Tissue-location-specific transcription programs drive tumor dependencies in colon cancer

Abstract

Cancers of the same tissue-type but in anatomically distinct locations exhibit different molecular dependencies for tumorigenesis. Proximal and distal colon cancers exemplify such characteristics, with BRAF^V600E predominantly occurring in proximal colon cancers along with increased DNA methylation phenotype. Using mouse colon organoids, here we show that proximal and distal colon stem cells have distinct transcriptional programs that regulate stemness and differentiation. We identify that the homeobox transcription factor, CDX2, which is silenced by DNA methylation in proximal colon cancers, is a key mediator of the differential transcriptional programs. Cdx2-mediated proximal colon-specific transcriptional program concurrently is tumor suppressive, and Cdx2 loss sufficiently creates permissive state for BRAF^V600E-driven transformation. Human proximal colon cancers with CDX2 downregulation showed similar transcriptional program as in mouse proximal organoids with Cdx2 loss. Developmental transcription factors, such as CDX2, are thus critical in maintaining tissue-location specific transcriptional programs that create tissue-type origin specific dependencies for tumor development.

Fig. 1. Abnormal Wnt pathway activation facilitates BRAF^V600E mutation-driven tumorigenesis in proximal and distal organoids derived from mouse colon.

a Schematic representation of gene editing of proximal or distal colon organoid using CRISPR-Cas9 targeting Apc or N-terminal of Ctnnb1 (β-catenin) gene, followed by induction of BRAF^V600E by Cre recombinase. b Representative composite images (phase-contrast and red fluorescent) of Wnt-factor independency analysis for proximal/distal colon organoid with/without abnormal Wnt-pathway activation (Apc-sgRNA, β-catenin-sgRNA group), and with Braf wild type or BRAF^V600E (growth assay data are representative of two biological replicates). c, d Subcutaneously grafted organoid-derived tumor growth curve of proximal (c)/distal (d) colon organoid with/without constitutive Wnt-pathway activation by mutant Apc or Ctnnb1 (β-catenin) (APC/ β-catenin-sgRNA) and with/without Cre-mediated BRAF^V600E activation injected subcutaneously into NSG mice. Error bars indicate means ± SD (n = 5 mice, right flank each mouse, one biological organoid replicate injected in the mice for assays). Source data are provided in FigurePlotsSourceData.xlsx. e Tumor volume formed by proximal versus distal colon organoids at day 22. Error bars indicate ± SD (n = 5 mice, right flank each mouse, one biological organoid replicate injected in the mice for assays). Two-sided Welch’s t-test. Source data are provided in FigurePlotsSourceData.xlsx.

Fig. 2. Loss of Cdx2 gene imparts transient Wnt-independent growth specifically to proximal colon organoids.

a Schematic representation showing proximal or distal colon organoids transduced with lentivirus containing Scramble-sgRNA, Sox17-sgRNA, Sfrp4-sgRNA, Cdkn2a-sgRNA, or Cdx2-sgRNA to individually target these genes using CRISPR-Cas9. b Representative images showing Wnt-factor-independent growth of proximal and distal colon organoids transduced with lentivirus expressing the Scramble-sgRNA, Sox17-sgRNA, Sfrp4-sgRNA, Cdkn2a-sgRNA, or Cdx2-sgRNA targeting these genes separately using CRISPR-Cas9 technology (growth assay data are representative of two biological replicates). c Quantitative real-time PCR analysis of stemness (Ephb2, Lgr5, and Ascl2) or differentiation (Muc2, Fabp2, Krt20, and Car1) markers for proximal colon organoids with Cdx2-sgRNA or Apc-sgRNA and cultured in WENR-plus or -minus medium. n = 3 biological replicates. Error bars indicate means ± SD. Two-sided Welch’s t-test. Source data are provided in FigurePlotsSourceData.xlsx.

