Oncogenic and tumor-suppressive forces converge on a progenitor-orchestrated niche to shape early tumorigenesis

Abstract

The transition from benign to malignant growth is a pivotal yet poorly understood step in cancer progression that marks the shift from a pathologically inert condition to a clinically lethal disease. Here, we integrate lineage tracing, single-cell and spatial transcriptomics to visualize the molecular, cellular and tissue-level events that promote or restrain malignancy during the tumor initiation in mouse models of pancreatic ductal adenocarcinoma (PDAC). We identify a discrete progenitor-like population of KRAS-mutant cells that co-activates oncogenic and tumor-suppressive programs-including p53, CDKN2A, and SMAD4-engaging senescence-like responses and remodeling their microenvironment, ultimately assembling a niche that mirrors invasive PDAC. KRAS inhibition depletes progenitor-like cells and dismantles their niche. Conversely, p53 suppression enables progenitor cell expansion, epithelial-mesenchymal reprogramming, and immune-privileged niche formation. These findings position the progenitor-like state as the convergence point of cancer-driving mutations, plasticity, and tissue remodeling-revealing a critical window for intercepting malignancy at its origin.

Figure 1.. Capturing spontaneous loss of p53 throughout the premalignant-to-malignant spectrum.

a. KP^LOH mouse model. Loss of GFP linked to the only wild-type p53 copy in the cell reports p53 loss of heterozygosity. b. Sampling strategy to characterize progression from premalignant to malignant states. c. Representative fluorescence image of a pre-tumor stage pancreas section. Arrowheads point to rare cells that lost GFP fluorescence upon p53 LOH. d. Force-directed layout (FDL) of single-cell transcriptional data from sorted Kras^G12D epithelial cells, colored by mouse stage and p53 status. e. Projection of transcriptional signatures of major subpopulations identified in premalignant pancreas. Multiple transcriptional signatures were used to annotate cell type (Methods). ADM, acinar-to-ductal metaplasia. f. Diffusion distance from pre-tumor p53-proficient or p53-deficient cells to the closest cancer-like cell. g. Single-cell karyotypes of pre-tumor p53-deficient cells inferred from scRNA-seq (Methods). Rows represent individual cells and columns represent genes, ordered by genomic position. Colors represent inferred loss or gain of genomic material. Chr, chromosome. h. FDL of pre-tumor p53-deficient cells, colored by genomic state.

Figure 2.. p53 is maximally active in rare premalignant cells during tumor initiation.

a. Expression of known p53 targets and markers of progenitor-like cells in pre-tumor p53-proficient and deficient cells, as a function of cell state. b. Representative smFISH image of pre-tumor stage pancreas, probing for p53 targets and the progenitor-like state marker Msn. c. FDL of Kras^G12D-positive epithelial cells along PDAC progression, colored by p53 average expression of p53 canonical targets shown in (a). d. Expression of tumor-suppressive and oncogenic gene signatures in pre-tumor p53-proficient cells or tumor p53-deficient cells (PDAC). p53 canonical signature derived from p53 targets in Fig. 2a. Other signatures shown are: p53 curated targets (Fisher); p53 TSAG, tumor suppression–associated genes; p53-restoration; Cdkn2a mRNA; TGFβ-dependent SMAD4 targets; HALLMARK EMT, epithelial-to-mesenchymal transition; senescence UP; UP in mouse PDAC (this work, see Methods); Kras/Fosl1; Kras injury; glycolysis/warburg (curated list, see Methods). e. UMAP of cells from healthy human pancreas and PDAC tissue in, colored by progenitor-like signature. PDAC cells and acinar cells are grayed out to facilitate visualization of duct-like and ADM cells from donors with and without cancer. Box highlights cells that exhibit highest progenitor-like signatures. f. Expression of tumor-suppressive and oncogenic signatures in pancreatic epithelial cells with and without donors. Different rows of ADM/duct-like cells correspond to PhenoGraph clusters. Colors represent z-score of average signature scores in each column. Box highlights the two PhenoGraph clusters with highest progenitor-like signatures in cells from donors without cancer.

Figure 3.. Transcriptional and morphological states undergo coordinated changes in the premalignant epithelium.

a. Experimental timeline for tissue collection after inducing caerulein-induced pancreatitis in KC mice. b. Fraction of progenitor-like epithelial cells in scRNA-seq data as a function of treatment condition and time. c. Representative images of smFISH staining for the progenitor-like marker Msn. The three fields of view are from the same tissue. Scale bars, 50 μm. d,e. Spatial representation (d) or single-cell transcriptional embedding (e) of Xenium data annotated by signatures of major premalignant subpopulations. f. Projection of gastric–progenitor diffusion component in transcriptional embedding of single-cells derived from Xenium data. g. Representative fields of view of premalignant epithelial lesions in Xenium data. Segmented nuclei are pseudo-colored by their gastric–progenitor DC value, using the colormap in (f). h–j. Distributions of lumen score (h), epithelial fraction in local spatial neighborhood (i) and lesion size (j) as a function of gastric–progenitor DC in epithelial cells (see Methods for details on the definition and quantification of morphological parameters). k. Schematic of the changes in lesion morphologies along the gastric-progenitor DC continuum.

