The starting point was a clinical puzzle. CHI3L1 is a secreted glycoprotein produced by many cell types, including macrophages, T cells, fibroblasts and tumor cells, and elevated levels in blood and tumors have long been linked to metastasis and poor outcomes across cancers. Yet in pancreatic tissue, most CHI3L1-expressing cells are barely detectable in normal epithelium, and the protein appears only as disease advances. In the team’s own patient cohort, immunohistochemistry showed minimal CHI3L1 in normal pancreatic epithelium but strong expression in tumor epithelial cells across stages IB through IIB, and higher CHI3L1 levels in pancreatic juice correlated with reduced disease-free survival. The question was how cancer cells acquire this expression in the first place.

Spatial transcriptomics provided the first clue. Using the GeoMx Digital Spatial Profiling platform on formalin-fixed tumor sections from three surgically resected patients, the researchers found that CHI3L1 transcripts in cancer cells were enriched in regions adjacent to fibroblasts and immune cells, marked by the proteins ACTA2 and CD45, compared with regions containing cancer cells alone. In the KPPC spontaneous mouse model of pancreatic cancer, which carries the KrasG12D mutation and loss of the tumor suppressor Trp53, intratumoral CHI3L1 rose progressively as tumors matured. Notably, cytokines previously reported to induce CHI3L1 in other contexts, including IL-4, IL-6, IL-13 and interferon-gamma, failed to induce the protein in pancreatic cancer cell lines, pointing to a regulatory mechanism unique to the multicellular microenvironment.

To dissect that mechanism, the team built laboratory stand-ins for the tumor microenvironment. In three-dimensional spheroids combining BxPC-3 cancer cells, pancreatic stellate cells and primary human macrophages in equal proportions, CHI3L1 expression in cancer cells rose markedly compared with cancer cells grown alone. In two-dimensional conditioned-media experiments, pancreatic cancer lines MiaPaCa-2 and PANC-1 were exposed to media collected from monocultures, double cultures or triple cultures of cancer cells, stellate cells and macrophages. Only the triple-culture media, designated CSM, drove robust, time-dependent increases in CHI3L1 mRNA and protein. Adding T cells to make a four-cell system changed nothing, indicating that the cancer-stroma-macrophage conversation is the essential driver.

Proteomic secretome profiling of the conditioned media identified the messengers. Heat-denaturing the media abolished CHI3L1 induction, confirming that heat-labile proteins were responsible, and mass spectrometry of concentrated supernatants pinpointed macrophage-derived chemokines CCL2 and CXCL10, together with the stellate-cell-derived matrix protein SPARC, as leading candidates. Strikingly, none of the three recombinant proteins could induce CHI3L1 on its own; only combined treatment worked, and the same cooperative response appeared in lung, esophageal and colorectal cancer cell lines as well. Blocking the corresponding receptors on cancer cells, CCR2 for CCL2, CXCR3 for CXCL10 and TGFβR1 for SPARC, suppressed the induction, but inhibiting only two of the three receptors was insufficient. The signal, in other words, is genuinely combinatorial.

The downstream wiring proved to be a two-layer mechanism. The three factors activated the transcription factor NF-κB, and chromatin immunoprecipitation showed that phosphorylated NF-κB bound directly to the CHI3L1 promoter after triple-culture stimulation. Luciferase reporter assays with promoter constructs and site-directed mutagenesis established that three of four predicted NF-κB binding sites were essential for this response. Meanwhile, the conditioned media also activated STAT3, which did not bind the promoter directly but instead accelerated the degradation of DNA methyltransferase 1, or DNMT1, an enzyme that maintains methylation marks. Cycloheximide chase assays showed faster DNMT1 decay in STAT3-overexpressing cells, and bisulfite pyrosequencing confirmed loss of CpG methylation across the CHI3L1 promoter after triple-culture education. With DNMT1 evicted from the promoter region encompassing the NF-κB sites, and NF-κB then binding, CHI3L1 is activated both transcriptionally and epigenetically. Re-expressing DNMT1 suppressed the induction, sealing the argument.