Fig. 3. Mutations in Sfrp4, Cdkn2a, and Sox17 synergize with Cdx2 deficiency to facilitate proximal colon-specific BRAF^V600E-driven tumorigenesis.

a Schematic representation showing proximal and distal colon organoids transduced with lentivirus expressing control gRNA, C2-gRNA (targeting Cdx2), SSC-gRNA (simultaneously targeting Sfrp4, Sox17, and Cdkn2a), and C2SSC-gRNA (simultaneously targeting Cdx2, Sfrp4, Sox17, and Cdkn2a) for targeted mutations in these genes using CRISPR-AsCpf1 and followed with induction of BRAF^V600E using Cre recombinase. b Representative composite images (phase-contrast and red fluorescent) of Wnt-factor independency analysis for proximal and distal colon organoids treated with control Scramble-gRNA, C2-gRNA, SSC-gRNA, and C2SSC-gRNA separately with or without of BRAF^V600E activation (growth assay data are representative of two biological replicates). c Representative images showing the ability of proximal colon-derived organoids with C2-gRNA and C2SSC-gRNA to form new organoids in WENR-minus medium upon BRAF^V600E induction. The organoids were cultured in WENR-minus medium for 7 days, separated into single cells, and challenged to grow in WENR-minus medium for a second passage (growth assay data are representative of two biological replicates). d Plot showing volume of subcutaneously grafted organoid-derived tumor/lesion formed by proximal colon organoids edited with C2-gRNA or C2SSC-gRNA combined with BRAF^V600E induction at 15th and 29th day after subcutaneous grafting. Error bars indicate means ± SD (n = 5 mice, right flank each mouse, one biological organoid replicate injected in the mice for assays). Two-sided Welch’s t-test. Source data are provided in FigurePlotsSourceData.xlsx. e Representative subcutaneously grafted organoid-derived tumors dissected from the groups in (d) at day 70. f Table summarizing subcutaneously grafted organoid-derived tumor/lesion growth of proximal and distal colon organoids with control gRNA, C2-gRNA, SSC-gRNA, or C2SSC-gRNA and with or without BRAF^V600E induction.

Fig. 4. Alterations to key pathways and stemness versus differentiation status in proximal colon organoids upon loss of Cdx2.

a Volcano maps representing the differentially expressed genes in proximal versus distal colon organoids, and those after Cdx2 loss, cultured in WENR-minus medium (n = 2 biological replicates each). Differential gene expression analyses were corrected for multiple hypothesis using BH-methods implemented in R/DESeq2. b Volcano maps showing differentially expressed genes in proximal and distal colon organoids upon Cdx2 loss compared to corresponding control organoids cultured in WENR-minus medium (n = 2 biological replicates each). Proximal organoids show greater number of genes with significantly altered expression compared to distal organoids. c Dot plots summarizing enriched KEGG pathways in the differentially expressed genes from comparing gene expression profiles of proximal versus distal colon organoids with Cdx2 loss (Cdx2-sgRNA) cultured in WENR-minus medium (n = 2 biological replicates each). d Dot plots summarizing enriched KEGG pathways in proximal and distal colon organoids upon Cdx2 loss compared to corresponding control organoids cultured in WENR-minus medium (n = 2 biological replicates each). e GSEA analysis showing Cdx2 deficiency (Cdx2-sgRNA) results in decreased differentiated cell (Crypt Top) gene expression signatures and an increase in stemness (Crypt Bottom) signature in proximal colon organoids cultured in WENR-minus medium, but not distal colon organoids (n = 2 biological replicates each). Gene signatures are from Kosinski et al..

Fig. 5. Cdx2 directs transcriptional program critical for maintenance of differentiation responsive genes in proximal colon stem cells.

