Xinqidi,Your biological research partner
Xinqidi Biotech Co.,Ltd,Wuhan,China 2008-2021
R&D 13th year

Oncogenic BRAF, unrestrained by TGFβ-receptor signalling, drives right-sided colonic tumorigenesis

Issuing time:2021-06-10 09:48


Right-sided (proximal) colorectal cancer (CRC) has a poor prognosis and a distinct mutational profile, characterized by oncogenic BRAF mutations and aberrations in mismatch repair and TGFβ signalling. Here, we describe a mouse model of right-sided colon cancer driven by oncogenic BRAF and loss of epithelial TGFβ-receptor signalling. The proximal colonic tumours that develop in this model exhibit a foetal-like progenitor phenotype (Ly6a/Sca1+) and, importantly, lack expression of Lgr5 and its associated intestinal stem cell signature. These features are recapitulated in human BRAF-mutant, right-sided CRCs and represent fundamental differences between left- and right-sided disease. Microbial-driven inflammation supports the initiation and progression of these tumours with foetal-like characteristics, consistent with their predilection for the microbe-rich right colon and their antibiotic sensitivity. While MAPK-pathway activating mutations drive this foetal-like signature via ERK-dependent activation of the transcriptional coactivator YAP, the same foetal-like transcriptional programs are also initiated by inflammation in a MAPK-independent manner. Importantly, in both contexts, epithelial TGFβ-receptor signalling is instrumental in suppressing the tumorigenic potential of these foetal-like progenitor cells.


Colorectal cancer (CRC)—a leading cause of cancer-related mortality worldwide—is a heterogeneous group of neoplasms arising from the epithelium lining the large intestine/colon. In recent years, the anatomical location/sidedness of colorectal tumours has emerged as an important determinant of disease progression, response to systemic therapy, and clinical outcome. Indeed, colorectal tumours that arise proximal (right) or distal (left) to the splenic flexure manifest profound differences in epidemiology, histopathogenesis, and molecular landscapes1,2,3.

Left-sided CRCs develop from benign adenomas through the conventional adenoma-carcinoma pathway typically entailing aberrant activation of Wnt signalling, most prominently via biallelic inactivation of the tumour suppressor APC, or activating mutations in CTNNB1, both of which result in the stabilization and nuclear translocation of β-catenin. Progression of premalignant tubular, villous, or tubulovillous adenomas to adenocarcinoma is underpinned by accumulation of mutations in oncogenes and tumour suppressor genes, such as KRAS, PIK3CA, TP53, and SMAD4, as well as chromosomal instability4,5. Right-sided CRCs (rCRCs) are thought to arise through an alternative serrated neoplasia pathway so-called because the precursor lesions—namely, hyperplastic polyps, sessile serrated adenomas/lesions (SSAs), and traditional serrated adenomas (TSAs)—are distinguished by a distinctive saw-tooth crypt morphology caused by marked infolding of the upper crypt epithelium6,7,8. SSAs, in particular, carry a significant risk for malignant progression6,7,8. In addition to the saw-tooth crypt architecture seen in longitudinal sections, right-sided tumours often exhibit other common histopathological features, including a flat/sessile morphology, and a mucinous histology, as well as a propensity towards aggressive local invasion with peritoneal metastases/carcinomatosis9,10,11,12,13. rCRCs carry a worse prognosis than their left-sided counterparts and are characterised by microsatellite instability (MSI) and aberrations in MAPK, TGFβ, and mismatch repair pathways9,10,11.

Oncogenic mutation of BRAF, resulting in a valine-to-glutamate substitution at residue 600 of the BRAF serine-threonine kinase (BRAFV600E), has been identified as a key initiating event in a sizeable subset of serrated premalignant lesions and rCRCs14,15. Independent of upstream RAS signalling, the BRAFV600E oncoprotein causes constitutive activation of the MAPK (RAS-RAF-MEK-ERK) signalling cascade, which plays key roles in cell survival, proliferation, differentiation, senescence, and apoptosis. Previous attempts to model rCRCs, utilising BrafV600E alone, resulted primarily in small intestinal tumours displaying Wnt-pathway activation16,17,18. However, these findings are discordant with data from patients harbouring BRAFV600Eright-sided colonic tumours that notably lack nuclear β-catenin positivity19. In respect of this apparent lack of Wnt-pathway activation, recent literature has implicated the Hippo-pathway effectors YAP/TAZ in the reprogramming and mobilization of foetal-like, Lgr5 epithelial populations during the early stages of colitis-associated regeneration20. In this context, Lgr5+ intestinal stem cell (ISC) signatures are suppressed with a concomitant increase in the expression of foetal intestinal markers, such as Ly6a/Sca1 and Anxa1, amongst others21,22. Similarly, YAP/TAZ signalling drives the transient foetal reprogramming that occurs during irradiation-induced regeneration, underpinned by suppression of Lgr5+ ISCs and attenuation of Wnt signalling23. Finally, this YAP/TAZ regenerative signalling has also been linked to the progression of APC-deficient foci to adenomas in the murine small intestine23. As chronic inflammation, and tissue injury and repair are associated with an elevated risk of CRC, it remains to be seen whether this YAP/TAZ-dependent, foetal-like, Lgr5 regenerative state plays a role in the development of right-sided colonic tumours lacking Wnt-pathway aberrations.

To date, the lack of tractable tumour models that faithfully recapitulate the topography and molecular landscapes of oncogenic BRAF-driven rCRCs has hampered efforts to decipher the underlying molecular mechanisms and develop targeted therapeutics. Here, we describe a human-like mouse model of right-sided colon cancer driven by oncogenic BRAF and loss of epithelial TGFβ-receptor signalling that develops proximal colonic tumours with a foetal-like progenitor phenotype (Ly6a/Sca1+) and, importantly, lack expression of Lgr5and its associated intestinal stem cell signature. We further ascribe a role for microbial-driven inflammation in the initiation of these tumours, consistent with the long-held contention that intestinal microbiota are a key factor in CRC development and the pathogenesis of right-sided disease in particular24,25. Finally, whether induced directly by oncogenic BRAF or by inflammation, we show that the tumorigenic potential of these foetal-like progenitor populations is fundamentally restrained by epithelial TGFβ signalling.


