Loss of BCL9/9l suppresses Wnt driven tumourigenesis in models that recapitulate human cancer.

Different thresholds of Wnt signalling are thought to drive stem cell maintenance, regeneration, differentiation and cancer. However, the principle that oncogenic Wnt signalling could be specifically targeted remains controversial. Here we examine the requirement of BCL9/9l, constituents of the Wnt-enhanceosome, for intestinal transformation following loss of the tumour suppressor APC. Although required for Lgr5+ intestinal stem cells and regeneration, Bcl9/9l deletion has no impact upon normal intestinal homeostasis. Loss of BCL9/9l suppressed many features of acute APC loss and subsequent Wnt pathway deregulation in vivo. This resulted in a level of Wnt pathway activation that favoured tumour initiation in the proximal small intestine (SI) and blocked tumour growth in the colon. Furthermore, Bcl9/9l deletion completely abrogated β-catenin driven intestinal and hepatocellular transformation. We speculate these results support the just-right hypothesis of Wnt-driven tumour formation. Importantly, loss of BCL9/9l is particularly effective at blocking colonic tumourigenesis and mutations that most resemble those that occur in human cancer.

Introduction

Deregulated Wnt signalling is a hallmark of colorectal cancer (CRC). This predominantly results from mutations in the tumour suppressor gene adenomatous polyposis coli (APC), which is found in 80% of the patients[1]. APC is a negative regulator of the canonical Wnt signalling pathway, forming part of the β-catenin destruction complex, with frequent mutation in CRC resulting in the hyperactivation of the pathway[2,3]. In the absence of Wnt signalling, APC associates with AXIN, Casein Kinase 1 (CK1) and glycogen synthase kinase 3 beta (GSK3β), which are required for the phosphorylation of β-catenin—marking it for ubiquitination and degradation[4]. Following Wnt activation or APC mutation, the complex is inactivated, whereby phosphorylated β-catenin can no longer be ubiquitinated, saturates the destruction complex and allows de novo synthesised β-catenin to translocate to the nucleus[5]. Nuclear β-catenin interacts with T-cell factor-1/lymphoid enhancer factor-1 (TCF/LEF1) transcription factors to drive target gene expression[6,7]. Additional transcriptional co-activators of β-catenin such as B-cell lymphoma 9 (BCL9)[8] and Pygopus[9] co-operate in β-catenin-mediated transcription, forming part of the Wnt enhanceosome[10].

The majority of APC mutations cluster in a specific region of the 5′ end of the gene, known as the mutation cluster region (MCR)[11]. The MCR encodes the 20 amino acid repeats (20AARs) which are required for β-catenin binding and degradation[12] and are truncated in CRC, leading to hyperactivated Wnt signalling. Interestingly, colon tumours retain on average two 20AARs[13], thought to result in a ‘just-right’ level of Wnt signalling, which may be sub-maximal[14]. There is evidence that the number of retained 20AARs influences CRC tumour location: proximal colonic tumours retained more than distal colonic tumours[15,16]. This tumour distribution could be influenced by the decreasing Wnt gradient that runs from the proximal to distal colon[15]. Leedham and colleagues proposed where tumours have high pathological Wnt signalling, proximal colonic tumour formation is unfavourable due to high underlying basal Wnt signalling levels in that region, instead distal colonic tumorigenesis is favoured[15]. Moreover, we recently showed that pharmacological reduction of Wnt signalling reduced intestinal stem cell (ISC) number, ISC competition and increased proximal small intestinal tumour formation in mice where Apc was deleted in the ISCs[17] These studies suggest that colon tumours select for APC mutations providing the optimal level of Wnt signalling and that Wnt signalling influences the size of the ISC pool as well as ISC competition.

There has been limited success in targeting Wnt signalling in CRC. Whilst some Wnt-driven cancers, such as those with RNF43 mutations or RSPO amplifications, appear sensitive to suppression of extracellular Wnt signalling using LRP6 blocking antibodies or Porcupine inhibition[18,19], these mutations are rare in CRC. Importantly, as the majority of CRCs carry APC mutations and are Wnt-ligand independent, there is a need to develop strategies that inhibit Wnt signalling in a ligand-independent manner[20]. This said, Tankyrase inhibitors, which stabilise AXIN, while exhibiting efficacy in CRC cell lines, have severe intestinal toxicity in vivo[21,22]. Additionally, cells that experience chronic Wnt signalling, including APC-mutant CRC cells are refractory to Tankyrase inhibition, due to the significant expression of BCL9l and LEF1 which shield β-catenin from AXIN-mediated destruction[23]. These studies highlight that successful Wnt-based therapies in CRC must act downstream of the destruction complex; disrupting binding of β-catenin to transcriptional activators. Importantly, proof of concept studies has shown that restoration of APC in aggressive carcinoma of mice causes tumour regression[24].

The β-catenin–TCF interface is large and dynamic, making it difficult to target[25]. However, significant interest exists in targeting the β-catenin–BCL9 association in the Wnt enhanceosome. BCL9 and BCL9l, functionally redundant mammalian homologues of the Drosophila gene legless, play a role in nuclear shuttling of β-catenin and promotion of β-catenin-dependent transcription[8,26,27]. While constitutive deletion of Bcl9 and Bcl9l is embryonically lethal[28], conditional deletion in the murine intestine is tolerated[29]. Deletion of Bcl9 and Bcl9l reduces colonic regeneration following acute colitis and decreases expression of Wnt target genes and ISC markers in colonic tumours generated by chemical carcinogenesis[29]. Hence, BCL9 and BCL9l have been proposed to regulate stemness within the intestinal crypts[30]. Furthermore, both are upregulated in human CRC[31,32] and overexpression of BCL9l significantly increased tumour formation in ApcMin/+ mice[33]. A number of small molecules targeting the β-catenin BCL9 interface have shown promise in APC-deficient cells both in vitro and in vivo[23,34], revealing that targeting of BCL9 and BCL9l may offer a therapeutic window in CRC.

As previous studies investigating BCL9 and BCL9l (BCL9/9l) were performed in colitis-associated cancer models, we wished to investigate the effect in models of cancer directly driven by activation of Wnt signalling either by Apc gene deletion or β-catenin stabilisation. We also sought to identify differences in the activation of oncogenic Wnt signalling when compared to homeostatic Wnt signalling to determine whether there was a therapeutic window for Wnt pathway inhibition following a mutation in the pathway. We report that deletion of Bcl9/9l sensitises the murine epithelium to perturbation of the Wnt pathway and impacts the Lgr5-ISC population. We show that BCL9/9l are required for the acute transformation of the intestine following homozygous deletion of Apc and for Wnt-driven transcriptional programmes associated with APC loss. Unexpectedly, we found that deletion of Bcl9/9l accelerated an APC-driven model of intestinal tumorigenesis and favoured adenoma formation within the proximal SI, but suppressed colonic tumour growth. However, if the β-catenin destruction complex is intact, BCL9/9l are absolutely required for mutant β-catenin-driven intestinal and hepatic transformation driven by mutant β-catenin. Moreover, Mieszczanek et al. (co-submitted manuscript) show that if mice carry a truncating mutation in Apc that is equivalent to human CRC, loss of Bcl9/9l makes these mice resistant to tumorigenesis. Crucially, we show that it is possible to reduce Wnt signalling to a level which prevents transformation in cancer cells which carry ligand-independent Wnt activating mutations, without disrupting normal homeostasis.

