The technological revolution behind this reframing is worth appreciating. For most of the history of cancer biology, fibroblasts were studied in bulk, and their heterogeneity was invisible. The advent of single-cell RNA sequencing allowed researchers to profile the gene expression of individual fibroblasts within a tumor, revealing a startling mosaic of states. Spatial transcriptomics then added geography to the picture, showing where each fibroblast subtype resides relative to cancer cells, blood vessels, and immune cells. Multi-omics approaches that layer epigenetic, proteomic, and metabolic data on top of these maps have now made it possible to trace how fibroblast identities shift as tumors evolve. What has emerged is a taxonomy of recurring CAF programs: myofibroblastic CAFs marked by alpha-smooth muscle actin and contractile machinery, inflammatory CAFs that secrete cytokines such as interleukin-6 and CXCL12, antigen-presenting CAFs that display MHC class II molecules, interferon-response CAFs, and vascular-associated CAFs that hug the tumor’s blood supply.

Each of these states maps onto specific steps of the metastatic cascade, the multi-stage journey that carries a tumor cell from its primary site to a distant organ. The first step, local invasion, depends heavily on extracellular matrix remodeling. Myofibroblastic CAFs are master builders and demolition crews of the matrix, depositing collagen and cross-linking it with enzymes such as lysyl oxidase. This stiffened, realigned matrix does not merely form a physical scaffold; it generates mechanical cues that activate signaling pathways like FAK, YAP/TAZ, and ROCK in neighboring cancer cells, promoting epithelial-mesenchymal transition, the program by which epithelial tumor cells acquire motile, invasive properties. In pancreatic ductal adenocarcinoma, one of the most fibroblast-dense malignancies known, this dense stroma creates both a barrier to immune infiltration and a highway for invasion, illustrating how a single fibroblast program can simultaneously enable multiple metastatic steps.

Once tumor cells begin to move, they must evade the immune system, and here the review highlights a particularly insidious role for inflammatory CAFs. Through secretion of chemokines such as CXCL12 and CCL5, these fibroblasts recruit immunosuppressive cells, including regulatory T cells and myeloid-derived suppressor cells, while excluding or exhausting cytotoxic T lymphocytes that would otherwise kill tumor cells. The JAK/STAT and NF-kappa-B signaling pathways are central to this inflammatory choreography. Antigen-presenting CAFs add another layer of complexity: by expressing MHC class II molecules and, in some contexts, immune checkpoints such as PD-L2, they can directly modulate T cell activity in ways that remain incompletely understood but appear to tilt the local immune balance toward tolerance rather than attack. The net effect is a tumor microenvironment in which the stroma itself acts as an immunological shield.

Angiogenesis and vascular dissemination represent the next stage of the journey, and vascular-associated CAFs sit at this frontier. Positioned adjacent to tumor blood vessels, these cells interact with endothelial cells through vascular endothelial growth factor signaling and other pathways, supporting the formation of the leaky, disorganized vasculature that allows circulating tumor cells to enter the bloodstream. The review also describes how fibroblasts contribute to pre-metastatic niche formation, a phenomenon in which the primary tumor sends molecular advance parties, often packaged in extracellular vesicles, to distant organs such as the lung, liver, and bone. These signals condition resident fibroblasts in future metastatic sites, prompting them to remodel the matrix, secrete inflammatory mediators, and lay down a welcoming bed before any cancer cell arrives. Neutrophil extracellular traps and other stromal components recruited in this process further prepare distant tissues for colonization.

Metastatic colonization, the final and often fatal step, is where fibroblast support may matter most. Circulating tumor cells that survive in the bloodstream face a hostile foreign microenvironment, and the success or failure of their settlement depends on the stromal soil they encounter. The review emphasizes that the relationship between CAF states and metastatic progression is bidirectional: not only do distinct fibroblast states regulate specific steps of the cascade, but tumor-derived cues, inflammation, vascular signals, mechanical stress, metabolic conditions, and even therapy itself can reshape the composition and behavior of the fibroblast population. Chemotherapy, for example, can induce fibroblast state transitions that render the microenvironment more permissive to regrowth, a form of therapy-induced plasticity that may explain some treatment failures.