BRAFV600E and TGFβ-receptor loss drive Wnt-low tumours in the right colon

Given the strong association of rCRCs with mutations in BRAF, defects in TGFβ-receptor signalling, and epigenetic silencing of the mismatch repair gene MLH12, we targeted these mutations to the murine intestinal tract. We used the tamoxifen-inducible intestinal epithelial cell-specific VillinCreER transgene to conditionally activate the BrafLSL-V600E knock-in allele, or to delete the floxed alleles of the genes encoding the TGFβ-receptor TGFBR1/ALK5 or the mismatch repair protein MLH1 (Tgfbr1/Alk5fl/fl or Mlh1fl/fl, respectively)26,27,28. VillinCreER; Alk5fl/fl mice (designated A) did not develop tumours as a result of the indicated induction regime (3 mg on one occasion), while a further cohort of 20 mice subject to a more stringent regime (3 mg, 2 mg on consecutive days) did not develop tumours by 475 days (Fig. 1a). In addition, VillinCreER; Mlh1fl/fl and VillinCreER; BrafLSL-V600E/+(designated B) mice developed tumours predominantly in the small intestine (Fig. 1a, Supplementary Fig. 1a, b).

Fig. 1: Epithelial-specific BrafV600E mutation combines with loss of TGFβ-receptor signalling to drive right-sided tumorigenesis without Wnt-pathway activation.

a Kaplan–Meier survival curves highlighting intestinal tumour-free survival (days after induction) of VillinCreER; BrafV600E/+ (B; n = 11; 3 mg tamoxifen induction), VillinCreER; Alk5fl/fl (A; n = 3; 3 mg tamoxifen induction), and VillinCreER; BrafV600E/+; Alk5fl/fl (BA; n = 18; 2 mg tamoxifen induction) mice. A mice were censored at 550 days post induction (indicated by tick marks on the Kaplan–Meier curve). (****p = 1.835e−7 for comparison of B vs BA; Mantel-Cox log-rank test, two-tailed). bRepresentative H&E staining of BA mouse colon (n = 8) (Swiss-roll). Right, higher magnification of dashed area in left panel, highlighting the proximal (right-sided) colonic location of adenocarcinomas. Scale bar, 500 μm. c Intestinal tumour location scoring in BA mice (n = 11 mice). Mean ± s.e.m. (*p = 0.0194, ***p = 0.0003, ****p = 8.5e−6; Mann–Whitney U-test, two-tailed). dSchematic representing the tumour distribution in the intestinal tract of BA mice, with circle size proportional to the regional tumour number scored in (c). Blue, caecum, and proximal colon; Red, left colon (descending) to rectum; Diagonal line, splenic flexure. e Representative images of mesenteric invasion, with tumour epithelium highlighted with a dotted outline (H&E; left panel), and mucin staining (Alcian Blue/PAS; right panel) in primary tumour sections from BA mice (n = 5). Scale bar, 500 μm. f Representative tumour sections, from VillinCreER; Apcfl/+, B, and BA mice (n = 5 mice per group), stained for β-catenin (immunohistochemistry) and the Wnt-pathway activation markers Lgr5, Axin2, and Notum (ISH). Scale bar, 100 μm. g Boxplots of the Wnt activation (WA) signature in WT vs BRAF-mutant CRCs from patients, right- vs left-sided CRCs, and across CMS subtypes. Median is shown with boxes which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5× the interquartile range. Left panel n = 579, middle n = 604 and right n = 467 independent samples (Left and middle panels: ***p < 0.001, T-test, two-tailed; Right panel: ***p < 2e−16, T-test, two tailed).

We reasoned that TGFβ-receptor inactivation in human cancer—occurring as a result of defective mismatch repair (due to a microsatellite sequence in TGFBR2, missing from the mouse genome)29—might synergise with mutant BRAF to initiate right-sided colonic tumours. Indeed, VillinCreER; BrafLSL-V600E/+; Alk5fl/fl (designated BA) mice developed tumours rapidly compared with B mice, with median survival times of 192 and 485 days, respectively (Fig. 1a). Crucially, BA mice developed adenocarcinomas predominantly in the proximal colon, which is equivalent to the right colon in humans (Fig. 1b–d). Further recapitulating rCRC, BA tumours exhibited a flat/sessile morphology, mucinous differentiation, and a pronounced ability to invade the mesentery (Fig. 1b, e).

To ensure that the preferential right-sidedness of tumours in BA mice was not due to a regional variation in recombination efficiency along the length of the intestinal tract, we scored tdTomato-reporter expression in VillinCreER; Rosa26LSL-tdTomato/+ mice following varying doses of tamoxifen. Indeed, at doses of 80 mg/kg, whilst emergent tumours were right-sided, the recombination efficiency was at its lowest in the colon and caecum compared with the small intestine, highlighting that the preferential right-sidedness of tumours in this model is a result of the specific genetic context (Supplementary Fig. 1c).

TGFβ growth-inhibitory signals are transduced through a hetero-tetrameric receptor complex comprised of TGFBR1/ALK5 and TGFBR2 homodimers. Given that mutations in TGFBR2 are more common than in TGFBR1 in human tumours30, we also generated VillinCreER; BrafV600E/+; Tgfbr2fl/fl mice31. These mice also developed primarily proximal colonic adenocarcinomas directly recapitulating the BA model (Supplementary Fig. 1d). Importantly, this shared phenotype between BA and VillinCreER; BrafV600E/+; Tgfbr2fl/fl mice ruled out the involvement of non-canonical receptor-specific TGFβ signalling in the suppression of right-sided tumorigenesis.

