CKI[agr] ablation highlights a critical role for p53 in invasiveness control

To investigate the physiological roles of CKIα, we generated mice in which Csnk1a1 was either deleted from the germline or loxP flanked (floxed) and therefore able to be conditionally deleted (Supplementary Fig. 1a). Whereas mice that were heterozygous for Csnk1a1 (Csnk1a1+/−) seemed to be normal, homozygous deficiency was embryonic lethal before embryonic day 6.5 (Supplementary Fig. 1b and Supplementary Table 1), suggesting a fundamental role for CKIα in embryogenesis. To study the role of CKIα in gut physiology, we crossed mice in which Csnk1a1 was floxed with mice expressing Vil1–Cre–ERT2, generating animals in which injection with tamoxifen causes deletion of the gene encoding CKIα exclusively in the intestinal epithelium (hereafter termed Csnk1a1Δgut mice). Within 5 days of treatment with tamoxifen, CKIα expression was largely abolished throughout the epithelia of the small bowel (Fig. 1a, b and Supplementary Fig. 2a) and colon (data not shown), and was absent for at least 2 weeks, indicating that intestinal progenitor cells had been targeted by the recombinase Cre. CKIα loss was accompanied by a mild increase in CKIδ expression (Supplementary Fig. 2b), while the priming phosphorylation of β-catenin on the serine residue at position 45 (S45) was abolished, eliminating the rest of the phosphorylation cascade (T41, S37 and S33) (Fig. 1b). Consequently, β-catenin was stabilized in the cytoplasm and nucleus, including in differentiated cells of the villus (Fig. 1a, b and Supplementary Fig. 2d). Using RKO cells (a human colorectal cancer cell line), we confirmed this non-redundant function of the α-isoform of CKI in vitro (Supplementary Fig. 2c). These findings indicate that CKIα is indispensable for initiating the β-catenin phosphorylation–degradation cascade in the gut epithelium. As expected, β-catenin accumulation in the gut of Csnk1a1Δgut mice was accompanied by robust activation of many Wnt target genes, including Myc, Axin2, Sox9, Cd44 and the genes encoding cyclin D1 and cyclin D2 (http://www.stanford.edu/~rnusse/pathways/targets.html) (Fig. 1c, d). Particularly striking was the distinct nuclear expression of cyclin D1, which spread into all villi throughout the small intestine. By contrast, in wild-type small bowel, cyclin D1 was restricted to the crypts. Likewise, CD44 and Myc were overexpressed in Csnk1a1Δgut (Fig. 1d). In other mouse models of Wnt hyperactivation, ectopic Paneth cells are common4, and these are clearly observed in small-bowel villi of Csnk1a1Δgut mice (Supplementary Fig. 2e). Thus, knockout of the gene encoding CKIα induced β-catenin stabilization and a massive Wnt response, comparable to other mouse models of Wnt activation and to colorectal cancers. Surprisingly, despite the robust activation of mitogenic Wnt target genes, gut homeostasis was preserved, and tumorigenesis was not observed. This is in stark contrast to Wnt activation in the mouse gut after deletion of the adenomatosis polyposis coli (Apc) gene, which resulted in immediate dysplastic transformation of the entire bowel and rapid death4. Instead, we found only mild atypia and minimal small-bowel crypt elongation, owing to an approximately twofold increase in the proliferating cell population (Supplementary Fig. 2e and data not shown). We therefore postulated that Csnk1a1 ablation elicits a simultaneous reaction that restrains the hyperproliferation and tumorigenesis that is expected on Wnt hyperactivation. Phenotypic changes in Csnk1a1Δgut mice might resemble the oncogene-induced senescence5 associated with DNA-replication stress, persistent DNA damage6 and apoptosis7. Accordingly, p19ARF, a hallmark of oncogene-induced senescence8, was found to be upregulated in Csnk1a1Δgut enterocytes (Fig. 2c). To assess the potential contribution of cellular senescence in protecting against proliferation and tumorigenesis, we sought signs of persistent DNA damage response (DDR), senescence and apoptosis on Csnk1a1 ablation in mouse gut, mouse embryonic fibroblasts (MEFs) and primary human fibroblasts. Csnk1a1 ablation triggered apoptosis, which was evident by