Results

BCL9/9l control intestinal Lgr5 expression

The ISC pool is regulated by Wnt signalling[17,35], with the highest levels defining the number of Lgr5 positive ISCs. Given that BCL9/9l are implicated in Wnt signalling and are required for colonic regeneration following acute colitis, we sought to determine whether Wnt signalling is perturbed in BCL9/9l-deficient intestines. To do this we generated VillinCreER Bcl9fl/fl Bcl9lfl/fl mice. Mice were injected with tamoxifen to induce intestinal Cre-mediated recombination, and harvested 4 days post-induction. Immunohistochemical staining confirmed accumulation of nuclear β-catenin at the base of small intestinal crypts following deletion of Bcl9/9l (Fig. 1a and Supplementary Figure 1a). Since BCL9/9l have been implicated in the nuclear shuttling of β-catenin, we performed a proximity ligation assay for E-cadherin: β-catenin complexes at the membrane. We observed that deletion of Bcl9/9l results in a significant increase in the number of E-cadherin: β-catenin complexes in small intestinal crypts compared with WT controls (Fig. 1b, c). Importantly, there was no difference in the expression of Ctnnb1 and Cdh1 between WT and BCL9/9l deficient crypts (Fig. 1d). These data suggest that there is a reduction in nuclear β-catenin following deletion of Bcl9/9l. To investigate whether this correlated with reduced Wnt signalling, we performed RNAseq on intestinal tissue from WT and VillinCreER Bcl9fl/fl Bcl9lfl/fl mice. Gene Set Enrichment Analysis (GSEA) revealed a negative enrichment for the Lgr5+ ISC gene signature (Fig. 1e and Supplementary Table 1). We confirmed reduced Lgr5 expression in BCL9/9l deficient crypts, while other Wnt target genes including Cd44, Axin2 and c-Myc or stem cell markers, Olfm4, were not altered (Fig. 1f; Supplementary Figure 1a & 1b). These data are consistent with our recent study and the work by the Kuo laboratory which showed that the Lgr5+ ISCs are the most dependent on the highest levels of Wnt signalling in the intestine[17,35]. This suggested that there may exist a therapeutic window for BCL9/9l loss in the intestine, with deletion preferentially impacting those cells most dependent upon higher levels of Wnt signalling, such as during regeneration or transformation.

Fig. 1

BCL9/9l control intestinal Lgr5 expression. a Representative β-catenin staining of small intestinal sections from Cre-induced WT and VillinCreER Bcl9fl/fl Bcl9lfl/fl mice sampled 4 days post tamoxifen injection. Red arrows indicate nuclear β-catenin at the base of the crypt. Scale bar = 50 µm. b Proximity ligation assay for E-cadherin: β-catenin complexes (red) from small intestinal sections of small intestinal sections from mice described in (a). Nuclei stained with DAPI (blue). Paneth cells appear red, but were excluded from analysis. Scale bar = 50 µm. c Quantification of proximity ligation assay, % positive area per crypt (E-cadherin: β-catenin) was quantified and 10 crypts were scored per mouse, n = 3 per group, one-way Mann–Whitney U test, P = 0.04. Data displayed as mean  ±SEM. d qPCR for Cdh1 and Ctnnb1 expression in small intestine of mice described in (a), n = 3 per group, one-way Mann–Whitney U test, P = 0.35 (Cdh1) and P = 0.2 (Ctnnb1). Data displayed as mean  ±SEM. e Gene Set Enrichment Analysis of RNAseq data obtained from small intestinal tissue from WT and VillinCreER Bcl9fl/fl Bcl9lfl/fl mice, n = 3 per group. f Representative Lgr5-RNAscope and β-catenin staining of small intestinal sections from mice described in (a). Red arrows indicate Lgr5-staining at the base of the crypt. Scale bar = 50 µm

BCL9/9l contribute to intestinal stem cell fitness

It has been shown that ISCs exhibit neutral drift dynamics[36,37] and that replacement of one stem cell by a neighbour is a stochastic process. Clonal fitness of ISCs has been investigated by tracing the clonal expansion of stem cells over time[38]. To determine whether the loss of the Lgr5 gene signature following deletion of BCL9/9l is functionally relevant in terms of ISC fitness, we induced Cre recombination in Lgr5-EGFP-CreER tdTomfl/+ and Lgr5-EGFP-CreER tdTomfl/+ Bcl9fl/fl Bcl9lfl/fl mice with 0.15 mg tamoxifen, so as to recombine, on average, in a single ISC per crypt, labelling that ISC with a tomato reporter (Fig. 2a). We observed a significant reduction in average clone size at 4 and 21 days post-induction following deletion of Bcl9/9l compared with WT ISCs (Fig. 2b). Moreover, we observed a profound increase in the number of partially fixed crypts at 21 days post-induction arising from BCL9/9l-null ISCs compared with WT ISCs (Fig. 2c). Interestingly, of fully fixed clones (completely tomato-positive crypts) from Lgr5-EGFP-CreER tdTomfl/+ Bcl9fl/fl Bcl9lfl/fl mice, 31 out of 35 retained Lgr5 expression, indicating that these crypts may have escaped recombination (Fig. 2d). Therefore, the apparent reduction in ISC fitness observed in BCL9/9l-deficient ISCs may be an underestimate of the true effect. Moreover, at day 21 there is a significant reduction in the number of tomato positive crypts in Lgr5-EGFP-CreER tdTomfl/+ Bcl9fl/fl Bcl9lfl/fl mice compared with controls, suggesting that BCL9/9l-deficient ISCs have been replaced by WT neighbours (Fig. 2e). These data suggest that BCL9/9l are required for the efficient function of Lgr5-positive stem cells, and that in their absence, ISCs have reduced fitness compared with WT ISCs. Importantly, despite this ISC phenotype, we showed that deletion of Bcl9/9l does not perturb intestinal homeostasis (Supplementary Figure 2a & b) in agreement with other studies[29].

