Smad4 and TGFβ1 dependent gene expression signatures in conditional intestinal adenoma, organoids and colorectal cancer

Abstract

TGF-β ligands suppress growth yet can paradoxically and potently promote cancer invasion and metastasis depending on downstream pathway mutational context, such as loss of Mothers against decapentaplegic homolog 4 (Smad4). Here, we characterised phenotypes and associated gene expression signatures in conditional murine intestinal adenoma with and without Smad4. Conditional Lgr5-CreER^T2 activation in Apc^fl/flSmad4^fl/fl mice resulted in homozygote floxed alleles (Apc^Δ/ΔSmad4^Δ/Δ) and adenoma formation. The adenoma phenotype was discordant, with reduced small intestinal adenoma burden yet development of large non-metastatic caecal adenoma with nuclear localisation of phospho-Smad2/3. Derived Apc^Δ/ΔSmad4^Δ/Δ adenoma organoids resisted TGF-β1 dose dependent growth arrest and cell death (IC₅₀ 534 pM) compared to Apc^Δ/ΔSmad4^+/+ (IC₅₀ 24 pM). TGF-β1 (390 pM) altered adenoma bulk mRNA expression most significantly for Id1^low and Spp1^high in Apc^Δ/ΔSmad4^Δ/Δ. Single cell RNAseq of caecal adenoma identified expansion of Lgr5^low, Pak3^high and Id1^low progenitor populations in Apc^Δ/ΔSmad4^Δ/Δ. Of the 76 Smad4 and TGF-β1 dependent genes identified in Apc^fl/flSmad4^fl/fl adenoma organoids, only 7 human equivalent genes were differentially expressed in SMAD4 mutated colorectal cancer (TCGA cohorts), including ID1^low. SMAD4^low, ID1^low SPP1^high and PAK3^high all correlated with poorer survival. Murine adenoma identified Smad4 dependent gene expression signatures that require further evaluation as functional biomarker classifiers of SMAD4 mutated cancer subtypes.

Keywords: Adenoma; Apc; Biomarkers; Colorectal cancer; Gene expression; Id1; Intestine; Pak3; Smad4; Spp1; TGF-β1.

Fig. 1

Adenoma growth following conditional disruption of Apc and Smad4 loxP alleles using Lgr5-CreER^T2. a Breeding strategy for experimental mice with combined loxP conditional alleles with Lgr5-CreER^T2 prior to 4-OHT injection in adult mice. Chromosome 18 and the location of the centromere (black dot) are shown for the location of both Apc (18qB1) and Smad4 (18qE2). Kaplan–Meier survival of Apc^fl/flSmad4^+/+, Apc^fl/flSmad4^fl/+ and Apc^fl/flSmad4^fl/fl mice post-Cre induction. Comparison between Apc^fl/flSmad4^+/+ and Apc^fl/flSmad4^fl/fl genotypes, *p = 0.011, Log-rank test. Caecal enlargement (arrow) from Apc^fl/flSmad4^fl/fl mice. Bar 1 cm. b Caecum (adenoma) weight, small intestine and colon total adenoma per mouse from Apc^fl/flSmad4^+/+ (n = 9), Apc^fl/flSmad4^fl/+ (n = 17) and Apc^fl/flSmad4^fl/fl (n = 13). Horizontal bars; mean. One-way ANOVA with Bonferroni’s multiple comparison test. ***p < 0.001. c Haematoxylin and Eosin (H&E) sections and anti-Smad4 antibody immunohistochemistry localisation in Apc^fl/flSmad4^+/+ and Apc^fl/flSmad4^fl/fl caecal adenoma. Arrows show crypt abscesses, upper panel. Smad4 localisation is reduced in adenoma and confined to adjacent normal colonic crypt cells in Apc^fl/flSmad4^fl/fl mice. Scale bar 100 µm. d Western blot of adenoma (T) and normal tissue (N) from Apc^fl/flSmad4^+/+ and Apc^fl/flSmad4f.^/fl with anti-phospho-Smad2 (p-Smad2) and anti-phospho-Smad1/5/8 antibodies. Densitometric analysis for p-Smad2 represented as ratio to β-actin (for Apc^fl/flSmad4^+/+ n = 3 and for Apc^fl/flSmad4^fl/fl n = 4 experimental replicates). Significant increase in p-Smad2 in adenoma (T) compared to normal tissues (N) in Apc^fl/flSmad4^fl/fl mice. (See also, Supplementary Information Western Blots Fig. 1d). ***p < 0.001. Non-parametric one-way ANOVA with Dunn’s multiple comparison test. Mean ± S.D.