TdT-mediated dUTP nick end labelling (TUNEL) assay and cleaved caspase-3 immunostaining (Fig. 2a and Supplementary Fig. 3a), as well as generated widespread signs of DDR (Fig. 2a, 53BP1) and senescence (Fig. 2b). Likewise, ablation of Csnk1a1 in MEFs and mRNA depletion of the corresponding gene, CSNK1A1, in human primary fibroblasts (IMR-90 cells) resulted in a persistent DDR and a senescence phenotype (Supplementary Fig. 3b–f). The ARF tumour suppressor (p19ARF in mice) and DDR are two major activation arms of the tumour-suppressor protein p53, targeting the two p53 antagonists MDM2 and MDMX (also known as MDM4), respectively9, 10. Furthermore, Mdmx ablation in the gut induced p53 activation and apoptosis, albeit with no tissue abnormalities11. Csnk1a1 ablation in the gut caused marked reduction of MDMX expression (Fig. 2d and Supplementary Fig. 4a). We therefore postulated that the combined effect of p19ARF induction and MDMX degradation might trigger p53 activation in Csnk1a1Δgut mice. In non-perturbed wild-type gut, p53 is normally inactive12. Indeed, p53 expression was undetectable in Csnk1a1+/Δgut (heterozygous) small or large bowel (Fig. 2d, e and data not shown). By contrast, Csnk1a1Δgut mice showed marked p53 expression (Fig. 2d, e). Transcriptome analysis revealed that, not only the Wnt cascade, but also the p53 pathway is strongly induced by Csnk1a1 ablation (Supplementary Table 2; P = 3.9 × 10−8). This was confirmed by monitoring the induction of several specific p53 target genes, including the anti-proliferative gene p21 (ref. 13) (Fig. 2d, e and Supplementary Fig. 4b). The elevated p53 expression in Csnk1a1Δgut enterocytes was apparently due to both transcriptional and post-transcriptional control mechanisms; although p53 (also known as Trp53) mRNA was elevated, the fold increase in p53 protein levels exceeded mRNA upregulation (Fig. 2d, e and Supplementary Fig. 7a). Thus, p53 protein stabilization occurred concurrently with β-catenin stabilization in CKIα-depleted RKO (colorectal) cells (Supplementary Fig. 4c). These data point to robust p53 activation in Csnk1a1Δgut epithelium as a possible means of counteracting the pro-tumorigenic effects of the Wnt pathway. As with Csnk1a1Δgut, intestinal Apc ablation results in Wnt hyperinduction and widespread cyclin D1 expression4, 14 (Supplementary Fig. 5a). However, the consequences of its ablation—with respect to both homeostasis and tumorigenesis—are profoundly different. One reason for this difference could be p53 activation. We observed only limited, sporadic activation of p53 and its target p21 in intestinal ApcΔgut epithelium compared with the substantial p53 expression in Csnk1a1Δgut epithelium (Supplementary Fig. 5a). Supporting these differences were our comparative transcriptome analysis of three enterocyte sources—Csnk1a1Δgut (Supplementary Table 2), ApcΔgut (ref. 4) and Apc1638N (ref. 15)—showing far greater enrichment of p53 target genes in Csnk1a1Δgut (P = 3.9 × 10−8) than in the APC-based mouse models (P = 9.4 × 10−3). Differential p53 activation by CKIα depletion was also observed in RKO cells harbouring wild-type p53 and Apc genes. Whereas β-catenin was strongly stabilized when either APC or CKIα was depleted using short hairpin RNAs (shRNAs), H2AX phosphorylation and p53 stabilization were observed only when CKIα was depleted (Supplementary Fig. 5b), indicating that DDR and p53 activation require more than just Wnt activation. Moreover, β-catenin depletion in CKIα-depleted RKO cells did not suppress DDR induction or p53 activation (Supplementary Fig. 5c), demonstrating that CKIα loss induces p53 independently of Wnt hyperinduction. Thus, the absence of intestinal tumours despite Wnt hyperactivation could be due to distinctive activation of p53 when Csnk1a1 is ablated. To assess the contribution of p53 to counteracting Wnt-driven tumorigenesis, we crossed Csnk1a1Δgut mice with p53Δgut mice. Remarkably, within 2 weeks of initiating the tamoxifen-induced co-ablation, the Csnk1a1Δgutp53Δgut (double-knockout) mice developed widespread high-grade dysplasia and numerous intramucosal carcinomas invading the lamina propria in most of the small-bowel crypts and