Fig. 2

BCL9/9l contribute to intestinal stem cell fitness. a Representative images from Lgr5-EGFP-CreER tdTomfl/+ (WT) and Lgr5-EGFP-CreER tdTomfl/+ Bcl9fl/fl Bcl9lfl/fl (BCL9 BCL9l) mice injected with 0.15 mg tamoxifen and sampled 10 days later, blue = DAPI (nuclei) and red = tomato. Scale bar = 100 µm. b Average clone size from Lgr5-EGFP-CreER tdTomfl/+ (WT) and Lgr5-EGFP-CreER tdTomfl/+ Bcl9fl/fl Bcl9lfl/fl (BCL9 BCL9l) mice injected with 0.15 mg tamoxifen and sampled 4, 10 and 21 days post-induction. 200 clones per mouse were scored, n = 3 per group at each time point, one-way Mann–Whitney U test, *P = 0.04. c Heat map for % distribution of clone size from mice described in (b). Blue = low and red = high. Pie charts represent fixation of crypts—red = tomato labelled ISC, green = unlabelled ISC. d Serial sections of RFP and Lgr5-RNAscope staining on small intestines from Lgr5-EGFP-CreER tdTomfl/+ (WT) and Lgr5-EGFP-CreER tdTomfl/+ Bcl9fl/fl Bcl9lfl/fl (BCL9 BCL9l) mice sampled 21 days post-induction. Scale bar = 50 μm. Red arrows indicate RFP and Lgr5 positive crypts, whilst blue arrows indicate RFP positive and Lgr5 negative crypts. e Quantification of the number of tomato-positive clones/field of view from mice described in (d), sampled 21 days post-induction, n = 3 per group, one-way Mann–Whitney U test, P = 0.04. Data displayed as mean  ±SEM

BCL9/9l are required for intestinal regeneration

We next addressed whether the loss of BCL9/9l could affect other phenotypes associated with Wnt signalling in the intestine. To determine whether Bcl9/9l deletion sensitises the SI to a further reduction of Wnt signalling, we treated VillinCreER Bcl9fl/fl Bcl9lfl/fl mice with the porcupine inhibitor LGK974, blocking Wnt ligand secretion[19]. This resulted in severe crypt atrophy and loss of proliferation after just 3 days (Fig. 3a, b; Supplementary Figure 3c). Vehicle-treated mice displayed no phenotype even after 30 days of treatment (Fig. 3a, b), with previous experiments demonstrating that LGK974 treatment is well tolerated in wildtype mice up to 50 days and beyond[17]. Wnt signalling is also essential for crypt/organoid culture in vitro. Importantly, BCL9/9l-deficient small intestinal crypts fail to establish in vitro (Supplementary Figure 3a & b).

Fig. 3

BCL9/9l are required for intestinal regeneration. a Survival plot for Cre-induced VillinCreER Bcl9fl/fl Bcl9lfl/fl mice treated with 5 mg/kg LGK974 or vehicle, twice daily starting 24 h after tamoxifen injection, n = 5 per group. Log-rank test, P-value = 0.0040. b Representative H&E (upper panel) and BrdU (lower panel) staining of mice described in (a), sampled at end-point (LGK974) or at a time-point (vehicle). Scale bars = 50 µm. c Wildtype (WT) and VillinCreER Bcl9fl/fl Bcl9lfl/fl mice were induced with tamoxifen and then 4 days post Cre-induction exposed to 10 Gy γ-irradiation and sampled 72 h later. The number of regenerating crypts per circumference in the small intestine was quantified and averaged per mouse, n = 4 per group, one-way Mann–Whitney U test, P = 0.0143. Data displayed as mean  ±SEM. d Representative H&E stains of small intestines from mice described in (c). Red arrows indicate regenerating intestinal crypts. Scale bar = 50 µm

Finally, the ability of the intestine to regenerate after insult is also Wnt-dependent. Therefore, we investigated whether BCL9/9l are required for intestinal regeneration following irradiation. To this end, Cre-induced VillinCreER Bcl9fl/fl Bcl9lfl/fl and wildtype mice were culled at a time-point 72 h post-irradiation (10 Gy γ-radiation). Histological analysis revealed that there was a significant reduction in the number of regenerating small intestinal crypts following deletion of Bcl9/9l (Fig. 3c, d), indicating their requirement for intestinal regeneration. Together, these data suggest that whilst dispensable for intestinal homeostasis, BCL9/9l are absolutely required following deregulation of the Wnt pathway in the intestine.

BCL9/9l are required for acute intestinal transformation

We next investigated the role for BCL9/9l in intestinal neoplasia, a process characterised by hyperactivated Wnt signalling following APC loss.

Homozygous deletion of Apc throughout the murine intestine leads to the acute transformation of the epithelium. The ensuing crypt-progenitor phenotype consists of significantly increased proliferation, perturbed differentiation and migration, and is maximally penetrant at 4 days post Cre-induced Apc deletion[39]. Therefore, we generated VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl animals. Coincident deletion of Bcl9/9l with APC loss strongly suppressed the crypt-progenitor phenotype, with a significant reduction in proliferation in both the SI and colon compared to controls (Fig. 4a–c and Supplementary Figure 4a–c). To determine whether this reduction in proliferation was due to a larger reduction in Wnt target gene expression, we performed RNAseq on intestinal tissue from Cre-induced VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice. This revealed downregulation of a number of Wnt target genes following Bcl9/9l deletion. Subsequent GSEA confirmed negative enrichment of genes upregulated following acute Apc deletion (Fig. 4d and Supplementary Table 2). These downregulated Wnt target genes including Lgr5, Axin2, Cd44 and SOX9,  were confirmed through RNA in situ hybridisation (RNAscope), immunohistochemical staining and qPCR (Fig. 4e–g; Supplementary Figure 4e). Importantly, many of these genes were unaffected in the normal intestine following Bcl9/9l deletion (e.g. Axin2, Cd44 and Sox9). Deletion of Apc disrupts the destruction complex, stabilising β-catenin, allowing translocation to the nucleus. Immunohistochemical staining confirmed accumulation of nuclear β-catenin within the small intestinal crypts of both Cre-induced VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice (Supplementary Figure 4d). This suggests that BCL9/9l play a critical role in the nucleus as part of the Wnt enhanceosome, alongside their role in shuttling β-catenin. Interestingly, RNAseq data from WT vs VillinCreER Bcl9fl/fl Bcl9lfl/fl and VillinCreER Apcfl/fl vs Cre-induced VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice revealed that BCL9/9l regulate many more genes following APC loss, with 129 genes differentially expressed between WT and BCL9/9l deficient intestinal epithelia and 655 differentially expressed genes between intestines from Cre-induced VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice (Supplementary Figure 4f). Indeed, only 57 differentially expressed genes are shared between the two datasets (Supplementary Figure 4f). Hence, in homeostasis the main role of BCL9/9l appears to be control of the Lgr5-positive ISC pool, while upon transformation of the epithelium following APC loss, BCL9/9l are required to drive oncogenic Wnt transcriptional programmes and consequently acquisition of the crypt-progenitor phenotype.

