significant difference in type I collagen deposition compared to unstimulated controls (Fig. 3a, CAF1 and CAF2). In contrast, only TGF-β1 stimulation induced a significant increase in PRO-C1 in CAF3 and CAF4, with no significant effect observed from PDGF-AB, IL-1α, or IL-6 (Fig. 3a, CAF3 and CAF4). PRO-C3 significantly increased with TGF-β1 stimulation across all CAFs, though other stimuli had minimal effect; only CAF1 exhibited a 1.5-fold increase in PRO-C3 levels upon IL-6 stimulation (P < 0.01) (Fig. 3b). Overall, PRO-C3 levels were notably higher in CAF1 and CAF2 compared to CAF3 and CAF4. PRO-C6 release showed differing responses across the CAFs: TGF-β1 stimulation significantly reduced PRO-C6 levels in CAF1 and CAF2, while it led to a 2-fold increase in PRO-C6 release in CAF3 and CAF4 (Fig. 3c). Stimulation with PDGF-AB induced a 1.5-2-fold increase of PRO-C6 levels across all CAFs. IL-1α stimulation led to a significant decrease in PRO-C6 release in CAF1 and a modest decrease in CAF2, with no observable effect in CAF3 and CAF4. Meanwhile, IL-6 had no impact on PRO-C6 release in any CAFs. Metabolic activity levels in CAF1 did not change significantly when stimulated with TGF-β1, PDGF-AB, and IL-1α compared to unstimulated cells, but decreased more than 2-fold upon stimulation with IL-6 (Fig. 3d). For CAF2, only PDGF-AB induced a modest decrease in metabolic activity compared to unstimulated cells (P < 0.05). In contrast, stimulation of CAF3 and CAF4 with TGF-β1, PDGF-AB, and IL-1α resulted in a 1.5-2-fold increase in metabolic activity, while no effect was observed with IL-6. Overall, CAF1 and CAF2 appear to be intrinsically active CAFs, exhibiting high baseline fibrotic activity and a limited response to TGF-β1 stimulation. On the other hand, CAF3 and CAF4 have a quiescent phenotype under baseline conditions and respond strongly to TGF-β1. Additionally, distinct stimuli demonstrated differential effects on collagen synthesis, with TGF-β1 primarily increasing PRO-C1 and PRO-C3, while PDGF-AB preferentially enhanced PRO-C6 release. Neither IL-1α nor IL-6 stimulation notably altered the PRO-C1, PRO-C3 and PRO-C6 levels in our model. Collagen formation in CAFs is modulated by ALK5i and Fresolimumab treatment Given the prominent role of TGF-β1 in modulating type I, III, and VI collagen synthesis, we next investigated whether blocking the TGF-β1 signaling pathway could further modify collagen production. To address this, CAFs were treated with ALK5i, a small molecule inhibitor targeting the type I receptor of the TGF-β1 superfamily. Additionally, cells were treated with Fresolimumab (GC1008), a human monoclonal antibody that neutralizes all TGF-β isoforms and has shown therapeutic potential in various cancers, including melanoma, renal carcinoma, and metastatic breast cancer50. Since CAF1 and CAF2, as well as CAF3 and CAF4, display similar phenotypes, CAF1/2 were classified as intrinsically active CAFs, characterized by high fibrotic activity even in the absence of TGF-β1 stimulation. In contrast, CAF3/4 were classified as quiescent CAFs, exhibiting low to minimal fibrotic activity under unstimulated conditions. Based on these baseline characteristics, fibroblasts were either left unstimulated or stimulated with TGF-β1. Co-treatments with ALK5i and Fresolimumab were initiated from day 0, with an additional treatment group receiving ALK5i starting from day 6. Figures 4 and 5 illustrate the release of PRO-C1, PRO-C3, and PRO-C6 by fibroblasts into cell culture media collected at multiple time points over a 12-day period (left graphs) or the accumulated release accumulated deposition over all four time points (right graphs). Intrinsically active CAFs did not exhibit increased PRO-C1 levels in response to TGF-β1 stimulation. However, treatment with ALK5i, either alone or in combination with TGF-β1, resulted in a 2-3-fold decrease in PRO-C1 compared to untreated cells, while ALK5i alone reduced PRO-C1 even more. For cells treated with ALK5i from day 6 onward, PRO-C1 levels declined significantly by day 12 relative to CAFs that were not treated with ALK5i (P < 0.01) (Fig. 4a). In contrast, quiescent CAFs showed a 2-fold increase in PRO-C1 following TGF-β1 stimulation. When ALK5i was applied alongside TGF-β1, PRO-C1 levels returned to baseline levels similar to untreated cells, while cells treated with ALK5i alone exhibited slightly reduced PRO-C1 release compared to untreated fibroblasts. Quiescent CAFs treated with ALK5i from day 6 also demonstrated decreased