villi (Fig. 3a (H&E) and Supplementary Fig. 6a), and they died shortly thereafter. Unlike Csnk1a1Δgut (knockout) mice, double-knockout mice showed massive proliferation throughout the crypt–villus axis, up to the tips of the villi (Fig. 3a (BrdU) and Supplementary Fig. 6b). This, however, was not attributable to further hyperactivation of Wnt signalling in the double-knockout mice: β-catenin stabilization and Wnt gene induction in double-knockout mice were comparable to those in Csnk1a1Δgut mice (Supplementary Fig. 6c–f). These data indicate that the tumour-suppressor function of p53 in Csnk1a1Δgut mice is not exerted through the suppression of Wnt signalling. Whereas the Bax induction and activation-associated cleavage of BAX protein seen in Csnk1a1Δgut mice were absent in double-knockout mice (Supplementary Fig. 7a, b), the extent of apoptosis in double-knockout mice was similar to that in Csnk1a1Δgut mice (Supplementary Fig. 7c, cleaved caspase 3), indicating that apoptosis by itself, whether p53-dependent or p53-independent, is ineffective at suppressing intestinal carcinogenesis. By contrast, almost no expression of p21 was observed in double-knockout mice (Supplementary Fig. 7b, c), suggesting that p21 might be a key factor in offsetting excessive proliferation and tumorigenesis in Csnk1a1Δgut mice. Human colorectal carcinogenesis is a protracted process. It entails multiple genetic mutations and epigenetic events, which propel tumour progression from an aberrant crypt focus through adenoma to invasive carcinoma3. In various mouse models, this process, which probably spans 10–15 years in humans, commonly occurs over many months and by a similar sequence of molecular and histological events16. The exceptionally rapid development of invasive intramucosal carcinomas in Csnk1a1Δgutp53Δgut (double-knockout) mice, bypassing the adenomatous phase, prompted us to validate the cancerous nature of the observed lesions. Intestinal villi of double-knockout mice were removed and transplanted under the kidney capsule of immunodeficient (NOD SCID) mice. Invasive adenocarcinomas of intestinal origin permeating the kidney tissue were observed in two of five NOD SCID recipients of double-knockout villi (Fig. 3b), whereas no viable intestinal tissue was detectable in recipients of either CKIα-deficient villi with one functional p53 allele (n = 3) or p53-deficient villi with one functional Csnk1a1 allele (n = 2) (data not shown). This confirmed the occurrence of bona fide carcinogenesis in double-knockout mice. Further evidence of the potent effect of combined loss of p53 and CKIα came from mice that were heterozygous for the gene encoding CKIα and lacked p53. Csnk1a1+/− and Csnk1a1−/fl (heterozygous) mice (where fl is a floxed allele) have a full lifespan without evidence of tumorigenesis (data not shown). Likewise, p53 knockout mice do not develop intestinal tumours, at least until the age of 6 months, by which time they succumb mostly to lymphomas and sarcomas17. By contrast, seven of nine mice homozygous for gut-ablated p53 and harbouring one active Csnk1a1 allele in the gut (Csnk1a1+/Δgutp53Δgut) developed invasive carcinomas that permeated the bowel muscular wall into the subserosal fat and some of which involved the serosal surface (stage T4) within 6 months of gene ablation. Remarkably, by immunohistochemistry analysis, all of these malignant lesions are CKIα-negative (Fig. 3c). DNA analysis of carcinoma tissue implicated loss of heterozygosity (LOH) at the Csnk1a1 locus as the source of carcinoma growth (Supplementary Fig. 8). Overall, our data point to a dual role for CKIα: it acts as a tumour suppressor by blocking Wnt-driven proliferation, and its absence activates p53 to counter carcinogenesis. Whereas p53 inactivation often coincides with the transition of benign colorectal tumour to invasive carcinoma in humans, existing mouse models for colorectal cancer do not clearly show how wild-type p53 suppresses carcinogenesis18, 19. An advantage of the Csnk1a1 ablation model is the robust, instantaneous emergence of intramucosal carcinoma, making it easier to track the origin of invasion. To elucidate the molecular basis of