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

Surakhy, Mirvat; Matheson, Julia; Barnes, David J.; Carter, Emma J.; Hughes, Jennifer; Bühnemann, Claudia; Sanegre, Sabina; Morreau, Hans; Metz, Paul; Imianowski, Charlotte J.; Hassan, Andrew Bassim

doi:10.1038/s41598-025-00908-4

Download PDF

Article
Open access
Published: 10 May 2025

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

Mirvat Surakhy¹,
Julia Matheson¹,
David J. Barnes^1,3,
Emma J. Carter¹,
Jennifer Hughes¹,
Claudia Bühnemann¹,
Sabina Sanegre¹,
Hans Morreau²,
Paul Metz¹,
Charlotte J. Imianowski¹ &
…
Andrew Bassim Hassan¹

Scientific Reports volume 15, Article number: 16330 (2025) Cite this article

2866 Accesses
1 Citations
Metrics details

Subjects

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.

SMAD2/3 mediate oncogenic effects of TGF-β in the absence of SMAD4

Article Open access 07 October 2022

SMAD4 induces opposite effects on metastatic growth from pancreatic tumors depending on the organ of residence

Article Open access 25 September 2025

LHPP suppresses colorectal cancer cell migration and invasion in vitro and in vivo by inhibiting Smad3 phosphorylation in the TGF-β pathway

Article Open access 04 October 2021

Introduction

Mutation of the downstream TGF-β-BMP ligand-dependent signalling pathway can result in context-dependent and paradoxical gain of function (GOF) of the cancer phenotype, i.e. the ligand activation of the normally functioning signalling pathway exerts a tumour suppressive effect, yet following downstream pathway mutation, ligand activation can then result in ‘oncogenic’ GOF activation, epithelial-mesenchymal transition (EMT) and potent promotion of the invasive and metastatic cancer phenotype¹. It is well established that human cancer associated variants (mutations) of the TGF-β-BMP-SMAD4 pathway are correlated with poorer prognosis in enriched sub-types of intestinal cancers such colorectal and pancreatic adenocarcinoma². Despite the incorporation of whole genome sequencing, gene level RNA sequencing and proteomic biomarkers into supervised machine learning prognostic classifiers in colorectal cancer, there remains a paucity of pathway-specific gene expression biomarkers more closely linked to molecular mechanisms and indicative of pathway activity^3,4. This background provided the impetus for further evaluation of the critically important TGF-β-BMP-SMAD4 pathway that when disrupted, is associated with the worse prognostic sub-types of intestinal cancer. Here, we used murine conditional Apc dependent intestinal models to generate in vivo and in vitro adenoma with and without Smad4 pathway modification to prospectively identify phenotypes and context specific gene expression.

The TGF-β1 ligand acts as a dimer to bring together specific heterodimeric cell surface type I and II serine/threonine kinase receptors to activate signals by phosphorylation of R-Smad transcription factors. For example, TGF-β1 binds to TGFBR1 and TGFBR2 to signal through p-Smad2/3, but can also bind to activin receptor type 1A (ACVR1) to transduce signal through p-Smad1/5^5,6. Following receptor activation, phospho-R-Smads form either heterotrimeric or heterodimeric complexes with Smad4, the co-Smad that continuously shuttles between the cytoplasm and nucleus^5,7. Smad4-R-Smad complexes bind with co-factors and DNA to co-ordinate gene expression⁶. Chromatin regions can alter quickly (within 10 min) following TGF-β1 stimulation, with assembled Smad complexes binding to cis-regulatory elements with AP-1 footprints^8,9. Smad4 and R-Smad proteins consist of two highly conserved globular domains, named Mad homology domains, both at the N-terminus (MH1) and C-terminus (MH2) separated by a less well-conserved serine and proline-rich linker region critical for protein–protein interactions and stability^6,9,10. The MH1 domain is involved in DNA binding through a conserved identical hairpin structure present in all members except Smad2^5,11,12. The majority of SMAD4 loss of function (LOF) cancer somatic mutations are missense, although nonsense, splice site, frameshift, and in-frame insertion/deletion mutations exist. In terms of protein localisation, frequent hot spot mutations occur within the conserved R-SMAD binding surface of the MH2 domain and in the MH1 domain along the DNA-binding interface¹³.

The specialised intestinal epithelium is comprised of crypt-villus functional cell units that provide an ideal experimental system for the investigation of specific components of the TGF-β-BMP-SMAD4 signalling pathway¹⁴. Differentiated cells in the crypt and villous comprise of four epithelial lineages: enterocytes, goblet, entero-endocrine and Paneth cells¹⁴. Leucine-rich repeat containing G-protein-coupled receptor (Lgr5) positive crypt base columnar cells (CBC) are stem cells that interact with Paneth cells to maintain the stem cell niche through Wnt activation and β-catenin stability^15,16. Lgr5 positive CBC cells divide, renew and create transit amplifying cells that undergo 4 to 5 cycles of cell division before differentiating as they ascend the villus. By using a Lgr5 driven Cre recombinase, genes with associated loxP sites can be randomly disrupted in CBCs once the Cre is activated. BMP signalling is active at the crypt-villous border towards the villous tip and is controlled by a ligand gradient to promote differentiation towards secretory and enteroendocrine cells^17,18,19. BMP concentrations are regulated in the intestinal crypts by several different BMP antagonists, such as Gremlin, Noggin and Chordin-like1²⁰. TGF-β type I/II receptors and ligands are present and expressed in the muscularis, lamina propria and the differentiated compartment of both the small intestine and colon²¹. TGF-β is also frequently present in the intestinal tumour microenvironment and is released either by cancer cells, tumour stromal cell or immune cells²². Intestinal adenoma can be conditionally generated when Apc loss of function no longer regulates the degradation of β-catenin, resulting in constitutive activation of transcription by β-catenin/Tcf4 and adenoma formation. Subsequent mutational events in humans are selected for during the progression of adenoma to invasive and metastatic carcinoma, e.g. RAS/MAPK GOF, PTEN and TP53 LOF, and later acquired LOF mutation of the TGF-β-BMP ligand activated receptor pathway, such a mutation of receptors and Smad2/3/4²³.

Here, we first set out to evaluate loss of the TGF-β pathway signal transduction regulator, Mothers against decapentaplegic homolog 4 (Smad4), in vivo and in vitro, with respect to a murine intestinal adenoma phenotype and dependent gene expression. In view of the complexity of the interacting pathways that can alter TGF-β1-Smad4 induced phenotypic effects, especially in immortalised human cancer cell lines, we sought to better define the contextual impact of Smad4 LOF to specifically evaluate the Smad4 and TGF-β1 ligand dependency on gene expression in murine models and to translate these findings with respect to human intestinal cancer. The conditional Apc^fl/fl model of intestinal adenoma using Lgr5CreER^T2 is well characterised, and so we first incorporated loxP dependent (LOF) of Smad4 (Smad4^fl/fl alleles) into this model.

Results

Combined Apc and Smad4 conditional disruption results in discordant adenoma phenotypes

To investigate the adult stage-specific genetic interactions between Apc and Smad4 in the murine intestine, we combined conditional alleles with a Cre driven by the promoter Lgr5 and a ROSA26-loxSTOP reporter allele (Lgr5-EGFP-IRES-CreER^T2 or Lgr5-CreER^T2)²⁴ (Fig. 1a). Heterozygote alleles were inter-crossed (Apc^fl/+Smad4^fl/+Lgr5-CreER^T2) on a C57Bl/6J background (backcrossed to C57Bl/6J for > 10 generations) to generate homozygotes and heterozygote littermates that differ in Smad4 conditional alleles^25,26. Conditional recombination followed tamoxifen injection (4-OHT) in adult animals that were then phenotypically analysed with respect to survival and adenoma phenotypes (Fig. 1, Supplementary Fig. 1). The floxed alleles of Smad4 (Smad4^Δ/Δ) resulted in loss of the MH1 DNA binding domain and the production of a truncated 43 kDa protein which was not capable of translocating to the nucleus (see below, Supplementary Fig. 2). The germline homozygote loxP alleles are denoted Smad4^fl/fl and Apc^fl/fl, and where somatic tissues have been directly tested for the presence of the resulting Cre-induced recombined alleles (floxed) using PCR, were then denoted as Smad4^Δ/Δ and Apc^Δ/Δ. Recombination rates in intestinal epithelial cells were lower overall with the Lgr5 transgene (5–6%), but highly specific for Lgr5 CBC stem cells^15,27. Rapid adenoma development occurred as expected following homozygous Apc^fl/fl loss of function (median 35–40 days post injection 4-OHT) and animals were culled immediately if they exceeded predefined humane endpoints. Comparison of overall survival (Kaplan–Meier) between Apc^fl/flSmad4^fl/flLgr5-CreER^T2 and Apc^fl/flSmad4^fl/+Lgr5-CreER^T2 with Apc^fl/flSmad4^+/+Lgr5-CreER^T2 controls surprisingly showed a modest but significant improvement in survival (p = 0.011, Fig. 1a). These observations suggest that introducing Smad4 LOF in Apc LOF intestinal adenoma does not result in overwhelming tumour progression and metastasis. However, analysis of intestinal adenoma distribution between the small intestine and colon showed a marked discordance in growth phenotypes in Apc^fl/flSmad4^fl/flLgr5-CreER^T2. A significant reduction in small intestinal adenoma burden (number of adenoma per mouse) was observed concurrent with paradoxical growth promotion of adenoma in the caecum (Fig. 1b). As the distribution of adenoma on a C57BL/6 background is predominantly in the small intestine rather than the colon and caecum, the multiple small intestinal adenoma would be expected to promote anaemia, small intestinal obstruction and decreased survival^27,28. The significantly reduced small intestinal adenoma burden in Apc^fl/flSmad4^fl/flLgr5-CreER^T2 compared to controls may have improved overall survival if this was the only phenotypic effect (Fig. 1a, b). Examination of the caecum and large intestine revealed markedly enlarged coalesced adenoma with associated mucus cysts where numerical analysis of single and separate adenoma was not possible, hence we reported caecal weight as a surrogate (Fig. 1b). The adenoma burden in heterozygote Apc^fl/flSmad4^+/flLgr5-CreER^T2 compared to controls was statistically non-significant and does not support either Smad4 haplo-insufficiency or a dominant negative effect of Smad4^fl/+. Histological grading of small intestinal adenoma showed low-grade dysplasia in all genotypes, whereas more advanced high grade dysplasia with cytonuclear and architectural changes including foci of microinvasion were more frequently observed in regions of Apc^fl/flSmad4^fl/flLgr5-CreER^T2 in the caecum (Supplementary Fig. 1, author HM pathological review). Further examination of H&E stained small intestinal adenoma tissue sections Apc^fl/flSmad4^+/+ mice had larger sessile adenomatosis appearance compared to Apc^fl/flSmad4^fl/fl that had mainly tubular adenomatous appearance. Crypt cystic regions (dirty necrosis) were detected in both genotypes but no muscularis tissue invasion, lymph node involvement and metastasis were observed within the relatively short median overall survival (Fig. 1c, Supplementary Fig. 1). The efficiency of Cre-mediated Smad4 allele deletion was examined by immunohistochemistry (IHC, Fig. 1c). Smad4 protein was clearly detected in Apc^Δ/ΔSmad4^+/+ adenoma and adjacent tissue as expected. Adenoma from Apc^Δ/Δ Smad4^Δ/Δ did not label strongly with anti-Smad4 antibodies, consistent with adenoma-specific disruption in both the small and large intestine (Fig. 1c). The presence of active TGF-β (BMP) signaling was supported by p-Smad2 and p-Smad1/5/8 immunoblots (Fig. 1d). Both p-Smad2 and p-Smad 1/5/8 labelling were detected in adenoma at higher levels from Apc^Δ/ΔSmad4^Δ/Δ compared to normal tissue and littermate control adenoma, consistent with either TGF-β alone or with BMP ligand local context specific activation in combination (p < 0.001, Fig. 1d).