Fig. 4

BCL9/9l are required for acute intestinal transformation. a Representative H&E staining of small intestines from WT, VillinCreER Bcl9fl/fl Bcl9lfl/fl, VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice sampled 4 days post Cre-induction. Red bars indicate the size of the proliferative crypt. Scale bar = 50 µm. b Representative BrdU staining of small intestines from mice described in (a). Mice were injected intraperitoneally with BrdU 2 h prior to being culled. Red bars indicate the size of the proliferative crypt. Scale bar = 50 µm. c Quantification of proliferation (BrdU positive cells) in the small intestines of mice described in (a). The number of BrdU-positive cells per half crypt was quantified, 25 crypts per mouse scored, n = 3–4 for each group, one-way Mann–Whitney U test, P = 0.014 for VillinCreER Apcfl/fl vs VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl and P = 0.35 for WT vs VillinCreER Bcl9fl/fl Bcl9lfl/fl. WT data from Supplementary Figure 2b. Data displayed as mean  ±SEM. d Gene Set Enrichment Analysis of RNAseq data from small intestinal tissue from Cre-induced VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice, n = 3 per group. e Representative Lgr5-RNAscope staining of small intestines from mice described in (a). Red arrows indicate Lgr5-staining at the base of WT crypts. Scale bar = 50 µm. f Representative SOX9 staining of small intestines from mice described in (a). Scale bar = 50 µm. g qPCR for Wnt target genes and intestinal stem cell markers from small intestinal tissue of Cre-induced VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice, n = 4–5 per group, one-way Mann–Whitney U test, **P < 0.01, P = 0.00795 (Bcl9), P = 0.00395 (Bcl9l), P = 0.00395 (Lgr5), P = 0.00395 (Cd44), P = 0.00395 (Axin2) and P = 0.00795 (c-Myc). Data displayed as relative to the mean of VillinCreER Apcfl/fl mice. Data displayed as mean  ±SEM

Deletion of Bcl9/9l alters intestinal tumour distribution

We demonstrated that inhibition of Porcupine resulted in reduced ISC number and a rapid fixation of mutant clones[17]. This accelerated tumorigenesis following APC loss due to rapid clonal fixation of crypts, and increased proximal small intestinal lesion number. Given that Bcl9/9l deletion caused a very similar phenotype, reduction of the Lgr5-positive ISC pool, one might predict they would phenocopy this data. However, since Bcl9/9l deletion was associated with a reduction of Wnt target gene expression following APC loss and a suppression of the crypt-progenitor phenotype, one might also predict reduced tumorigenesis. We examined whether loss of BCL9/9l modified tumorigenesis in VillinCreER Apcfl/+ mice. Deletion of Bcl9/9l significantly accelerated intestinal tumorigenesis and reduced survival (Fig. 5a). This reduction in survival was due to a significant increase in tumour burden—with mice developing hundreds of small lesions (Fig. 5b–d). Moreover, deletion of Bcl9/9l resulted in a profound alteration in tumour distribution. Tumours were uniformly distributed between the proximal and distal SI in Cre-induced VillinCreER Apcfl/+ mice, while upon deletion of Bcl9/9l there was a significant increase in the number of proximal intestinal tumours (Fig. 5e). The tumours that arose were deficient for BCL9, and despite being positive for nuclear β-catenin, were also negative for Lgr5 (Fig. 5f). Next, we sampled Cre-induced VillinCreER Apcfl/+ and VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl mice at 50 days post-induction to confirm an increased rate of tumour formation. The number of both macroscopic and microscopic lesions was significantly increased in BCL9/9l-null intestines, with a concomitant shift towards proximal small intestinal lesion formation also observed (Supplementary Figure 5a & b).

Fig. 5

Deletion of Bcl9/9l alters intestinal tumour distribution. a Survival curve for VillinCreER Apcfl/+ and VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl mice aged until clinical end-point, n = 16 for VillinCreER Apcfl/+ (3 censors—1 mouse had lymphoma, 1 with elongated teeth and another displayed rapid weight loss without significant tumour burden) and n = 19 for VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl, Log-rank test, P = 0.0002. b Plot for total intestinal and colonic tumours from mice described in (a), n = 12 for VillinCreER Apcfl/+ and n = 16 for VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl, one-way Mann–Whitney U test, P < 0.0001. Data displayed as mean  ±SEM. c Plot for average tumour size from intestines and colons from mice described in (a), n = 12 for VillinCreER Apcfl/+ and n = 16 for VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl, one-way Mann–Whitney U test, P < 0.0001. Data displayed as mean  ±SEM. d Plot for total intestinal and colonic tumour burden from mice described in (a), n = 12 for VillinCreER Apcfl/+ and n = 16 for VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl, one-way Mann–Whitney U test, P < 0.0002. Data displayed as mean  ±SEM. e Plot of tumour distribution along the small intestine and colon of mice described in (a), n = 7 for VillinCreER Apcfl/+ and n = 14 for VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl, one-way Mann–Whitney U test, P < 0.0001 (proximal SI and distal SI) and P = 0.37 for colon. Data displayed as % of total tumours/mouse. Data displayed as mean  ±SEM. f Representative staining for BCL9 (top panel), β-catenin (middle panel) and Lgr5-RNAscope (bottom panel) of small intestinal tumours from mice described in (a). Scale bars = 100 µm. g Representative images from a colonoscopy of VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice induced with a single injection of 4-hydroxytamoxifen into the colonic sub-mucosa. Red arrows indicate tumours. h Quantification of colonic tumour growth (normalised to the size of the lumen) of mice described in (g), n = 6 for VillinCreER Apcfll/fl and n = 4–5 for VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl, one-way Mann–Whitney U test, P = 0.13 (2 weeks) and P = 0.04 (4 weeks). Data displayed as mean  ±SEM

To determine a role for BCL9/9l in colonic tumour growth we injected 4-hydroxy tamoxifen into the colonic submucosa of VillinCreER Apcfl/fl and VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice to induce local recombination[40]. Mice were imaged via colonoscope at 2 and 4 weeks post-induction. Small tumours were visible in both groups 2 weeks post-induction; however after 4 weeks, tumours of VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice were significantly smaller than those of VillinCreER Apcfl/fl mice (Fig. 5g, h). This reduced rate of colonic tumour formation translated into a survival benefit with Bcl9/9l deficiency (Supplementary Figure 6a). The colonic tumours that formed in VillinCreER Apcfl/fl Bcl9fl/fl Bcl9lfl/fl mice retained expression of both Bcl9 and Bcl9l (Supplementary Figure 6b), suggesting that BCL9/9l are required for colonic tumour growth following the loss of APC.

A possible explanation for the increased rate of tumour formation is that deletion of Bcl9/9l accelerated loss of the second copy of Apc. This could be achieved through DNA damage and loss of heterozygosity (LOH). We examined the abundance of γH2AX, a marker of DNA damage through IHC staining of intestinal tissue or tumours either proficient or deficient in BCL9/9l and found no difference (Supplementary Figure 7a). To examine LOH, we designed a high sensitivity in situ hybridisation probe (Basescope) to specifically detect exon 14 of Apc[17]. Given that this exon is specifically deleted in the Apcfl allele following Cre-mediated recombination, it follows that if the Apc locus undergoes LOH during tumour formation the probe will be undetected in resulting tumours. We observed that staining was negative in tumours, but positive in the adjacent normal epithelium, from both Cre-induced VillinCreER Apcfl/+ and VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl mice (Supplementary Figure 7b), suggesting that in both cases, the second copy of Apc is lost via LOH.

