Diet-induced rewiring of the Wnt gene regulatory network connects aberrant splicing to fatty liver and liver cancer in DIAMOND mice

Several preclinical models have been recently developed for metabolic associated fatty liver disease (MAFLD) and associated hepatocellular carcinoma (HCC) but comprehensive analysis of the regulatory and transcriptional landscapes underlying disease in these models are still missing. We investigated the regulatory and transcriptional landscape in fatty livers and liver tumours from DIAMOND mice that faithfully mimic human HCC development in the context of MAFLD. RNA-sequencing and ChIP-sequencing revealed rewiring of the Wnt/β-catenin regulatory network in DIAMOND tumours, as manifested by chromatin remodelling and associated switching in the expression of the canonical TCF/LEF downstream effectors. We identified splicing as a major mechanism leading to constitutive oncogenic activation of β-catenin in a large subset of DIAMOND tumours, a mechanism that is independent on somatic mutations in the locus and that has not been previously shown. Similar splicing events were found in a fraction of human HCC and hepatoblastoma samples.

One tumour not excluding exon 3 ("Inclusion") and one tumour with exon 3 exclusion transcripts ("Exclusion") from RNA-seq data were used as controls.b) Alignment of Sanger-sequenced Ctnnb1 transcripts to genomic region.Normal Ctnnb1 transcripts for 1 RD control liver (green) and 1 FL tissue (blue) are displayed in upper two rows, and exon 3 exclusion transcripts (orange) are displayed below.c) Alignment of genomic sequence in the Ctnnb1 exon 3 region for 1 RD liver (green), 1 FL tissue (blue) and the Ctnnb1 exon 3 exclusion tumour tissue (orange) from b).Note that only two out of 8 exon 3 exclusion tumours have genomic deletions in the region (T7 and T8).For these, both the wild type allele and the deletion are depicted.

Inventory of supplementary tables provided as excel-files
Table S1.Differentially expressed genes between RD, FL and HCC samples with corresponding enriched GO and KEGG terms.Genes were categorised into early, gradual, tumour-specific, fatty liver-specific, and switching.
Table S2.Differentially expressed Wnt-related genes between RD, FL and HCC samples and list of enriched Wnt-related GO term for each gene category.Subset of Table S1 only displaying genes, GO and KEGG terms related to Wnt-signalling.
Table S3.Differential H3K27ac regions in RD, FL and HCC samples and GO and KEGG term enrichment for proximal genes.Regions were categorised into early, gradual, tumour-specific, fatty liver-specific, and switching.
Table S4.Differential H3K27me3 regions in RD, FL and HCC samples and GO and KEGG term enrichment for proximal genes.Regions were categorised into early, gradual, tumourspecific, fatty liver-specific, and switching.
Table S5.Differential Wnt-related H3K27ac regions in RD, FL and HCC samples, with GO and KEGG term enrichment for proximal genes.Subset of Table S3 only displaying genes, GO and KEGG terms related to Wnt-signalling.
Table S6.Differential Wnt-related H3K27me3 regions in RD, FL and HCC samples, and significantly enriched Wnt-related GO and KEGG terms.Subset of Table S4 only displaying genes, GO and KEGG terms related to Wnt-signalling.
Table S7.Complete list of differential alternative splice events between RD, FL and HCC with associated GO and KEGG terms.
Table S8.Table of Wnt-related differential alternative splice events between RD, FL and HCC together with significantly enriched Wnt-related GO terms for each comparison.Subset of Table S7 only displaying genes, GO and KEGG terms related to Wnt-signalling.

Figure S1 .
Figure S1.Obesity and insulin resistance in DIAMOND mice.a) Body weight change over time in mice fed RD (Control mice, green line) and mice fed WD (DIAMOND mice, orange line).Mean ± SD are indicated.Difference in body mass is statistically significant at all time points from T=7days.For simplification only 5 time points are depicted in the graph.*p< 0.05, ***p<0.001.b) Body mass composition, c) fasted blood glucose (left), and fasted insulin (right) at start of the experiment (T=0) and after 42 weeks of diet (T=42w) for controls (green circles) and DIAMOND (orange squares).Individual data points, mean ± SD are indicated.****p<0.0001,ns = not significant (Student´s t-test).Blood glucose levels during d) Insulin tolerance test (ITT), e) Glucose Tolerance Test (GTT), and f) insulin plasma profile during Glucose-Stimulate Insulin Secretion test (GSIS) for controls (green) and DIAMOND (orange).Mean ± SD are indicated.*adj.p<0.05, **adj.p<0.01,ns = not significant (Student´s t-test).

Figure S4 .
Figure S4.DIAMOND tumours display Ctnnb1 exon 3 exclusion independent on genomic mutations in the region.a) Agarose gel electrophoresis of PCR amplified cDNA from RD liver, FL and HCC tissue.Arrows indicate exon 3 exclusion amplicon.One tumour not excluding exon 3 ("Inclusion") and one tumour with exon 3 exclusion transcripts ("Exclusion") from RNA-seq data were used as controls.b) Alignment of Sanger-sequenced Ctnnb1 transcripts to genomic region.Normal Ctnnb1 transcripts for 1 RD control liver (green) and 1 FL tissue (blue) are displayed in upper two rows, and exon 3 exclusion transcripts (orange) are displayed below.c) Alignment of genomic sequence in the Ctnnb1 exon 3 region for 1 RD liver (green), 1 FL tissue (blue) and the Ctnnb1 exon 3 exclusion tumour tissue (orange) from b).Note that only two out of 8 exon 3 exclusion tumours have genomic deletions in the region (T7 and T8).For these, both the wild type allele and the deletion are depicted.

Figure S5 .
Figure S5.Uncropped images of gels for agarose gel electrophoresis in Figure S4.