PRO-C1 levels by day 12 (P < 0.05) (Fig. 4b). The response in PRO-C3 mirrored that of PRO-C1 across the CAFs. For intrinsically active CAFs, ALK5i treatment reduced PRO-C3 levels below those in untreated cells (2-fold decrease) (Fig. 4c), whereas in quiescent CAFs, PRO-C3 levels returned to baseline when treated with ALK5i (Fig. 4d). In both CAF phenotypes, ALK5i treatment from day 6 resulted in decreased PRO-C3 levels by day 12 (P < 0.0001). For collagen VI production, intrinsically active CAFs showed a tendency towards reduced PRO-C6 levels under ALK5i treatment compared to untreated cells. Additionally, when ALK5i was administered on day 6, no significant reduction in PRO-C6 was detected on days 9 and 12 (Fig. 4e). Quiescent CAFs displayed neutralized PRO-C6 levels under continuous ALK5i treatment and a decrease in PRO-C6 at day 12 when fibroblasts were treated with ALK5i from day 6 (P < 0.0001) (Fig. 4f). Fig. 4Targeted inhibition of the TGF-β1 pathway by ALK5i alters fibrotic activity in CAFs. CAF1-4 were cultured under unstimulated conditions or stimulated with TGF-β1. Furthermore, cells were either treated with ALK5i alone, or additionally to TGF-β1 either from day 0 or from day 6. Collagen production of type I (a, b), type III (c, d), and type VI (e, f) were quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. Biomarkers of ECM synthesis were measured at days 3, 6, 9, and 12 (left graphs) or as accumulated deposition over all four time points (right graphs). All data is shown as mean ± SD of 1 experiment, with each 4–8 technical replicates. Statistical significance between treatment groups across different time points was assessed using an ordinary two-way ANOVA followed by Tukey’s multiple comparisons test, comparing the means of all groups within each timepoint with each other. For accumulated deposition analyses, statistical significance between treatment groups was evaluated using an ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, comparing the means of all treatment groups with each other. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Targeted inhibition of the TGF-β1 pathway by ALK5i alters fibrotic activity in CAFs. CAF1-4 were cultured under unstimulated conditions or stimulated with TGF-β1. Furthermore, cells were either treated with ALK5i alone, or additionally to TGF-β1 either from day 0 or from day 6. Collagen production of type I (a, b), type III (c, d), and type VI (e, f) were quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. Biomarkers of ECM synthesis were measured at days 3, 6, 9, and 12 (left graphs) or as accumulated deposition over all four time points (right graphs). All data is shown as mean ± SD of 1 experiment, with each 4–8 technical replicates. Statistical significance between treatment groups across different time points was assessed using an ordinary two-way ANOVA followed by Tukey’s multiple comparisons test, comparing the means of all groups within each timepoint with each other. For accumulated deposition analyses, statistical significance between treatment groups was evaluated using an ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, comparing the means of all treatment groups with each other. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Additionally, cells were treated with Fresolimumab, either alone or in combination with TGF-β1. PRO-C1 levels in intrinsically active CAFs were similar between untreated cells, TGF-β1-stimulated cells, and those additionally treated with Fresolimumab. Furthermore, cells treated exclusively with Fresolimumab did not exhibit a reduction in PRO-C1 levels compared to untreated cells (Fig. 5a). In contrast, elevated PRO-C1 levels upon TGF-β1 treatment decreased modestly for quiescent CAFs with Fresolimumab treatment (P < 0.05). However, Fresolimumab alone did not change PRO-C1 levels in quiescent CAFs compared to untreated fibroblasts (Fig. 5a). Regarding collagen III formation, TGF-β1 treatment resulted in a 1.5-2-fold increase in PRO-C3 levels in intrinsically active CAFs and treatment with Fresolimumab only resulted in a trend toward reduced PRO-C3 levels (Fig. 5c). In contrast, Fresolimumab significantly reduced PRO-C3 levels in quiescent CAFs that were treated with TGF-β1 (Fig. 5d). No significant difference was detected between CAFs treated with Fresolimumab alone and those treated with both TGF-β1 and Fresolimumab. Regarding type VI collagen, intrinsically active CAFs exhibited stable PRO-C6 levels across all conditions. However, there was