Combined Apc and Smad4 disruption modifies growth of adenoma organoids

In order to generate an organoid culture, we isolated Apc^Δ/ΔSmad4^Δ/Δ (Smad4^Δ/Δ) and Apc^Δ/ΔSmad4^+/+ (Smad4^+/+) caecal adenoma and generated in vitro cultures in Matrigel using published conditions^24,29. Using a ROSA26-LSL-YFP reporter line bred with Apc^fl/flSmad4^fl/flLgr5-CreER^T2, we confirmed that 40HT exposure resulted in uniform YFP expression in adenoma organoid culture that was also TGF-β1 independent (Fig. 2a). In culture, Smad4^Δ/Δ adenomas appeared smaller and grew slower compared to Smad4^+/+ adenomas, as measured by both area and the alamarBlue (AB) growth assay (Supplementary Fig. 3). Noggin is a BMP pathway inhibitor that normally prevents enterocyte differentiation and is required for crypt culture maintenance^23,24,29. Noggin addition had no effect on adenoma growth (Supplementary Fig. 3). SB-505124, a TGF-βR1 small molecule inhibitor also had no significant inhibitory effect on Smad4^+/+ and Smad4^Δ/Δ adenoma growth (1 μM) without the exogenous addition of TGF-β1 (Supplementary Fig. 3). These data suggest that the adenoma culture conditions appeared to be independent of endogenous TGF-β supply.

To exclude whether there may have been a genotype dependent difference in the floxing efficiency between Apc and Smad4 that may have accounted for the two-fold differences in growth in culture, we also generated adenoma with an alternative conditional Cre (Villin-CreER^T2). Here, cultured small intestinal organoids from healthy animals bred with non-floxed alleles, Apc^fl/flSmad4^+/+Villin-CreER^T2Rosa26-YFP and Apc^fl/flSmad4^fl/flVillin-CreER^T2Rosa26-YFP genotypes, were cultured in organoid media containing R-spondin-1, Noggin and EGF. Intestinal organoids that formed a normal crypt-villous morphology were exposed to 4-OHT followed by withdrawal of R-spondin-1. This resulted in adenoma organoid formation in both Apc^Δ/ΔSmad4^+/+ and Apc^Δ/ΔSmad4^Δ/Δ (Supplementary Fig. 4). Allele recombination was confirmed by PCR product of 259 bp and 512 bp for Apc and Smad4 alleles, respectively. Both the Apc^fl/fl and Smad4^fl/fl alleles recombined at the same time and rate following 4-OHT treatment, and were detected as early as day 2 post treatment (Supplementary Fig. 4). Thus, floxing efficiency appeared similar between alleles.

Deletion of MH1 domain of Smad4 following Cre-mediated recombination

To evaluate the expression of Smad4 mRNA in Smad4^Δ/Δ and Smad4^+/+ in vitro, we used bulk RNA-seq of adenoma organoids and discovered two different Smad4 mRNA transcripts. Smad4 gene tracking with Cuffdiff revealed transcripts from the intact Smad4 gene, and a second transcript from the Smad4 gene with exon 2 excised following loxP mediated recombination (Supplementary Fig. 2). Immunoblots confirmed expression of Smad4 full-length protein (61 kDa) in both Smad4^+/+ and Smad4^+/Δ adenomas, with a smaller band of (43 kDa) detected in protein extracts from Smad4^+/Δ and Smad4^Δ/Δ adenomas (Supplementary Fig. 2). In silico analysis revealed an open reading frame in exon 5 of the Smad4 sequence and mapping with ExPASy revealed that exon 2 deletion results in a truncated protein in which the Smad4 MH1 DNA binding domain is selectively deleted. The predicted size of the truncated protein was 43 kDa, comprising the linker and Smad4 MH2 domain (Supplementary Fig. 2). The full-length Smad4 protein in Smad4^+/+ adenoma showed nuclear localisation following exposure of 390 pM TGF-β1 for 2 h. Under the same conditions, no nuclear localisation of Smad4 (43 kDa) was detected in Smad4^Δ/Δ adenomas, consistent with loss of DNA binding function (Supplementary Fig. 2). Taken together, Cre-mediated recombination of the Smad4 gene resulted in selective MH1 deletion and a truncated protein that did not undergo nuclear localisation following TGF-β1 stimulation. These data, and the apparent reduction of detection using the same anti-Smad4 antibody in IHC of Smad4^Δ/Δ adenoma, are consistent with non-DNA binding and presumably soluble truncated SMAD4 protein that depletes from of cells during tissue processing resulting in reduced detection signals.

TGF-β1 dose dependency in Smad4^Δ/Δ adenoma organoids

We next tested the dose and timing effects of exogenous addition of TGF-β1 to adenoma organoid cultures on adenoma phenotypes and the subsequent localisation of R-Smads. TGF-β1 induced dose-dependent growth inhibition of both the Smad4^+/+ and Smad4^Δ/Δ adenomas, but with a 20-fold differential sensitivity (Fig. 2b). Smad4^+/+ adenomas were more sensitive to TGF-β1, with an IC₅₀ of 24 pM and threshold of 0.39 pM. For Smad4^Δ/Δ, the TGF-β1 IC₅₀ was 534 pM with a threshold between 39-390 pM. The decrease in adenoma viability with TGF-β1 dosing was prevented when the adenoma cultures were pre-incubated with SB-505124 (Fig. 2b). For subsequent experiments, a TGF-β1 concentration of 390 pM was used to differentiate the TGF-β1 context-dependent sensitivity effects between Smad4^+/+ (100 fold higher than the threshold) and Smad4^Δ/Δ (at threshold but below IC₅₀) adenomas. TGF-β1 addition (390 pM) resulted in the detection of phosphorylated p-Smad2 and p-Smad3 in both the Smad4^+/+ and Smad4^Δ/Δ adenomas as expected (Fig. 2c). Despite the differential sensitivity, the p-Smad2 kinetics of response to TGF-β1 stimulation were similar between genotypes, with p-Smad2 maximally detected after 1–2 h and associated with nuclear localisation (Fig. 2d). These data also suggest that the increased levels of pSmad2 detected in the in vivo adenoma of were most likely due to increased local TGF-β-BMP ligand supply in Apc^Δ/ΔSmad4^Δ/Δ (Fig. 1d). Within 24 h of TGF-β1 administration, > 90% of the Smad4^+/+ adenoma cells either had either morphological changes of apoptosis with organoid retraction, or changes in shape toward a spindle or mesenchymal phenotype (see Supplementary Movies 1–4). These features appeared delayed to 72 h in Smad4^Δ/Δ adenomas but were still detectable (Fig. 2e). The reduced viability and altered morphology were accompanied by significantly decreased proliferation and increased apoptosis in Smad4^+/+ adenoma relative to Smad4^Δ/Δ adenomas when quantified by EdU and cleaved caspase 3 (CC3) labelling, respectively (Fig. 2f, g).

TGF-β1 and Smad4-dependent Id1 and Spp1 gene expression in adenoma organoids

To assess the transcriptome response to TGF-β1 that might inform differences between Smad4^+/+ and Smad4^Δ/Δ adenoma organoids, bulk paired-end RNA-seq detected 16,143 gene transcripts that we utilised for the analysis of differentially expressed genes (DEG)(Fig. 3a–h). Three separate mice per genotype provided adenoma biological replicates and were harvested at different times t = 0, t = 1 and t = 12 h after exposure to a 390 pM TGF-β1 (Fig. 3a). Principle component analysis (PCA) of gene expression showed a clear separation of the samples based on genotype and duration of TGF-β1 exposure (Fig. 3b). Comparison of DEG at baseline between Smad4^Δ/Δ and Smad4^+/+ at t = 0 (no TGF-β1) showed 111 upregulated genes (log₂FC > 1.5) and 51 downregulated genes (log₂FC < -1.5) (Fig. 3c, Supplementary Table 1). The most important observation was the magnitude of the genotype dependent expression of a group of genes that included Id1 (Inhibitor of DNA binding 1), Ddit4, Serpine1, Skil, Mucins, Myc, JunB and E2f. target genes (Fig. 3c, d, Supplementary Fig. 5). Gene ontology (GO) analysis of biological processes showed enrichment for epithelial cell proliferation and response to TGF-β1 (Supplementary Fig. 5, Supplementary Table 2). Gene set enrichment analysis (GSEA) using hallmark gene sets with “tmod” in Smad4^Δ/Δ showed downregulation of Myc target genes, differential regulation of TGF-β signalling genes, mainly by upregulation, and enrichment of other pathways including IL2-Stat5 and TNF signalling (AUC > 0.5, padj < 0.0001, Fig. 3e, Supplementary Table 2). Importantly, mucins in Smad4^Δ/Δ adenomas were among the DEGs with Muc4, Muc20, Muc2 being upregulated and Muc6 being downregulated (padj p < 0.05, supplementary Table 1, supplementary Fig. 6). Subsequent Mucin 2 immunolabeling also confirmed protein overexpression in Smad4^Δ/Δ adenoma tissue sections (Supplementary Fig. 6).

For subsequent timepoints post-TGF-β1, DEGs were mainly observed in Smad4^+/+ adenoma organoids and included Skil, Junb, Serpine 1, Ddit4 and Id1 (Fig. 3d, Supplementary Fig. 5, Supplementary Table 1). The differential suppression of Id1 in Smad4^Δ/Δ compared to Smad4^+/+ after 1 h achieved a 2.83 log₂FC difference. By 12 h, Id1 maintained suppression (1.2 log2FC, Fig. 3d, Supplementary Table 1). Despite the repression of expression, Id1 protein abundance and immune-localisation did not appear to be significantly altered between adenoma genotypes (Supplementary Fig. 7). We also observed that Spp1 (Secreted phosphoprotein 1 or osteopontin), a monocyte/macrophage and extracellular matrix gene, was the only gene significantly upregulated gene following TGF-β1 exposure (12 h) in Smad4^Δ/Δ (log₂FC > 1.5 threshold, Fig. 3c, d, h, Supplementary Table 3).