BCL9/9l are required for mtCtnnb1 intestinal transformation

The Apc allele used in the previous experiments is truncated at codon 580 and therefore lacks any β-catenin binding activity[41], whereas humans mutated APC frequently retains some binding. In the co-submitted article, Mieszczanek et al. recapitulate the increase in proximal intestinal tumour initiation following Bcl9/9l deletion in the SI of ApcMin/+ mice—an allele which lacks β-catenin binding activity[42], while demonstrating that deletion almost completely blocks tumour formation in Apc1322T/+ mice—an allele which retains some β-catenin binding activity. Hence, we hypothesised that the larger the APC protein and therefore more β-catenin binding activity that is retained, the more dependent cancer cells would become upon BCL9/9l for Wnt driven transformation.

To investigate this, we used a model that expresses a mutant form of β-catenin that cannot be degraded, resulting in intestinal transformation[15]. We hypothesised that in this instance, deletion of Bcl9/9l would slow transformation. Expression of a single copy of mutant β-catenin (Ctnnb1ex3/+) throughout the murine intestine and colon transformed the SI leading to a crypt-progenitor phenotype reminiscent of Apcfl/fl intestines, after approximately 21 days post Cre-induction. Deletion of Bcl9/9l significantly extended survival (Fig. 6a) and at endpoint, these mice had developed tumours as opposed to displaying the crypt-progenitor phenotype like the control cohort (Supplementary Figure 8a & b). Importantly, these tumours were escapers that retained expression of both Bcl9 and Bcl9l (Supplementary Figure 8c & d). Sampling VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice 21 days post Cre-induction confirmed that deletion of Bcl9/9l prevents the expansion of Ctnnb1 mutant crypts. Intestinal crypts from Cre-induced VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice are significantly smaller and exhibited a significant reduction in the number of BrdU positive cells when compared to VillinCreER Ctnnb1ex3/+ mice (Fig. 6b, c). There was a significant decrease in expression of a number of Wnt target genes including Lgr5, Axin2, SOX9 and CD44 in crypts from Cre-induced VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice compared with Cre-induced VillinCreER Ctnnb1ex3/+ mice (Fig. 6c–e). This complete suppression of the mutant β-catenin-driven phenotype by deletion of Bcl9/9l was consistent with our proposed model in which oncogenic Wnt signalling above a threshold of Wnt activation is required for transformation.

Fig. 6

BCL9/9l are required for mtCtnnb1 intestinal transformation. a Survival curve for Cre-induced VillinCreER Ctnnb1ex3/+ and VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice aged until clinical endpoint, n = 5 for VillinCreER Ctnnb1ex3/+ and n = 8 for VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl, log-rank test, P = 0.0001. b Quantification of proliferation (BrdU positive cells) in the small intestines of Cre-induced VillinCreER Ctnnb1ex3/+ sampled at end-point and VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice sampled 21 days post Cre-induction. The number of BrdU-positive cells per half crypt was quantified, 25 crypts per mouse scored, n = 3 for each group, one-way Mann–Whitney U test, P = 0.04. Data displayed as mean  ±SEM. c Representative H&E (upper panels—red bar indicates the size of the proliferative zone), BrdU (middle panels) and Lgr5-RNAScope (lower panels) staining of small intestinal sections from Cre-induced VillinCreER Ctnnb1ex3/+ (at end-point) and VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl (sampled at day 21) mice. Mice were injected with BrdU intraperitoneally 2 h prior to being culled. Scale bars = 50 µm. d Representative β-catenin (upper panel), Bcl9-RNAscope (middle panel) and Bcl9l-RNAscope (lower panel) staining of small intestinal sections from mice described in (b). Scale bars = 50 µm. e Representative Axin2-RNAscope (upper panel), SOX9 (middle panel) and CD44 (lower panel) staining of small intestinal sections from mice described in (b). Scale bars = 50 µm

To understand the contribution of BCL9/9l to the transformation of the SI driven by Ctnnb1 mutations, we generated VillinCreER Ctnnb1ex3/ex3 Bcl9fl/fl Bcl9fl/fl mice. These mice were induced and sampled 4 days post Cre-induction. Bcl9/9l deletion significantly reduced proliferation in both the SI and colon of Cre-induced VillinCreER Ctnnb1ex3/ex3 mice (Supplementary Figure S9a–d and Supplementary Figure 10a–c). Following transcriptional profiling of small intestinal tissue from Cre-induced VillinCreER Ctnnb1ex3/ex3 and VillinCreER Ctnnb1ex3/ex3 Bcl9fl/fl Bcl9fl/fl mice, GSEA demonstrated that gene programmes upregulated following Apc deletion[39] or enriched in human CRC[43] are suppressed upon Bcl9/9l deletion (Supplementary Figure 9g and Supplementary Table 3). Moreover, Bcl9/9l deletion also significantly reduced the expression of a number of Wnt target genes, including Lgr5, Axin2, Cd44 and c-Myc (Supplementary Figure 9e & f). The relative fold-change in expression of a subset of Wnt target genes including Cd44 and c-Myc following Bcl9/9l deletion was more pronounced in the VillinCreER Ctnnb1ex3/ex3 setting (6- and 4.9-fold, respectively), compared to VillinCreER Apcfl/fl (1.9- and 1.7-fold, respectively) (Supplementary Figure 11a & b). This supports the notion that cells which maintain a functional APC protein with significant β-catenin binding capacity have a higher dependency upon BCL9/9l for Wnt-driven transformation.

BCL9/9l loss increases membrane β-catenin in mtCtnnb1 crypts

To understand the relative dependencies of β-catenin mutant and APC-deficient tumours on BCL9/9l, we examined the intracellular distribution of β-catenin. Whilst in both cases we observed nuclear β-catenin staining (Figs. 5f and 6d) there was an increase in membranous β-catenin in crypts from VillinCreER Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice sampled at 21 days post Cre-induction compared with Cre-induced VillinCreER Ctnnb1ex3/+ mice (Supplementary Figure 12a). From this we infer that the level of nuclear β-catenin is reduced following deletion of BCL9/9l. Interestingly, there was no difference in membrane-associated β-catenin in tumours from Cre-induced VillinCreER Apcfl/+ and VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl mice (Supplementary Figure 12b). This would argue that the larger the APC protein retained, the more dependent the cell is on BCL9/9l for the efficient shuttling of β-catenin from the cytoplasm and the membrane to the nucleus.

