Figures and data
![](https://prod--epp.elifesciences.org/iiif/2/97279%2Fv1%2Fcontent%2F589235v1_fig1.tif/full/max/0/default.jpg)
DKK2 is required for Lgr5 positive stem cells-driven liver metastasis of colorectal cancer.
(A) Isolated cells from AKP tumor organoids expressing the tdTomato reporter gene were dissociated and injected into the spleen of wild type C57BL/6 mice. 8 weeks after the injection, metastatic tumor growth was measured by in vivo imaging analysis. (B) Representative pictures of in vivo imaging analysis. Ctrl: AKP control organoids transduced with scrambled guide RNA, KO: Dkk2 knockout AKP organoids. (C-D) Statistic analysis of liver metastasis (C) and survival (D). Ctrl (n=10), KO (n=15), n represents the number of mice. (E) Quantitative gene expression analysis of Dkk2, Lgr5 and Hnf4α1 in liver metastasized colon cancer cells. Ctrl (n=11), KO (n=12), n represents the number of isolated cancer nodules. (F) Representative flow cytometry analysis data of c-Src phosphorylation (pSrc) and Lgr5 expression in metastasized colon cancer cells. (G) Statistic analysis of Lgr5 expression and pSrc in (E). The average of mean fluorescence intensity in control samples was set as 1 and relative fluorescence intensity was calculated. (H) Statistic analysis of the percentile of Lgr5 high (Lgr5 and pSrc double positive) cells in metastasized colon cancer in (F). Each symbol represents an individually isolated cancer nodule. n.s = not significant, **P<0.01, ****P<0.0001; two-tailed Welch’s t-test (C, E, G, H). Error bars indicate mean ± s.d. Log-rank test (D). Results are representative of three independent experiments.
![](https://prod--epp.elifesciences.org/iiif/2/97279%2Fv1%2Fcontent%2F589235v1_fig2.tif/full/max/0/default.jpg)
DKK2 is indispensable for the generation of cancer cells with Paneth cell properties in colon cancer organoids.
(A) A schematic diagram of the generation of KO organoids using CRISPR technique and the reconstitution of DKK2 in organoids by recombinant mouse DKK2 protein (rmDKK2) treatment. Lysozyme (LYZ) expression is highlighted for the followings. (B) A volcano plot of RNA-seq analysis comparing KO versus AKP organoids. Paneth cell marker genes are highlighted as blue circles (AKP=3 and KO=5 biological replicates were analyzed). (C) Confocal microscopy analysis of LYZ positive cells in AKP or KO organoids in a time-dependent manner using anti-Lysozyme (LYZ) antibody. (D) Quantitative real time PCR analyses of Lyz1 and Lyz2 in 8 days cultured colon cancer organoids. KO organoids were cultured in the presence of 1 μg/ml of recombinant mouse DKK2 protein. *P<0.05, **P<0.01, ***P<0.001; two-tailed Welch’s t-test. Error bars indicate mean ± s.d. Results are representative of three biological replicates.
![](https://prod--epp.elifesciences.org/iiif/2/97279%2Fv1%2Fcontent%2F589235v1_fig3.tif/full/max/0/default.jpg)
Cancer cells harboring Paneth cell properties were reduced in Dkk2 knockout metastasized colon cancer tissues in mice. Control AKP or KO colon cancer organoids were transplanted via splenic vein as described in Figure 1. 3 weeks after transplantation, mice were sacrificed to analyze metastatic tumor growth in liver. (A) Single cell RNA sequencing analysis (scRNA-seq) of liver metastasized colon cancer tissues. The uniform manifold approximation and projection (UMAP) plot clustered epithelial, endothelial, liver and immune cells in metastasized cancers based on transcriptome analysis. The cancer epithelial cell cluster was sub-clustered to identify cells with Paneth cell properties (cluster 6, Paneth positive). (B) The correlation between Dkk2 and Lgr5 expression in the cluster of epithelial cells by Pearson r test. (C) Ingenuity pathway analysis (IPA)-suggested top 5 canonical pathways of the scRNA-seq data of Lgr5 positive epithelial cells in KO compared to AKP. z-Scores indicate activation or inhibition of the suggested pathways. The significance values for the pathways are calculated by the right-tailed Fisher’s exact test. (D) Box plots show expressions of the genes involved in the glycolysis I pathway in (C). ns = not significant, *<P<0.05, **P<0.01, ***P<0.001, ****P<0.0001; Wilcoxon signed-rank test.
![](https://prod--epp.elifesciences.org/iiif/2/97279%2Fv1%2Fcontent%2F589235v1_fig4.tif/full/max/0/default.jpg)
Identification of colon cancer cells harboring Paneth cell properties in humans.