Further significant time dependent differences in gene expression occurred by 12 h post-TGF-β1. For example, 35 DEGs were observed in Smad4^Δ/Δ compared to 689 in Smad4^+/+ adenomas at t = 12 h vs t = 0 h (Fig. 3d, Supplementary Table 1). Pathway enrichment analysis using tmod also showed that E2f., G2m, Myc and mTORC targets were more significantly downregulated in Smad4^+/+ adenomas compared to Smad4^Δ/Δ adenomas (Fig. 3e, Supplementary Table 2). Time series analysis using the likelihood ratios showed differences in expression of genes including Fos, Lama3, Aldh1a3, Col7a1 and Doks associated with increased induction by t = 12 h, whereas E2f8, Exo1, Clspn, Kif14 and Kif12 all were decreased in Smad4^+/+ adenomas. Overall, the suppression of cell cycle genes was evident in Smad4^+/+ compared to Smad4^Δ/Δ adenomas, as shown by Metascape overrepresentation (Fig. 3f, g, Supplementary Table 3). Finally, TRRUST (Transcriptional Regulatory Relationship Unravelled by Sentence based Text mining) enrichment analysis implemented in Metascape, also suggested potential regulation by transcription factors, with Smad4-dependent downregulated genes being co-regulated by Sp1, Smad3 and Jun, and upregulated genes by Rbl2, Tp53 and E2f1 (Supplementary Fig. 5). Several gene signatures were therefore determined that were Smad4 and TGF-β1 dependent, with the two genes with the greatest magnitude of differential expression being Id1 (low) and Spp1 (high). A summary of gene specific expression of Id1, Fn1, Myc, Pak3, Sema5a, Spp1 and Slc14a1 are shown with comparison of Apc^Δ/ΔSmad4^+/+ and Apc^Δ/ΔSmad4^Δ/Δ with respect to time post-TGF-β1 exposure (Fig. 3h).

TGF-β1 dependent gene expression and protein correlation

Smad4 LOF did not fundamentally alter the range of TGF-β1 dose dependent phenotypic responses, but did change the overall magnitude and relative gene expression. Whilst TGF-β1 administration in Smad4^+/+ adenomas resulted in a > 4.5 Log₂FC relative increase in expression of Fn1, we did not observe a similar change in Fn1 protein by Western blot (fibronectin 1, Supplementary Fig. 8a). Similar findings were observed for N-cadherin and when N-cadherin was compared relative to β-catenin, even when EMT associated actin reorganisation was evident (Supplementary Fig. 8b, c). In addition, TGF-β1 exposure resulted in a 0.25 log₂FC reduction in Survivin (Birc5) expression after 12 h in Smad4^Δ/Δ compared to a 2.55 log₂FC reduction in Smad4^+/+ adenoma. Anti-survivin antibody western blots and nuclear localisation suggested detectable reduction in protein levels in Smad4^+/+ adenoma by 18 h post- TGF-β1 (Supplementary Fig. 9). The Wnt dependent intestinal stem cell marker Lgr5 was downregulated in Smad4^+/+ following TGF-β1 (12 h) compared to Smad4^Δ/Δ adenoma (2 log₂FC, Supplementary Table 1), also suggesting modification in Wnt pathway activity. E-cadherin and β-catenin proteins remained immunolocalised at the adherens junctions and in nuclei in both adenoma genotypes as expected (Supplementary Fig. 10). Immunoblots of adenomas from both genotypes showed similar multiple molecular weight bands for β-catenin and E-cadherin in the context of disruption of the β-catenin destruction complex by LOF of Apc^Δ/Δ (Supplementary Fig. 10b, c). Gene expression for Myc was not markedly different between genotypes, yet c-Myc protein levels appeared increased in Smad4^Δ/Δ (12–18 h) compared to Smad4^+/+ adenomas (Supplementary Fig. 10c, Supplementary Table 1). Overall, there did not appear to be significant direct correlation between changes in gene expression and protein localisation and abundance between genotypes. This implies that protein abundance, regulated by numerous and variable processes independent of gene expression, may not be a universal and reliable surrogate for mRNA gene expression in this context.

Single cell RNA-seq of Smad4^fl/fl caecal adenoma are enriched in Pak3^Hi epithelial progenitor cells

In view of the differential changes in bulk mRNA expression, we next sought to quantify mRNA expression at the single cell level in the large adenoma directly harvested from the caecum. Epithelial cells (Epcam cluster) consisted of the Lgr5 stem cell (Lgr5 ISC) and seven stem cell-like (SC-L) clusters based on differential expression of Lgr5 and other stem-cell markers such as Lrig1, Smoc2, Hopx as previously defined^30,31,32,33 (Fig. 4a, b. Supplementary Fig. 11, Supplementary Table 4). Differential enrichment in stem cell clusters was observed between the two genotypes, consisting of 61.8% of the total cell population in Smad4^fl/fl adenoma compared to 47.3% in Smad4^+/+ (Fig. 4c, d, Supplementary Table 3). A Pak3^High cluster (Pak3; P21 Rac1/CDC42 activated kinase 3) was the most enriched in Smad4^fl/fl (10.5%) compared to Smad4^+/+ (2.5%) control adenoma. Pathway enrichment analysis for DEGs between Smad4^fl/fl and Smad4^+/+ for the Pak3^high cluster showed an enrichment for the Wnt pathway and pluripotency markers (Supplementary Fig. 12). The same comparison in adenoma organoids between genotypes and all timepoints post TGF-β1 showed a non-significant difference in expression of Pak3, suggesting this was an in vivo adaptation that was not reproduced in organoids (Supplementary Table 1, Fig. 3h). The Slc14a1^Highcluster was also enriched in Smad4^fl/fl (7%) compared to Smad4^+/+ (3.5%), where expression of the stem cell marker Ly6a was enriched (Sac1). Conversely, the Lgr5 ISC cluster was significantly enriched in Smad4^+/+ (5.8%) compared to Smad4^fl/fl adenoma (2.7%) (Fig. 4d). Smad4^fl/fl showed significantly lower proportions of enterocytes (EC) as a consequence (7.8% Smad4^fl/fl vs 23.4% Smad4^+/+) yet Fmnl2^Hgh, Cachd1^High and Slc12a2^High showed similar proportions between genotypes. Note that the SC-L and the Lgr5 ISC clusters originate from the same node when viewed with cluster tree (0.2 resolution) (Fig. 4a). The Tacstd2^Hgh cluster (murine foetal intestinal progenitor gene³⁴) and the Cl^high cluster (Clusterin marker of the revival stem cell³⁵) are SC-L clusters with a similar proportions and originate from the same node (0 at 0.2 resolution). Markers for non-stem cells that lacked stem cell expression markers, included ribosomal gene expression³³ in transient amplifying cells (TA) and the Paneth cells marker Pnliprp1 (pancreatic lipase-related protein 1)³⁰, enteroendocrine precursor Gadd45g³³, Mt3 and Pnliprp2 in secretory precursor cells (secretory pr).

Using both the 1.2 and 2.2 resolution cluster tree, we identified two groups of cells; the enterocyte early precursor (ECepr) expressing Apoa1 Dmbt1, Reg4, and Reg3g, and the enterocyte late precursor (EClpr) expressing Reg3b, Mt1, and Mt2³³. In Smad4^fl/fl, EClpr (0.26% vs 4.1%) and ECepr (1.3% vs 6.5%) were depleted compared to Smad4^+/+ adenoma. The mature enterocyte clusters, EC-1 and EC-2, showed expression of Krt19, Krt8, Lgals3, Lypd8, Emp1, and Krt20 with differences in Mxd1, Cdkn1a and Fam3b expression. Significantly, 2.2% of Smad4^fl/fl were EC-1 cells compared to 5.4% for Smad^+/+ (Fig. 4a, b, c, d). Thus, scRNA-seq of in vivo adenomas identified Pak3^high and Slc14a1^high as genes with increased expression and relatively differentially enriched in progenitor epithelial cells in Smad4^fl/fl caecal adenoma. 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) confirms the distribution changes for Pak3 and Id1 (Fig. 4e). Moreover, comparison of combined scRNAseq data using a pseudo-bulk approach, also confirmed significant downregulation of Id1 and upregulation of Pak3, consistent with the bulk-organoid RNA-Seq (Fig. 3h, Fig. 4f). In addition, Id1 and Sema5a were the uniquely deregulated genes that were common between organoid bulk RNA-seq and the primary pseudo-bulk caecal adenoma scRNA-seq data (Supplementary Fig. 13). For Spp1, the high overall mRNA expression in adenoma derived organoids was not matched in the caecal derived single cell mRNA in epithelial cells, presumably as expression was confined mainly to the monocyte-macrophage lineage not present in organoid cultures (Fig. 4f).

Smad4 dependent ID1, SPP1 and PAK3 expression in human colorectal cancer

Having identified TGF-β1-BMP-Smad4 pathway dependency of Id1^low and Spp1^high gene expression, and Pak3^high expressing cell enrichment in primary caecal intestinal adenoma, we next explored whether any of these functional gene expression biomarkers are also present in human colorectal cancer (Fig. 5, Fig. 6, Supplementary Table 5). Data was derived from 382 individual cases from the human Cancer Genome Atlas project (TCGA; COAD, Colon Adenocarcinoma, READ, Rectal adenocarcinoma) that had both mutational profiles in c-BioPortal and RNA-seq data. We tested whether the co-existing mutational context with and without SMAD4 variants in colorectal cancer altered SMAD4 dependent gene expression and key regulatory gene expression pathways. We first assembled common mutational subgroups of colorectal cancer based on WNT-APC-β-catenin, RAS-MAPK, PI3K, and TP53 pathway variants (Fig. 5a). This test matrix derived four variant based sub-groups named Group 1–4 (Fig. 5a). Firstly, colorectal cancers with confirmed WNT-APC-β-catenin variants were selected and subsequently sub-divided. Group 1 and Group 2 cases were SMAD4 wild-type and Group 3 and 4 cases had SMAD4 pathogenic variants. Groups 1&3, and 2&4 differed depending on the presence or absence of second pathway variants in either the RAS-MAPK, PI3K or TP53 pathway, respectively. We next utilised the 76 DEGs identified in murine adenoma organoids and applied the equivalent genes to the human colorectal RNA-seq data. Volcano plots revealed only 7 of these genes were significantly differentially expressed based on the presence (n = 44) or absence (n = 191) of a SMAD4 pathogenic variant (Fig. 5b). These genes were BCAS1, CACNB2, CREB3L1, ID1, ID3, RASGRF2 and SLC30A2. Further comparison of the subgroups based on these genes along with second pathway variants revealed 26 gene-context pairs that were significantly upregulated and 15 gene-context pairs that were significantly downregulated (Fig. 5c, d). Despite the correlation established for Spp1 and Pak3 in mouse adenoma organoids, these two genes did not appear significantly differentially expressed in human colorectal cancer with respect to SMAD4 variants (Fig. 5c). Importantly, the upregulated expression of CREB3L1 and BCAS1 and downregulated expression of ID1 and ID3 appeared least impacted by the co-existing 2^nd (cancer) pathway variants based on visualisation of the heat maps. However, analysis based on our three predefine criteria (see Methods) of the comparison of the increased expression of CREB3L1 and BCAS1, and repression of ID1 and ID3, only ID1^low met the significance criteria. Analysis of the TCGA cohort with respect to either the presence or absence of any additional TGF-β/BMP pathway variants also showed that ID1 expression was also significantly downregulated with co-existing SMAD4 and other TGF-β pathway downstream variants compared to normal control tissues (Supplementary Table 5).