invasiveness in Csnk1a1Δgutp53Δgut mice, we compared the transcriptome of the doubly ablated enterocytes with that of singly ablated (p53Δgut or Csnk1a1Δgut) enterocytes. Among the upregulated genes, there is a subset, which we denote PSIS (p53-suppressed invasiveness signature), associated with diverse invasiveness functions, including loss of enterocyte polarity and adhesion (for example, Prox1), tissue remodelling (for example, the Ifitm gene family, Plat and Mmp7) and facilitation of motility (for example, Pls3) (Supplementary Table 3 and Supplementary Fig. 9a). Ifitm gene family members are new biomarkers for colorectal cancer20 and are associated with invasion at early stages of human head and neck cancer21. PROX1 overexpression characterizes highly dysplastic colorectal adenoma and carcinoma, and it promotes colonic tumorigenesis by modulating polarity and adhesion22. We found that Prox1 and several Ifitm gene family members are strongly induced in Csnk1a1Δgutp53Δgut enterocytes (Supplementary Fig. 9a). These genes harbour TCF-binding sites23, indicating that they are transcriptionally induced by Wnt signals only if p53-mediated repression is abolished. Immunohistochemistry analysis indicates that PROX1, IFITM2 and IFITM3 proteins are abundantly expressed at foci of Csnk1a1Δgutp53Δgut intramucosal carcinomas (Fig. 3d), whereas there is little or no expression of these genes in the gut of Csnk1a1+/Δgut or Csnk1a1Δgut mice. Likewise, expression of PROX1, IFITM2 and IFITM3 is plentiful in the invasive carcinoma that develops following LOH at the Csnk1a1 locus on a p53-null background and after transplantation of double-knockout villi to NOD SCID mice (Supplementary Fig. 9b). Consistent with these findings, the highest expression of PROX1 in human colorectal adenomas is observed in areas of severe dysplasia22, which are often enriched in p53 mutations24. Together, these data point to cells that express PSIS being more capable of invading the lamina propria than neighbouring epithelial cells that do not express PSIS genes. To confirm direct participation of a PSIS gene in invasiveness control, we depleted PROX1 in COLO 205 cells, a human colorectal cancer cell line in which APC and p53 are mutated. COLO 205 cells show Matrigel invasion on stimulation with phorbol myristate acetate (PMA). Depletion of PROX1 mediated by shRNA (Fig. 4a) resulted in a threefold reduction of Matrigel invasion (Fig. 4b, c). In addition, depletion of p53 in MCF 10A (mammary) cells (which harbour wild-type p53) resulted in substantial PROX1 elevation and Matrigel invasion, which was suppressed by co-depletion of p53 and PROX1 (Supplementary Fig. 9c–e). These results indicate that PROX1 expression is a key component of the p53-controlled invasive phenotype. In addition to its well-known tumour-suppressive role through transcriptional activation of cell cycle inhibitors and pro-apoptotic genes13, p53 can also mediate transcriptional repression, either through direct association with other transcription factors (for example, NFY25) or through the activation of its target gene p21 (ref. 26). The p21 protein can repress transcription by various mechanisms, including indirect inactivation of E2F transcription factors through the Rb protein pathway, as well as direct binding to E2F1 and Myc26. NFY-, E2F- and Myc-binding motifs are abundant among the promoters of genes upregulated in double-knockout mice, and nearly all of the PSIS genes harbour at least one of these binding sites (Supplementary Fig. 10a, b). To examine whether p21 can suppress PSIS and invasiveness, we overexpressed p21 concomitantly with green fluorescent protein (GFP) in COLO 205 cells. The resultant downregulation of PROX1 protein expression (Fig. 4d) and decrease in number of GFP-positive invading cells in a Matrigel assay (Fig. 4e, f) implied PSIS suppression by p21. PROX1 repression by p21 was not due to cell cycle arrest, as p21-overexpressing COLO 205 cells continued to proliferate (Fig. 4d, pHH3). These in vitro findings indicated that the p53–p21 pathway may control tissue invasion of epithelial cells through suppression of PSIS-gene expression. Of note, PROX1 expression is