Perhaps the most consequential message of the review concerns the context-dependent duality of CAF function. Fibroblasts are not uniformly villainous. Certain fibroblast states and signals have been shown to restrain tumor growth, maintain tissue architecture, and support anti-tumor immunity. The authors point to clinical lessons from attempts at indiscriminate stromal depletion: in some settings, eliminating fibroblasts or blocking their hallmark signaling pathways worsened outcomes, unleashing more aggressive tumor behavior. The Sonic hedgehog pathway in pancreatic cancer became a cautionary tale in this regard, as its inhibition depleted stroma yet failed to improve, and in some models worsened, disease. Vitamin D receptor signaling, by contrast, has been associated with a tumor-restraining fibroblast phenotype, suggesting that reprogramming fibroblasts toward benign states may be more productive than killing them.

This insight underpins the precision therapeutic strategies the review advocates. Rather than viewing the stroma as a target for demolition, the authors propose selective modulation: identifying which fibroblast states dominate a given tumor, at a given metastatic stage, and intervening to disable their tumor-promoting functions while preserving or enhancing their tumor-restraining ones. Candidate approaches include inhibiting fibroblast activation protein on specific CAF subsets, blocking TGF-beta signaling in myofibroblastic CAFs, reprogramming inflammatory CAFs through agents such as all-trans retinoic acid, and exploiting metabolic vulnerabilities unique to particular fibroblast states. Emerging technologies such as CAR-T cells engineered against CAF-specific antigens and photodynamic therapy aimed at stromal compartments are also evaluated, with the authors stressing that each strategy must account for the risk of pushing plastic fibroblasts into more dangerous states.

The framework proposed by Li and colleagues integrates four dimensions that have often been studied in isolation: cellular state, spatial niche, metastatic stage, and phenotypic plasticity. In this view, a myofibroblastic CAF at the invasive front of a primary tumor, an inflammatory CAF clustered near excluded T cells, and a vascular-associated CAF lining a metastatic niche in the liver are not variations on a theme but distinct therapeutic targets requiring distinct interventions. The authors acknowledge that major questions remain unresolved, including how the reciprocal feedback loops between tumor cells and fibroblasts determine therapeutic response, and how best to monitor CAF state composition in patients over time. Single-cell and spatial technologies are increasingly being applied to clinical samples, raising the prospect that fibroblast-state profiling could one day guide stromal-targeted treatment decisions just as molecular profiling of tumor cells guides targeted therapy today.

For a field that has spent decades focused almost exclusively on the cancer cell itself, the message of this synthesis is quietly radical: the most abundant cell in many tumors is not a passive bystander but an active, plastic, and heterogeneous participant in metastasis, and the tools to map and manipulate it are finally at hand. If the metastasis-oriented framework proposed here proves correct, the future of anti-metastatic therapy may depend less on attacking tumor cells directly and more on negotiating with the fibroblasts that build their roads, guard their borders, and prepare their landing sites. Turning these cellular accomplices from collaborators into inhibitors, the review suggests, could transform how medicine confronts the spread of cancer, the process responsible for the majority of cancer deaths worldwide.

Subject of Research: Heterogeneity of cancer-associated fibroblasts and their roles in tumor metastasis and targeted therapy

Article Title: Heterogeneity of cancer-associated fibroblasts in tumor metastasis: mechanisms and precision therapeutic strategies

Article References: Li, Y., Chen, Y., Shen, X., Lou, J., Zhang, L., Qin, J., Pan, Y., & Wang, S. (2026). Heterogeneity of cancer-associated fibroblasts in tumor metastasis: mechanisms and precision therapeutic strategies. Molecular Cancer. https://doi.org/10.1186/s12943-026-02805-4

Image Credits: AI Generated

DOI: 10.1186/s12943-026-02805-4

Keywords: cancer-associated fibroblasts, tumor metastasis, tumor microenvironment, extracellular matrix remodeling, single-cell RNA sequencing, spatial transcriptomics, pre-metastatic niche, epithelial-mesenchymal transition, angiogenesis, immune evasion, CAF plasticity, stromal-targeted therapy