Compared to non-SMAD4 associated TGF-β/BMP pathway pathogenic variants (e.g. in Type 1 receptors, other SMADs or in the BMP pathway), SMAD4 variants specifically appeared to have the greatest impact on overall ID1 repression (Fig. 6a, p < 0.0001). In the context of all TGF-β/BMP pathway mutations combined, however, SPP1 mRNA did appear to be expressed significantly higher compared to cases without variants in the TGF-β and BMP pathway (p = 0.044, respectively, Dunn test, Fig. 6a). For PAK3, no significant overall expression differences were observed between SMAD4 mutational groups (Fig. 6b). Having established that the Smad4 dependent gene expression signature in mouse adenoma only overalapped with the SMAD4 dependent gene expression in human colorectal cancer with respect to ID1^low, we next evaluated whether any of the expression biomarkers impacted on the survival outcome of colorectal cancer. By using maximum selected rank statistics to define a single cut-point (Fig. 6b), Smad4^low expression (p = 0.0019) and ID1^low expression (p = 0.0062) as well as SPP1^high (p = 0.0072) and PAK3^high (p = 0.012), all appeared to be associated with significantly poorer overall survival in the TCGA cohort (Kaplan–Meier, Fig. 6b). The less significant impact of PAK3^high may reflect the relative low proportion of PAK3^high expressing sub-populations of cells within tumours.

Discussion

The identification of cancer cells that have disruption of the TGF-β-BMP-SMAD4 pathway may have important implications for risk assessment for invasion and metastasis. Validated models are required in order to discover mechanism and pathway specific gene biomarkers². Moreover, validated RNA based molecular prognostic classifiers for colorectal cancer have rapidly developed in line with the now clinically applied breast cancer equivalents. Such bulk RNA based machine learning classifiers have identified several clusters (e.g. CRISB) that include a stromal based signature that correlates with poor prognosis in colorectal cancer^3,36,37. A supervised machine learning CMS1-4 classifier (Consensus Molecular Classifier) has correlated gene level expression signatures with four prognostic groups, where the poor prognostic group CSM4 was correlated with EMT and TGF-β dependency, yet the classifier does not utilise specific RNA expression that are validated readouts of TGF-β-BMP-SMAD4 pathway disruption. With respect to the latter, human colorectal polyp derived organoids have recently been used to discover TGF-β dependent gene sets in human serrated adenomas combined with a BRAF^V600E mutations, where there was an induced mesenchymal phenotype, yet this model did not address specific context of Smad4 loss of function³⁸. Further analysis of classifiers of the human expression data has also newly identified three classes of pathway derived subtypes (PDS1-3) in colorectal cancer with and without KRAS mutations, where CSM4 and CSM1 combine in the PDS2 subtype where TGF-β-BMP-SMAD4 pathway expression profiles appear enriched⁴. These subtypes appeared independent of KRAS driver gene variant classes and phenotypic appearance, and for PDS1 and 2, correlated with RNA expression data from mouse genetically modified models. Despite the likely pathway specific variations, gene-pathway specific gene expression signatures are becoming better understood and being directly applied from mouse model systems to human intestinal cancer classifiers. In this murine system, we sought to evaluate gene-pathway specific gene expression signatures for Smad4.

We discovered that the generation of homozygous Smad4 with Apc LOF alleles was associated with a discordant intestinal phenotype, with adenoma suppression in the small intestine and adenoma progression in the caecum. By deriving organoid cultures from the adenoma, we discovered that the high sensitivity to TGF-β1 induces similar EMT and cell death phenotypes in a dose dependent manner, with a relative tenfold dose resistance to TGF-β1 in Smad4^Δ/Δ adenoma. Following TGF-β1 exposure, we identified genes with extremes of differential expression by genotype, including reduced Id1 and increased Spp1 gene expression as consistent features in Smad4^Δ/Δ murine adenoma organoids. The gene expression changes as consistent with Smad4 independent activity of the R-Smads, in this case this is likely to be mainly Smad3^39,40,41. Single cell RNA sequencing of the primary caecal adenomas also detected enrichment of the Pak3^High expressing progenitor epithelial cell population in Smad4^Δ/Δ but no changes in bulk expression in adenoma organoids. We assume that these differentially expressed genes are the main underlying differences that account for functional discordant progression of adenoma, for example Id1^Low, although this will need to be directly tested in future experiments.

Initial models of Smad4 LOF in the intestine have shown Smad4 dependent progression phenotypes, that have been further evaluated using conditional disruption⁴². For example, Smad4^+/- mice spontaneously developed juvenile polyposis (JP) at 1.5 years, and invasive adenocarcinoma when crossed with Apc^∆716/+ mice^42,43,44,45. In addition, an Apc^1638N/+Smad4^fl/flK19-CreERT2 model appeared to show an increased small intestinal and colonic adenoma burden⁴⁶. SMAD4 LOF also blocked differentiation during tumour progression in an orthotopic organoid transplantation model mimicking an intestinal adenoma-carcinoma transition⁴⁷. Additional murine models incorporating disruption of the TGF-β-BMP pathway (Tgfbr2, Smad3) with other contextual drivers (e.g. Apc, Kras, inflammation) have also shown increased intestinal invasion and metastasis that are strongly pathway dependent^42,48,49. Selective disruption of Apc and Smad4 in the caecal and colonic epithelium using a carbonic anhydrase CreER also resulted in pathway dependent invasive adenoma⁵⁰. CRISPR/Cas9-mediated genome editing of human intestinal epithelial organoid cultures introduced mutations in APC, TP53, KRAS and SMAD4 which, when combined and selected, developed aggressive invasive carcinomas²³. Moreover, these findings may explain the selection of deletions of chromosome arms 18p and 18q (including SMAD4) in regions of human adenoma and 66% of invasive colorectal cancers^51,52. The generation of homozygous floxed alleles in this conditional model results in rapid adenoma formation within 30 days, as it avoids stochastic LOH of the WT allele in the Apc truncated variant alleles of previous models. As a result, temporal development of further invasion and metastasis are restricted compared to models that are either dependent on time for LOH of the Apc allele or have lower recombination events. In our case, strict use of humane endpoints because of anaemia and intestinal obstruction are likely to happen before any prospect of the detection of invasive and metastatic sites. The discordant pattern of adenoma development in Apc^fl/flSmad4^fl/fl suggests that the caecum and colon may be more prone to adenoma initiation and progression when both genes are disrupted. As with similar mouse models, this is most likely to be due to the caecal inflammatory microenvironment and the local TGF-β-BMP ligand supply. The anatomy of the mouse appendix and colon are markedly different from that of the human, even so the model may be relevant to the rare mucinous type adenocarcinomas that can arise in the human appendix. In terms of the inflammatory microenvironment, we show downregulation of Pla2g2a (2.5log₂FC) in Smad4^Δ/Δ adenoma. Pla2g2a is considered to be a tumour suppressor independent of the Wnt/β-catenin signalling pathway, and intraluminal secretion of Pla2g2a is a host defence mechanism during the active phase of ulcerative colitis⁵³. Moreover, Reg3γ and Reg3β were also downregulated (1.2log₂FC) in Smad4^Δ/Δ adenomas and as both genes are important for epithelial defence against bacterial invasion^54,55. Thus, adenoma Smad4 specific alteration in genes expression may indirectly alter the inflammatory and innate immune environment. A further confounder to the phenotype may be modifier alleles co-segregating with LoxP alleles, although this was mitigated by backcrossing to > 10 generations of all conditional alleles to C57Bl/6J²⁷. Excess TGF-β-BMP may also directly promote adenoma formation independent of Apc LOF, as targeted deletion of Bmpr1a can also lead to crypt expansion, fused villi and adenoma formation due to an increase in β-catenin⁵⁶. Overall, adenoma Smad4 LOF may stimulate the conditions for increased local TGF-β1 ligand supply and further stimulation of the pathway, contributing to differentiation arrest, chronic inflammation and subsequent adenoma progression.

The dose dependency of TGF-β1 ligand supply with respect to adenoma organoid cell growth phenotype has not been previously described to our knowledge, but may contribute functionally to the manifestation of differential phenotypes with Smad4 LOF; at a given dose, pathway gene expression positively modifies differentiation and EMT function, but at the same time there is resistance to cell death and cell cycle arrest. The time-dependent TGF-β1 on gene expression used a fixed concentration of 390 pM which exceeds the TGF-β1 IC₅₀ of Smad4^+/+ adenoma, resulted in differential expression of genes that were Smad4 and TGF-β1 dependent, with the most significant being decreased expression of Id1. Id1 codes for a non-DNA binding inhibitor of basic-helix-loop-helix (bHLH) transcription factors (e.g. GATA4) to inhibit multiple cell processes including differentiation. ChiP-Seq has shown that Smad3 and Smad4 can bind the Id1 promoter to directly stimulate early induction following TGF-β1 exposure⁴⁰. Feedback Id1 repression is then thought to be also pSmad3 dependent but through feedback activation of ATF3, that then directly represses Id1 expression⁵⁷. Id1 over-expression in pancreatic ductal adenocarcinoma in mouse and human is also associated with more invasive and undifferentiated cells, a mechanism proposed to be dependent on EIF2 signalling⁵⁸ and independent of the TGF-β pathway⁵⁹. Id1 proteins detected by IHC may show increased expression in colorectal cancer and mouse models, and one now retracted report proposed that a genetic knock-out of Id1 results in adenoma suppression in the small intestine of Apc^Min/+, but not when colitis is chemically induced. Id1 protein localisation may not correlate with gene expression in tissues, as we have observed, and so is likely to context dependent at the protein level⁶⁰. Recently, a Nur77 based molecular switch has also been proposed to regulate Id1, where stabilisation of TGF-β induced Smad3 degradation leads to an increase in Id1 expression, whereas in the absence of TGF-β, Nur77 enhances Smad3 degradation and results in reduces Id1 expression⁶¹. The signalling of TGF-β and Id1 associated gene expression may not be solely dependent on Smad4, as Smad1/5 dual signalling through TGFBR1 and ACVR1 also results in increased Id1 expression and EMT⁶⁰.

Spp1 (osteopontin) expression normally occurs in osteoclasts and macrophages, but has also been specifically detected in tumour associated M2 macrophages, with senescence associated secretary phenotype (SASP) epithelial tumour cells and at the invasive front of human colorectal tumours^{62,63,64,65,66}. As well as extensive circumstantial data associating high SPP1 expression in human colorectal cancer, Spp1 was also previously reported to be a key Smad4 dependent expressed gene functionally associated with murine prostatic cancer invasion in a Pten prostate deletion model⁶⁷. High SPP1 expression may have therapeutic implications, as immune exclusion correlates in the tumour microenvironment with cancer associated fibroblasts resulting in resistance to immune checkpoint therapy^68,69.

Pak3 is a well-known downstream effector of Cdc42 and Rac1 GTPases that mediates motility and MAPK activation associated with cancer invasion and metastasis^70,71,72. In a recent adenovirus Cre-induced Kras^G12D, p53^fl/fl, Smad4^fl/fl model of lung cancer, increased metastasis was associated with Smad4 LOF and Pak3 activation, the latter identified as a downstream effector of Smad4 via the PAK3-JNK-Jun pathway⁷². Interestingly, the proposed mechanism appeared to be due to attenuation of Smad4 dependent transcription of miR-495 and miR-543 that can directly bind to the PAK3 3’UTR. Overexpression of PAK family members are also associated with CRC outcome, transition of adenoma to carcinoma and stabilisation of β-catenin^{73,74,75,76,77,78,79}.