a tendency towards higher levels of PRO-C6 when cells were treated with Fresolimuab for these CAFs (Fig. 5e). Quiescent CAFs showed a 2–3-fold increase in PRO-C6 levels following TGF-β1 treatment. Treatment with Fresolimumab reduced PRO-C6 levels back to baseline (Fig. 5f). Fig. 5Targeted inhibition of the TGF-β1 pathway by Fresolimumab alters fibrotic activity in CAFs. CAFs (CAF1–4) were cultured under unstimulated conditions or stimulated with TGF-β1. Furthermore, cells were either treated with ALK5i alone, or additionally to TGF-β1 either from day 0 or from day 6. Collagen production of type I (a, b), type III (c, d), and type VI (e, f) were quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. Biomarkers of ECM synthesis were measured at days 3, 6, 9, and 12 (left graphs) or as accumulated deposition over all four time points (right graphs). All data is shown as mean ± SD of 1 experiment, with each 4–8 technical replicates. Statistical significance between treatment groups across different time points was assessed using an ordinary two-way ANOVA followed by Tukey’s multiple comparisons test, comparing the means of all groups within each timepoint with each other. For accumulated deposition analyses, statistical significance between treatment groups was evaluated using an ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, comparing the means of all treatment groups with each other. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Targeted inhibition of the TGF-β1 pathway by Fresolimumab alters fibrotic activity in CAFs. CAFs (CAF1–4) were cultured under unstimulated conditions or stimulated with TGF-β1. Furthermore, cells were either treated with ALK5i alone, or additionally to TGF-β1 either from day 0 or from day 6. Collagen production of type I (a, b), type III (c, d), and type VI (e, f) were quantified using the PRO-C1, PRO-C3, and PRO-C6 assays, respectively. Biomarkers of ECM synthesis were measured at days 3, 6, 9, and 12 (left graphs) or as accumulated deposition over all four time points (right graphs). All data is shown as mean ± SD of 1 experiment, with each 4–8 technical replicates. Statistical significance between treatment groups across different time points was assessed using an ordinary two-way ANOVA followed by Tukey’s multiple comparisons test, comparing the means of all groups within each timepoint with each other. For accumulated deposition analyses, statistical significance between treatment groups was evaluated using an ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test, comparing the means of all treatment groups with each other. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. For intrinsically active CAFs, metabolic activity levels were comparable between treatment groups and only decreased when fibroblasts were treated with ALK5i alone, compared to untreated cells (Fig. 6a). In contrast, for quiescent CAFs, treatment with TGF-β1 resulted in nearly a 2-fold increase in metabolic activity compared to untreated fibroblasts (Fig. 6b). Furthermore, quiescent cells that were treated with both TGF-β1 and ALK6i starting from day 6 also showed the tendency towards a higher metabolic activity. Fig. 6Metabolic activity of CAFs treated with ALK5i and Fresolimumab. Intrinsically active CAFs (a) and quiescent CAFs (b) were cultured either under unstimulated conditions or stimulated with TGF-β1. Cells were further treated with ALK5i or Fresolimumab, either alone or in combination with TGF-β1. Treatments were added from day 0 or, in the case of ALK5i, additionally from day 6. Metabolic activity was assessed on day 12 (at the end of the experiment). All data is shown as mean ± SD of 1 experiment, with each 4–8 technical replicates. To assess statistical significance between treatment groups an ordinary one-way ANOVA with Dunnett’s multiple comparisons test was performed, comparing the means of all treatment groups with each other. **P < 0.01. Metabolic activity of CAFs treated with ALK5i and Fresolimumab. Intrinsically active CAFs (a) and quiescent CAFs (b) were cultured either under unstimulated conditions or stimulated with TGF-β1. Cells were further treated with ALK5i or Fresolimumab, either alone or in combination with TGF-β1. Treatments were added from day 0 or, in the case of ALK5i, additionally from day 6. Metabolic activity was assessed on day 12 (at the end of the experiment). All data is shown as mean ± SD of 1 experiment, with each 4–8 technical replicates. To assess statistical significance between treatment