Spatial Transcriptomics of Intraductal Papillary Mucinous Neoplasms of the Pancreas Identifies NKX6-2 as a Driver of Gastric Differentiation and Indolent Biological Potential

Abstract

Intraductal papillary mucinous neoplasms (IPMN) of the pancreas are bona fide precursor lesions of pancreatic ductal adenocarcinoma (PDAC). The most common subtype of IPMNs harbors a gastric foveolar-type epithelium, and these low-grade mucinous neoplasms are harbingers of IPMNs with high-grade dysplasia and cancer. The molecular underpinning of gastric differentiation in IPMNs is unknown, although identifying drivers of this indolent phenotype might enable opportunities for intercepting progression to high-grade IPMN and cancer. We conducted spatial transcriptomics on a cohort of IPMNs, followed by orthogonal and cross-species validation studies, which established the transcription factor NKX6-2 as a key determinant of gastric cell identity in low-grade IPMNs. Loss of NKX6-2 expression is a consistent feature of IPMN progression, while reexpression of Nkx6-2 in murine IPMN lines recapitulates the aforementioned gastric transcriptional program and glandular morphology. Our study identifies NKX6-2 as a previously unknown transcription factor driving indolent gastric differentiation in IPMN pathogenesis.

Significance: Identification of the molecular features driving IPMN development and differentiation is critical to prevent cancer progression and enhance risk stratification. We used spatial profiling to characterize the epithelium and microenvironment of IPMN, which revealed a previously unknown link between NKX6-2 and gastric differentiation, the latter associated with indolent biological potential. See related commentary by Ben-Shmuel and Scherz-Shouval, p. 1768. This article is highlighted in the In This Issue feature, p. 1749.

Figure 1.. Spatially resolved transcriptomic analyses of IPMN tissue samples.

A) Uniform Manifold Approximation and Projection (UMAP) showing clustering of spots based on gene expression profiles for all 13 samples. B) Dot plot including cell type scores per cluster obtained with the robust cell type decomposition (RCTD) algorithm. Color represents the normalized average cell proportion and dot size represents the maximum cell proportion per cluster. Representative spatial maps showing correlation between cell composition and localization provided by RCTD, spatial gene expression, cluster assignments, and histology. C) Mapping of epithelial cells, EPCAM expression, and localization of spots from cluster 11 for a HG IPMN sample. D) Localization and proportion of naïve CD4T and naïve B cells, and visualization of cluster 2 spots from a HG IPMN sample with co-occurring cancer. A tissue area corresponding to a lymphocyte aggregate is outlined and a zoomed-in image is provided. (E) Plasma cells, IGLC2 expression, and cluster 4 shown for a LG IPMN sample. IPMN areas are outlined in black in the optical images from H&E stained tissues.

Figure 2.. Histologically directed spot selection reveals area-specific gene expression changes associated with IPMN grade and PDAC.

A) Selection of epilesional, juxtalesional, and perilesional areas shown in a LG IPMN sample. B) Heatmaps plotting expression of genes differentially expressed between LG and HR (high-risk) in epilesional (left) and juxtalesional (right) spots. Transcripts exclusively showing differential expression in the jutxalesional and not the epilesional compartments were included in the juxtalesional heatmap. Violin plot showing gene expression levels for selected features for spots in C) epilesional and D) juxtalesional areas.

Figure 3.. Spatial organization of the tumor microenvironment in IPMN samples.

A) Representative spatial maps showing proportion of cell types in the ST spots as predicted by RCTD. The image at the top shows the spots colored according to their region. B) Violin plot showing per spot proportions of plasma and mast cells grouped by sample code and colored by type. Representative spatial maps highlighting the proportion and localization of epithelial, plasma (C), and mast (D) cells in the IPMN samples. On the right side, spots are colored according to their corresponding region (epilesional, jutxalesional, and perilesional). E) Percentage of cells staining positive for CD27 (left) and TPSAB1(right) in the adjacent microenvironment to the IPMN lesions in the samples analyzed by the COMET imaging system. A dash line connects the LG and HG IPMN lesions that correspond to the same patient and tissue section. F) COMET images showing co-staining of CD27 (top) and TPSAB1 (bottom) with EPCAM and DAPI in LG and HG IPMN samples.

Figure 4.. NKX6-2 expression is elevated in LG IPMN and associated with a distinct transcriptional profile.