There are significant limitations in comparing mouse adenoma derived gene expression signatures and human colorectal cancer in this study, hence we would consider the comparison with respect to the evolutionary coserved TGF-β-BMP-SMAD4 pathway as being exploratory. The numerous challenges that restrict the phenocopy of the mouse models to human colorectal cancer concern the the developmental, physiology and regulatory gene expression profiles are fundamentally different between human and mouse, best exemplified by the different distributions to adenoma formation when the Apc gene is disrupted, small intestine in the mouse and colon in the human. In terms of the comparison of pathway dependent gene expression signatures, of the 76 differentially expressed TGF-β-BMP-SMAD4 gene-pathway dependent candidate gene expression signatures in mouse, only 7 of these DEGs were also identified in human, several of which (CREB3L1, BCAS1, ID1 and ID3) also appeared independent of common co-existing pathogenic variants in the RTK-RAS, TP53 and PI3K pathways, but only ID1^low remained significant. SMAD4, ID1, SPP1 and PAK3 mRNA expression also appeared to have prognostic implications for colorectal cancer using the TCGA cohorts, and with further evaluation required in extended human cohorts, are candidate functional readouts associated with disruption of the TGF-β-BMP-SMAD4 pathway. In summary, we identified TGF-β and SMAD4 gene expression signatures in mouse adenoma, both for bulk and single cell sequencing, that require future prospective evaluation of pathway based prognostic-predictive classifiers across intestinal cancer subtypes⁴.

Materials and methods

Conditional Cre intestinal adenoma model

Animal work was performed under a UK Home Office licence, and approved by University of Oxford Animal ethics committee. All methods were carried out in accordance with relevant guidelines and regulations. Home Office licence approvals include planned experimental design, statistical considerations, procedures (including schedules of humane killing in this case using enflurane anaesthesia prior to neck dislocation) and outcomes and are in full accordance with the Arrive guidelines (https://arriveguidelines.org). Mouse lines were obtained from Elisabeth Robertson, Sir William Dunn School, Oxford University, U.K. (Smad4^fl/fl and RosaYFP^fl/fl)²⁵ and Alan Clarke (deceased), Cardiff University, U.K. (Lgr5-CreER^T2, Apc^fl/fl and Vil-CreER^T2)²⁴. The mouse lines were backcrossed to C57Bl/6J as reported including Smad4^fl/fl (Smad4^tm1Rob) (Supplementary Material and Methods). Genotyping PCRs were briefly (5'-3', forward-F and reverse-R): Apc F-GTTCTGTATCATGGAAAGATAGGTGGTC and R-CACTCAAAACGCTTTTGAGGGTTG (GAGTACGGGGTCTCTGTCTCAGTGAA for recombined product), Smad4 F-CTTTTATTTTCAGATTCAGGGGTTC (F-AAAATGGGAA-AACCAACGAG for recombined product) and R-TACAAGTGTATGTCTTCAGCG. Genotype combinations were generated by experimental breeding. Smad4^fl/flRosaYFP^fl/fl and Apc^fl/flSmad4^fl/flRosaYFP^fl/fl were crossed with either Lgr5-CreER^T2 or Vil-CreER^T2 mice. Apc^fl/flSmad4^+/+RosaYFP^fl/fl Lgr5-CreER^T2/Vil-CreER^T2 animals were used as controls. Mice bred with Vil-CreER^T2 mice were used for an in vitro experiment only. To generate combined heterozygote Apc^fl/+ Smad4^fl/+loxP alleles, heterozygote Apc^fl/+ was bred with homozygote Smad4^fl/fl and combined heterozygotes (both loxP alleles located on mouse Chromosome 18) were inter-crossed to generate homozygote alleles and the experimental mice as in Fig. 1a. For Lgr5-CreER^T2 activation, adult mice age ≥ 8 weeks had an intra-peritoneal injection of 5 mg (200 mg.kg⁻¹) tamoxifen dissolved in corn oil. Mice were monitored daily, small intestinal and colonic adenomas were scored for number, diameter, distribution and weight as described²⁷.

Intestinal adenoma organoids

Adenoma formed in vivo from Smad^fl/fl or Smad4^+/+ Lgr5 mice were cultured as described^27,80. For crypt culture, the small intestines were collected from Smad4^fl/flRosaYFP^fl/fl and Apc^fl/flSmad4^fl/flRosaYFP^fl/ Vil-CreER^T2 mice and processed as described⁸⁰. Briefly, crypts were cultured in Matrigel and supplemented with ADF2 and crypt growth factors, ENR (8.1 nM EGF (E), 500 ng ml-1 R-spondin1 (R), and 2.16 nM Noggin (N)) and incubated at 37 °C, 5% as described. At D3, crypts generated from transgenic mice were incubated with 500 nM 4-hydroxy-tamoxifen (4-OHT) (Millipore, 579,002) dissolved in 100% EtOH (test), vehicle (negative control) with 5 µl EtOH for 12 h at 37 °C. The test and the vehicle treated organoids were supplemented with EN only (no R) and the normal crypts with ENR. Proteins were extracted from organoids as described⁸¹.

Cell labelling and imaging

Neutral buffered formalin fixed tissues were paraffin embedded and 5 µm sections were processed either by immunohistochemistry (IHC) or immunofluorescence (IF). The antibodies that used in this study are listed (Supplementary Materials and Methods). Organoids growing on coverslips in 24 well plates were used for IF. Organoid proliferation used the thymidine analogue 5-ethynyl-2’-deoxyuridine (EdU) for 2 h using Click-iT EdU Alexa Fluor-555 Imaging Kit (Invitrogen, C10338). H&E and IHC slides were viewed by brightfield-fluorescent microscopy (Olympus BX60). Images were acquired by CCD camera and Nuance 2.10 software (CRI, Massachusetts). Confocal images were captured with an Olympus FluoView FV1000 using 40 × 1.3NA lens and associated software. Labelled slides (EdU, IHC or IF) were analysed using ImageJ 1.48 software. AlamarBlue (AB) (AbD Serotec, BUF012B), was used to quantify adenoma growth. Absorbance after 24 h was read at 570 nm and 600 nm using a Spectramax M5 plate reader (MTX Lab Systems). Live cell imaging was conducted with a Zeiss Axiovert 200 inverted microscope with a BSI Prime CMOS camera with a Neofluar 10 × 0.3NA lensfor adenoma organoids treated with TGF-β for 24 h. MetaMorph software (v 7.8.12.0) multidimensional Z-stack acquisition with 100 ms exposure acquired at regular intervals of 15 min over an extended duration of 6–18 h culture at 37.5 °C, and 5% CO₂ in air. Figure images were assembled following auto adjustment of tone, colour and brightness/contrast.

TGF-β1 induced gene expression bulk RNASeq

Early passage adenomas seeded at 9,000 cells per 25 µl well were used for RNA-Seq. Adenomas were grown in ADF2 medium supplemented with 8.1 nM EGF, 10 µM Y-27632 and 1 µM SB-505142. The medium was changed on D2, the cells were washed, EGF and 390 pM TGF-β1 added and samples extracted from three different time points (t = 0, t = 1, t = 12 h post TGF-β1 treatment). Three biological replicates from each genotype and RNA was pooled from 8 wells per treatment. RNA extraction was performed using Quick-RNA MiniPrep (Zymo Research, R1054), assessed using a Nanodrop-1000 spectrophotometer and RNA Clean & concentrator™-5 (Zymo Research, R1015). RNA-Seq was performed at the Wellcome Trust Center for Human Genetics, University of Oxford. Briefly, directional multiplexed mRNA and miRNA Libraries were prepared. All mRNA was ribo-depleted and converted to cDNA. Paired-end sequencing (100nt) were run on a HiSeq 2500 (Illumina) for mRNA-Seq.

Transcript abundances from RNA-Seq data were quantified using the Tuxedo pipeline⁸². Fastq reads were aligned to the mouse genome (mm10/GRCm38) with the TopHat-Bowtie2 aligner, versions 2.1.0 and 2.3.4, respectively and expression of transcripts was quantified with Cufflinks 2.2.1, as fragments per kilobase per million mapped reads (FPKM). Counts were also generated using the feature ‘Counts’ from the Subread package. Pairwise differential expression analysis between groups was performed in R using DESeq2 version 1.28⁸³. Genes that have less than 5 reads in at least 3 samples were removed. Genes with adjusted p-values (FDR) < 0.05 were considered significant. Shrunk log2fold change (log2FC) was calculated using as a shrinkage estimator. Genes considered as differentially regulated between 2 conditions have ± 1.5log2FC. Time-series analysis was performed using the likelihood ratio test implemented in DESeq2. Gene ontology term enrichment analysis for the significant DEG genes performed using “clusterProfiler” where we evaluated biological processes and molecular functions⁸⁴. For easier visualisation, we used “tmod”package for enrichment analysis (https://peerj.com/preprints/2420v1). In tmod, we applied CERNOtest, Coincident Extreme Ranks in Numerical Observations (CERNO) after converting mouse genes to human orthologs. We performed the enrichment analysis using Hallmark gene sets from MSigDB v7.2 (Broad Institute) and the blood signature from LI modules in tmod, gene sets with the area under the curve (AUC) ≥ 0.5 and padj < 0.05 were considered positive. Metascape website was utilised for pathway analysis of certain gene lists, and to identify the transcription regulatory network based on TRRUST ontology⁸⁵.

Early TGF-β target genes were classified based on log2FC ± 1.5 expression and adjusted p < 0.05 after 1 h compared to t = 0 in Smad4^+/+ adenomas. Transient early upregulated/downregulated genes were the genes with opposite directions to their initial regulation when compared to late time points (t = 12 vs t = 0 and/or t = 12 vs t = 1 h) in Smad4^+/+ adenomas. Late TGF-β upregulated/downregulated genes were classified according to their upregulation/downregulation at 12 h in Smad4^+/+ adenomas compared to t = 0 and/or t = 1 h with log2FC ± 1.5 and adjusted p < 0.05. Genes were considered Smad4-independent if they were significantly expressed in both Smad4^+/+ and Smad4^Δ/Δ adenomas with TGF-β1 treatment compared to their corresponding untreated baseline. However, if genes were differentially expressed in an opposite direction to that of TGF-β target genes in Smad4^Δ/Δ vs Smad4^+/+ at t = 0, they were classified as Smad4-dependent genes and excluded from the Smad4^Δ/Δ vs Smad4^+/+ t = 0 list. TGF-β target genes that were only upregulated in Smad4^+/+ adenomas were classified as Smad4-dependent genes. Genes that were expressed in Smad4^Δ/Δ vs Smad4^+/+ at t = 0 h and were not among TGF-β target genes and considered for baseline comparisons.