We next sought to determine whether BA tumours exhibited Wnt-pathway activation. Unlike APC-deficient (VillinCreER; Apcfl/+) or B tumours, BA tumours did not show nuclear β-catenin accumulation (Fig. 1f). Furthermore, we detected diminished expression of select Wnt-target genes (Lgr5, Axin2, and Notum) and our previously reported Wnt-target gene signature32, confirming that these right-sided colonic tumours lack Wnt-pathway activation (Fig. 1f, Supplementary Fig. 2). Using RNA-sequencing (RNA-seq) data from VillinCreER; Apcfl/fl intestinal tissues, harvested on day 4 post induction, we next derived a Wnt activation (WA) signature (Supplementary Fig. 3a), filtered to retain only epithelial genes that cluster close to AXIN2 (a prominent transcriptional target of β-catenin-dependent Wnt signalling) in bulk human gene expression data33,34,35. Importantly, this signature was significantly downregulated in patient tumours with BRAF-mutant status compared with BRAF wild-type counterparts, right-sided vs left-sided location, and the CRC consensus molecular subtype (CMS) 1 (enriched for MSI and inflammatory/immune gene expression) vs CMS2 (enriched for genes associated with canonical Wnt signalling; Fig. 1g)36. Collectively, these findings demonstrate that although BRAF activation alone is not sufficient to drive tumour formation in the right colon, concurrent loss of epithelial TGFβ-receptor signalling cooperates in the genesis of proximal colonic tumours with histological features of human rCRCs and attenuated Wnt/β-catenin signalling.

BA tumours exhibit foetal-like differentiation from early initiation

Due to the notable absence of Lgr5+ cells in BA tumours, we next ascertained whether these tumours recapitulated the Lgr5-independent, foetal spheroid signature21 associated with the regenerative response that ensues following ionizing radiation or dextran sulphate sodium salt (DSS) treatment20,23. RNA-seq and gene set enrichment analysis (GSEA) revealed a strong association of BA tumours with this foetal-like signature compared with VillinCreER (WT) controls (Fig. 2a–c, Supplementary Table 1). Of note, these foetal markers included Ly6a and Anxa1, two genes highly expressed during DSS-induced regeneration20. We also found enrichment for genes associated with an inflammatory response within BA tumours (Fig. 2a, right panel)37. Given that a number of foetal markers may be expressed by stromal cells, particularly the immune infiltrate, we confirmed epithelial cell-specific expression of select foetal markers (Ly6a and Anxa1) via in situ hybridization (ISH) (Fig. 2d). Interestingly, cytokeratin 7—a keratin not normally expressed in adult intestinal epithelium, but a component of the foetal gene program—is strongly expressed in BA colonic tumour epithelium (Supplementary Fig. 1e). Indeed, CK7 has been put forth as an independent marker of the serrated pathway in CRC38.

Fig. 2: Wnt-low, right-sided tumours express a foetal-like signature, present from early initiation.

a GSEA plots showing enrichment of foetal spheroid and HALLMARK inflammatory response signatures in endpoint BA colonic tumours (n = 4) vs WT proximal colon (n = 3). NES normalised enrichment score, FDR false discovery rate. (p = 0.0, represents p < 0.001 calculated empirically.) bHeatmap showing Z-score-transformed relative expression levels of selected foetal signature genes differentially expressed between endpoint BA intestinal tumours (n = 4) and WT proximal colonic tissue (n = 3). c qRT-PCR of indicated foetal markers in BA intestinal tumours vs WT proximal colon (n = 4 mice per group). Mean ± s.e.m. (*p = 0.0286, compared with WT; Mann–Whitney U-test, two-tailed). d Representative ISH of BA intestinal tumour epithelium stained for foetal markers Ly6a and Anxa1 (n = 5 mice). Scale bar, 100 μm. e GSEA plots showing positive enrichment of the foetal spheroid signature (left panel) and negative enrichment of the Lgr5+ ISC signature (right panel) in BA proximal colonic tissue vs WT control tissue, 30 days post induction (n = 5 per group). NES normalised enrichment score, FDR false discovery rate. (p = 0.0, represents p < 0.001 calculated empirically.) f Heatmap showing Z-score-transformed relative expression levels of selected foetal signature genes in BA proximal colonic tissues (n = 5), compared with WT (n = 4), 30 days post induction. g qRT-PCR of indicated foetal markers in A, B, and BA vs WT proximal colonic tissues, 30 days post induction (n = 4 per group). Mean ± s.e.m. (*p = 0.0286, compared with WT; Mann–Whitney U-test, two-tailed). h Representative ISH of WT and BA proximal colonic tissue, stained for foetal markers Ly6a and Anxa1, 30 days post induction (n = 5 per group). Scale bar, 100 μm.

We next addressed whether the foetal-like gene signature emerges during tumorigenesis or whether it is present early on, from tumour initiation. For this, we harvested proximal colonic tissue from BA and WT mice, on day 30 (d30) post induction, and used RNA-seq to examine the expression of the same foetal markers at this early timepoint. Indeed, we found a significant enrichment of the foetal-like gene signature, as well as suppression of the Lgr5+ISC signature22 (Fig. 2e–g, Supplementary Table 1) in d30 BA tissues relative to the WT proximal colon. Again, we confirmed epithelial cell-specific expression of Ly6a and Anxa1 by ISH (Fig. 2h).