Extracellular vesicles, or EVs, have become one of the most actively pursued frontiers in liquid biopsy. Nearly every cell type sheds them, and they ferry proteins, lipids and nucleic acids between cells, effectively rewiring distant tissues. Tumor-derived EVs are particularly consequential: previous work has shown they can sculpt the so-called pre-metastatic niche, preparing far-flung organs to receive arriving cancer cells. In an earlier study, the same group demonstrated that EV-associated TRAIL, or tumor necrosis factor-related apoptosis-inducing ligand, promotes pre-metastatic niche formation and that measuring it by enzyme-linked immunosorbent assay, or ELISA, could predict postoperative lung metastasis in hepatocellular carcinoma. But pancreatic cancer posed a different and harder analytical problem, one that forced the researchers to rethink how such measurements should be made.

The difficulty lies in abundance. The PDAC tumor microenvironment is dominated by fibroblasts and dense extracellular matrix, with actual tumor cells making up only a small fraction of the tissue mass. That architecture translates into a bloodstream where tumor-derived vesicles are a minority population drowned in a sea of vesicles from other sources, along with lipoproteins and protein complexes that overlap in size and biophysical behavior. When the team applied their established ELISA workflow to PDAC plasma, the results were sobering: EV-associated TRAIL readings in PDAC patients showed no significant elevation over healthy controls and were markedly lower than in hepatocellular carcinoma. Most optical density values clustered near the lower end of the standard curve, close to the assay’s limit of quantification, where measurements become compressed and unreliable. Nanoparticle tracking analysis confirmed the underlying biology, revealing only a modest increase in total plasma EV concentration in PDAC compared with healthy donors.

The solution the researchers pursued was to move from bulk measurement to single-particle enumeration using nanoscale flow cytometry on the widely available CytoFLEX platform. Detecting objects smaller than 200 nanometers by flow cytometry is notoriously tricky, and the team’s first task was choosing the right trigger channel, the signal the instrument uses to decide that a particle is present. Testing fluorescent beads of 100, 200 and 300 nanometers, they found that violet side scatter, excited by the 405-nanometer laser, detected the smallest particles most sensitively, while conventional 488-nanometer side scatter offered the best separation between bead populations and instrument noise. Their workflow therefore uses a dual-channel configuration: VSSC for triggering events and 488-SSC for analysis, maximizing sensitivity without sacrificing resolution.

Equally critical was taming the swarm effect. When particle concentrations are high, multiple vesicles pass through the laser interrogation point simultaneously and are recorded as a single event, artificially deflating counts and inflating fluorescence through signal summation. By performing serial dilutions of 100-nanometer fluorescent beads and monitoring event rate, side scatter and fluorescence in parallel, the researchers defined a linear acquisition window of roughly 10^7 particles per milliliter, corresponding to about 3,000 to 6,000 events per second at a flow rate of 60 microliters per minute. Outside this window, event rates deviated from proportionality and scatter signals crept upward, the signature of coincidence. All subsequent EV measurements were confined to this calibrated range, and the team showed that acquiring overly concentrated samples produced false increases in apparent CD63-positive and TRAIL-positive events, a cautionary demonstration for anyone adapting similar protocols.

Specificity demanded equally rigorous controls. All buffers were filtered through 0.02-micrometer filters to minimize background particles, antibody aggregates were spun down before staining, and a post-staining wash step removed free fluorophore-conjugated antibodies that would otherwise shift the reference-noise fluorescence distribution rightward and mask true positive events. Positive gates were set using matched isotype controls, and detergent lysis experiments confirmed that detected signals came from membrane-bound vesicles rather than free protein complexes. Using vesicles from TRAIL-overexpressing HEK293T cells as a positive model, the workflow resolved distinct CD63 single-positive, TRAIL single-positive and CD63/TRAIL double-positive subsets, with good inter-assay reproducibility of roughly 11 to 13 percent coefficient of variation and the ability to detect TRAIL-positive vesicles at approximately 1 percent abundance within the total EV population.