Caecal adenoma single cell mRNA sequencing

Pooled single-cell suspensions from were obtained from caecal tumours from three Apc^fl/fl,Smad4^+/+, Lgr5Cre^ERT2 and five Apc^fl/fl, Smad4^fl/fl, Lgr5Cre^ERT2 mice. Cell suspensions underwent library preparation using the Chromium Single Cell 3’ kit (10X Genomics) and sequencing using Illumina HiSeq 2 × 150 bp (Azenta, USA), with approximately 25,000 reads per cell. Sample demultiplexing, barcode processing and single-cell 5′ unique molecular identifier (UMI) counting (Cell Ranger Software Suite v7.0.1) with FASTQ alignment to the mouse reference transcriptome mm10-3.0.0. The following criteria were then applied to each cell: gene number between 250 and 7,000, UMI count > 500 and mitochondrial gene percentage < 0.15, log10 Genes Per UMI > 0.80. Doublets were removed using the scDblFinder library. Genes with ten read counts or less were removed. A total of 20,886 cells were analysed (11,212 for Smad4^fl/fl and 9674 for Smad4^+/+). Dimensionality reduction and clustering used the filtered gene-barcode matrix normalised using the SCTransform with Seurat (v4.3.0)⁸⁶. Variables including mitochondria Ratio, nUMI, nGene and the cell cycle gene difference were regressed and the top 3000 variable genes were used. The filtered gene-barcode matrix of all samples were then integrated, k nearest neighbour (k-NN) generated using FindNeighbors (Seurat) using the top 40 principal components and clusters identified using FindCluster function (Louvain algorithm). We used resolution 1.2 to classify the different cell types and resolution 2.2 to determine the enterocyte precursors based on clustree (v0.5.1)⁸⁷. Gene markers that identified each cluster relative to all other clusters was determined with FindAllMarkers function in Seurat using the Wilcox test (log-fold change threshold 0.25). Manual cell labelling of the top marker genes for each cluster utilised previous scRNA-seq studies^30,32,88. Significant differences (p values, and confidence intervals via bootstrapping) in cell type proportion between each genotype was determined with the permutation test implemented in scProportionTest package in R (V4.2.3). Data was visualised using Uniform Manifold Approximation and Projection (UMAP) dimensional reduction and plots were generated using Seurat and the scCustomize library. Each cluster was determined using FindAllMarkers function in Seurat (Wilcox test). The log fold change threshold was set to 0.25. Manual cell labelling utilised gene clusters in previous studies^30,31,32,33, the Lgr5 intestinal stem cell signature³² and the CellMarker 2.0 database⁸⁸. The significant differences in cell type proportion used the permutation test implemented in scProportionTest package in R which calculate a p-value for each cluster, and a confidence interval for the magnitude difference via bootstrapping⁸⁹. All analysis was performed in R V4.2.3. Differential gene expression analysis for each cluster between Smad4^fl/fl and Smad4^+/+ were performed using MAST⁹⁰ with FindAllMarkers function in Seurat. Differentially expressed genes were selected based on log fold change threshold and expressed in at least in 25% in one of the groups. Pathway analysis was performed using Metascape⁸⁵. Pseudo-bulk expression analysis was performed using MAST cell implemented in Seurat’s FindMarkers function with the first and second identities using all Smad4^+/+ and Smad4^fl/fl cells, respectively.

TCGA colorectal cancer analysis

Normalised gene expression from RNA-seq (illumine hiseq Level 3 RSEM normalized genes) retrieved from http://firebrowse.org/?cohort=COADREAD accessed October 2019.

Genes and pathways that are frequently mutated in colorectal cancer, retrieved from Cancer Genome Atlas project (TCGA)⁵¹, include the Wnt/β-catenin pathway (APC, CTNNB1, DKK1, DKK2, DKK3, DKK4, LRP5, FZD10, FAM123B:AMER1, AXIN2, TCF7L2, FBXW7, and ARID1A), TP53 pathway (ATM, TP53). PI3K (IGF2, IRS2, PTEN, PIK3R1, and PIK3CA), RTK-RAS pathway (NRAS, KRAS, BRAF, ERBB2, and ERBB3). TGF-β pathway (ACVR1B, ACVR2A, SMAD2, SMAD3, SMAD4, TGFBR1 and TGFBR2) and BMP pathway (SMAD1, SMAD9, SMAD5, BMPR1A, BMPR1B and ACVR1). Genomic variants (missense mutation, homozygote deletion, amplification or fusion gene) in these genes for COAD and READ TCGA cohorts were retrieved from cBioPortal (https://www.cbioportal.org), accessed December 2019. In COAD and READ TCGA cohorts the variant status was available for 56 patients with SMAD4 variants. 98.1% of these patients had alterations in the Wnt/β-catenin pathway, 59.3% had alteration in the TP53 pathway, 46.3% in the PI3K pathway and 64.8% had variants in the RTK-RAS pathway. Normalized RNA-seq was available for 51 normal tissues and 326 patient tumours with variant data. In TCGA COAD-READ cohorts, 44 patients with probable pathogenic and possibly pathogenic SMAD4 variants, but with no other variants in the TGF-β pathway, had RNA-seq data. One patient did not have a variant in the Wnt pathway and was excluded from the analysis. Patients with known variants in the TGF-β pathway were excluded from the control group. Using this patient group, we validated the top 76 DEG in the mouse adenomas, genes with ± 2log2FC and padj < 0.05 from Smad4^Δ/Δ vs Smad4^+/+ (t = 0 h), either in the presence or absence of variants in the PI3K, TP53, PTK-RAS pathways in the context of SMAD4 variant. Statistical analysis was performed using Kruskal–Wallis test followed by FDR adjustment. Genes with p < 0.05 following FDR adjustment were considered significant. Multiple comparisons were performed using Dunn test followed by Holm adjustment to detect SMAD4-dependent genes. The expression between 2 groups were considered positive if adjusted p < 0.05. We considered genes to be SMAD4 dependent genes based on three criteria. First, there needs to be a significant difference in gene expression between the groups (Group 1-Group 3, and Group 1-Group 4, Group 2-Group 3 and Group 2-Group 4) in the context of the three second pathways with and without variants. Second, there should be no significant difference in gene expression between Group 1-Group 2, and between Group 3 -Group 4, respectively. Third, any change in gene expression should follow the same direction as observed in the mouse adenomas. Note that 84 genes differentially regulated at t = 0 with absolute log2FC > = 2 and adjusted p < = 0.05. Of those, 76 genes had human orthologs, and 73 genes had RNA data in the TCGA cohorts. Survival data for the COAD-READ TCGA cohort were retrieved (Table S1, Tab TCGA-CDR, accessed May 2019⁹¹).

Statistical analysis

Statistical and graphical data analyses were performed using either Graph-Pad Prism version 6.0. software or R version 4.0.2. Images were analysed using ImageJ 1.48 software. Final Figures were generated using Photoshop (CS5). Immunoblot analysis was performed as follow; loading control and test from scanned immunoblots was aligned horizontally on Photoshop (CS5) and analyzed using ImageJ 1.48 software. For TCGA COAD-READ expression data, the cut-off point between high and low mRNA expression levels for each gene was computed based on the maximally selected rank statistics (maxstat) using the surv_cutpoint function from the ‘survminer’ package in R.

Data availability

Data relating to this paper are available in Supplementary Tables 1–5. Gene expression data generated in the current study are deposited in the European Nucleotide Archive (https://www.ebi.ac.uk/ena/browser/home); PRJEB85309 for bulk RNA-Seq and PRJEB85413 for scRNA-seq.