BCL9/9l are required for mtCtnnb1 hepatocyte transformation

Hepatocellular carcinoma (HCC) is characterised by activating mutations of β-catenin in approximately 30% of patients[44]. To investigate a role for BCL9/9l in liver transformation, we utilised an adeno-associated virus (AAV) system to express Cre recombinase specifically in hepatocytes to induce recombination of target genes[45]. We chose a titre of AAV8-TBG-Cre virus that gives near constitutive recombination of the adult liver. Acute deletion of Bcl9/9l did not perturb liver homeostasis or proliferation (Supplementary Figure 13a & b), though aged AAV8-TBG-Cre Bcl9fl/fl Bcl9lfl/fl mice did have a small yet significant reduction in liver-to-body weight ratio compared to controls (Supplementary Figure 13c). Moreover, we observed a reduced expression of the zonation marker and Wnt target gene Glutamine Synthetase after 140 days post-induction (Supplementary Figure 13a). We investigated the role for BCL9/9l in β-catenin-driven hepatocyte transformation through generation of Ctnnb1ex3/+ and Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice and treated with AAV8-TBG-Cre virus. Acute hepatic expression of a single copy of mutant β-catenin leads to an increase in the expression of a number of Wnt target genes, including Lgr5, Axin2, Bcl9 and Bcl9l 4 days post-induction, in a BCL9/9l dependent manner (Supplementary Figure 14a–d). AAV8-TBG-Cre Ctnnb1ex3/+ mice develop hepatomegaly within 2 weeks post-induction, and develop a liver failure phenotype which was completely suppressed through deletion of Bcl9/9l, where mice were aged up to 140 days before being euthanised (Fig. 7a and Supplementary Figure 14e). Interestingly after 140 days these mice did exhibit small liver lesions, deficient for Bcl9/9l expression (Supplementary Figure 14f).

Fig. 7

BCL9/9l are required for mtCtnnb1 hepatocyte transformation. a Survival curve for AAV8-TBG-Cre induced Ctnnb1ex3/+ and Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice aged until clinical end-point or time-point, n = 5 per group, Log-rank test, P = 0.0008. b Liver-to-body weight ratio (%) of AAV8-TBG-Cre induced WT (sampled at day 4), Ctnnb1ex3/+ (sampled at end-point) and Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl (sampled at day 14), n = 3–5 per group, one way Mann–Whitney U test, P = 0.018 AAV8-TBG-Cre WT vs AAV8-TBG-Cre Ctnnb1ex3/+ and P = 0.008 for AAV8-TBG-Cre Ctnnb1ex3/+ vs AAV8-TBG-Cre Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl. Data displayed as mean  ±SEM. c Quantification of BrdU positive hepatocytes from AAV8-TBG-Cre Ctnnb1ex3/+ (sampled at end-point) and Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl (sampled at day 14). Number of BrdU positive hepatocytes scored per ×20 objective field of view (FOV), 10 FOVs scored per mouse, n = 4 per group. One-way Mann–Whitney U test, P = 0.014. Data displayed as mean  ±SEM. d Representative staining for SOX9, Axin2-RNAscope, Lgr5-RNAscope, Glutamine Synthetase (red arrows indicate the central vein and blue arrows indicate the portal tract areas, respectively), β-catenin and BrdU in liver sections from AAV8-TBG-Cre induced Ctnnb1ex3/+ and Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice aged until clinical end-point or time-point. Scale bar = 50 µm

Given the survival extension provided by deletion of Bcl9/9l, we chose to compare livers from AAV8-TBG-Cre Ctnnb1ex3/+ and AAV8-TBG-Cre Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice sampled at day 14. Here, there was a significant reduction in liver-to-body weight ratio of AAV8-TBG-Cre Ctnnb1ex3/+ Bcl9fl/fl Bcl9lfl/fl mice compared with AAV8-TBG-Cre Ctnnb1ex3/+ mice (Fig. 7b). The reduced liver-to-body weight ratio was also accompanied by a significant reduction in the number of BrdU positive hepatocytes (Fig. 7c). We also observed downregulation of a number of Wnt targets, including SOX9, Lgr5 and Axin2 following deletion of Bcl9/9l (Fig. 7d). Additionally, mutant β-catenin drove the expansion of the Glutamine Synthetase positive zone around the central vein; which was suppressed by deletion of Bcl9/9l (Fig. 7d).

Discussion

Since the finding that restoration of APC expression can cause regression of aggressive CRC, there has been renewed interest in targeting Wnt signalling in APC deficient CRC[24]. Although an excellent proof of concept, there is a paucity of strategies to target the Wnt pathway in ligand independent cancers that have mutations in either APC or CTNNB1. Our work highlights the possibility that inhibiting BCL9/9l could provide an excellent strategy. Moreover, we elucidate key threshold levels of Wnt signalling which differentiate normal homeostasis from transformation. Most importantly, our study and the co-submitted study by Mieszczanek et al. highlight that the APC mutations associated with human CRC that retain β-catenin binding will be most sensitive to BCL9/9l inhibition.

Key among our conclusions is the elucidation of two different transcriptional programmes driven by BCL9/9l during homeostasis and transformation. We find that BCL9/9l are dispensable for intestinal homeostasis, although required for the expression of the ISC marker Lgr5 as previously suggested[29]. Importantly not only did acute deletion of Bcl9/9l lead to the loss of the Lgr5+ ISC gene signature, it functionally reduced ISC fitness compared to WT ISCs. This perturbation in the Lgr5+ ISC pool following Bcl9/9l deletion may explain why BCL9/9l-null intestinal crypts have reduced capacity to regenerate following irradiation since Lgr5+ ISCs are required for intestinal regeneration[46]. We observed that there was a significant increase in the E-cadherin bound β-catenin following deletion of Bcl9/9l, agreeing with previous studies highlighting a role for these two proteins in the nuclear shuttling of β-catenin.

Despite hyperactive Wnt signalling being a hallmark of CRC, efficacious therapies against the pathway are limited. This is due to the lack of agents that act downstream of the destruction complex. As BCL9/9l are required for a subset of β-catenin-mediated transcriptional targets in the normal intestine and are dispensable for intestinal homeostasis we investigated whether they play a role in intestinal epithelial transformation. Acute deletion of both copies of Apc rapidly transforms the murine intestine within 4 days[39]. This hyperproliferative phenotype along with the expression of a large number of Wnt target genes was suppressed following deletion of Bcl9/9l. GSEA revealed that BCL9/9l are required for the expression of a transcriptional programme that is induced following APC loss in the mouse intestine[39] and Wnt target genes that are upregulated in human CRC[43]. This suggests that BCL9/9l are required for an oncogenic β-catenin-mediated transcriptional programme which permits the acute transformation of the murine intestine following APC loss. It is important to note that with the exception of Lgr5, many Wnt target genes that are upregulated following APC loss, including Axin2, Cd44 and c-Myc were specifically reduced when Bcl9/9l were deleted concurrently with Apc, as opposed to in the WT intestinal epithelium where these genes were unaffected.