Having demonstrated low Wnt-pathway activation in rCRCs and our BA model (Fig. 1f, g, Supplementary Fig. 2), we next generated an analogous epithelial-specific gene signature (designated BA) from our endpoint tumour and d30 BA RNA-seq data (Supplementary Fig. 3b). Using this BA signature to interrogate CRC-patient datasets34,35, we identified a significant correlation with BRAF mutation, right-sidedness, and the CMS1 subtype in CRC-patient tumours (Fig. 3a–c). We also found a significant inverse correlation between our BA and WA signatures and their associated CMS subtypes in patient samples as well as a positive correlation between our BA signature and the foetal spheroid signature of Mustata et al.21 (Fig. 3d, Supplementary Fig. 3c). Additionally, we observed a significant association between our BA signature and more advanced primary tumour stage (T3–4), consistent with the locally aggressive behaviour of our BA model (Fig. 3e). Accordingly, our BA signature also correlated with the shorter survival of relapsing patients and showed a trend towards an increased frequency of peritoneal disease upon relapse (Fig. 3f, Supplementary Fig. 3d), consistent with the pattern of metastatic spread in patients with rCRCs12,13.

Fig. 3: Poor-prognosis BRAF-mutant, right-sided patient CRCs are Wnt-low and express the BA signature from inception.

a Boxplot showing expression of the BA signature in WT vs BRAF-mutant CRCs from patients (n = 579). Median is shown with boxes, which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5× the interquartile range (***p = 3.58e−5; T-test, two-tailed). b Boxplot showing expression of the BA signature of right- vs left-sided CRCs from patients (n = 604). Median is shown with boxes, which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5× the interquartile range (***p = 1.28e−9; T-test, two-tailed). c Boxplot showing expression of the BA signature across CMS subtypes (n = 461). Median is shown with boxes, which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5x the interquartile range (***p = 2.16e−13; T-test, two-tailed). d Scatter plot of WA and BA signature values from human tumours showing a negative correlation (Pearson’s r = −0.232, p = 8.36e−9, two-sided) (n = 467). CMS plots are overlain highlighting the differential expression of the signatures between CMS1 and CMS2 subtypes. CMS1 (MSI-high, immune): orange; CMS2 (canonical Wnt signalling): blue; CMS3 (metabolic dysregulation): pink; CMS4 (mesenchymal): green. Median is shown with boxes, which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5x the interquartile range. e Expression of the BA signature is higher in T3–4 human tumours than T1–2 (***p = 0.0002; T-test, two-tailed) (n = 604). Median is shown with boxes, which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5× the interquartile range. f Patients with BA signature-high tumours (red), defined as above the median value, have shorter survival after relapse compared with those harbouring BA signature-low tumours (blue) (***p = 0.00016; Wald test, two-tailed) (n = 197). g Expression of the WA signature is higher in tubulovillous adenomas (TVA) (n = 29) than sessile serrated adenomas (SSA) (n = 15) (***p = 1.538e−05; T-test, two-tailed). Median is shown with boxes, which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5× the interquartile range. h Expression of the BA signature is higher in sessile serrated adenomas (SSA) than tubulovillous adenomas (TVA) (***p = 1.498e−05; T-test, two-tailed). Median is shown with boxes, which extend from the 25th to the 75th percentile and whiskers which extend an additional 1.5× the interquartile range. i Higher values of the BA signature are significantly associated with lower expression values of the WA signature in pooled TVA/SSA human precancerous lesions (Pearson’s r = −0.66, p = 1.11e−06). The regression line is presented with 95% confidence intervals (shaded area).

We also observed a significant enrichment of the BA signature in sessile serrated adenomas/lesions (SSAs) compared with tubulovillous adenomas (TVAs), the precancerous lesions of CRCs associated with mutations in BRAF and Wnt-pathway components, respectively (Fig. 3g, h)39. Furthermore, we found a striking negative correlation between the WA and BA signatures in these same precancerous lesions implying a fundamental difference in their underlying biology from inception (Fig. 3i)39. These data support the notion that cells expressing the foetal-like BA signature are present from the early stages of human right-sided tumorigenesis, in similarity with our findings in the BA model where this signature was present from early tumorigenesis.

BRAFV600E promotes YAP-dependent foetal-like differentiation

Given the central role of the transcriptional coactivator YAP in foetal reprogramming during injury-induced regeneration20,23, we next examined our endpoint tumour and d30 RNA-seq data to determine whether YAP activity is associated with the foetal-like phenotype in our BA model. Utilising the YAP target-gene signature, generated by Gregorieff et al., we identified a strong correlation with YAP activation in both the endpoint tumour and d30 BA colonic tissues (Fig. 4a, b, Supplementary Table 1)23. Top-ranking genes in this YAP signature notably include foetal markers, such as Ly6a and Anxa1, as well as classic YAP targets such as Ctgf23.

Fig. 4: BRAFV600E drives foetal-like differentiation via YAP activation, underpinning cell viability in the proximal colon.