References

David, C. J. & Massague, J. Contextual determinants of TGFbeta action in development, immunity and cancer. Nat. Rev. Mol. Cell. Biol. 19, 419–435. https://doi.org/10.1038/s41580-018-0007-0 (2018).
Article CAS PubMed PubMed Central Google Scholar
Massague, J. & Sheppard, D. TGF-beta signaling in health and disease. Cell 186, 4007–4037. https://doi.org/10.1016/j.cell.2023.07.036 (2023).
Article CAS PubMed PubMed Central Google Scholar
Dunne, P. D. & Arends, M. J. Molecular pathological classification of colorectal cancer-an update. Virchows Arch. https://doi.org/10.1007/s00428-024-03746-3 (2024).
Article PubMed PubMed Central Google Scholar
Malla, S. B. et al. Pathway level subtyping identifies a slow-cycling biological phenotype associated with poor clinical outcomes in colorectal cancer. Nat. Genet. https://doi.org/10.1038/s41588-024-01654-5 (2024).
Article PubMed PubMed Central Google Scholar
Heldin, C.-H. & Moustakas, A. Role of Smads in TGFβ signaling. Cell Tissue Res. 347, 21–36 (2012).
Article CAS PubMed Google Scholar
Macias, M. J., Martin-Malpartida, P. & Massague, J. Structural determinants of Smad function in TGF-beta signaling. Trends Biochem. Sci. 40, 296–308. https://doi.org/10.1016/j.tibs.2015.03.012 (2015).
Article CAS PubMed PubMed Central Google Scholar
Inman, G. J., Nicolas, F. J. & Hill, C. S. Nucleocytoplasmic shuttling of Smads 2, 3, and 4 permits sensing of TGF-beta receptor activity. Mol. Cell 10, 283–294. https://doi.org/10.1016/s1097-2765(02)00585-3 (2002).
Article CAS PubMed Google Scholar
Guerrero-Martinez, J. A., Ceballos-Chavez, M., Koehler, F., Peiro, S. & Reyes, J. C. TGFbeta promotes widespread enhancer chromatin opening and operates on genomic regulatory domains. Nat. Commun. 11, 6196. https://doi.org/10.1038/s41467-020-19877-5 (2020).
Article CAS PubMed PubMed Central Google Scholar
Inman, G. J. & Hill, C. S. Stoichiometry of active smad-transcription factor complexes on DNA. J. Biol. Chem. 277, 51008–51016. https://doi.org/10.1074/jbc.M208532200 (2002).
Article CAS PubMed Google Scholar
Massague, J. TGFbeta signalling in context. Nat. Rev. Mol. Cell Biol. 13, 616–630. https://doi.org/10.1038/nrm3434 (2012).
Article CAS PubMed PubMed Central Google Scholar
Gaarenstroom, T. & Hill, C. S. TGF-β signaling to chromatin: How Smads regulate transcription during self-renewal and differentiation. Seminars in Cell and Developmental Biology 32, 107–118. https://doi.org/10.1016/j.semcdb.2014.01.009 (2014).
Massagué, J., Seoane, J. & Wotton, D. Smad transcription factors. Genes Dev. 19, 2783–2810 (2005).
Article PubMed Google Scholar
Fleming, N. I. et al. SMAD2, SMAD3 and SMAD4 mutations in colorectal cancer. Cancer Res. 73, 725–735. https://doi.org/10.1158/0008-5472.CAN-12-2706 (2013).
Article CAS PubMed Google Scholar
Gehart, H. & Clevers, H. Tales from the crypt: new insights into intestinal stem cells. Nat. Rev. Gastroenterol. Hepatol. 16, 19–34. https://doi.org/10.1038/s41575-018-0081-y (2019).
Article PubMed Google Scholar
Sato, T. et al. Paneth cells constitute the niche for Lgr5 stem cells in intestinal crypts. Nature 469, 415–418. https://doi.org/10.1038/nature09637 (2011).
Article CAS PubMed Google Scholar
Tan, S. H. et al. A constant pool of Lgr5(+) intestinal stem cells is required for intestinal homeostasis. Cell Rep. 34, 108633. https://doi.org/10.1016/j.celrep.2020.108633 (2021).
Article CAS PubMed Google Scholar
Beumer, J. et al. Enteroendocrine cells switch hormone expression along the crypt-to-villus BMP signalling gradient. Nat. Cell Biol. 20, 909–916. https://doi.org/10.1038/s41556-018-0143-y (2018).
Article CAS PubMed PubMed Central Google Scholar
Beumer, J. & Clevers, H. Cell fate specification and differentiation in the adult mammalian intestine. Nat. Rev. Mol. Cell Biol. 22, 39–53. https://doi.org/10.1038/s41580-020-0278-0 (2021).
Article CAS PubMed Google Scholar
Haramis, A.-P.G. et al. De novo crypt formation and juvenile polyposis on BMP inhibition in mouse intestine. Science 303, 1684–1686 (2004).
Article CAS PubMed Google Scholar
Zhang, Y. & Que, J. BMP signaling in development, stem cells, and diseases of the gastrointestinal tract. Annu. Rev. Physiol. 82, 251–273. https://doi.org/10.1146/annurev-physiol-021119-034500 (2020).
Article CAS PubMed Google Scholar
Winesett, M. P., Ramsey, G. W. & Barnard, J. A. Type II TGF(beta) receptor expression in intestinal cell lines and in the intestinal tract. Carcinogenesis 17, 989–995. https://doi.org/10.1093/carcin/17.5.989 (1996).
Article CAS PubMed Google Scholar
Hahm, K. B. et al. Loss of transforming growth factor beta signalling in the intestine contributes to tissue injury in inflammatory bowel disease. Gut 49, 190–198. https://doi.org/10.1136/gut.49.2.190 (2001).
Article CAS PubMed PubMed Central Google Scholar
Drost, J. et al. Sequential cancer mutations in cultured human intestinal stem cells. Nature 521, 43–47. https://doi.org/10.1038/nature14415 (2015).
Article CAS PubMed Google Scholar
Barker, N. et al. Crypt stem cells as the cells-of-origin of intestinal cancer. Nature 457, 608–611. https://doi.org/10.1038/nature07602 (2009).
Article CAS PubMed Google Scholar
Chu, G. C., Dunn, N. R., Anderson, D. C., Oxburgh, L. & Robertson, E. J. Differential requirements for Smad4 in TGFbeta-dependent patterning of the early mouse embryo. Development 131, 3501–3512. https://doi.org/10.1242/dev.01248 (2004).
Article CAS PubMed Google Scholar
Costello, I., Biondi, C. A., Taylor, J. M., Bikoff, E. K. & Robertson, E. J. Smad4-dependent pathways control basement membrane deposition and endodermal cell migration at early stages of mouse development. BMC Dev. Biol. 9, 54. https://doi.org/10.1186/1471-213X-9-54 (2009).
Article CAS PubMed PubMed Central Google Scholar
Matheson, J. et al. Epithelial-mesenchymal transition and nuclear beta-catenin induced by conditional intestinal disruption of Cdh1 with Apc is E-cadherin EC1 domain dependent. Oncotarget 7, 69883–69902. https://doi.org/10.18632/oncotarget.11513 (2016).
Article PubMed PubMed Central Google Scholar
Hassan, A. B. & Howell, J. A. Insulin-like growth factor II supply modifies growth of intestinal adenoma in Apc(Min/+) mice. Cancer Res. 60, 1070–1076 (2000).
CAS PubMed Google Scholar
Sato, T. et al. Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett’s epithelium. Gastroenterology 141, 1762–1772. https://doi.org/10.1053/j.gastro.2011.07.050 (2011).
Article CAS PubMed Google Scholar
Haber, A. L. et al. A single-cell survey of the small intestinal epithelium. Nature 551, 333–339. https://doi.org/10.1038/nature24489 (2017).
Article CAS PubMed PubMed Central Google Scholar
Yan, K. S. et al. Non-equivalence of Wnt and R-spondin ligands during Lgr5(+) intestinal stem-cell self-renewal. Nature 545, 238–242. https://doi.org/10.1038/nature22313 (2017).
Article CAS PubMed PubMed Central Google Scholar
Munoz, J. et al. The Lgr5 intestinal stem cell signature: robust expression of proposed quiescent “+4” cell markers. EMBO J 31, 3079–3091. https://doi.org/10.1038/emboj.2012.166 (2012).
Article CAS PubMed PubMed Central Google Scholar
Grun, D. et al. Single-cell messenger RNA sequencing reveals rare intestinal cell types. Nature 525, 251–255. https://doi.org/10.1038/nature14966 (2015).
Article CAS PubMed Google Scholar
Mustata, R. C. et al. Identification of Lgr5-independent spheroid-generating progenitors of the mouse fetal intestinal epithelium. Cell Rep. 5, 421–432. https://doi.org/10.1016/j.celrep.2013.09.005 (2013).
Article CAS PubMed Google Scholar
Ayyaz, A. et al. Single-cell transcriptomes of the regenerating intestine reveal a revival stem cell. Nature 569, 121–125. https://doi.org/10.1038/s41586-019-1154-y (2019).
Article CAS PubMed Google Scholar
Bramsen, J. B. et al. Molecular-subtype-specific biomarkers improve prediction of prognosis in colorectal cancer. Cell Rep. 19, 1268–1280. https://doi.org/10.1016/j.celrep.2017.04.045 (2017).
Article CAS PubMed Google Scholar
Isella, C. et al. Stromal contribution to the colorectal cancer transcriptome. Nat. Genet 47, 312–319. https://doi.org/10.1038/ng.3224 (2015).
Article CAS PubMed Google Scholar
Fessler, E. et al. TGFbeta signaling directs serrated adenomas to the mesenchymal colorectal cancer subtype. EMBO Mol. Med. 8, 745–760. https://doi.org/10.15252/emmm.201606184 (2016).
Article CAS PubMed PubMed Central Google Scholar
Fink, S. P., Mikkola, D., Willson, J. K. & Markowitz, S. TGF-beta-induced nuclear localization of Smad2 and Smad3 in Smad4 null cancer cell lines. Oncogene 22, 1317–1323. https://doi.org/10.1038/sj.onc.1206128 (2003).
Article CAS PubMed Google Scholar
Liang, Y. Y., Brunicardi, F. C. & Lin, X. Smad3 mediates immediate early induction of Id1 by TGF-beta. Cell Res. 19, 140–148. https://doi.org/10.1038/cr.2008.321 (2009).
Article CAS PubMed Google Scholar
Zhang, Y. et al. High throughput determination of TGFbeta1/SMAD3 targets in A549 lung epithelial cells. PLoS ONE 6, e20319. https://doi.org/10.1371/journal.pone.0020319 (2011).
Article CAS PubMed PubMed Central Google Scholar
Takaku, K. et al. Intestinal tumorigenesis in compound mutant mice of both Dpc4 (Smad4) and Apc genes. Cell 92, 645–656. https://doi.org/10.1016/s0092-8674(00)81132-0 (1998).
Article CAS PubMed Google Scholar
Kitamura, T. et al. SMAD4-deficient intestinal tumors recruit CCR1+ myeloid cells that promote invasion. Nat. Genet. 39, 467–475. https://doi.org/10.1038/ng1997 (2007).
Article CAS PubMed Google Scholar
Takaku, K. et al. Gastric and duodenal polyps in Smad4 (Dpc4) knockout mice. Cancer Res. 59, 6113–6117 (1999).
CAS PubMed Google Scholar
Takaku, K., Wrana, J. L., Robertson, E. J. & Taketo, M. M. No effects of Smad2 (madh2) null mutation on malignant progression of intestinal polyps in Apc(delta716) knockout mice. Cancer Res. 62, 4558–4561 (2002).
CAS PubMed Google Scholar
Freeman, T. J. et al. Smad4-mediated signaling inhibits intestinal neoplasia by inhibiting expression of beta-catenin. Gastroenterology 142, 562-571 e562. https://doi.org/10.1053/j.gastro.2011.11.026 (2012).
Article CAS PubMed Google Scholar
Fumagalli, A. et al. Genetic dissection of colorectal cancer progression by orthotopic transplantation of engineered cancer organoids. Proc. Natl. Acad. Sci. U S A 114, E2357–E2364. https://doi.org/10.1073/pnas.1701219114 (2017).
Article CAS PubMed PubMed Central Google Scholar
Oshima, H. et al. Suppressing TGFbeta signaling in regenerating epithelia in an inflammatory microenvironment is sufficient to cause invasive intestinal cancer. Cancer Res. 75, 766–776. https://doi.org/10.1158/0008-5472.CAN-14-2036 (2015).
Article CAS PubMed Google Scholar
Sakai, E. et al. Combined mutation of Apc, Kras, and Tgfbr2 effectively drives metastasis of intestinal cancer. Cancer Res. 78, 1334–1346. https://doi.org/10.1158/0008-5472.CAN-17-3303 (2018).
Article CAS PubMed Google Scholar
Tetteh, P. W. et al. Generation of an inducible colon-specific Cre enzyme mouse line for colon cancer research. Proc. Natl. Acad. Sci. U S A 113, 11859–11864. https://doi.org/10.1073/pnas.1614057113 (2016).
Article CAS PubMed PubMed Central Google Scholar
Cancer Genome Atlas, N. Comprehensive molecular characterization of human colon and rectal cancer. Nature 487, 330–337. https://doi.org/10.1038/nature11252 (2012).
Article CAS Google Scholar
Lips, E. H. et al. Progression and tumor heterogeneity analysis in early rectal cancer. Clin. Cancer Res. 14, 772–781. https://doi.org/10.1158/1078-0432.CCR-07-2052 (2008).
Article CAS PubMed Google Scholar
Fijneman, R. J. et al. Expression of Pla2g2a prevents carcinogenesis in Muc2-deficient mice. Cancer Sci. 99, 2113–2119. https://doi.org/10.1111/j.1349-7006.2008.00924.x (2008).
Article CAS PubMed PubMed Central Google Scholar
Dessein, R. et al. Toll-like receptor 2 is critical for induction of Reg3 beta expression and intestinal clearance of Yersinia pseudotuberculosis. Gut 58, 771–776. https://doi.org/10.1136/gut.2008.168443 (2009).
Article CAS PubMed Google Scholar
Haapamaki, M. M. et al. Gene expression of group II phospholipase A2 in intestine in ulcerative colitis. Gut 40, 95–101. https://doi.org/10.1136/gut.40.1.95 (1997).
Article CAS PubMed PubMed Central Google Scholar
He, X. C. et al. BMP signaling inhibits intestinal stem cell self-renewal through suppression of Wnt-beta-catenin signaling. Nat. Genet. 36, 1117–1121. https://doi.org/10.1038/ng1430 (2004).
Article CAS PubMed Google Scholar
Kang, Y., Chen, C. R. & Massague, J. A self-enabling TGFbeta response coupled to stress signaling: Smad engages stress response factor ATF3 for Id1 repression in epithelial cells. Mol. Cell 11, 915–926. https://doi.org/10.1016/s1097-2765(03)00109-6 (2003).
Article CAS PubMed Google Scholar
Tang, Y., Zhang, S., Li, J., Wu, C. & Fan, Q. ID1 marks the tumorigenesis of pancreatic ductal adenocarcinoma in mouse and human. Sci. Rep. 12, 13555. https://doi.org/10.1038/s41598-022-17827-3 (2022).
Article CAS PubMed PubMed Central Google Scholar
Huang, Y. H. et al. ID1 mediates escape from TGFbeta tumor suppression in pancreatic cancer. Cancer Discov. 10, 142–157. https://doi.org/10.1158/2159-8290.CD-19-0529 (2020).
Article CAS PubMed Google Scholar
Ramachandran, A. et al. TGF-beta uses a novel mode of receptor activation to phosphorylate SMAD1/5 and induce epithelial-to-mesenchymal transition. Elife https://doi.org/10.7554/eLife.31756 (2018).
Article PubMed PubMed Central Google Scholar
Niu, B. et al. Interplay between transforming growth factor-beta and Nur77 in dual regulations of inhibitor of differentiation 1 for colonic tumorigenesis. Nat. Commun. 12, 2809. https://doi.org/10.1038/s41467-021-23048-5 (2021).
Article CAS PubMed PubMed Central Google Scholar
Li, N. et al. Mapping and modeling human colorectal carcinoma interactions with the tumor microenvironment. Nat. Commun. 14, 7915. https://doi.org/10.1038/s41467-023-43746-6 (2023).
Article CAS PubMed PubMed Central Google Scholar
Lin, D. et al. Identification of a novel macrophage-related prognostic signature in colorectal cancer. Sci. Rep. 14, 2767. https://doi.org/10.1038/s41598-024-53207-9 (2024).
Article CAS PubMed PubMed Central Google Scholar
Lv, H. et al. Comparative analysis of single-cell transcriptome reveals heterogeneity and commonality in the immune microenvironment of colorectal cancer and inflammatory bowel disease. Front. Immunol. 15, 1356075. https://doi.org/10.3389/fimmu.2024.1356075 (2024).
Article CAS PubMed PubMed Central Google Scholar
Nakajima, T. et al. Osteopontin expression in the invasive front stroma of colorectal adenocarcinoma is associated with tumor budding and prognosis. Pathol. Res. Pract. 240, 154190. https://doi.org/10.1016/j.prp.2022.154190 (2022).
Article CAS PubMed Google Scholar
Ozato, Y. et al. Spatial and single-cell transcriptomics decipher the cellular environment containing HLA-G+ cancer cells and SPP1+ macrophages in colorectal cancer. Cell Rep. 42, 111929. https://doi.org/10.1016/j.celrep.2022.111929 (2023).
Article CAS PubMed Google Scholar
Ding, Z. et al. SMAD4-dependent barrier constrains prostate cancer growth and metastatic progression. Nature 470, 269–273. https://doi.org/10.1038/nature09677 (2011).
Article CAS PubMed PubMed Central Google Scholar
Luo, H. et al. Pan-cancer single-cell analysis reveals the heterogeneity and plasticity of cancer-associated fibroblasts in the tumor microenvironment. Nat. Commun. 13, 6619. https://doi.org/10.1038/s41467-022-34395-2 (2022).
Article CAS PubMed PubMed Central Google Scholar
Qi, J. et al. Single-cell and spatial analysis reveal interaction of FAP(+) fibroblasts and SPP1(+) macrophages in colorectal cancer. Nat. Commun. 13, 1742. https://doi.org/10.1038/s41467-022-29366-6 (2022).
Article CAS PubMed PubMed Central Google Scholar
Gao, Z. et al. PAK3 promotes the metastasis of hepatocellular carcinoma by regulating EMT process. J. Cancer 13, 153–161. https://doi.org/10.7150/jca.61918 (2022).
Article CAS PubMed PubMed Central Google Scholar
Ma, T., Li, H., Yang, W., Liu, Q. & Yan, H. Over-expression of miR-193a-3p regulates the apoptosis of colorectal cancer cells by targeting PAK3. Am. J. Transl. Res. 14, 1361–1375 (2022).
CAS PubMed PubMed Central Google Scholar
Tan, X. et al. Loss of Smad4 promotes aggressive lung cancer metastasis by de-repression of PAK3 via miRNA regulation. Nat. Commun. 12, 4853. https://doi.org/10.1038/s41467-021-24898-9 (2021).
Article CAS PubMed PubMed Central Google Scholar
Carter, J. H. et al. Pak-1 expression increases with progression of colorectal carcinomas to metastasis. Clin. Cancer Res. 10, 3448–3456. https://doi.org/10.1158/1078-0432.CCR-03-0210 (2004).
Article CAS PubMed Google Scholar
Gong, W. et al. P21-activated kinase 5 is overexpressed during colorectal cancer progression and regulates colorectal carcinoma cell adhesion and migration. Int. J. Cancer 125, 548–555. https://doi.org/10.1002/ijc.24428 (2009).
Article CAS PubMed Google Scholar
Huynh, N. et al. Depletion of p21-activated kinase 1 up-regulates the immune system of APC(∆14/+) mice and inhibits intestinal tumorigenesis. BMC Cancer 17, 431. https://doi.org/10.1186/s12885-017-3432-0 (2017).
Article CAS PubMed PubMed Central Google Scholar
King, H., Nicholas, N. S. & Wells, C. M. Role of p-21-activated kinases in cancer progression. Int. Rev. Cell Mol. Biol. 309, 347–387. https://doi.org/10.1016/B978-0-12-800255-1.00007-7 (2014).
Article CAS PubMed Google Scholar
Li, C. et al. High-level expression of P21-Cdc/Rac-activated kinase 7 is closely related to metastatic potential and poor prognosis of colon carcinoma. Oncotarget 7, 46042–46055. https://doi.org/10.18632/oncotarget.10017 (2016).
Article PubMed PubMed Central Google Scholar
Li, X. & Li, F. p21-activated kinase: Role in gastrointestinal cancer and beyond. Cancers (Basel) https://doi.org/10.3390/cancers14194736 (2022).
Article PubMed PubMed Central Google Scholar
Song, B., Wang, W., Zheng, Y., Yang, J. & Xu, Z. P21-activated kinase 1 and 4 were associated with colorectal cancer metastasis and infiltration. J. Surg. Res. 196, 130–135. https://doi.org/10.1016/j.jss.2015.02.035 (2015).
Article CAS PubMed Google Scholar
Sato, T. et al. Single Lgr5 stem cells build crypt-villus structures in vitro without a mesenchymal niche. Nature 459, 262–265. https://doi.org/10.1038/nature07935 (2009).
Article CAS PubMed Google Scholar
Lee, G. Y., Kenny, P. A., Lee, E. H. & Bissell, M. J. Three-dimensional culture models of normal and malignant breast epithelial cells. Nat. Methods 4, 359–365. https://doi.org/10.1038/nmeth1015 (2007).
Article CAS PubMed PubMed Central Google Scholar
Trapnell, C. et al. Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks. Nat. Protoc. 7, 562–578. https://doi.org/10.1038/nprot.2012.016 (2012).
Article CAS PubMed PubMed Central Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550. https://doi.org/10.1186/s13059-014-0550-8 (2014).
Article CAS PubMed PubMed Central Google Scholar
Yu, G., Wang, L. G., Han, Y. & He, Q. Y. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 16, 284–287. https://doi.org/10.1089/omi.2011.0118 (2012).
Article CAS PubMed PubMed Central Google Scholar
Zhou, Y. et al. Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat. Commun. 10, 1523. https://doi.org/10.1038/s41467-019-09234-6 (2019).
Article CAS PubMed PubMed Central Google Scholar
Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888-1902 e1821. https://doi.org/10.1016/j.cell.2019.05.031 (2019).
Article CAS PubMed PubMed Central Google Scholar
Zappia, L. & Oshlack, A. Clustering trees: a visualization for evaluating clusterings at multiple resolutions. Gigascience https://doi.org/10.1093/gigascience/giy083 (2018).
Article PubMed PubMed Central Google Scholar
Hu, C. et al. Cell Marker 2.0: an updated database of manually curated cell markers in human/mouse and web tools based on scRNA-seq data. Nucleic Acids Res 51, D870–D876. https://doi.org/10.1093/nar/gkac947 (2023).
Article CAS PubMed Google Scholar
Miller, S. A. et al. LSD1 and aberrant DNA methylation mediate persistence of enteroendocrine progenitors that support BRAF-mutant colorectal cancer. Cancer Res. 81, 3791–3805. https://doi.org/10.1158/0008-5472.CAN-20-3562 (2021).
Article CAS PubMed PubMed Central Google Scholar
Finak, G. et al. MAST: a flexible statistical framework for assessing transcriptional changes and characterizing heterogeneity in single-cell RNA sequencing data. Genome Biol 16, 278. https://doi.org/10.1186/s13059-015-0844-5 (2015).
Article CAS PubMed PubMed Central Google Scholar
Liu, J. et al. An Integrated TCGA pan-cancer clinical data resource to drive high-quality survival outcome analytics. Cell 173, 400-416 e411. https://doi.org/10.1016/j.cell.2018.02.052 (2018).
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

