Enhancer Reprogramming Promotes Pancreatic Cancer Metastasis

Abstract

Pancreatic ductal adenocarcinoma (PDA) is one of the most lethal human malignancies, owing in part to its propensity for metastasis. Here, we used an organoid culture system to investigate how transcription and the enhancer landscape become altered during discrete stages of disease progression in a PDA mouse model. This approach revealed that the metastatic transition is accompanied by massive and recurrent alterations in enhancer activity. We implicate the pioneer factor FOXA1 as a driver of enhancer activation in this system, a mechanism that renders PDA cells more invasive and less anchorage-dependent for growth in vitro, as well as more metastatic in vivo. In this context, FOXA1-dependent enhancer reprogramming activates a transcriptional program of embryonic foregut endoderm. Collectively, our study implicates enhancer reprogramming, FOXA1 upregulation, and a retrograde developmental transition in PDA metastasis.

Keywords: FOXA1; enhancer; metastasis; organoid; pancreatic cancer; pancreatic ductal adenocarcinoma.

Figure 1. Paired Tumor- and Metastasis-derived Organoids Provide a Model to Study Mechanisms of PDA Progression

(A) Schematic diagram demonstrating establishment of paired primary tumor- and metastasis-derived organoid cultures from the KPC mouse model of pancreatic cancer. (B) Hematoxylin and eosin (H&E) staining of murine primary tumor and metastases derived from KPC mice (top) and bright field images of established organoids (bottom). Scale bars, 100 μm (top) and 1 mm (bottom). (C and D) Representative bright field images (C, left) and H&E staining (C, right) of lung and quantification (D) 2 months after tail vein injection of T³ and M^1L organoids (5 × 10⁵ cells per each injection, n = 4) in C57Bl/6J mice. Scale bars, 5 mm. (E and F) Representative bright field images (E) and quantification (F) one month after portal vein injection of T³ and M^1L organoids (5 × 10⁵ cells per each injection, n = 4) in C57Bl/6J mice. (G) Quantification of subcutaneous tumors 6 weeks after injection of T³ and M^1L organoids (5 × 10⁵ cells per each injection, n = 5) in C57Bl/6J mice. (H) Quantification of tumor weight 4 weeks after peritoneal wall injection of T²³ and M^10P organoids (5 × 10⁵ cells per each injection, n = 4) in C57Bl/6J mice. For bar charts, each dot represents a mouse and the mean +/- SEM is shown. p-value was determined by Student’s t-test. See also Figure S1.

Figure 2. Massive enhancer reprogramming accompanies the metastatic transition

(A) Distribution of gained H3K27ac peaks across the indicated P, T and M organoids. From a total of 34,852 H3K27ac ChIP-seq peaks, gained H3K27ac peaks are defined by more than 10-fold increase of H3K27ac signal relative to the average of N⁵ and N⁶ (N^avg). Bottom brackets indicate the pairs of matching T and M organoids. (B) Heatmap representation of GAIN regions based on H3K27ac occupancy in organoids. The 857 GAIN regions were identified from the union of all gained H3K27ac peaks described in Figure 2A. 10-Kb around the center of GAIN regions are displayed for each organoid. Each row represents a single region (n = 857). Each column represents an individual organoid (n = 16). (C) Representative H3K27ac ChIP-seq profiles of GAIN regions at Lbh and Acan loci in N, P, T and M organoids. (D) Heatmap representation of unsupervised hierarchical clustering based on H3K27ac occupancy at total H3K27ac ChIP-seq peaks. Samples were clustered based on Spearman correlation coefficient with average linkage. (E) Pie chart showing the genomic annotations of 857 GAIN regions according to the location of a given peak. TSS, ‘−1-kb to +100-bp’ of transcription start sites. TTS, ‘−100-bp to +1-Kb’ of transcription termination sites. UTR includes both 5′ and 3′ UTRs. (F) Metagene representation of the mean ChIP-seq signal for the indicated histone marks (left and middle) or the mean ATAC-seq signal (right) across GAIN regions in the indicated organoids. Metagenes are centered on the middle of GAIN regions and 10-Kb around the center of GAIN regions are displayed. To compare relative signal changes (denoted as value-x), the mean signal of each biological replicate was determined by averaging signals of 2-Kb (H3K27ac or H3K4me1 ChIP-seq) or 1-Kb (ATAC-seq) around the center of GAIN regions. p-value was determined by Kolmogorov–Smirnov test. See also Figure S2 and Table S1 and S2.