Published colorectal cancer patient scRNA-seq data (REF) was analyzed to identify the presence of cancer cells with Paneth cell properties. (A) The UMAP plot of total 31 clusters. (B) Normal cells, primary tumor cells and liver metastasized cells (Metastasis) are shown in the UMAP plot clusters. (C-D) Expression levels of lysozyme (LYZ) and Paneth cell module scores are displayed in the UMPA plot. (G) Based on the analysis in (C) and (D), cancer cells harboring Paneth cell properties are indicated as red dots (LYZ+ cancer cells). (E) Lysozyme expression and Paneth cell module scores of single cells in Normal, Primary Tumor and Metastasis samples are presented by dot plots. (F) Dot plots in (E) are overlayed. (H) The percentiles of cancer cells with Paneth cell properties (LYZ+ cancer cells) are shown.
![](https://prod--epp.elifesciences.org/iiif/2/97279%2Fv1%2Fcontent%2F589235v1_fig5.tif/full/max/0/default.jpg)
Colon cancer cells with Paneth cell properties contribute to glycolysis.
Colorectal cancer patient scRNA-seq data were further analyzed by Gene set enrichment analysis (GSEA) and Ingenuity Pathway Analysis (IPA) (A) GSEA of cancer cells with Paneth cell properties (LYZ+ cancer cells) compared to all other cancer cells (LYZ-cancer cells), shown in Fig.4 (G). (B) Representative gene expressions in the hallmark of glycolysis pathway are shown. (C) Upstream regulators suggested by IPA in Primary Tumor and Metastasis are presented. Activation z-scores are indicated in the bar. IPA predicted activation or inhibition of the upstream regulators are colored as red and cyan, respectively.
![](https://prod--epp.elifesciences.org/iiif/2/97279%2Fv1%2Fcontent%2F589235v1_fig6.tif/full/max/0/default.jpg)
DKK2-driven reduction of HNF4α1 protein in murine colon cancer organoids promotes cancer cells with Paneth cell properties.
(A) Ingenuity pathway analysis (IPA)-suggested upstream regulators of the RNA-seq data shown in Figure 2A. Z-scores are presented in the bar. IPA predicted activation or inhibition of the upstream regulators are colored as red and cyan, respectively. (B-C) List of top 5 transcriptional factors (TF) in the motif enrichment analysis of ATAC-seq data comparing normal colonic organoids, AKP and KO cancer organoids. * indicates possible false positive. (D) Chromatin immunoprecipitation sequencing (Chip-seq) analysis of KO organoids using an anti-HNF4A antibody. (E) Quantitative expression of Sox9 in colon cancer organoids in the presence or absence of DKK2 (rmDKK2: recombinant mouse DKK2 protein added in organoid culture). (F) Analysis of Sox9 expression after knockdown HNF4α1 in KO colon cancer organoids (KO HNF4α1 KD). ns = not significant, *P<0.05, **P<0.01, ***P<0.001; two-tailed Welch’s t-test. Error bars indicate mean ± s.d. (G) Representative images of confocal microscopy analysis of Lyz-stained cancer cells with Paneth cell properties in DKK2 KO HNF4α1 KD organoids. (H-I) Control AKP or KO colon cancer organoids were transplanted via splenic vein as described in Figure 1. 3 weeks after transplantation, mice were sacrificed to analyze metastatic tumor growth in liver. Quantification of cancer cells with Paneth cell properties in metastasized tumor nodules by flow cytometry for Lysozyme (LYZ) and HNF4α1. Tumor cells were initially gated by the tdTomato reporter expression. Representative images of flow cytometry are shown (H). Statistic analyses of the percentiles of LYZ positive cells (% of upper left) and HNF4α1 positive cells (% of lower right) in tumor nodules (I). **P<0.01, ****P<0.0001; two-tailed Welch’s t-test. Error bars indicate mean ± s.d. Data are representative of two independent experiments. (J-K) Reduced HNF4A regulation activity with enhanced SOX9 regulation activity in LYZ+ cancer cells in human colorectal cancer scRNA-seq data. Box plots represent the regulon activity of HNF4A and SOX9 in LYZ+ cancer cells. ****P<0.0001; Wilcoxon signed-rank test (J). Z-scaled regulon activities of HNF4A and SOX9 in human colon cancer cells is displayed by heatmap (K).
![](https://prod--epp.elifesciences.org/iiif/2/97279%2Fv1%2Fcontent%2F589235v1_fig7.tif/full/max/0/default.jpg)
The suggested mechanism of DKK2 in the formation of colon cancer cells with Paneth cell properties.
In the absence of DKK2, Sox9 expression is inhibited by HNF4A. Our previous report has shown that DKK2 activates Src followed by degradation of HNF4A protein (28). HNF4A deficiency leads to Sox9 expression in colon cancer cells that induces Paneth cell properties including the expression of lysozymes and defensins. Formation of cancer cells with Paneth cell properties by DKK2 contributes the outgrowth of metastasized colon cancer cells in the liver.