Figure 4.. Continuous cellular and molecular remodeling events during the assembly of the progenitor niche.

a. Representative section of premalignant pancreas harvested 2 days post-injury, analyzed using the Xenium platform and colored by cell type. b. Projection of gastric–progenitor DC in premalignant cells. Epithelial cells not categorized as gastric-like or progenitor-like are outlined in dark blue, but not pseudo-colored. c. Niches, comprising all cells within a 60-μm radius of a central anchor epithelial cell, are ordered by the average gastric–progenitor DC of their constituent epithelial cells. d. Location of individual mRNA molecules associated with select myofibroblasts and monocyte/macrophage subpopulations in niches depicted in (c). e. Contour plots denote the density of niche epithelial cells in select bins along the average gastric-progenitor DC (top) and corresponding shifts in the density of microenvironment cells (bottom). f. Average niche expression of select genes in (left) epithelial, (middle) fibroblast or (right) myeloid immune cells along the average gastric-progenitor DC. Niches are ordered by average gastric–progenitor DC value, divided into 100 equal bins. Dotted lines indicate DC value at which epithelial cells begin expressing progenitor-like markers. Communication genes associated with progenitor niches are highlighted in red. g. Average expression of wound-healing response genes (GO Biological Processes) from our Xenium panel. Each dot represents a single biological replicate (n=15 mice). Values denote the average z-scored expression of wound-healing genes in the specified cellular compartments, in either gastric or progenitor niches. P-values, two-tailed Wilcoxon Rank Sums Test. h. Immunofluorescence staining for cellular states enriched in the progenitor niche, as a function of PDAC progression. Scale bars, 100 μm.

Figure 5.. Intercellular communication modules define the progenitor niche.

a. (Top) Gene-gene correlation matrix of communication gene niche expression in distinct cellular compartments. (Middle) Average expression of communication genes in canonical gastric of progenitor niches. (Bottom) Communication modules identified through graph-based community detection. Boxes highlight communication modules associated with the progenitor niche in each compartment (see Methods). b-d. (Top) Each dot is a single mRNA detected in a specific cellular compartment. Scatter plots show colocalization of cognate ligand-receptor pairs from (b) IL-18, (c) GM-CSF signaling, or (d) Tgfb1 produced in different cellular compartments (see Fig. S8c for markers of progenitor-like cells and associated niche cells in the same tissue region). (Bottom) Niche expression of cognate ligands and receptor pairs from (b) IL-18, (c) GM-CSF, or (d) TGFβ signaling. Each dot denotes the average niche gene expression of the specified communication genes in a specific bin along the gastric-progenitor niche continuum. e. Schematic of multicellular interaction circuits enabled by engagement of communication modules in the progenitor niche.

Figure 6.. Consequences of acute oncogenic Kras inhibition in the premalignant pancreas.

a. Timeline of acute oncogenic Kras inhibition in the premalignant pancreas. b. Representative images and quantification of HMGA2 staining in KP^LOH mice treated with vehicle (n = 3) or MRTX1133 (n = 4). Tissue was collected 2 days after the first inhibitor or vehicle dose. Scale bars, 50 μm. Each dot in quantification corresponds to an individual mouse; bar corresponds to average value. c. Two-dimensional density representation of Xenium single-cell gene expression data from mice treated with vehicle (n = 2) or MRTX1133 (n = 4). Purple contours, density of transcriptional states in the vicinity of progenitor-like epithelial cells; pink contours, density after MRTX1133 treatment. d. Differential abundance analysis of transcriptional neighborhoods of fibroblasts or myeloid cells in MRTX1133-treated compared to vehicle-treated samples. Each dot represents a transcriptional neighborhood as defined by MiloR (see Methods). Color represents the enrichment of progenitor-like cells in the spatial vicinity of cells in the transcriptional neighborhood. Significantly enriched or depleted transcriptional neighborhoods are outlined in black. e. Representative images and quantification of TNC staining in KP^LOH mice treated with vehicle (n = 3) or MRTX1133 (n = 4). Tissue was collected 2 days after the first inhibitor or vehicle dose. f. Representative images of Xenium data from vehicle or MRTX1133 treated mice. Each dot is a cell centroid, and colors represent select cell states. Scale bars, 250 μm. b,e. Scale bars, 50 μm. Each dot in quantification corresponds to an individual mouse; bar corresponds to average value. P-value, Two Tailed Wilcoxon Rank Sums test.

Figure 7.. Consequences of p53 knockdown in the premalignant pancreas.

a. Mouse model for doxycycline-inducible knockdown of p53 in the premalignant pancreatic epithelium. b. Representative images and quantification of HMGA2 staining 3 weeks post-pancreatitis in shp53 (n = 11) or shRen (n = 10) mice. c. Two-dimensional density representation of single-cell transcriptomes from shRen (n = 4) or shp53 (n = 5) Kras^G12D+ pancreatic epithelial cells. d. Randomly sampled shp53 or shRen cells (top) and average score of expression signatures in shRen or shp53 cells (bottom) binned along the gastric–progenitor DC (bins <10 cells are not plotted). e,f. Representative images and quantification of VIM staining 3 weeks post-pancreatitis in shp53 (n = 10) or shRen (n = 10) mice. g. Xenium-based quantification of microenvironment subpopulations associated with the progenitor niche, as a function of the fraction of progenitor-like cells in the tissue. Each dot is a single biological replicate. h. Representative images of Xenium data from KC^shRen or KC^shp53 mice 3 weeks post pancreatitis. Each dot is a cell centroid, and colors represent select cell states. Side panels show the abundance of Ccn2+ myofibroblasts or Itgax+/Cd274^high macrophage/monocytes in their associated sample. Scale bars, 500 μm. b,e. Scale bars, 50 μm. Each dot in quantification corresponds to an individual mouse; bar corresponds to average value. P-value, Two-tailed Wilcoxon Rank Sums test.