a GSEA plots showing enrichment of YAP up- and down-regulated gene signatures in endpoint BA tumours (n = 4) vs WT proximal colon (n = 3). NES normalised enrichment score, FDR false discovery rate. (p = 0.0, represents p < 0.001 calculated empirically.) b GSEA plots showing positive enrichment of YAP up- and down-regulated gene signatures in BA proximal colonic tissue vs WT control tissue, 30 days post induction (n = 5 per group). NES normalised enrichment score, FDR false discovery rate. (p = 0.0, represents p < 0.001 calculated empirically.) c qRT-PCR of indicated foetal markers in proximal colonic tissue from vehicle- and FAKi-(VS-4718)-treated BA mice, 30 days post induction (n = 5 mice per treatment). Mean ± s.e.m. (ns not significant; **p = 0.0079; Mann–Whitney U-test, two-tailed). d qRT-PCR of Lgr5 and indicated foetal markers in BA organoids, induced in vitro with 4-hydroxytamoxifen and sampled 5 days post induction (n = 3 organoid lines from 3 separate uninduced BA mice). Fold change is shown relative to uninduced organoids (EtOH-treated). Mean ± s.e.m. (*p = 0.05; Mann–Whitney U-test, two-tailed). e qRT-PCR of Ly6a in BA organoids, induced in vitro with 4-hydroxytamoxifen and sampled 48 h post drug-treatment, 5 days post induction. Drugs were reconstituted in DMSO vehicle (0.1%) at the indicated final concentrations: verteporfin (YAPi; 3 μM), AZD6244 (MEKi; 100 nM), ERKi (100 nM), eCF506 (SRCi; 50 nM), and VS-4718 (FAKi; 1 μM). n = 3 organoid lines from 3 separate uninduced BA mice. Fold change is shown relative to DMSO-treated, induced organoids. Mean ± s.e.m. (*p = 0.05; Mann–Whitney U-test, two-tailed). f Representative H&E-staining of the proximal colon and caecum from BrafV600E/+ Alk5fl/fl Yapfl/fl mice 6 days post induction, highlighting severe epithelial loss and necrosis compared with BA controls (n = 5 mice per group). Scale bar, 500 μm. gImmunofluorescent staining of Matrigel ENRW-cultured BA and VillinCre; Apcfl/+ tumour spheroids for SCA1/LY6A (green) and YAP1 (red). Nuclei were counterstained with DAPI (blue). Merged images are shown on the right. Representative images of n = 3 biological replicates. Scale bar, 100 μm. hImage analysis scoring of the percentage of BrdU+ cells in the proximal colon epithelium of WT (n = 6), A (n = 4), B (n = 6), and BA (n = 6) mice, 30 days post induction. Mean ± s.e.m. (ns not significant; Mann–Whitney U-test, two-tailed). i Image analysis scoring of percentage of cleaved caspase 3+ cells in the proximal colon epithelium of WT (n = 6), A (n = 4), B (n = 6), and BA (n = 6) mice, 30 days post induction. Mean ± s.e.m. (**p = 0.0076, Mann–Whitney U-test, two-tailed).

We next used well-characterised and efficacious pharmacological agents to interrogate the signalling pathways eliciting YAP activation in our BA model. Previous studies have suggested that the activation of YAP, triggered upon remodelling of the extracellular matrix during DSS-induced colitis and regeneration, is dependent on FAK signalling20. However, we were unable to suppress the foetal program in vivo using the FAK-inhibitor VS-471840,41(FAKi), suggesting that mechanotransduction is not involved in the activation of YAP in our BA model (Fig. 4c).

Evaluation of intestinal FITC-dextran uptake in both B and BA mice revealed a significant intestinal barrier defect (Supplementary Fig. 4b), which was associated with pronounced proprial neutrophil infiltration in the caecum and proximal colon (Supplementary Fig. 4c). This locoregional inflammation, which co-localized with the formation of tumour foci, suggested a possible role for microenvironmental cues in triggering the foetal-like phenotype in our BA model. We therefore generated organoids from uninduced BA and B mice and induced gene recombination in vitro. Importantly, under these sterile culture conditions and in the absence of stromal influences, induction of mutant Braf alone was sufficient to activate the foetal-like program, accompanied by loss of Lgr5 and upregulation of foetal markers (Fig. 4d, Supplementary Fig. 5a). In the organoid setting, induction of foetal marker expression was not affected by selective inhibitors of FAK (VS-4718)40,41 or SRC (eCF506)42, but it was significantly suppressed by inhibitors of MEK (AZD6244)43,44, ERK45, and YAP (verteporfin)20 (Fig. 4e), with efficacy of MEK and ERK inhibitors demonstrated by suppression of the canonical MAPK target Dusp6 (Supplementary Fig. 5b). Notably, treatment of BA mice with the MEK-inhibitor AZD624443,44 (MEKi) markedly suppressed the expression of foetal markers confirming the role of the MAPK pathway in the elaboration of the foetal-like phenotype in vivo (Supplementary Fig. 4a). Together, these results suggest that MAPK-activating mutations directly engage with YAP signalling and, consequently, induce foetal reprogramming in an epithelial cell-intrinsic, ERK-dependent manner. In this regard, the suppression of the foetal-like phenotype through YAP inhibition in vitro raised the possibility that the foetal-like differentiation program may be an adaptive survival mechanism directly initiated in Braf-mutant epithelium. To test the role of YAP in vivo using a genetics approach, we generated VillinCreER; BrafV600E/+; Alk5fl/fl; Yapfl/fl mice. Upon tamoxifen-induction, these mice rapidly developed weight loss and abdominal hunching, and had to be sampled on day 6 post induction due to these clinical signs. Histological evaluation of VillinCreER; BrafV600E/+; Alk5fl/fl; Yapfl/fl intestinal tissues confirmed the necessity of YAP for intestinal epithelial cell survival since, in its absence, we detected marked epithelial necrosis that was most pronounced in the proximal colon, although it was also evident throughout the rest of the intestine (Fig. 4f). Consistent with previous literature, tamoxifen-induced VillinCreER; Yapfl/fl mice showed no discernible phenotype46, confirming that YAP is dispensable for normal intestinal homeostasis.

We also observed that whilst Braf mutant organoids (B and BA) induced with tamoxifen in vitro grew in Matrigel, they could only be successfully passaged over the longer term in the presence of PGE2, a potent inflammatory mediator that has previously been shown to be a key component of the culture medium used to support the passage of foetal intestinal spheroids47. Endpoint tumour BA spheroids also exhibited this PGE2-dependency and maintained the same foetal-like phenotype in long-term culture as evidenced by LY6A expression and YAP1 nuclear localization (Fig. 4g). Given the locoregional inflammation in our BA model of right-sided colonic tumorigenesis, the requirement of exogenous PGE2 for long-term organoid passage is likely significant in light of the recently described requirement for pericryptal stromal production of PGE2 during post-injury wound healing48. This suggests that the combined inhibition of YAP and COX2, the enzyme that synthesizes PGE2, may hold therapeutic promise for the treatment of mutant BRAF-driven rCRCs.