Fig. 2

TGF-β1 and Smad4 dependent signalling in intestinal adenoma organoids. a Expression of YPF reporter, E-cadherin localisation and DNA (DAPI) in Apc^Δ/ΔSmad4^+/+ and Apc^Δ/ΔSmad4^Δ/Δ adenoma organoids with and without exposure to TGF-β1 (390 pM). b TGF-β1 (range: 0.039- 3900 pM) induced growth inhibition of adenoma organoids quantified by alamarBlue (measured at 24 h. Data normalised to untreated control (100%) (n = 4 experimental replicates). Growth inhibition of 390 pM TGF-β1 rescued by pre-treatment with SB-505124 (ALK4,5,7 inhibitor) for 2 h (1 mM). Error bars ± S.D. c Representative immunoblots of Smad4, total R-Smads and p-R-Smads after TGF-β1 (2 h) from Apc^Δ/ΔSmad4^+/+ and Apc^Δ/ΔSmad4^Δ/Δ adenoma organoids with and without SB-505124. Graphs of p-Smad2/ Smad2 and p-Smad3/ Smad3 densitometry relative to β-actin (n = 3 experimental replicates). Non-parametric one way-ANOVA with Dunn’s multiple comparison. (See also, Supplementary Information Western Blots Fig. 2c). d Confocal immunofluorescence images of Smad2 and Smad3 nuclear localisation (DAPI) following TGF-β1 (390 pM) exposure and SB-505124 inhibition. Scale bar 50 µm. Associated Immunoblots of p-Smad2 at different time points (h) following TGF-β1 (390 pM), densitometry relative to β-actin. (See also, Supplementary Information Western Blots Fig. 2d). e Bright field confocal images of organoids 24–72 h following TGF-β1 exposure. Red arrows intact spherical (circular) organoid morphology, yellow arrows show shape change, red dotted line indicates organoids with marked disruption, cell spreading and cellular transformation. Quantification shows relative preservation of spheroid (circular) organoids in Apc^Δ/ΔSmad4^Δ/Δ. One-way ANOVA with Sidak’s multiple comparison test ****p < 0.0001. f Suppression of organoid proliferation following TGF-β1 treatment assessed with EdU incubation for 2 h. Confocal images of EdU incorporation (green) in nuclei (DAPI- blue) from Apc^Δ/ΔSmad4^+/+ and Apc^Δ/ΔSmad4^Δ/Δ and quantification using ImageJ (n = 10 organoids). One-way ANOVA with Sidak’s multiple comparison test *** p < 0.001, ****, p < 0.0001, respectively. g Cleaved Caspase 3 (CC3) activation following 24–72 h of TGF-β1 (390 pM). CC3 (arrowheads) labelling in confocal images supported by immunoblot labelling (18 h). Arrows indicate margin of organoids. (See also, Supplementary Information Western Blots Fig. 1g). Scale bar 50 µm.

Fig. 3

TGF-β1 and Smad4 dependent bulk RNA expression signatures in adenoma organoids. a Adenoma organoids derived from 3 mice per genotype of Apc^Δ/ΔSmad4^+/+ and Apc^Δ/ΔSmad4^Δ/Δ were exposed to TGF-β1 (390 pM) (0, 1 and 12 h, eight wells per time point), were then pooled for paired-end bulk RNA-Seq (total of 18 RNA-Seq samples). b Principal components of transcriptional clusters according to genotype and time (vst-transformed). c Venn diagram classifies TGF-β1 and Smad4 dependent down-regulated and upregulated genes. Threshold utilised 1.5 log2FC and adjusted p-value < 0.05. d Volcano plots of differentially expressed genes comparing time and genotype (^+/+ = Apc^Δ/ΔSmad4^+/+, ^Δ/Δ = Apc^Δ/ΔSmad4^Δ/Δ). In all volcano plots, genes with -log(p-adj) = 30 are shown. Cut-off values for log2FC (1.5) and adjusted-p-value (1e-3) are marked with dashed lines. Apc^Δ/ΔSmad4^Δ/Δ DEGs are much reduced in number following TGF-β1 treatment. e Enriched gene modules of DEG using the hallmark gene set enrichment (GSEA with tmod package). Comparisons shown in each column. Red and blue indicate the proportion of genes in a module that are either upregulated or downregulated, respectively. Width of each box relates to its effect size, where lighter, less saturated colours indicate lower significance with p-values. f Heatmap showing time-dependent changes in expression (log2FC) for the 50 genes with the lowest adjusted p-values (Likelihood ration test) and g pathway enrichment for these genes are shown generated using Metascape. h Summary of gene specific expression comparing Apc^Δ/ΔSmad4^+/+ and Apc^Δ/ΔSmad4^Δ/Δ with time post TGF-β1 exposure. Statistical comparison performed using the Wald test in Deseq2, followed by Benjamini and Hochberg multiple hypothesis correction.