groups an ordinary one-way ANOVA with Dunnett’s multiple comparisons test was performed, comparing the means of all treatment groups with each other. **P < 0.01. In summary, our results demonstrate that the fibrotic activity of CAFs can be effectively modulated by inhibiting distinct steps of the TGF-β1 pathway. Both small molecules and biologicals proved effective in this system, allowing for varied therapeutic approaches to be tested. Additionally, we observed distinct responses to treatment across different CAF types, highlighting the heterogeneity in fibrotic behavior and personalized treatment. Differences in collagen modifications among CAFs reflect overall net-collagen content Next, we aimed to assess whether the selective modulation of collagen types I, III, and VI synthesis, as measured by PRO-C1, PRO-C3 and PRO-C6 in the collected cell culture media, also manifested in the overall collagen content in the culture well. To assess the total collagen content within the well, the ECM was decellularized on day 12, and the overall collagen deposition was visualized with Picro Sirius Red staining. Intrinsically active CAFs exhibited high collagen deposition under untreated conditions, while quiescent CAFs showed minimal collagen staining (Fig. 7a). Following TGF-β1 treatment, collagen staining remained comparable to the untreated condition in intrinsically active CAFs but was markedly increased for quiescent CAFs (Fig. 7b). This increase aligns with biomarker measurements for PRO-C1, PRO-C3, and PRO-C6, which also rose for quiescent CAFs after TGF-β1 stimulation (Fig. 3). Intrinsically active CAFs treated with ALK5i or Fresolimumab displayed slightly reduced collagen deposition compared to untreated cells, while staining for quiescent CAFs were comparable to untreated levels (Fig. 7c and d). Also here, the overall collagen content reflects what was measured by collagen-based biomarkers (Figs. 4 and 5). Fig. 7Differences in collagen modifications among CAFs reflect overall net-collagen content. On day 12 of the SiaJ, the ECM of CAFs was decellularized and stained with Picro Sirius Red to visualize the overall net-collagen content in each well of either untreated CAFs (a), CAFs treated with TGF-β1 alone (b) and in combination with ALK5i (c) or Fresolimumab (d). Images were taken using a bright field microscope. Differences in collagen modifications among CAFs reflect overall net-collagen content. On day 12 of the SiaJ, the ECM of CAFs was decellularized and stained with Picro Sirius Red to visualize the overall net-collagen content in each well of either untreated CAFs (a), CAFs treated with TGF-β1 alone (b) and in combination with ALK5i (c) or Fresolimumab (d). Images were taken using a bright field microscope. Overall, matrix staining closely mirrored the collagen formation biomarkers. Intrinsically active CAFs exhibited high baseline levels of PRO-C1 and PRO-C3, which corresponded to the high net collagen content detected at a later time (day 12). In contrast, quiescent CAFs showed relatively low baseline levels of these biomarkers, with both increasing upon TGF-β1 stimulation, reflected by a corresponding rise in total collagen content at day 12. Treatment with either ALK5i or Fresolimumab reduced PRO-C1 and PRO-C3 levels across all CAFs, aligning with the observed decrease in net collagen deposition. Discussion In this study, we characterized CAFs in vitro with a particular focus on their collagen synthesis profiles and responses to potential profibrotic and antifibrotic stimuli. By quantifying translational biomarkers of type I (PRO-C1), type III (PRO-C3), and type VI (PRO-C6) collagens, we demonstrate pronounced heterogeneity among CAF populations with respect to both intrinsic collagen production and responsiveness to external signaling cues. Notably, we observed that two CAFs (CAF1 and CAF2) exhibited high baseline production of fibrillar collagens under untreated conditions, whereas CAF3 and CAF4 required stimulation with TGF-β1 to induce a comparable fibrotic response. These findings highlight the existence of CAF subtypes with distinct fibrotic activity states in the TME13,51. CAF heterogeneity is increasingly recognized as a defining feature of the TME, reflecting differences in cellular origin, activation state, and functional specialization. CAFs have been implicated in a wide range of tumor-promoting processes, including cancer cell proliferation and survival, invasion and metastasis, angiogenesis, immune modulation, and therapy