Importantly, the ability of mutant BRAF alone to drive foetal-like differentiation demonstrated that TGFβ signalling does not suppress the foetal phenotype per se but, instead, independently restrains tumorigenesis. Scoring of d30 proximal colonic tissues for BrdU-positivity clearly showed that ablation of epithelial TGFβ-receptor signalling has no effect on epithelial cell proliferation (Fig. 4h). The elevated levels of apoptosis induced by mutant BRAF in vivo, however, were suppressed following epithelial-specific loss of TGFβ-receptor signalling, indicating reduced sensitivity to the initiation of cell death (Fig. 4i) as has previously been reported following loss of TGFβ-receptor signalling49. Taken together, these findings suggest that loss of TGFβ-receptor signalling allows colonic epithelial cells to evade apoptosis rather than augmenting their proliferative capacity.

Right-sided tumorigenesis is supported by microbial-driven inflammation

Our results indicate that mutant BRAF can initiate the foetal-like phenotype in organoids, via the activation of YAP, without additional genetic events or input from contextual cues. Nevertheless, the proximal colonic location of BA tumours—a region heavily colonized by intestinal microflora—and the presence of locoregional inflammation, evident from the early stages of tumour initiation, raised the possibility that local microenvironmental factors may influence tumorigenesis. Indeed, our data showed that the integrity of the mucosal barrier is compromised in BA tissues, likely allowing exposure of the epithelium to gut microbiota and toxic metabolites that can promote tumorigenesis by provoking inflammation24,50. To test this, we treated BA mice with broad-spectrum antibiotics from the time of tamoxifen-induction (Supplementary Fig. 6a). Antibiotics significantly extended survival and reduced the tumour number in BA mice, but this effect was only significant for right-sided tumours (Fig. 5a, b, Supplementary Fig. 6b). Comparison of RNA-seq analyses between antibiotic- and vehicle-treated d30 BA mice confirmed that eleven of the top fifteen pathways, significantly downregulated by antibiotic treatment, were associated with inflammation and the immune response (Supplementary Fig. 6c). Although still present, the foetal-like phenotype was dampened, with a reduction in the expression of foetal markers, including Ly6a, Ly6d, Krt7, and Anxa1, both in the tissue as a whole and in the epithelial compartment specifically (Fig. 5c, Supplementary Fig. 6d). These findings support the hypothesis that, although BRAF mutation alone drives foetal-like differentiation, microbial-driven inflammation is an additional key factor contributing to early tumour formation in the BA model, and that the local microbial milieu supports tumour initiation, right-sidedness, and disease progression. Our studies have, therefore, established a synergistic relationship between epithelial BrafV600E mutation, loss of TGFβ-receptor signalling, and the local inflammatory microenvironment in driving right-sided, foetal-like tumours.

Fig. 5: Inflammation and epithelial TGFβ-receptor loss cooperate in the formation of foetal-like tumours.

a Kaplan–Meier survival curves highlighting tumour-free survival (days after tamoxifen-induction) of vehicle-treated (n = 28) and antibiotic-treated (n = 28) BA mice (***p = 0.0006; Mantel-Cox log-rank test, two-tailed). b Total number of intestinal tumours in vehicle-treated (n = 28) and antibiotic-treated (n = 28) BA mice. Mean ± s.e.m. (****p = 2.9e−7; Mann–Whitney U-test, two-tailed). c qRT-PCR of indicated foetal markers in proximal colonic tissue from vehicle- and antibiotic-treated BA mice, 30 days post induction (n = 4 mice per treatment). Fold change is shown relative to vehicle-treated BA mice Mean ± s.e.m. (*p = 0.0286; Mann–Whitney U-test, two-tailed). d Kaplan–Meier survival curves highlighting survival (days after induction) of sulindac-treated A (n = 5) and B (n = 3) mice. e Intestinal tumour location scoring of sulindac-treated A mice (n = 5). Mean ± s.e.m. (ns not significant; *p = 0.0397, **p = 0.0079; Mann–Whitney U-test, two-tailed). f Representative ISH of sulindac-treated A proximal colonic tumours, stained for foetal markers Ly6a and Anxa1 (n = 5 mice). Scale bar, 100 μm. g GSEA plots of RNA-seq data from sulindac-treated A vs WT proximal colonic tissues, 30 days post induction, using the foetal spheroid (left panel) and Lgr5+ ISC (right panel) gene signatures (n = 4 mice per group). NES normalised enrichment score, FDR false discovery rate. (Left panel—p = 0.0, represents p < 0.001, Right panel—p = 0.008, calculated empirically.) h qRT-PCR of indicated foetal markers in proximal colonic tissue from sulindac-treated A mice vs WT counterparts, 30 days post induction (n = 4 mice per group). Fold change is shown relative to untreated WT mice. Mean ± s.e.m. (*p = 0.0286; Mann–Whitney U-test, two-tailed).

Inflammation drives colonic tumorigenesis in a MAPK-independent manner

We next sought to model the effects of wounding and inflammation to determine whether these were sufficient to initiate and drive tumorigenesis in B or A mice. High-dose sulindac treatment has previously been shown to result in proximal colonic epithelial damage51. Therefore, to provide a brief window of inflammation/wounding, we administered sulindac for 14 days immediately post induction (Supplementary Fig. 7a). This acute insult was sufficient to cause tumour formation in the proximal colon of A but not B mice (Fig. 5d). Tumours in sulindac-treated A mice were almost exclusively right-sided (Fig. 5e), consistent with the location of the injury51. Similarly to the BA model, sulindac-treated A tumours were invasive, exhibited a mucinous histology, and stained strongly positive for foetal transcripts in the epithelial compartment (Fig. 5f, Supplementary Fig. 7b). In addition, proximal colonic tissue harvested from sulindac-treated A mice on d30 post induction (16 days after cessation of sulindac treatment) showed a significant positive enrichment for the foetal-like signature and a negative enrichment for the Lgr5+ ISC signature, as observed in BA mice (Fig. 5g, h, Supplementary Table 1).