Figure 3. Elevated FOXA1 Expression and Occupancy is Associated with PDA Metastasis and GAIN Enhancer Activation

(A) Representation of motifs enriched at GAIN regions. Mouse promoters were used as the comparison library. p-values correspond to the corrected p in the output. AP1 motif is ranked first; SOX motif is ranked second; FOXD1 and FOXA2 motifs are ranked third and fourth, respectively. Position weight matrices for FOXD1 and FOXA2 are represented as Forkhead motifs. (B) RNA-seq based Foxa1 mRNA expression in organoids. Each dot represents an independent organoid culture and the mean +/− SEM is shown. p-value was determined by Student’s t-test. N, normal; P, PanIN; T, primary tumor; M, metastasis; rpkm, reads per kilobase per million mapped reads. (C) Representative FOXA1 ChIP-seq profiles of GAIN regions at Lbh and Acan loci in the indicated organoids. (D) Metagene representation of the mean FOXA1 ChIP-seq signal across GAIN regions in the indicated organoids. The mean signal of each biological replicate was determined by averaging signals of 1-Kb around the center of the indicated regions. p-value was determined by Kolmogorov–Smirnov test. (E) RNA fluorescence in situ hybridization (RNA-FISH) of the indicated tissue sections including diaphragm metastasis from KPC mice using a Foxa1-specific probe (red). Two independent KPC primary tumor and metastatic lesions were examined. White arrows indicate areas of focal Foxa1 mRNA expression in primary PDA tumors. Counterstain, DAPI (blue). Scale bars, 100 μm. (F) Microarray based FOXA1 mRNA expression in publicly available PDA patient data. Each dot represents individual patient. The mean +/− SEM is shown. p-values were determined by Student’s t-test. (G) FOXA1 expression during pancreatic differentiation of human embryonic stem cells (hES) relative to foregut stage. (H) Gene set enrichment analysis (GSEA) of posterior foregut versus hES RNA-seq using a signature of GAIN genes. Normalized enrichment score (NES) and nominal p-value (p) were provided according to GSEA. (I) GSEA of averaged six M (M^avg) versus averaged six T (T^avg) RNA-seq using posterior foregut signature genes. See also Figure S3 and Table S3.

Figure 4. FOXA1 Promotes Enhancer activation and Metastatic Capabilities in 2D PDA Cell Lines

(A) Western blot analysis with the indicated antibodies and whole cell lysates prepared from KPC-2D cells stably expressing FOXA1 cDNA (KPC-2D/FOXA1) or control (KPC-2D/empty). (B) Metagene representation of the mean FOXA1 (left) or H3K27ac (right) ChIP-seq signal across GAIN regions in the indicated organoids. The mean signal of each biological replicate was determined by averaging signals of 1-Kb (FOXA1) or 2-Kb (H3K27ac) around the center of the indicated regions. p-value was determined by Kolmogorov–Smirnov test. (C) GSEA of averaged KPC-2D/FOXA1 (FOXA1^avg) versus averaged KPC-2D/empty (empty^avg) RNA-seq using GAIN genes (left), and posterior foregut signature genes (right). (D) Anchorage-independent sphere formation assay of KPC-2D/empty and KPC-2D/FOXA1 cells (n = 4). Bright field images were taken from 4x field after 2 weeks of culture and the size of spheres were measured for quantification. The mean +/− SEM is shown and p-value was determined by Student’s t-test. Scale bars, 500 μm. (E) Boyden-chamber invasion assay of KPC-2D/empty and KPC-2D/FOXA1 cells (n = 4). Invaded cells were stained with Syto13 (GFP) 24 hours after cell seeding. Fluorescent microscopic images were taken from 4× field (right) to quantify the number of invaded cells (left). The mean +/− SEM is shown and p-value was determined by Student’s t-test. Scale bars, 500 μm. (F) Tail vein injection of KPC-2D/empty and KPC-2D/FOXA1 cells (2.5 × 10⁵ cells) into C57Bl/6J mice to assess lung colonization at 4 weeks post-injection. Representative images of H&E staining are shown (right). Percent tumor area per lung lobe (%) was determined by dividing the whole surface area of the largest lung lobe with lung tumor area of the given lobe. The mean +/− SEM is shown and p-value was determined by Student’s t-test (n = 15). Scale bars, 5 mm. (G) Western blot analysis with the indicated antibodies and whole cell lysates prepared from hT2-2D cells stably expressing FOXA1 cDNA (hT2-2D/FOXA1) or control (hT2-2D/empty). (H) Metagene representation of H3K27ac ChIP-seq signal across GAIN regions in hT2-2D cells. The mean signal of each biological replicate was determined by averaging signals of 2-Kb H3K27ac around the center of the indicated regions. p-value was determined by Kolmogorov– Smirnov test. (I) GSEA of hT2-2D/FOXA1 versus hT2-2D/empty using a signature of GAIN genes (left) and posterior foregut signature genes (right). Normalized enrichment score (NES) and nominal p-value (p) were provided according to GSEA. (J) Anchorage-independent sphere formation assay of hT2-2D/empty and hT2-2D/FOXA1 cells (n = 3). Bright field images were taken from 4x field after 2 weeks of culture and the size of spheres were measured for quantification. The mean +/− SEM is shown and p-value was determined by Student’s t-test. Scale bars, 500 μm. See also Figure S4 and Table S4–S5.