resistance8. Beyond serving as structural components, collagens produced by CAFs actively shape the biochemical and biomechanical properties of the TME52. Elevated deposition of fibrillar collagens such as type I and III contribute to matrix stiffening, altered integrin signaling, enhanced cancer cell invasiveness, and immune cell exclusion, while type VI collagen has been associated with tumor cell survival and stress resistance53,54. In this context, the distinct collagen production profiles observed among the CAFs in our study suggest that different CAF populations may differentially influence tumor progression through their ECM output, with intrinsically active CAFs potentially contributing to a more aggressive, desmoplastic tumor phenotype. TGF-β1 is a central regulator of fibroblast activation and ECM remodeling and is secreted by both tumor and stromal cells in many cancer types43,44,55,56. Consistent with this, we observed a strong induction of type I, III and VI collagen synthesis in CAF3 and CAF4 following TGF-β1 stimulation, supporting its role as a key driver of CAF-mediated fibrosis. In contrast, CAF1 and CAF2 showed only a limited response to TGF-β1, suggesting that these fibroblasts may rely on alternative profibrotic mechanisms or exist in a pre-activated state. CAF activation is also influenced by receptor tyrosine kinase ligands such as PDGF and FGF, inflammatory cytokines, and biomechanical cues such as ECM stiffness57–60, which may contribute to the intrinsic fibrotic activity observed in CAF1 and CAF2. In addition to differences in activation state, our data indicates that distinct signaling pathways preferentially regulate the synthesis of specific collagen types. While TGF-β1 was the most potent inducer of type I and III collagen synthesis, PDGF-AB predominantly enhanced type VI collagen production, particularly in CAF1 and CAF243. Interestingly, TGF-β1 reduced type VI collagen formation in these CAFs, whereas both TGF-β1 and PDGF-AB increased type VI collagen synthesis in CAF3 and CAF4. These findings underscore the complexity of CAF signaling and suggest that effective modulation of tumor fibrosis may require simultaneous targeting of multiple pathways. Although IL-6 and IL-1α are known activators of CAFs and have been associated with pro-tumorigenic functions, they did not increase the production of the collagens analyzed here43,61,62, indicating that their profibrotic effects may involve other ECM components or matrix-modifying processes. Given the prominent role of fibrosis in limiting drug penetration and immune infiltration in solid tumors, there has been considerable interest in targeting CAFs and ECM remodeling as therapeutic strategies. In this study, we evaluated two TGF-β–targeted antifibrotic agents, ALK5i and Fresolimumab, both of which effectively reduced collagen synthesis in vitro, particularly in CAFs with high fibrotic activity (Fig. 8)63,64. ALK5i was especially potent, completely blocking TGF-β1–induced collagen production and suppressing intrinsic collagen synthesis in highly active CAFs. Importantly, delayed administration of ALK5i also reduced collagen synthesis, suggesting that established fibrotic phenotypes may be partially reversible. However, further validation of this effect requires extending the SiaJ model beyond 12 days. Fresolimumab also reduced collagen production but was less effective, potentially reflecting incomplete neutralization of TGF-β signaling at the concentrations tested. Interestingly, metabolic activity did not correlate directly with collagen production across CAF populations. CAF3 exhibited the highest metabolic activity while producing relatively low levels of type I and III collagen, indicating that increased fibroblast proliferation or metabolic activity does not necessarily equate to a fibrotic phenotype. This observation further emphasizes the functional heterogeneity of CAFs and highlights the importance of directly assessing ECM production. Overall, these results indicate that targeting the TGF-β pathway could provide a viable therapeutic strategy to mitigate fibrosis in cancer patients. Indeed, several antifibrotic drugs are currently being evaluated in preclinical and clinical trials65–69. However, despite promising clinical data, several antifibrotic drug candidates, including Fresolimumab, have entered clinical trials in oncology but were ultimately discontinued due to limited efficacy, safety concerns, strategic setbacks and the inherent complexity and heterogeneity of solid