Given the capacity of BRAFV600E-driven MAPK activation to engage a foetal-like program directly in vitro, we next addressed whether MAPK activation was responsible for the foetal-like differentiation emerging through sulindac-driven inflammation in vivo. Utilising the MEK-inhibitor AZD6244, we demonstrated that the instigation of the foetal-like program is likely not dependent on MAPK activation in this inflammatory context in WT mice (Supplementary Fig. 7c).

The ability of sulindac-induced inflammation to drive right-sided tumorigenesis in A mice mirrored the topography of emergent tumours in BA mice. Therefore, we next addressed whether we could influence the location of tumour formation by guiding inflammation to a different site. DSS has previously been shown to drive tumorigenesis in TGFβ-receptor knockout mice52. Using this model, we confirmed that DSS-induced inflammation elicited tumour formation in A mice. As expected, the emergent colonic tumours were invasive, displayed a mucinous histology, lacked Wnt-pathway activation, and expressed foetal transcripts in their epithelium. Importantly, these tumours were located more distally in the colon, as would be expected with DSS treatment (Supplementary Fig. 7d–f). Collectively, these findings suggest that, in the presence of inflammatory stimuli driving foetal-like epithelial differentiation, loss of TGFβ-receptor signalling is permissive for the emergence of invasive colonic tumours at the site of inflammation.


An improved understanding of the molecular pathogenesis of mutant BRAF-driven rCRCs will inform the development of effective preventative and therapeutic strategies for this aggressive CRC subset. Preclinical models that faithfully recapitulate the salient features of disease initiation and progression, as bestowed by different oncogenic drivers, are an important tool in this endeavour. Here, we describe a mouse model of serrated intestinal cancer wherein oncogenic BRAF and loss of epithelial TGFβ-receptor cooperatively promote the formation of right-sided colonic tumours with a Wnt-low, foetal-like signature. Emergent tumours in this model recapitulate the topography, molecular landscape, and metastatic tropism of human right-sided disease, lending important insights into the pathogenesis of mutant BRAF-driven rCRCs that arise via the serrated neoplasia pathway.

While mutant BRAF-driven serrated tumours show a clear predilection for the right side of the colon in humans, most existing mouse models develop lesions primarily in the small intestine rather than the colon16,17,18. Interestingly, Rad et al. found that BRAFV600E induced widespread serrated hyperplasia throughout the intestinal tract. Yet, the vast majority of hyperplastic polyps that progressed to dysplasia in this model were located in the small intestine, with lesions exhibiting features of TSAs, rather than SSAs, which the authors attributed to the fact that SSAs typically emerge in the colon16. By contrast, BA tumours develop in the proximal/right colon and carry a Wnt-low, foetal-like signature which correlates with BRAF-mutant status, right-sidedness, MSI-high, CMS1 CRCs as well as pre-neoplastic SSAs in humans.

In further contrast to the BA model presented here, previous serrated tumorigenesis models exhibit deregulated Wnt signalling16,17,18. Notably, however, patient-derived right-sided serrated precursor lesions rarely harbour mutations in APC or CTNNB1, though mutations in lesser-known Wnt-pathway components can be acquired during the progression to dysplasia/more advanced stages53,54,55. Importantly though, BRAFV600E right-sided colonic tumours in patients have been shown to display membranous not nuclear β-catenin localisation19, and we therefore suggest that our BA model better recapitulates this tumour group.

Previous studies employing an inducible human BRAFV600K transgene resulted in epithelial serration and exhaustion/depletion of Lgr5+ ISCs, underpinned by their conversion to transit-amplifying (TA) progenitor cells18. By contrast, the depletion of Lgr5+ ISCs in the BA model is accompanied by the appearance of cells with a foetal-like phenotype. There are several differences between these models that may account for this discrepancy. First, the BA model develops SSA-like tumours in the proximal colon, whereas the transgenic BRAFV600K model develops hyperplasia and TSA-like dysplastic polyps in the small intestine. Aside from region-specific differences in basal proliferation rates and crypt dynamics, the signals and niche cell-types that determine cell-lineage allocation—i.e., the interconversion between adult and foetal-like stem cell populations and TA-progenitor cell states/fates—vary profoundly lengthways along the intestinal tract and up the vertical crypt-axis56. Accordingly, BRAFV600E-mutant small intestinal and colonic organoids differ profoundly in basal MAPK-pathway activity, with small intestinal organoids able to tolerate higher pathway activity upon BRAFV600E expression57. Notably, though, the transcriptomic profiles of BRAFV600E colonic organoids more closely resemble human BRAFV600E-mutant CRCs57. Second, the BA model expresses physiological levels of a knock-in BRAFV600E mutant driven by the endogenous Braf promoter, whereas the transgenic model relies on the overexpression of a human BRAFV600K transgene, which attains expression of MAPK-target genes akin to levels in more advanced dysplastic tumours. A reduction in ISC markers was also found in the hyperplastic intestinal tissue of BRAFV637E knock-in mice, which harbours comparable levels of MAPK activity to the transgenic model, although a TA cell fate was not expressly demonstrated16,18.