a ChIP-seq heatmaps showing differential binding of Cdx2 in proximal or distal colon organoids cultured in WENR-plus versus WENR-minus medium (the two panels on the left), and in proximal versus distal colon organoids cultured in WENR-plus or -minus medium (the two panels on the right) (n = 3 biological replicates). b PCA representation of Cdx2 enrichment at genomic loci. Each dot of the same color represents 3 replicates from independent mice-derived proximal and distal colon organoids cultured in WENR-plus or -minus medium. c Annotation with respect to gene structure for regions bound by Cdx2 at sites across the genome in proximal or distal colon organoids cultured in WENR-plus or -minus medium, and annotation for those regions showing gain or loss of Cdx2 binding in proximal versus distal colon organoid cultured in WENR-plus or -minus medium. d Violin maps summarizing the expression changes of genes whose promoters are directly bound by Cdx2 in proximal and distal colon organoids with or without Cdx2 deficiency cultured in WENR-plus or -minus medium (n = 2 biological replicates for gene expression data). e GSEA results from comparing gene expression of genes whose promoters are bound by Cdx2 in proximal and distal organoids. Comparisons are made between Cdx2 deficient (Cdx2-sgRNA) vs. control (Scramble-sgRNA) for proximal and distal organoids grown in WENR-plus (top panel) or WENR-minus (middle panel) (n = 2 biological replicates for gene expression data). Schematic model diagram showing that Cdx2-bound genes in proximal organoids are dependent on Cdx2 for their expression while in distal organoids these genes are less dependent on Cdx2. f Left panel shows ChIP-seq enrichment traces showing Cdx2 binding in Hnf4a gene locus for the three replicates of proximal and distal colon organoids cultured in WENR-plus or -minus medium. Right panel shows histogram summarizing Hnf4a mRNA expression changes in Cdx2 deficient (Cdx2-sgRNA) versus control proximal and distal colon organoids cultured in WENR-plus or -minus medium. Barplots depict mean log2-fold change differences in gene expression from DESeq2 analyses. (n = 2 biological replicates for each comparison).

Fig. 6. Cdx2 loss in mouse proximal colon organoid recapitulates gene expression patterns of human colon cancers with low CDX2 expression.

a Violin plots showing expression of CDX2, HNF4A, SATB2, and VIL1 in normal proximal colon tissue, proximal colon cancer, normal distal colon tissue and distal colon cancer in human colon cancers (TCGA-COAD) database (p-values from one-sided Games-Howell test for post hoc analysis for comparing multiple groups with unequal variances is shown) (normal proximal colon samples = 21; proximal colon cancer samples = 293; normal distal colon samples = 20; distal colon cancer samples = 180). b Violin plots showing expression of CDX2 in proximal versus distal colon cancer using GEO database (GSE39582) (p-values from two-sided Welch’s t-test shown) (proximal colon cancer samples = 176; distal colon cancer samples = 267). c Histogram on the left shows distribution of proximal and distal tumors classified into CDX2-positive (n = 372) and CDX2-negative (n = 71) expression groups. Histogram on the right shows distribution of colon cancers with BRAF mutation or wild type for BRAF classified into the same CDX2-positive (n = 328) and CDX2-negative (n = 64) expression groups. Data were analyzed from GEO database (GSE39582). Chi-Squared test p-values shown. d GSEA results from comparing gene expression of mouse gene orthologues of the human genes whose expression is correlated with CDX2 expression in human colon cancers (TCGA-COAD). Comparisons are made between Cdx2 deficient (Cdx2-sgRNA) vs. control (Scramble-sgRNA) for proximal and distal organoids grown in WENR-minus (top panel) or WENR-plus (bottom panel) (n = 2 replicates each). The set of genes whose expression is correlated with CDX2 expression in human proximal colon cancers is used for comparing gene expression relationship in the proximal organoids (left panels), while the genes correlated with CDX2 expression in human distal colon cancers are used for comparing gene expression relationship in the distal organoid (right panels). Human colon cancer gene expression dataset is from the TCGA. e Correlation of CDX2, HNF4A, SATB2, and VIL1 gene expression in proximal colon cancer in the TCGA-COAD database. P-values from Pearson’s correlation test are shown (proximal colon cancer samples = 293). f Heat map showing gene expression profiles in Cdx2-KO versus Scramble-sgRNA in proximal or organoids cultured in WENR-minus or -plus medium (n = 2 replicates each case). Plot shows expression values for mouse gene orthologues of the human genes whose expression is correlated with CDX2 expression in human colon cancers (TCGA-COAD).