Figure 5. FOXA1 Cooperates with GATA5 to Activate GAIN Enhancers and Promote PDA Progression in vivo

(A) Western blot analysis of T³/empty or T³/FOXA1-GATA5 organoids with the indicated antibodies. (B) Metagene representation of the mean H3K27ac ChIP-seq signal across GAIN regions in the indicated organoids. Mean H3K27ac signal of two parental T organoids (T³ and T²³) and their matched M organoids (M^1L and M^10P) are displayed (left). T³ and T²³ organoids stably expressing control, FOXA1, GATA5 or FOXA1-GATA5 construct were subjected to H3K27ac ChIP-seq to compare with parental T (blue dotted line) or matched M (red dotted line) organoids. Parental T organoids were compared with matched M organoids, and T/FOXA1 or T/FOXA1-GATA5 organoids were compared with T/empty organoids. The mean signal of each biological replicate was determined by averaging signals of 2-Kb around the center of the indicated regions. p-value was determined by Kolmogorov–Smirnov test. (C) Representative H3K27ac ChIP-seq profiles of GAIN regions at Lbh and Acan loci in the indicated organoids. (D) GSEA of averaged T/FOXA1-GATA5 (FOXA1-GATA5^avg) versus averaged T/empty (empty^avg) RNA-seq using GAIN genes (left) or posterior foregut signature genes (right). (E-F) Quantification of the number of mice with primary tumors at injection site (pancreas, E) and metastases (F) after orthotopic injection of T³/empty or T³/FOXA1-GATA5 organoids in athymic nu/nu mice (n = 5). 5 × 10⁵ cells were orthotopically injected and mice were sacrificed at 12 weeks post-injection. Frequency of PDA and microinvasive PDA development in orthotopic injection is 40% (n = 2/5) and 20% (n = 1/5), respectively for T³/empty and 80% (n = 4/5) and 20% (n = 1/5), respectively for T³/FOXA1-GATA5. (G) H&E staining of lung micrometastases (n = 4/5, arrows) upon orthotopic injection of T³/FOXA1-GATA5 organoids. Scale bars, 200 μm. (H) Tumor development after subcutaneous injection of T³/empty and T³/FOXA1-GATA5 organoids. 5 × 10⁵ of cells were subcutaneously injected into C57Bl/6J mice and mice were sacrificed at 11 weeks post-injection. Note, four of T3/empty-injected mice (n = 5) did not show apparent subcutaneous tumor formation in a syngeneic background. See also Figure S5 and Table S4–S5.

Figure 6. Suppression of FOXA1 Deactivates GAIN Enhancers and Diminishes the Aggressiveness of Metastasis-derived PDA Organoids

(A) Western blotting of M^1L organoids stably expressing shRen (M^1L/shRen) or shFoxa1.2959 construct (M^1L/shFoxa1) with the indicated antibodies. (B) Representative H3K27ac ChIP-seq profiles of GAIN regions at Lbh and Acan loci in the indicated organoids. (C) Metagene representation of the mean FOXA1 (left) or H3K27ac (right) ChIP-seq signal across GAIN regions in the indicated organoids. The mean signal of each biological replicate was determined by averaging signals of 1-Kb (FOXA1) or 2-Kb (H3K27ac) around the center of the indicated regions. p-value was determined by Kolmogorov–Smirnov test. (D) GSEA of M^1L/shFoxa1 versus M^1L/shRen using GAIN genes (left) or posterior foregut signature genes (right). (E) Quantification of primary tumor weight from injection site (pancreas) after orthotopic injection of M/shRen or M/shFoxa1 organoids in athymic nu/nu mice (n = 10). 5 × 10⁵ cells were orthotopically injected and mice were sacrificed at 6 weeks post-injection. Each dot represents a mouse and the mean +/− SEM is shown. p-value was determined by Student’s t-test. (F) Quantification of the number of mice bearing metastasis upon orthotopic injection of M/shRen or M/shFoxa1 organoids (n = 10). See also Figure S6 and Table S4–S5.