The conversion of the entire Lgr5+ ISC pool into TA progenitor cells, in the transgenic BRAFV600K model18, is reminiscent of normal crypt homeostasis. Therein, reduced EGFR/MAPK activity at the crypt base prevents the quiescent ISC pool from differentiating prematurely, while elevated MAPK signalling in the upper crypt dictates a proliferative TA progenitor cell fate58. Although the determinants underlying the different fates of BRAF-mutant cells in these two models remain unclear, the implications are profound: While reversion of Lgr5+ ISCs to a foetal-like state, in the absence of TGFβ signalling, fuels aggressive BA tumour growth, conversion to a TA-like state leads to exhaustion of the cancer stem cell pool and clonal extinction. This likely explains the finding that, in the transgenic model, the progression of dysplastic foci was frequently accompanied by the acquisition of Wnt/β-catenin-pathway mutations, which restored ISC marker expression18. Indeed, this is consistent with studies showing that BRAFV600E promotes loss of ISCs but, paradoxically, triggers differentiation; here, oncogenic progression is contingent on mutations that reverse differentiation (e.g. ablation of the differentiation-promoting transcription factors CDX2 or SMAD4) to restore stemness potential59. Accordingly, aging-induced acquisition of epigenetic alterations predisposes to BRAFV600E-driven transformation60, with several studies implicating cooperating inactivation of Cdkn2a/p16INK4a16, Cdx259,61, or both atypical PKCs (PKCζ and PKCλ/ι)62 in BRAFV600E-induced tumorigenesis. Similarly, our studies identify epithelial TGFβ-receptor signalling as an important gatekeeper to BRAFV600E-driven right-sided tumorigenesis. Indeed, we also show that the tumorigenic potential of foetal-like populations in the right colon is suppressed by epithelial TGFβ-receptor signalling, independent of the factors driving the foetal-like differentiation program, be it an acute inflammatory insult or aberrant MAPK activation. Indeed, this strongly suggests that the co-occurrence of TGFBR2 and BRAF mutations in patient rCRCs30provides foetal-like right-sided epithelial cells with a selective advantage.

Our study further shows that the MAPK pathway, via ERK, is fundamentally capable of driving a foetal-like differentiation program in the colonic epithelium in a YAP-dependent manner, and that this is further augmented in vivo by microbial-driven inflammation. Controversy surrounds the role of YAP in intestinal tumorigenesis, with both tumour-suppressive63,64 and oncogenic roles23,65,66,67 ascribed in different contexts. Several studies have demonstrated that YAP can inhibit Wnt signalling during intestinal regeneration and tumorigenesis23,63,64,68, consistent with the suppression of Lgr5 and the ISC signature in our BA model. In strong support of our findings, while this paper was under review, Reischmann et al. reported that the expression of BRAFV600E in colonic organoids elicits a foetal-like dedifferentiation program, associated with a transcriptional signature enriched for Hippo-pathway targets that resembles human BRAFV600E-driven CRC57. Interestingly, activation of the ERK and YAP pathways has also been detected in an inflammation-driven model of serrated intestinal tumorigenesis arising through atypical-PKC deficiency, independently of Braf (and Kras) mutations62, signifying that YAP dysregulation may commonly underlie distinct serrated neoplasia pathways independently of the driver mutation.

As multiple Lgr5+ and Lgr5 populations exhibit phenotypic plasticity and tumour-initiating potential within the intestine, questions remain as to the likely cell(s)-of-origin of the emergent tumours in the BA model and the serrated neoplasia pathway. A recent study identified a rare, quiescent population of revival stem cells, distinguished by elevated expression of clusterin (Clu) as well as Anxa1, Cxadr, and Basp1, that is mobilized and expanded following damage to the intestinal epithelium, regenerating all the major intestinal cell types, including Lgr5+ ISCs, in a YAP1-dependent manner69. It remains unclear whether different crypt progenitors and/or mature cell types are able to dedifferentiate to a foetal-like state in response to BRAF activation, or whether homeostatic colonic crypt subpopulations, such as the revival stem cells, acquire Braf mutation and initiate serrated tumorigenesis independently of Wnt signalling: the answer awaits lineage-tracing and single-cell genomics approaches in our BA model.

Our data have allowed us to identify and highlight the diverse nature of left and right-sided CRC in patients, showing that they are underpinned throughout tumour initiation and progression by different signalling pathways, which lead to either a Wnt high, Lgr5+ tumour phenotype or a BRAF-mutant, foetal-like, Lgr5-/Wnt-low phenotype, respectively. rCRCs that harbour oncogenic BRAF, however, continue to pose a major therapeutic challenge, with patients deriving little to no benefit from prevailing treatment regimens. While BRAF inhibitors have shown considerable efficacy in melanoma70, BRAF-mutant rCRCs are refractory to such treatment owing to EGFR-driven feedback activation of the MAPK pathway71,72. Of note, the YAP-driven regenerative response in the injured intestinal epithelium promotes cell survival by activating EGF signalling23. Accordingly, elevated YAP expression is predictive of intrinsic and acquired resistance to BRAF- and MEK-inhibitors in human BRAFV600E CRCs73, lending clinical relevance to our finding that YAP is a survival factor in BA tumours, and advocating for concurrent inhibition of YAP and BRAF-MAPK signalling in this setting. Moreover, a recent study in compound mutant organoids, with Rspo3-fusions and Kras/Braf, Trp53, and Smad4 mutations, implicated YAP/TAZ foetal reprogramming and lineage reversion in the acquisition of resistance to Wnt-targeted therapy. Significantly, emergent organoids were addicted to YAP/TAZ signalling and were sensitive to YAP/TAZ inhibition74. Encouragingly, foetal-like colonic BA tumour epithelium also maintains a dependence on therapeutically targetable, microenvironmental-derived inflammatory factors, such as PGE2. Additionally, the absolute requirement for YAP activation in the intestinal epithelium after BRAF mutation provides a further novel potential therapeutic target downstream of BRAF activation.

Our BA model recapitulates the topography, molecular landscape, and metastatic tropism of human Wnt-low, BRAF-mutant rCRCs and offers potential actionable therapeutic targets for this aggressive therapy-resistant subtype.

Article classification: Biological abstract
Share to:
Add:Room A11-329, 1st Floor, No.1, SBI Venture Street, Optics Valley, East Lake
New Technology Development Zone, Wuhan, China.
Certificate NO.:U18Q28010569R0S