A key observation from the tumour models driven by APC loss was that deletion of Bcl9/9l favoured adenoma formation within the proximal SI, whilst colonic tumour growth was suppressed. It has previously been proposed that human CRC tumours have a ‘just-right’ level of Wnt signalling, which may in fact be sub-maximal due to truncated APC proteins that retain β-catenin binding sites[13,14]. Furthermore, a decreasing Wnt gradient has been described from the proximal SI to the distal colon[15]. Together, the basal level of Wnt signalling within the intestinal epithelium along with intra-tumoural Wnt signalling has been proposed to influence the regional distribution of tumours in both humans and mice. For instance, ApcMin/+ and Apc1322T/+ mice, which harbour different truncating Apc mutations, have distinct tumour distributions along their SI due to the different levels of Wnt signalling within their tumours[15,42]. Tumours from ApcMin/+ mice have high levels of Wnt signalling which is not permissive for proximal SI tumour formation; hence tumours form in the distal SI. Conversely, Apc1322T/+ tumours retain two β-catenin binding domains and consequently have sub-maximal Wnt signalling, permitting tumour formation in the proximal intestine[15,42]. This in turn may explain the tumour distribution we observe in VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl mice, since our Apc580S allele lacks any β-catenin binding sites[41]. Here deletion of Bcl9/9l reduces intratumoural Wnt signalling, favouring tumour formation in the proximal SI, but is not permissive for colonic tumour formation due to the relatively low underlying basal Wnt signalling; observations which support the ‘just-right’ Wnt signalling hypothesis. Here our data is very consistent with the submission of Mieszczanek et al. They show that the ApcMin/+ mouse also develops a large number of small proximal intestinal tumours when Bcl9/9l is deleted, whilst Apc1322T/+ mice are resistant to tumorigenesis. Therefore loss of BCL9/9l now reduces Wnt signalling activation to a level that can no longer transform the intestine. Our parallel studies using an activated mutant β-catenin allele supports this hypothesis. Here there is almost a complete abrogation of both intestinal and liver transformation driven by mutant β-catenin when BCL9/9l are lost. Interestingly, BCL9 expression has been shown to increase as HCC progresses and that those patients with high BCL9 expression have a worse prognosis[47,48], highlighting the human relevance of our findings.

Whilst the data presented here are consistent with the ‘just-right’ hypothesis of Wnt signalling, it is important to note it remains correlative. Other cell-extrinsic factors such as niche factors, the microbiome or nutrient availability vary greatly throughout the length of the intestine and these in turn may impact the regional distribution of tumours within VillinCreER Apcfl/+ Bcl9fl/fl Bcl9lfl/fl mice. Therefore, further studies are required to confirm the ‘just-right’ hypothesis. It is notable that upon comparison of gene expression profiles following Apc deletion to those following expression of two copies of mutant β-catenin concomitant with loss of BCL9/9l, we find that while some canonical Wnt targets such as Axin2 and Lgr5 are equally downregulated, others, including c-Myc and Cd44 are more substantially reduced following β-catenin mutation. This demonstrates that in the context of specific Wnt activating mutations, BCL9/9l loss preferentially affects specific Wnt target genes, such as c-Myc, which in turn may be crucial for transformation.

Mechanistically our data supports a model that in normal cells, BCL9/9l is redundant for most Wnt target genes and is only required for genes dependent upon the highest level of Wnt signalling, such as Lgr5+ ISC genes. This is consistent with work using Wnt inhibitors in vivo, where the ISC genes are the most sensitive to Wnt inhibition[35]. Following the loss of BCL9/9l there is more β-catenin at cell junctions and less β-catenin available for optimal target gene expression of ISC genes. Following APC loss there is a global role for BCL9/9l in the Wnt enhanceosome to allow optimal expression of Wnt target genes. In the colon, reducing expression of these target genes is sufficient to suppress tumorigenesis. In the Wnt high proximal SI this level of Wnt is now very efficient for tumour initiation. However, if the truncated APC protein can still bind β-catenin, this leads to a further reduction of Wnt signalling with more β-catenin at cell junctions and a concomitant reduction of Wnt target gene expression. This leads to inefficient tumorigenesis in either Apc1322T/+ or Ctnnb1ex3/+ mice (Fig. 8). These observations raise the possibility that reduction rather than ablation of Wnt signalling can be efficacious in human tumours carrying these mutations.

Fig. 8

BCL9/9l play a context-dependent role in β-catenin mediated transcription. a In the WT setting, BCL9/9l are required for Lgr5 transcription and expression of the Lgr5-ISC signature. Upon deletion of Bcl9/9l, there is increased E-cadherin bound β-catenin and a reduction in ISC fitness. b Following the loss of APC, there is an increase in nuclear β-catenin and concomitant increase in the expression of a number of Wnt target genes in a BCL9/9l-dependent manner, such as Lgr5, Axin2, Sox9, Cd44 and c-Myc. Deletion of Bcl9/9l significantly reduces the expression of many of these Wnt target genes, however this global reduction in Wnt signalling is permissive for tumour formation in the proximal small intestine, but not in the colon. c Expression of a mutant copy of β-catenin that can no longer be phosphorylated induces a Wnt transcriptional programme even in the presence of an intact destruction complex. Upon deletion of Bcl9/9l, there is increased membrane-associated β-catenin and a subsequent failing to induce a transcriptional programme that is sufficient mutant for β-catenin-driven intestinal and hepatocyte transformation

In summary, we have uncovered a role for BCL9/9l downstream of the β-catenin destruction complex, mediating a β-catenin-driven oncogenic transcriptional programme required for Wnt-mediated intestinal and hepatocyte transformation. Continued development of BCL9/9l inhibitors may yield a therapeutic window for CRC and β-catenin-driven HCC.

Methods

Mouse experiments

Mouse colonies: All experiments were performed according to UK Home Office regulations (licence 70/8646), and reviewed by local ethical review committee at the University of Glasgow. Male and female C57BL/6J >20 g mice were induced  with tamoxifen from 6 to 12 weeks of age. The alleles used were as follows: VillinCreER  [49], Apc580S  [41], Bcl9fl, Bcl9lfl [28], Lgr5CreER  [50], R26R-LSL-tdTomato (tdTomfl)[51] and Ctnnb1ex3  [52]. Recombination in the acute models was induced using a single intraperitoneal injection of 80 mg/kg tamoxifen for 2 consecutive days and mice sacrificed 4 days post-induction. VillinCreER Apcfl/+ mice were aged until they showed clinical signs (anaemia, hunching and/or weight loss).

Intracolonic Cre inductions were administered under general anaesthesia—a single 70 µl 100 nM dose of 4-hydroxy tamoxifen (Merck Millipore, Cat# 579002-5MG) was injected into the colonic sub-mucosa via a colonoscope. Colonic tumour growth was then monitored via a colonoscope. Tumour volume was measured relative to lumen size using ImageJ.

For regeneration experiments, mice were exposed to γ-irradiation from a caesium-137 source. This delivered γ-irradiation at 0.423 Gy min−1.

The Porcupine inhibitor LGK974 was administered in a concentration of 5 mg/kg BID (oral gavage) in a vehicle of 0.5% Tween-80/0.5% methylcellulose.

AAV mediated recombination was performed as previously described[53]. Briefly, viral particles (2 × 1011 genetic copies/mouse) of AAV8.TBG.PI.Cre.rBG (UPenn Vector Core, Catalogue number: AV-8-PV1091) were injected via tail vein in 100 µl PBS. Mice were sacrificed and analysed at the indicated timepoints or aged until clinical endpoint—weight loss, hunching and a swollen abdomen.

In accordance with the 3Rs, the smallest sample size was chosen that could give a significant difference. Given the robust phenotypes of the Apcfl/fl model, and our prediction that BCL9 and BCL9l were essential, the minimum sample size assuming no overlap in control vs. experimental is three animals. No randomisation was used and the experimenter was blinded to genotypes.