We thank Elisabeth Robertson for the conditional Smad4 model, Alan Clarke (deceased) for the conditional intestinal models for and Caroline Hill for discussion. Linda Randall and Nicky Hamp for mouse technical assistance.

Funding

We thank Cancer Research UK (A429-ABH), Clarendon-Islamic Development Bank scholarship programme (MS), Rhodes Trust (JM) and Dunn School (ABH) for funding.

Author information

Authors and Affiliations

Oxford Molecular Pathology Institute, Sir William Dunn School of Pathology, University of Oxford, South Parks Road, Oxford, OX1 3RE, UK
Mirvat Surakhy, Julia Matheson, David J. Barnes, Emma J. Carter, Jennifer Hughes, Claudia Bühnemann, Sabina Sanegre, Paul Metz, Charlotte J. Imianowski & Andrew Bassim Hassan
Department of Pathology, Leiden University Medical Centre, Leiden, The Netherlands
Hans Morreau
Institute of Cancer and Genomic Sciences, University of Birmingham, Birmingham, B15 2TT, UK
David J. Barnes

Authors

Mirvat Surakhy
View author publications
Search author on:PubMed Google Scholar
Julia Matheson
View author publications
Search author on:PubMed Google Scholar
David J. Barnes
View author publications
Search author on:PubMed Google Scholar
Emma J. Carter
View author publications
Search author on:PubMed Google Scholar
Jennifer Hughes
View author publications
Search author on:PubMed Google Scholar
Claudia Bühnemann
View author publications
Search author on:PubMed Google Scholar
Sabina Sanegre
View author publications
Search author on:PubMed Google Scholar
Hans Morreau
View author publications
Search author on:PubMed Google Scholar
Paul Metz
View author publications
Search author on:PubMed Google Scholar
Charlotte J. Imianowski
View author publications
Search author on:PubMed Google Scholar
Andrew Bassim Hassan
View author publications
Search author on:PubMed Google Scholar

Contributions

Experiment conceived by ABH. Mouse breeding and phenotypic analysis by MS and JM. Imaging by MS, SB, and CB. bulk gene expression MS and DB. Genotyping by MS, EC, and JH. Adenoma histopathology HM. Evaluation of organoid model systems MS, PM and CJI. Single cell sequencing analysis MS and ABH. TCGA analysis, MS and ABH. Figures MS and ABH. Manuscript was written by ABH, MS and DB. All authors approved the manuscript.

Corresponding author

Correspondence to Andrew Bassim Hassan.

Ethics declarations

Competing of Interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Supplementary Video 1.

Supplementary Video 2.

Supplementary Video 3.

Supplementary Video 4.

Supplementary Information 1.

Supplementary Information 2.

Supplementary Information 3.

Supplementary Information 4.

Supplementary Information 5.

Supplementary Information 6.

Supplementary Information 7.

Supplementary Information 8.

Supplementary Information 9.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Surakhy, M., Matheson, J., Barnes, D.J. et al. Smad4 and TGFβ1 dependent gene expression signatures in conditional intestinal adenoma, organoids and colorectal cancer. Sci Rep 15, 16330 (2025). https://doi.org/10.1038/s41598-025-00908-4

Download citation

Received: 09 December 2024
Accepted: 02 May 2025
Published: 10 May 2025
DOI: https://doi.org/10.1038/s41598-025-00908-4