Single-cell RNA sequencing of six pancreatic datasets spanning adjacent normal tissue, primary tumors and metastases added a patient-level dimension. Macrophages emerged as the predominant source of CCL2 and the exclusive producers of CXCL10, while a fibroblast subpopulation expressed SPARC. Among cancer cells, a CHI3L1-high subcluster, termed CHI_T, was present in primary tumors and reached its highest proportion in metastatic samples, and it preferentially expressed the receptors CCR2, CXCR3 and TGFBR1. Predicted cell-cell communication scores showed increasing signaling from macrophages and SPARC-positive fibroblasts toward this subcluster as disease progressed from normal tissue to primary tumor to metastasis. The authors caution that their pseudotime ordering reflects transcriptional similarity across spatial niches rather than a confirmed temporal sequence, since the cross-sectional data cannot establish lineage or chronology.

The cellular choreography behind the three factors was itself revealing. Transwell assays confirmed that CCL2, whether recombinant or produced by cancer cells, recruits CCR2-expressing macrophages, an effect abolished by CCR2 knockdown. Recombinant CCL2, but not CXCL10, upregulated SPARC production in stellate cells, showing that macrophage-derived CCL2 primes the stroma. CXCL10, despite appearing macrophage-specific in single-cell data, was undetectable in any monoculture and emerged only under triple-culture conditions, meaning it requires the full multicellular interaction. When the team knocked down CCL2 and CXCL10 in macrophages and SPARC in stellate cells simultaneously, the resulting conditioned media lost most of their ability to induce CHI3L1 in cancer cells, whereas single knockdowns had little effect. In the KPPC mouse model, CCL2 rose earliest, CXCL10 and SPARC peaked around week seven, and CHI3L1 climbed through weeks seven to eight.

Clinically, multiplex immunofluorescence on 57 patient tumors allowed stratification into a group with high expression of all four markers, CHI3L1, CCL2, CXCL10 and SPARC, comprising 23 patients, versus the remaining 34. Kaplan-Meier analysis showed a trend toward shorter overall and disease-free survival in the four-high group, with univariate hazard ratios of 1.672 and 1.684 respectively. After adjusting for age, T stage and N stage, the four-high signature was independently associated with shorter overall survival, with an adjusted hazard ratio of 1.852 and a P value of 0.040, while the disease-free survival association reached borderline significance at a hazard ratio of 1.778 and P equal to 0.050. The authors note that the modest sample size and borderline statistics warrant validation in larger cohorts.

The therapeutic test came in immunocompetent mice bearing orthotopic KPPC-luc tumors. Seven groups received vehicle, single inhibitors of CCR2, CXCR3 or TGFβR1, a triple-inhibitor regimen, or the triple inhibitors plus a CHI3L1-neutralizing antibody. Tumor incidence at sacrifice fell from six of seven control mice to one of seven in both the triple-inhibitor and combination groups, though bioluminescent quantification reached statistical significance only for the combination group versus control. No treatment-related toxicity appeared on serum biochemistry or organ histology. Flow cytometry of draining lymph nodes showed that both multi-pathway regimens significantly increased perforin-positive CD8 T cells, a marker of cytotoxic potential. Because so few tumors remained evaluable in the treated groups, tumor-level immune findings, including lower frequencies of PD-1-positive and LAG3-positive exhausted CD8 T cells in the combination-treated tumor, are presented descriptively rather than statistically. The authors conclude that combining upstream stromal blockade with CHI3L1 neutralization is a promising combinatorial strategy for pancreatic cancer immunotherapy that now requires adequately powered confirmation, and that the framework of stromal signals driving epigenetic reprogramming in cancer cells may extend to other proteins that ramp up during tumor progression.

Subject of Research: Stromal CCL2, CXCL10 and SPARC signaling driving CHI3L1 expression and therapeutic targeting in pancreatic ductal adenocarcinoma

Article Title: Stromal CCL2/CXCL10/SPARC axes drive pro-tumor CHI3L1 expression for microenvironmental modulation in pancreatic cancer progression and therapeutic innovation

Article References: Su, P.-C., Shan, Y.-S., Hou, Y.-C., Huang, P.-H., Liao, P.-C., & Wang, Y.-C. (2026). Stromal CCL2/CXCL10/SPARC axes drive pro-tumor CHI3L1 expression for microenvironmental modulation in pancreatic cancer progression and therapeutic innovation. Journal of Biomedical Science, 33(1), Article 93. https://doi.org/10.1186/s12929-026-01295-2

Image Credits: AI Generated

DOI: 10.1186/s12929-026-01295-2

Keywords: pancreatic cancer, CHI3L1, tumor microenvironment, CCL2, CXCL10, SPARC, NF-κB, STAT3, DNMT1, macrophages, pancreatic stellate cells, immunotherapy