Immunohistochemistry

IHC was performed on formalin-fixed intestinal sections. Standard IHC techniques were used throughout this study. Primary antibodies used for immunohistochemistry were as follows: BrdU (1:200, BD Biosciences #347580), SOX9 (1:500, Chemicon #AB5535), β-catenin (1:50, BD Biosciences #610154), BCL9 (1:500, Abnova #H00000607-MO1), Glutamine Synthetase (1:200 BD Biosciences #610518), γ-H2AX (1:50, Cell Signalling Technologies #9718) and CD44 (1:50 BD Biosciences #550538). For each antibody, staining was performed on at least three mice of each genotype, representative images are shown for each staining. For nuclear β-catenin staining Tris-EDTA based antigen retrieval was used.

Immunofluorescence

IF was performed on formalin-fixed intestinal sections. Citrate buffer antigen retrieval was used on all sections. Primary antibodies used for IF were as follows: β-catenin (1:200, BD Biosciences #610154), E-cadherin (1:200, Cell Signalling technologies #3195). Sections were stained with DAPI before mounting. Sections were imaged on Zeiss LSM confocal microscope with a ×40 objective.

RNAscope

In situ hybridisation detection for Lgr5 (312178), Olfm4 (311838), Axin2 (400338), Bcl9 (529268) and Bcl9l (466698) mRNA (All Advanced Cell Diagnostics) was performed using RNAscope 2.5 LS (Brown) Detection Kit (Advanced Cell Diagnostics) on a Bond Rx autostainer (Leica) strictly according to the manufacturer’s instructions. Basescope (Advanced Cell Diagnostics) ApcEx14 #701641 (detects wild-type Apc exon 14) was used according to the manufacturer’s instructions.

Proliferation and regeneration

Proliferation levels were assessed by measuring BrdU incorporation. Mice were injected with 250 µl of BrdU (Amersham Biosciences) 2 hours before being sacrificed. Immunohistochemical staining for BrdU was then performed using an anti-BrdU antibody. For each analysis, 25 half crypts were scored per mouse from at least three mice of each genotype. Regenerating crypts were scored from H&E-stained sections as previously described[54]. The number of regenerating crypts per circumference of a small intestinal section was scored from at least 5 different sections per mouse. A minimum of three mice per genotype was scored. For liver sections, the number of BrdU positive fields of view were quantified, at least 10 views were scored per mouse from at least three mice of each genotype.

Quantitative PCR (qRT-PCR)

Whole pieces of small intestinal tissue or liver were used for RNA purification using RNeasy Mini Kit (QIAGEN, #74104) according to the manufacturer’s instructions. 1 µg of RNA was reverse transcribed using DyNAmo cDNA Synthesis Kit (Thermo Scientific, #F-470L) according to the manufacturer’s instructions, cDNA was diluted 1:10 in RNase-free water. qPCR was performed on each sample in technical duplicate, and with at least three biological replicates per genotype, in a 20 µl reaction mixture containing 10 µl of 2XDyNAmo HS master mix (Thermo Scientific), 0.5 µM of each of the primers (detailed later) and 3 µl cDNA generated previously. The reaction mixture without a template was run in duplicate as a control. The reaction conditions were as follows: 95 °C for 15 min, followed by 40 cycles of three steps consisting of denaturation at 95 °C for 15 s, primer annealing at 60 °C for 30 s, and primer extension at 72 °C for 30 s. A melting curve analysis was performed from 65 to 95 °C in 0.5 °C intervals. Gapdh was used to normalise for differences in RNA input. Primer sequences are described in Supplementary Table 4.

Crypt culture

Mouse small intestines were isolated from wildtype and Cre-induced VillinCreER Bcl9fl/fl Bcl9lfl/fl mice sacrificed 4 days post tamoxifen injection, and opened longitudinally and washed with PBS. Crypts were isolated as previously described[55]. Isolated crypts were mixed with 20 µl of Matrigel (BD Bioscience), plated in 24-well plates in Advanced DMEM/F12 supplemented with penicillin–streptomycin, 10 mM HEPES, 2 mM glutamine, N2, B27 (all from Gibco, Life Technologies), 100 ng ml−1 Noggin and 50 ng ml−1 EGF (both Peprotech). Wild-type crypts were also supplemented with R-spondin conditioned medium. Growth factors were added every 2 days.

RNAseq

Whole tissue from the small intestine was used for RNA purification. RNA integrity was analysed with a NanoChip (Agilent RNA 6000 Nanokit #5067-1511). A total of 2 µg of RNA was purified via Poly-A selection. The libraries were run on the Illumina Next Seq 500 using the High Output 75 cycles kit (2 × 36 cycles, paired-end reads, single index). Analysis of the RNAseq data was carried out as previously described in ref. [56].

Gene set enrichment analysis (GSEA)

GSEA analysis was performed using the GSEA v2.0 software (Broad Institute). The comparison gene sets were obtained from published sources; Lgr5 high vs low study A, Lgr5 cross-platform[57], progenitor cluster[58], genes upregulated following APC-KO[39], Wnt target genes that are increased in human CRC[43], direct and functional β-catenin targets in SW480 cells[59].

Proximity ligation assay (PLA)

PLA was performed on tissue samples fixed at 4 °C for <24 h in 10% formalin prior to processing using the Duolink Detection kit (Sigma) according to the manufacturer’s instructions. Briefly, after citrate buffer-mediated antigen retrieval, the slides were incubated with goat E-cadherin (1:200, R&D Systems AF748) and mouse β-catenin (1:2000, #610154, BD Biosciences) overnight. Detection was performed with PLA probes (anti-goat and anti-mouse) conjugated to oligonucleotides. After ligation, amplification detection with a fluorescent probe, slides were imaged on a Zeiss LSM confocal microscope. Z-stacks with ×40 objectives were taken. PLA dots in crypts were analysed with ImageJ calculated as area fraction.

Clonal counting

Lgr5-EGFP-CreER tdTomfl/+ and Lgr5-EGFP-CreER Bcl9fl/fl Bcl9lfl/fl mice were induced with 0.15 mg tamoxifen and then aged for 4, 10 and 21 days. The proximal small intestine was isolated flushed and opened longitudinally and then fixed in 4% paraformaldehyde at room temperature for 4 hours. Tissue was subsequently stored in PBS at 4 °C ahead of processing. Tissue was incubated with DAPI overnight and imaged lumen side down using a Zeiss LSM confocal microscope on a ×10 objective. On average 20 images were acquired from each mouse and the proportion of Tomato positive crypts was determined and a minimum of 200 crypts scored per mouse.

Liver enzyme detection

Plasma was separated from blood obtained via cardiac puncture after collection into heparin. Liver biochemistry was performed using the Siemens Dimension Expand clinical chemistry system and compatible kits (Siemens Diagnostics, USA).

Statistical analysis

Statistical analysis was performed with GraphPad Prism V6 Software (La Jolla, CA, USA) using one-tailed Mann–Whitney tests or otherwise stated. For individual value plots, data displayed as mean  ±standard error of the mean (SEM).