Single-cell and spatial profiling map a conserved axis in gastroenteropancreatic neuroendocrine tumors
Gastroenteropancreatic neuroendocrine tumors sit in a biologic gray zone: clinically heterogeneous, often slow-growing, and still difficult to sort by behavior using routine histology alone. A multiomic analysis of 38 well-differentiated pancreatic neuroendocrine tumors and small-intestinal neuroendocrine tumors now shows that this heterogeneity is not random. Instead, the malignant compartment organizes along a conserved neuronal-to-secretory spectrum that is visible at the level of transcription, chromatin accessibility, and tissue architecture.
The study integrated single-nucleus RNA sequencing, single-nucleus ATAC sequencing, and spatial transcriptomics from archival flash-frozen tumors and matched formalin-fixed paraffin-embedded slides. That design matters because it lets gene expression be read alongside regulatory DNA accessibility and physical location within the tumor. The result is a more complete view of how these cancers behave than any single assay could provide.
Two malignant programs dominated the landscape: a neuronal-like program and a secretory neuroendocrine program. These were not isolated, binary states. They spanned a continuous transcriptional spectrum, implying that individual tumors can occupy intermediate positions rather than sit neatly in one category or the other. For a disease family known for protean clinical courses, that continuum offers a plausible biologic explanation for why two tumors with similar morphology can behave very differently at the bedside.
Distinct regulatory circuitry differed by organ site
Matched chromatin accessibility profiles showed that the regulatory logic was not identical in the two anatomic settings. In small-intestinal tumors, the accessible chromatin favored MAX:MYC and MITF transcription factor motifs. In pancreatic tumors, ISL1 and TFAP4 motifs were more prominent. Those differences point to organ-specific epigenetic control rather than a single uniform malignant program imposed across all gastroenteropancreatic neuroendocrine tumors.
The biological implication is straightforward: the same broad disease class is being steered by different upstream transcriptional networks depending on tissue of origin. In practice, this means a pancreatic lesion and a small-intestinal lesion may share neuroendocrine identity while still relying on different developmental circuitry and chromatin states to sustain malignancy. Such divergence is exactly where precision therapy could eventually separate from broad cytotoxic logic.
One sentence captures the core message plainly: these tumors are not only defined by where they arise, but by how their epigenetic switches are wired.
Spatial organization linked malignant states to stromal niches
Spatial transcriptomic mapping revealed a sharp anatomic split between the two malignant programs. Regions enriched for the neuronal-like state localized to densely cellular tumor zones with relative depletion of stromal infiltration. In contrast, regions enriched for the secretory neuroendocrine program occupied fibrovascular and stromal niches populated by endothelial cells, fibroblasts, and myeloid cells.
This is the kind of architecture that can shape therapy resistance, metastatic potential, and local invasion. A densely cellular compartment with less stromal interference may support one set of growth or survival behaviors, while a fibrovascular niche offers a different mixture of growth factors, extracellular matrix, and immune-interacting signals. The microenvironment is not merely surrounding the tumor here; it appears to be participating in the malignant state itself.
Three ligand-receptor programs stood out in the secretory-high regions: TGFB1-ITGB1, VEGFA-FLT1, and LAMA2-ITGA1. Each of these points toward a biologic setting rich in fibrosis, angiogenesis, and matrix signaling. In a tumor that often depends on both secretory phenotype and vascular support, this coupling between state and niche is highly plausible mechanistically.
Metastatic lesions carried the profibrotic and proangiogenic program
Across both pancreatic and small-intestinal tumors, one common malignant program, labeled cNMF2 in the study, was enriched in metastatic lesions. It also carried signatures of profibrotic and proangiogenic biology. That combination suggests a more invasive and microenvironmentally interactive state, one that may help tumors disseminate while adapting to supportive stromal and vascular signals at distant sites.
For clinicians who manage these tumors over years rather than weeks, the practical consequence is in how metastatic biology may differ from primary disease. If a specific malignant program preferentially tracks with metastasis, then the program itself becomes a candidate biomarker for risk stratification and, eventually, therapeutic targeting. The study does not claim clinical implementation, but it does provide a mechanistic map linking a transcriptional state to the metastatic compartment.
A framework for NET heterogeneity that connects biology to therapy targets
The central advance is conceptual as much as technical. Gastroenteropancreatic neuroendocrine tumors can now be viewed as organized along a conserved neuronal-to-secretory axis, with that axis shaped by distinct epigenetic programs and embedded in specific stromal niches. This framework unifies pancreatic and small-intestinal disease around shared malignant logic while preserving the organ-specific differences that matter for biology.
The therapeutic relevance lies in the pathway dependencies exposed by the analysis. Regions dominated by secretory neuroendocrine biology were linked to endothelial, fibroblast, and myeloid signaling, while the metastatic program carried profibrotic and proangiogenic features. Those findings point toward vulnerabilities in tumor-stroma communication, angiogenic signaling, and matrix-associated pathways rather than a single uniform target across all NETs.
A final point is hard to ignore: the spatial context of a tumor state can be as important as the state itself. In this cohort, malignant identity, regulatory circuitry, and microenvironmental position were tightly coupled, and that coupling may help explain why gastroenteropancreatic neuroendocrine tumors remain so difficult to classify by conventional methods alone.
The study’s limitation is also its boundary. The work defines architecture and candidate biology, but it does not establish which of these programs will predict treatment response in the clinic or which can be safely disrupted without collateral harm to normal neuroendocrine tissue. Those questions now belong to prospective validation in larger cohorts and, eventually, in treatment-linked studies.