Two conserved malignant cell states may reorganize how GEP-NETs are classified
Gastroenteropancreatic neuroendocrine tumors have long been grouped by organ of origin, morphology, and proliferation index, yet those labels often flatten biologic differences that matter at the bedside. In the current JCI issue, Karam, Hoffman, and colleagues are placed in a different framework altogether: one centered on malignant cell state rather than simply where the tumor began or how fast it divides. The work examined 38 pancreatic and small-intestinal NETs with joint single-nucleus RNA and ATAC sequencing, then projected the malignant programs onto matched spatial transcriptomes. Two conserved programs emerged repeatedly — neuronal-like and secretory neuroendocrine — and those programs were not confined to one organ site.
The finding is consequential because it reframes heterogeneity as a continuum rather than a set of rigid bins. Tumors from the pancreas and small intestine both occupied this axis, which suggests that tissue of origin alone does not capture the biology driving clinical behavior. The two programs were also associated with different transcription factor motifs, pointing to distinct regulatory circuitry rather than a superficial change in gene expression. In practical terms, the study moves NET classification closer to the molecular logic that already guides decision-making in other solid tumors.
Secretory program scores trended higher in metastatic disease.
That single observation gives the analysis immediate clinical gravity. If the secretory state continues to track with dissemination in larger cohorts, it could become a molecular signal for risk stratification beyond the usual anatomic and histologic labels. The work does not establish a validated prognostic test, but it does identify a plausible axis by which aggressive behavior may be recognized earlier, before standard indices fully separate indolent from dangerous disease.
Spatial niche differences suggest biology is shaped by the tumor microenvironment
The study does more than sort tumors into two molecular buckets. By integrating spatial transcriptomics, the investigators showed that the cell states occupy distinct microenvironmental niches. That matters because malignant programs are rarely autonomous; they are shaped by oxygen gradients, stromal contacts, immune surveillance, and local signaling cues that can reinforce or suppress differentiation. A tumor cell adopting a secretory phenotype in one region and a neuronal-like phenotype in another may not merely reflect clonal variation. It may also reflect how the surrounding tissue architecture channels lineage programs.
This spatial dimension helps explain why NETs can appear phenotypically similar under the microscope yet behave differently in patients. Well-differentiated tumors often retain organ-specific features, but morphology alone can miss the transcriptional wiring that determines secretion, proliferation, and dissemination. A state-based model gives pathologists and translational teams a way to think about these tumors as dynamic systems rather than static lesions. It also raises a mechanistic possibility: if the microenvironment helps maintain one state over another, then the niche itself could become a therapeutic target.
Transcription factor motifs point to distinct regulatory programs
The two conserved malignant states were enriched for largely different transcription factor motifs. In biologic terms, that indicates the states are not interchangeable expressions of the same disease process; they likely arise from different regulatory inputs that lock cells into different phenotypes. For oncology practice, this matters because transcriptional states often map more directly to drug sensitivity than broad organ labels do. A tumor operating through one program may respond differently to pathway inhibition, epigenetic modulation, or microenvironmental targeting than a tumor in the alternate state.
The study does not claim therapeutic targets or treatment responses. It does, however, provide the kind of molecular scaffolding needed to identify them. The combination of single-nucleus RNA and ATAC sequencing is especially powerful here because it captures both expression and accessible chromatin, allowing investigators to infer which regulatory regions may be driving the programs. That level of resolution is exactly what has been missing from traditional NET classification schemes, which are useful for staging but limited in predicting biologic diversity among closely matched tumors.
What this means for prognosis and treatment selection
For patients with GEP-NETs, the immediate promise is not a new approved biomarker but a more informative way to stratify disease. If neuronal-like and secretory states can be validated prospectively, clinicians may eventually use them alongside anatomic site, morphology, and proliferation markers to identify patients at higher risk for metastatic progression. The current analysis also suggests that seemingly similar tumors may not deserve the same treatment assumptions. A state-driven framework could help explain why some NETs remain slow-growing while others advance despite comparable histology.
The study also points toward a more nuanced research agenda. Future work will need to determine whether these cell states are stable over time, whether they shift after therapy, and whether metastatic lesions preserve the same program seen in the primary tumor. Those questions are central to clinical utility. A classification scheme only becomes useful when it can survive the realities of tumor evolution, sampling bias, and treatment pressure. Without that, even a biologically elegant model stays in the lab.
For now, the most important contribution is conceptual. NET heterogeneity is no longer confined to a discussion of size, site, and proliferation. It can be read as a cell-state axis linked to tissue architecture, regulatory motifs, and possible metastatic propensity. That is a sharper lens for a disease group that has outgrown its older categories, and it may eventually help explain why two tumors that look nearly identical on routine pathology can behave very differently at the bedside.