insufficient for inducing the invasion phenotype in vivo22; rather, concerted action of several PSIS genes may be required to promote tissue invasion. To elucidate the role of p21 in Csnk1a1Δgut mice in vivo, we generated Csnk1a1Δgutp21−/− mice and examined the consequences of dual gene ablation. Much like Csnk1a1Δgutp53Δgut mice, Csnk1a1Δgutp21−/− mice showed Wnt hyperinduction and massive proliferation, as evident from cyclin D1 and 5-bromodeoxyuridine (BrdU) staining spreading into the villi compartment (Supplementary Fig. 11a), indicating that p21 indeed mediates the p53-dependent growth arrest. Csnk1a1Δgutp21−/− mice expressed p53 and its typical target genes at Csnk1a1Δgut-comparable levels (Supplementary Fig. 11b, c); however, like Csnk1a1Δgutp53Δgut mice, Csnk1a1Δgutp21−/− mice developed severe dysplasia and intramucosal carcinomas throughout the small bowel (Supplementary Fig. 11a, H&E), providing direct genetic evidence that p21 is a mediator of p53-dependent invasiveness control. This was also emphasized by gene expression array analysis showing repression of PSIS genes by p21 (Supplementary Fig. 11d and Supplementary Table 3). Interestingly, the proliferation rate of the intramucosal carcinoma cells is significantly reduced compared with cells of the adjacent non-invasive epithelium: intramucosal carcinomas upregulate the PSIS marker PROX1 (Fig. 4g and Supplementary Fig. 11a (PROX1)) yet are mostly BrdU-negative (Fig. 4g). Therefore, p21 is a key factor in the suppression of p53-mediated invasiveness, independently of its role in cell cycle control. Our studies implicate p53 as the guardian of invasiveness. In contrast to a recent report on cell lines27, we detected no epithelial-to-mesenchymal transition in the Csnk1a1Δgutp53Δgut bowel (data not shown), distinguishing the invasiveness control function of p53 in intestinal carcinogenesis from other tumour-suppressor mechanisms. Furthermore, there is probably little in common between the invasiveness promotion of a gain-of-function p53 mutant, which functions like an oncogene28, and the tumour-suppressive property of wild-type p53 described here. For instance, whereas epidermal growth factor receptor signalling is augmented in cells harbouring p53 gain-of-function mutants28, it does not differ between Csnk1a1Δgut and Csnk1a1Δgutp53Δgut enterocytes (data not shown). Our Csnk1a1Δgut and Csnk1a1Δgutp53Δgut knockout models recapitulate critical aspects of human colorectal carcinogenesis: Wnt hyperactivation, DDR and senescence, and p53 loss. Whereas Wnt activation, DDR and senescence are early molecular events in human colorectal tumorigenesis, p53 inactivation is typically a late event, concurrent with the transition of benign tumours to invasive colorectal cancer3. At this transition point, the control of tissue invasion—rather than the well-established cell-autonomous functions of p53 (apoptosis induction, cell cycle arrest and senescence7)—is probably the critical tumour-suppressive function of p53. Notably, in the alternative colorectal carcinogenesis model—the ‘de novo colorectal carcinogenesis model’29, in which, unlike the traditional adenoma–carcinoma sequence3, invasiveness precedes other aspects of neoplastic growth—p53 loss is an early carcinogenesis event. This alternative model is most common in human colorectal cancer associated with inflammatory bowel diseases24, possibly indicating that environmental cues, such as inflammatory mediators, may synergize with p53 loss in upregulating PSIS genes, inducing rapid invasiveness. An inflammatory reaction sustaining cellular senescence30 could fulfil a similar function in the Csnk1a1 knockout model. We propose that invasion control, which could be triggered early or late in carcinogenesis, is a cardinal tumour-suppressive function of p53 in Wnt-driven tumours. A CKIα expression vector was constructed on the basis of pGEM-11Zf(+), to which an XbaI- and SalI-digested fragment of a neomycin cassette flanked by two loxP sites was inserted from a pL2-neo expression vector. Exons 1 and 2 of the mouse Csnk1a1 gene were cloned into the vector, flanked by loxP sites using a third loxP site. Short (1-kilobase) and long (5-kilobase) homology sequences were cloned upstream and downstream of the targeted exons, respectively. All genomic fragments were amplified by PCR from 129/SvJ mouse DNA. The vector was linearized with SalI and purified using phenol–chloroform extraction and ethanol precipitation methods. R1 embryonic stem (ES) cells (129/SvJ-mouse derived) were electroporated and cultured on a feeder layer of MEFs using DMEM supplemented with 15% ES-cell-tested FBS and 1,000 U ml−1 ESGRO (Chemicon). Neomycin selection was done in 0.2 mg ml−1 G418 (Sigma). pCA–NLS–Cre was used as a Cre expression vector for transient transfection of Cre into ES cells. Selection was done in 2 μg ml−1 puromycin. R1 ES cells were aggregated to CD-1 mouse morulae. Chimaeric mice were bred with CD-1 mice to check for germline transmission. Pgk1–Cre transgenic mice31 were used as a deleter strain for the generation of germline Csnk1a1 deletion. For generation of conditional Csnk1a1 knockout mice, Cre was transiently expressed in ES cells to produce specific deletion of the neomycin cassette and an intact floxed Csnk1a1 allele. A second aggregation was done, and Vil1–Cre–ERT2 (ref. 32) transgenic mice were used as a gut-specific deleter to generate a conditional Csnk1a1 knockout mouse. Inducible double-knockout mice (p53 and Csnk1a1) were generated by crossing mice in which Csnk1a1 was floxed with mice in which p53 was floxed33. p21−/− mice34 were purchased from the Jackson Laboratory (stock #003263). Double-knockout mice (p21 and Csnk1a1) were generated by crossing mice in which Csnk1a1 was floxed with p21−/− mice. Mice were kept under specific pathogen-free conditions at The Weizmann Institute of Science and at the Hadassah Medical School of the Hebrew University. All mouse experiments were performed in accordance with guidelines of the relevant institution’s ethics committee. For mouse genotyping, DNA from the tail or ear of 4-week-old pups was extracted by means of standard protocols. For embryo genotyping, a small section of the embryo was taken for DNA preparation. The following primers were used for simultaneous detection of wild-type and null Csnk1a1 alleles: 5′-AACAAGATGGCGGCCTCG (forward primer for wild-type and null Csnk1a1); 5′-CGCACCAGTTTGTATTTTCC (reverse primer for wild-type Csnk1a1); and 5′-GGGCGAATTCTGCAGATATC (reverse primer for null Csnk1a1). The following primers were used for simultaneous detection of wild-type and floxed Csnk1a1 alleles: 5′-CGTGACGCCGACAGAG (forward primer for wild-type Csnk1a1); 5′-ATAAGTGGGGGAGGCTGCTA (reverse primer for wild-type Csnk1a1); 5′-GCAGGAAGTGGCAGTGAAAC (forward primer for floxed Csnk1a1); and 5′-GGGCGAATTCTGCAGATATC (reverse primer for floxed Csnk1a1). Other primers that were used for genotyping: 5′-CACAAAAAACAGGTTAAACCCAG (forward primer for wild-type and floxed p53); 5′-AGCACATAGGAGGCAGAGAC (reverse primer for wild-type and floxed p53); 5′-AGCAATTCACACGTATTTGG (forward primer for wild-type p21); 5′-TGACGAAGTCAAAGTTCCACC (reverse primer for wild-type p21); 5′-AAGCCTTGATTCTGATGTGGGC (forward primer for null p21); 5′-GCTATCAGGACATAGCGTTGGC (reverse primer for null p21); 5′-ATGTCCAATTTACTGACCGTACACC (forward primer for cre); and 5′-CGCCTGAAGATATAGAAGATAATCG (reverse primer for cre). Tamoxifen (Sigma) was dissolved in corn oil (Sigma), and mice were injected either subcutaneously (100 mg kg−1 per injection), five to seven injections every other day for 10–14 days, or intraperitoneally (120 mg kg−1) on two consecutive days. At 3–5 days after the last injection, each mouse was injected intraperitoneally with 10 μl g–1 BrdU (GE Healthcare) and killed 2 h later. The jejunum, ileum and the entire large intestine were flushed with ice-cold PBS, cut open longitudinally and subjected to fixation in 4% formaldehyde and paraffin embedding (FFPE). Small pieces of jejunum were embedded in Tissue-Tek O.C.T. Compound (Sakura) and frozen at −80 °C. Intestinal epithelial cells (IECs) were isolated from the middle part of the small intestine, as described previously35 but with slight modifications: intestinal cells were separated into single cells in Hank’s balanced salt solution containing 10 mM HEPES, 5 mM EDTA and 0.5 mM