Retinoic acid signaling drives differentiation toward the absorptive lineage in colorectal cancer

Abstract

Retinoic acid (RA) signaling is an important and conserved pathway that regulates cellular proliferation and differentiation. Furthermore, perturbed RA signaling is implicated in cancer initiation and progression. However, the mechanisms by which RA signaling contributes to homeostasis, malignant transformation, and disease progression in the intestine remain incompletely understood. Here, we report, in agreement with previous findings, that activation of the Retinoic Acid Receptor and the Retinoid X Receptor results in enhanced transcription of enterocyte-specific genes in mouse small intestinal organoids. Conversely, inhibition of this pathway results in reduced expression of genes associated with the absorptive lineage. Strikingly, this latter effect is conserved in a human organoid model for colorectal cancer (CRC) progression. We further show that RXR motif accessibility depends on progression state of CRC organoids. Finally, we show that reduced RXR target gene expression correlates with worse CRC prognosis, implying RA signaling as a putative therapeutic target in CRC.

Keywords: Biological sciences; Cancer; Cell biology; Functional aspects of cell biology; Oncology.

Graphical abstract

Figure 1

Retinol metabolism is enriched in differentiated cells of the intestine (A) Gene set enrichment analysis (GSEA) of gene expression dynamics in enterocyte-enriched mouse small intestinal organoid cultures (EN) versus wt organoids (ENR). Significantly enriched KEGG gene sets (false discovery rate [FDR] <0.001) associated with EN gene expression have a positive normalized enrichment score (NES), whereas gene sets associated with ENR gene expression have a negative NES. Genes associated with retinol metabolism are enriched in EN organoids. (B and C) (B) Expression of retinol metabolism-associated genes in ENR and EN organoids. The heatmaps are row-matched, and all the genes that are shown are dynamically expressed at the protein level. The left heatmap shows mRNA dynamics, whereas the right heatmap shows protein dynamics of the indicated genes. Relative expression is visualized as log2 fold change over the row mean (RMS). Genes associated with the retinol metabolism are upregulated in enterocyte-enriched organoids on RNA and protein levels. (C) GSEA of gene expression dynamics in STAR-negative versus STAR-positive cells of human colon and colorectal cancer organoids (Oost et al., 2018). STAR-negative cells represent differentiated cells and are significantly enriched for retinol metabolism-associated gene expression (NES = 1.98, FDR = 0.001).

Figure 2

Modulation of RA signaling alters intestinal differentiation (A) Schematic overview of drugs used in this study. Biological ligands for RAR (ATRA) and RXR (9-cis RA) were used, as well as a synthetic RAR antagonist (AGN193109), a synthetic RXR antagonist (HX531), and two synthetic RXR agonists (LG100268, NRX1942014). (B) Bright-field microscopic photographs of mouse small intestinal organoids treated with drugs indicated in (A). HX531 treatment results in cystic organoids with thin, elongated crypts. Colored rectangles refer to colors used in (A). (C) Quantification of organoid circularity. HX531-treated organoids are more circular, meaning that their area to perimeter ratio is larger. Number of organoids measured is indicated in the plot; mean ± SD is shown. ∗∗∗p < 0.001 (ANOVA followed by Dunett's post-hoc test). Only HX531 significance is shown; comparisons of other treatments with control all result in p > 0.05. (D) Bubble plot showing GSEA parameters normalized enrichment score (NES, color) and FDR (size) for cell type-specific gene sets from Haber et al. (2017) in organoids treated with compounds indicated in (A). All enrichment scores were calculated taking DMSO as reference. Inhibition of either RAR or RXR leads to dramatic reduction of enterocyte-specific genes (n = 1 per condition). (E) Alkaline phosphatase/nuclear fast red staining for DMSO and HX531-treated organoids. Alkaline phosphatase expression is abundantly clear at the apical membrane and in the lumen of control organoids, but greatly reduced in HX531-treated organoids. Three images shown for each condition; scale bar, 100μm. (F) Alcian blue/nuclear fast red staining for DMSO and HX531-treated organoids. Goblet cells containing mucous are blue and have been marked with arrows. The amount of goblet cells in HX531-treated organoids is increased. Scale bar, 100μm. (G) Quantification of (F), Alcian blue-positive (AB+) cells over area (arbitrary units) is shown for DMSO and HX531-treated organoids. n = 10 images for each condition, mean ± SD is shown. ∗p < 0.05 (Mann-Whitney U test).

Figure 3

Inhibition of RXR disrupts differentiation in CRC organoids (A) Bubble plot showing GSEA parameters NES (color) and FDR (size) for cell type-specific gene sets from Wang et al. in AKP and AKPS organoids treated with indicated compounds. All enrichment scores were calculated using corresponding organoid line treated with DMSO as reference. Inhibition of RXR dimerization by HX531 treatment leads to severe reduction of enterocyte-specific genes in AKP and a modest reduction of these genes in AKPS organoids (n = 3 per condition). (B) Volcano plot showing genes that are differentially expressed in AKP control and HX531-treated organoids. Red dots indicate significantly differential genes (FDR <0.05, n = 740), most prominent differentially expressed (DE) genes are labeled. Among genes severely downregulated upon HX531 addition are several canonical enterocyte marker genes such as ALPI and ANPEP. (C) Bubble plot showing GO terms enriched in DMSO-treated organoids compared with HX531-treated organoids. Fold enrichment is plotted on the x axis; p value is indicated by color, and size of gene set by size. Intestinal absorption and other terms associated with mature enterocytes are significantly enriched in DMSO-treated organoids compared with HX531-treated organoids. (D) GSEA plot showing STAR-positive genes in AKP and AKPS. STAR-positive genes are significantly more enriched in AKPS, indicating that these organoids have more stem cells. (E) Same as (D), but for STAR-negative genes. These genes show enrichment in AKP, indicating that these organoids are more differentiated.

Figure 4

RXR-mediated differentiation disarms colorectal cancer (A) Heatmap displaying GimmeMotifs (maelstrom) Z score using differential ATAC-seq peaks (FDR<0.05). Higher Z score indicates that a motif is more accessible in a given sample. Differentiation-associated motifs such as HNF4A and HNF4G are more accessible in AKP. A motif that can be bound by RXRA (arrow) is also more accessible in AKP. (B) GSEA plot showing enrichment of genes that have an RXRA motif in their promoter in AKP and AKPS. Promoters were defined as 500 bp surrounding transcription start site. The sequence logo of the motif that was used is shown in Figure S3A. These genes are significantly enriched in AKP, indicating that accessibility of the motif corresponds to higher RNA expression. (C) Kaplan-Meier curve showing survival probability of patients with colorectal cancer based expression of selected genes. A g set was used composed of all genes that were significantly downregulated (p value < 0.01) in either AKP or AKPS organoids upon treatment with HX531. Blue line shows the 25% of patients with the highest expression of these genes, termed “Control-like.” Red line shows the 25% of patients with the lowest expression of these genes, termed “HX531-like.” Patients with HX531-like gene expression have significantly lower survival probability than patients with Control-like gene expression.