To validate the approach in the messy reality of blood plasma, the researchers spiked known quantities of fluorescently labeled HEK293T-derived vesicles into unstained plasma, co-isolated them with endogenous particles by ultracentrifugation, and measured recovery by nano-flow cytometry. Although absolute recovery was modestly below input, DiR-positive counts scaled linearly with the spiked input across the dilution series, demonstrating that the workflow faithfully captures relative changes in specific EV subsets even within the lipoprotein-rich plasma matrix. Characterization of the isolated PDAC plasma vesicles by transmission electron microscopy revealed the classic cup-shaped morphology, nanoparticle tracking analysis showed a mean diameter of 157 nanometers at concentrations near 10^11 particles per milliliter of plasma, and immunoblotting confirmed the presence of the canonical EV marker TSG101.

The clinical payoff came when the workflow was applied to patient cohorts. In a first cohort of 80 PDAC patients, 47 without liver metastasis and 33 with, plasma EV-associated TRAIL was significantly elevated in the metastatic group, and a logistic regression model built on the percentage of TRAIL-positive vesicles among total plasma EVs discriminated metastatic from non-metastatic disease with an area under the ROC curve of 0.766. More strikingly, in an independent validation cohort of 85 patients who underwent surgical resection with no radiologically detectable metastasis at the time of surgery, preoperative EV-associated TRAIL levels predicted liver metastatic recurrence within the first two postoperative years, with AUC values of 0.718 at one year and 0.681 at two years. Because blood was drawn on the day of surgery, the measurement effectively functioned as an early warning system, flagging occult micrometastatic disease that imaging could not yet see.

Biologically, the findings also carry mechanistic interest. PDAC plasma vesicles showed a marked increase in the TRAIL single-positive fraction, roughly 3.1 percent compared with about 1.3 percent in HEK293T-derived vesicles, consistent with recent evidence that TRAIL incorporation into EVs proceeds predominantly through ESCRT-dependent biogenesis routes rather than the tetraspanin-enriched, CD63-associated compartments conventionally associated with exosomes. This reinforces a growing appreciation that EVs arise from multiple, partially distinct biogenetic pathways, and that epitope-defined subpopulations may carry information invisible to bulk assays. It also explains why single-particle analysis outperformed ELISA here: in tumor types with low circulating tumor-derived EV content, averaging across the whole vesicle pool can bury clinically meaningful signals beneath the quantification floor.

The authors are candid about limitations. Ultracentrifugation, the workhorse of EV enrichment, can co-isolate lipoproteins and protein aggregates of similar size and density, potentially inflating background, and future implementations may benefit from orthogonal purification such as size-exclusion chromatography or immunoaffinity capture. Fluorescence compensation in the low-signal regime of nanoscale cytometry also demands careful single-stained controls. Nevertheless, by implementing the workflow on a routine clinical cytometer with explicitly reported settings and quality-control criteria, the study lowers the barrier for other laboratories to reproduce and extend the approach. If validated in larger, prospective cohorts, a simple blood draw measuring TRAIL-positive vesicles could give pancreatic cancer surgeons and oncologists something they have long lacked: a molecular head start on the metastasis that most often decides this disease’s course.

Subject of Research: Detection of plasma extracellular vesicle-associated TRAIL by nanoscale flow cytometry for predicting liver metastasis in pancreatic ductal adenocarcinoma

Article Title: Detection of plasma EV-associated TRAIL by nanoscale flow cytometry for liver metastasis prediction in PDAC

Article References: Huang, C.-X., Jian, J.-H., Hao, J.-S., Zhou, Z.-W., Li, Z.-Q., Kuang, D.-M., & Wu, C.-Y. (2026). Detection of plasma EV-associated TRAIL by nanoscale flow cytometry for liver metastasis prediction in PDAC. Advanced Biotechnology, 4(1), Article 6. https://doi.org/10.1007/s44307-026-00102-1

Image Credits: AI Generated

DOI: 10.1007/s44307-026-00102-1

Keywords: extracellular vesicles, nanoscale flow cytometry, TRAIL, pancreatic ductal adenocarcinoma, liver metastasis, liquid biopsy, biomarker, ELISA, pre-metastatic niche, plasma, CytoFLEX, swarm effect