A) Spatial maps of NKX6-2 expression and localization from ST analyses of four representative samples. Optical images for the corresponding H&E images are provided below, with the neoplastic epithelium outlined. B) Violin plot showing NKX6-2 expression in the IPMN epilesional spots grouped by sample. C) Percentage of NKX6-2 positive nuclei in IPMN epithelium determined by COMET cyclic IF staining (p = 0.0039). D) Localization of NKX6-2 in the nuclei of LG IPMN epithelium showed by co-staining with EPCAM, DAPI, and NKX6-2. An H&E image of the area stained is provided in the bottom left. E) NKX6-2 staining of LG and HG IPMN lesions from the same tissue section showing higher NKX6-2 levels in the LG IPMN nuclei. H&E images of the areas stained are provided below. F) Representative gene transcripts observed at significantly higher levels in spots of high NKX6-2 (left panel) or low NKX6-2 levels (right panel) in the ST dataset from LG IPMN. G) Co-staining of GATA6 and NKX6-2 showing co-expression in the LG IPMN nuclei. H) Median GATA6 intensity obtained from the COMET staining in NKX6-2+ versus NKX6-2- nuclei (p = 0.03).

Figure 5.. Enrichment of the GIP signature in NKX6-2 expressing IPMN.

A) Enrichment of gastric pit and isthmus gene sets in the NKX6-2 associated signature obtained from the ST LG IPMN dataset. B) Representative COMET images showing co-staining of NKX6-2, MUC5AC, and TFF1 in three different LG IPMN samples. C) Heatmap showing expression of NKX6-2 and the GIP signature in 23 archival IPMN tissue areas from the NCI Cancer Moonshot PCAPP dataset (Semaan et al, bioRxiv 2022). Samples are ranked by NKX6-2 expression from right to left. Heatmap values represent normalized RNA-seq expression values. D) Dot plot showing normalized enrichment scores and size for gastric pit and isthmus gene sets in both the PCAPP and ST dataset by GSEA. E) Regulon activities scored for NKX6-2+ (green) and NKX6-2- (light yellow) spots based on SCENIC analyses.

Figure 6.. Nkx6-2 expression in a genetically engineered model of IPMN.

ST analyses of murine cystic lesions collected from Kras;Gnas mice. A) From top to bottom: H&E stained tissue samples. Spots corresponding to clusters of high epithelial contribution were selected and colored according to Nkx6-2 expression, red = Nkx6-2+, blue = Nkx6-2 –. Zoomed-in images of areas presenting with high percentage of Nkx6-2 + spots. B) Expression of gene transcripts associated with Nkx6-2 in the Kras;Gnas ST dataset shown in a violin plot. (C) Representative images of Kras;Gnas mouse tissues stained with NKX6-2 (red), P120 (green) and DAPI (blue). Arrows indicate enlarged area. (D) Quantification of the percentage of NKX6-2 positive cells in P120+DAPI+ epithelial cells. Three images per mouse captured under 40X magnification were used to count the number of cells and each dot shows the average percentage of cells in each mouse.

Figure 7.. Enforced expression of Nkx6-2 in a cell line derived from Kras;Gnas mice.

A) Heatmap showing normalized expression values of genes differentially expressed in the Kras;Gnas^Nkx6-2+ versus control cell line that were also found to be differentially expressed in Nkx6-2 + vs Nkx6-2 – spots in the ST dataset from Kras;Gnas mice. Expression data is provided for three RNA-seq replicates. Note that genes with average raw counts < 10 from were not included in the heatmap. B) Enrichment of the gastric isthmus and pit cell gene sets in the Nkx6-2 overexpressing cell line (Kras;Gnas^Nkx6-2+). C) Cell proliferation assay using the Kras;Gnas^Nkx6-2+ or control cells. * P < 0.05. D) Western blot of cell lysates obtained from the control and Kras;Gnas^Nkx6-2+ cells, stained with NKX6-2, E-Cadherin, Vimentin, ZEB1, ZEB2 and β-actin antibodies. E) Representative H&E images of the tumors derived from the control or Kras;Gnas^Nkx6-2+ cells at day 14 after orthotopic injection into pancreas of littermate (Kras;Gnas) or immunodeficient nude mice (NuJ). Images were captured under 10X or 40X (inset) magnification.