Fig. 4

Smad4 dependent adenoma cell populations using single cell RNA-seq from in vivo caecal adenoma. a Pooled caecum adenoma cells derived from different mice, Apc^fl/flSmad4^+/+ n = 3 and Apc^fl/flSmad4^fl/fl n = 5, were subject to scRNA-Seq. Cluster trees show the resolution from 0 to 1.2 with the EClpr and ECepr clusters identified at resolution 2.2. b Heatmap of the scaled expression of the top 5 genes identified in each cluster (genes selected based on the biggest difference between the cluster compared to all clusters). Lgr5 ICS = Lgr5 Intestinal stem cell; EC = enterocyte; EClpr = enterocyte late precursor; ECepr = enterocyte early precursor; TA = transit amplifying; CAF = cancer associated fibroblast; Secretory Pr = secretory precursor; hi = high referring to specific genes. c UMAP plot of the cell-type clusters from Apc^fl/flSmad4^+/+ (9674, n = 3) and Apc^fl/flSmad4^fl/fl (11,212, n = 5) based on the expression of known marker genes. d Bar graph; frequencies of clusters of the single cells in Apc^fl/flSmad4^+/+ and Apc^fl/flSmad4^fl/fl adenoma. Point range plot; the confidence interval for the absolute log2FC (obs_log2FC) for the different identified cell types. Significant results are labelled in red. Note enrichment of Pak3^high, Slc14a1^high, CAF and neutrophil populations and the reduction in ECepr and EClpr populations in Apc^fl/flSmad4^fl/fl adenoma. Permutation test with FDR adjustment for p-value. e UMAP plots comparing genotypes for single cell expression of ID1 (suppression in all cell types in Apc^fl/flSmad4^fl/fl), Spp1 (monocytes and macrophage expression) and Pak3 (increased expression in progenitor enterocyte population of Apc^fl/flSmad4^fl/fl). f Violin plots of pseudobulk RNA expression of Id1, Spp1 and Pak3. Statistical analysis using MAST with Bonferroni adjustment for p-value.

Fig. 5

SMAD4 pathogenic variant dependent expression biomarkers for the Smad4-TGF-β gene-pathway in human colorectal cancer (TCGA). a Flow diagram of TCGA bulk somatic DNA and RNA NGS sequencing data (COAD = colon adenocarcinoma, READ = rectal adenocarcinoma) with known WNT pathway driver variants (n = 326). Pathogenic variants for TGF-β, activin or BMP receptors and Smad 2/3/4/5/7/9 were identified (n = 135) from wild-type cases (n = 191). Sub-groups (G1-4) were then assembled with or without co-existing common pathway pathogenic variants in either RTK-RAS, TP53 or PI3K pathways. b The human equivalents of the 76 significantly differentially expressed genes obtained from the comparison of bulk mRNA expression in the mouse adenoma were then compared in the human TCGA mRNA data set. Volcano plot of median log₂ fold-change against adjusted p-value (-Log10) using Mann–Whitney U test (non-normal distribution using Shapiro–Wilk test) identified 7 human genes (in red) significantly different in expression between Smad4 wild-type (n = 191) and Smad4 pathogenic variants (n = 44) [BCAS1, CACNB2, CREB3L1, ID1, ID3, RASGRF2 and SLC30A2]. c, d Comparison between sub-groups with and without context dependent variants G1-G4 are shown as heat maps only for the significantly differentially expressed genes (median RSEM expression log₁₀). Non-parametric Kruskal–Wallis test (FDR adjustment p < 0.05) was followed by Dunn’s test and Holm adjustment for multiple comparisons. 26 gene expression-context pairs were identified to be upregulated in the TCGA cohort (c) and 15 gene expression-context pairs were significantly down regulated (d). Spp1 and Pak3 median expression are shown separately but were non-significant between groups (right heatmap). Of the comparison of the increased expression of CREB3L1 and BCAS1 and repression of ID1 and ID3, only ID1 met the significance criteria.

Fig. 6

SMAD4, ID1, SPP1 and PAK3 expression biomarkers for the Smad4-TGF-β gene-pathway and human colorectal cancer (TCGA) survival. a Comparison of ID1, SPP1 and PAK3 mRNA expression in cases with somatic pathogenic variants in either SMAD4 alone, SMAD4 with other genes in the TGF-β/BMP pathway (TGF-β SMAD4 mutated), single gene variants in either the TGF-β or BMP pathway (TGF-β or BMP mutated), more than one variant in TGF-β /BMP (TGF-β and BMP mutated) or variants of BMP pathway genes alone (BMP mutated). Statistical analysis was performed using Kruskal–Wallis test (p < 0.0001) followed by Dunn test multiple comparison test adjusted by Holm test (shown in the plot). Note the significant repression of ID1 and increase in SPP1 in SMAD4 mutated cases. b Kaplan–Meier survival curves showing survival differences between high and low gene expression of SMAD4, ID1, SPP1 and PAK3 (left). Cut points were determined using the maximally selected rank statistics (maxstat) method and are shown for each gene (right). Statistical significance between the High vs low expression groups determined with the Log-rank test (p values shown).