tumors and ECM-producing CAFs70. Furthermore, the tumor stroma can exhibit both pro-tumorigenic and anti-tumorigenic roles, meaning that complete stromal ablation may inadvertently promote more aggressive cancer phenotypes71. A major reason for these failures may be the lack of appropriate patient selection. Treating unselected populations likely masks therapeutic benefits in tumors that are truly TGF-β-driven. Tumors with a dense fibrotic stroma may represent a more responsive subgroup, as this microenvironment reflects active TGF-β signaling. Moreover, our findings reveal that the efficacy of TGF-β–targeted treatments vary among CAF populations, highlighting the need for personalized approaches to targeting fibroblast-mediated fibrosis. Another significant trend in oncology is the repurposing of antifibrotic drugs, initially developed for conditions like IPF, for cancer treatment72,73. Sara et al. previously demonstrated the efficacy of Nintedanib and Pirfenidone, two drugs used to treat IPF, in reducing collagen synthesis in a lung fibroblast SiaJ model44. It would be interesting to investigate whether these compounds can similarly modulate collagen deposition in CAFs. Additionally, new drugs, such as PXS-5505, a first-in-class highly selective and potent pan-lysyl oxidase inhibitor, are under development and already showed promising results in pancreatic tumors, by decreasing chemotherapy-induced tumor desmoplasia, reduces cancer cell invasion and metastasis and enhances efficacy of chemotherapy in a mice model74. Another emerging pillar in cancer therapy are antibody-drug conjugates (ADCs), which have gained increasing attention and shown clinical efficacy across various cancer types as summarized by Shastry et al.75 and Riccardi et al.76. However, Zhang et al. demonstrated that CAFs contribute to resistance against ADCs by forming a physical barrier that limits drug penetration, leaving a subset of patients refractory to treatment. Targeting focal adhesion kinase (FAK) - a key regulator of CAF maintenance - with the small molecule inhibitor IN10018 successfully disrupted this barrier and improved ADC delivery to the tumor, independent of the specific antigen targeted by the ADCs22. These advances raise hope that more anti-stromal therapies will be under investigation and development in the future, to enhance therapy efficacy and improve survival of cancer patients. Collagen biomarkers, such as PRO-C3 and PRO-C6, provide prognostic value and can guide patient selection for antifibrotic therapy36,38,77. In our model, biomarker measurements closely mirrored total collagen deposition assessed by Picro Sirius Red staining, supporting their utility as surrogate readouts of ECM formation. Combined with the SiaJ model, these biomarkers provide a rapid and translational platform for evaluating antifibrotic drug efficacy in vitro. Despite these promising findings, several limitations warrant consideration. Our study focused on a limited number of CAFs derived from specific types of cancer. Given the heterogeneity of CAFs between and within tumor niches, more CAFs from various donors and indications need to be tested. Although we tested CAFs from different types of organs, we cannot draw further tissue-specific conclusions, as we each only tested one donor. Furthermore, our 2-D in vitro model does not fully capture the complexities of the TME. Even though we culture fibroblasts in Ficoll-rich media, allowing for enhanced deposition of collagens, several cell types such as cancer cells or immune cells are missing to fully recapitulate the collagen turnover and remodeling during tumor fibrogenesis78–80. Therefore, co-culture models of CAFs with different cancer cells or/and immune cells are necessary, as these cell types synthesize important proteases, including matrix metalloproteinases (MMPs), that shape the stroma in the TME in cancer15,81,82. Another limitation is that we did not directly quantify endogenous TGF-β production, activation, or receptor-level dynamics. As a result, the proposed model in Fig. 8 reflects differences in TGF-β signaling activity but does not distinguish between variations in ligand availability, activation state, receptor sensitivity, or downstream signaling. Fig. 8Proposed summary figure. CAFs represent a heterogeneous population with variable fibrotic activity. Quiescent CAFs exhibit low fibrotic activity, whereas intrinsically active CAFs display enhanced collagen deposition associated with sustained TGF-β signaling activity. Exogenous TGF-β1 stimulation induces a strong increase in collagen deposition in quiescent CAFs, while intrinsically active CAFs show only a limited additional response, consistent with a pre-activated signaling state. Inhibition of TGF-β signaling via ALK5 reduces collagen production to baseline levels in quiescent CAFs and below baseline in intrinsically active CAFs, indicating dependence on ongoing TGF-β pathway activity in the latter. In contrast, TGF-β neutralization with Fresolimumab has a limited effect on fibrotic activity and restores collagen levels toward baseline, suggesting that ligand neutralization does not fully suppress TGF-β signaling. Created in https://BioRender.com. Proposed summary figure. CAFs represent a heterogeneous population with variable fibrotic activity. Quiescent CAFs exhibit low fibrotic activity, whereas intrinsically active CAFs display enhanced collagen deposition associated with sustained TGF-β signaling activity. Exogenous TGF-β1 stimulation induces a strong increase in collagen deposition in quiescent CAFs, while intrinsically active CAFs show only a limited additional response, consistent with a pre-activated signaling state. Inhibition of TGF-β signaling via ALK5 reduces collagen production to baseline levels in quiescent CAFs and below baseline in intrinsically active CAFs, indicating dependence on ongoing TGF-β pathway activity in the latter. In contrast, TGF-β neutralization with Fresolimumab has a limited effect on fibrotic activity and restores collagen levels toward baseline, suggesting that ligand neutralization does not fully suppress TGF-β signaling. Created in https://BioRender.com. Future studies should focus on validating our findings in vivo and investigating additional CAF populations to deepen the understanding of their functional diversity. A previous study has already highlighted the translational potential of PRO-C3 and PRO-C6 in a pancreatic CAF culture model38; similar approaches are needed to extend this research to other cancer types. Important to mention is also, that we only investigated the formation of three types of collagens on the protein level. Even though type I and III collagen are mainly associated with architectural changes in the TME83, a study that investigated the collagen expression patterns based on RNA-sequencing data in tumor samples across different types of cancer showed that many more genes of different types of collagens are upregulated in the TME. These include type V, type VIII, type X, type XI or type XII collagen, among others31. Also Cords et al.84 saw an increased expression of COL10A1, COL11A1, COL8A1 COL1A2, COL12A1, COL3A1, COL8A1 and COL5A2 by myCAFs in breast cancer. Therefore, investigating the formation of additional collagens, including those that are less abundant or considered minor, has the potential to provide deeper insights into the complexity of collagen production in CAFs85. Moreover, analyzing the corresponding mRNA expression levels will further enhance our understanding of the regulatory mechanisms underlying collagen synthesis and the contribution of specific collagen types to ECM remodeling and cancer progression. In conclusion, this study highlights the heterogeneity and complexity of CAFs, with a particular focus on their distinct roles in ECM remodeling. We demonstrate that different CAFs exhibit varying intrinsic and extrinsic fibrotic activity levels, which critically influences their responses to both profibrotic and antifibrotic stimuli, including TGF-β1 induction and inhibition. These findings emphasize the importance of considering both intrinsic and extrinsic fibrotic activity when studying CAFs and their role in tumor fibrosis. By leveraging collagen-based translational biomarkers, we can assess CAF fibrotic activity levels in the SiaJ model, enabling us to guide and enhance antifibrotic drug development and finding patients most likely to benefit from antifibrotic therapies. This approach not only enhances our understanding of CAF-targeted interventions but also underscores the relevance of translational biomarkers in optimizing anti-TGF-β1 strategies for cancer treatment. The insights gained here could pave the way for developing targeted therapies aimed at modulating the fibrotic response in cancer, ultimately improving patient outcomes. Future investigations should continue to explore the complex signaling networks involved in CAF activation and the therapeutic potential of targeting these pathways in diverse cancer settings.