dithiothreitol, at 37 °C for 30 min. Sections (5 μm) were cut for haematoxylin and eosin (H&E) staining and immunohistochemistry analysis. For immunohistochemistry, sections were incubated with antibodies detecting CKIα (C-19; 1/1,000; Santa Cruz Biotechnology), p21 (F-5; 1/50; Santa Cruz Biotechnology), c-Myc (N-262; 1/100; Santa Cruz Biotechnology), PROX1 (1/200; R&D Systems), EPHB2 (1/200; R&D Systems), β-catenin (1/200; BD Transduction), BrdU (Ab3; 1/100; NeoMarkers), cyclin D1 (SP4; 1/125; Lab Vision), CD44 (IM781; 1/200; eBioscience), p53 (CM5; 1/200; Novocastra), cleaved caspase 3 (1/100; Cell Signaling Technology), IFITM2/3-Fragilis (1/400; Abcam) and lysozyme (1/5,000; Dako). Secondary antibodies were horseradish peroxidase (HRP)-polymer anti-mouse, anti-rabbit, anti-goat and anti-rat antibodies (Nichirei, Dako and Biocare). 3,3′-Diaminobenzidine (DAB) chromogen (Lab Vision) was used for detection. For dual detection of apoptosis and DDR, TUNEL (Roche) was performed according to the manufacturer’s instructions, followed by overnight incubation with anti-53BP1 antiserum (1/200; Bethyl) and detection with goat anti-rabbit antibodies conjugated to Alexa Fluor 647 (1/1,000; Molecular Probes). For double immunofluorescence of BrdU and PROX1, FFPE slides were blocked with 3% bovine serum albumin (BSA), 5% donkey serum and 0.1% Triton X-100 in Tris-buffered saline with Tween 20 (TBST) for 1 h, incubated for 16 h at 4 °C with primary antibodies as described, and then exposed to donkey anti-mouse antibodies conjugated to Alexa Fluor 488 and donkey anti-goat antibodies conjugated to Alexa Fluor 647 (1/1,000; Molecular Probes). Hoechst stain (1 μg ml−1; Molecular Probes) was used for nuclear counterstaining. For MDMX immunofluorescence, 5 μm sections were cut from O.C.T.-embedded frozen tissues and fixed in cold acetone for 10 min. Sections were blocked and incubated overnight at 4 °C with anti-MDMX antibody (clone 82; 1/500; Sigma) and subjected to goat anti-mouse antibody conjugated to Alexa Fluor 647 (1/1,000; Molecular Probes). In situ hybridization with digoxigenin (DIG; Roche)-labelled cryptdin probe14 was carried out on FFPE sections. Following deparaffinization and rehydration, slides were immersed in 2 M HCl for 15 min, digested with proteinase K (Roche) for 15 min at 37 °C, fixed for 10 min in 4% paraformaldehyde, acetylated with acetic anhydride in 0.1 M triethanolamine solution, pH 8, pre-incubated at 70 °C for 1 h in hybridization buffer (saline sodium citrate (SSC; pH 4.5), 50% formamide, 2% blocking powder (Roche), 5 mmol l−1 EDTA, 50 μg ml−1 yeast transfer RNA, 0.1% Tween 20, 0.05% 3[3-cholaminopropyl diethylammonio]-1-propane sulphonate and 50 μg ml−1 heparin), and incubated with the probe in a humidity chamber for 24–48 h at 70 °C. Following three washes in 2×SSC with 50% formamide, five washes in TBST and blocking for 1 h at 4 °C in TBST containing 0.5% blocking powder (Roche), slides were incubated for 16 h at 4 °C with anti-DIG Fab–alkaline phosphatase (1/2,000; Roche) in blocking solution (Roche). Following five washes in TBST and two rinses in 0.1 M Tris, pH 8.8, containing 0.1 M NaCl, slides were stained with nitroblue tetrazolium/5-bromo-4-chloro-indolyl-phosphate (NBT/BCIP) solution (BM purple; Roche) at 25° C for 1–2 h, dehydrated and mounted. For senescence-associated β-galactosidase (SA-β-gal) staining, 10 μm sections were cut from O.C.T.-embedded frozen tissue and allowed to adhere to coated slides at 25° C for 1 min before fixation for 15 min in PBS containing 0.5% glutaraldehyde (Sigma). Sections were rinsed with PBS, pH 5.5, containing 1 mM MgCl2 and incubated at 37 °C for 16 h in pre-warmed and filtered 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-gal) solution (0.1% X-gal (Ornat) dissolved in PBS, pH 5.5, containing 1 mM MgCl2, 5 mM potassium ferrocyanide and 5 mM potassium ferricyanide). Sections were rinsed with PBS, post-fixed in 95% ethanol, rehydrated, counterstained with nuclear fast red, dehydrated and mounted. Protein was extracted by whole-cell-extract protocols from cell pellets in protein lysis buffer containing protease and phosphatase inhibitors (p-nitrophenyl phosphate (PNPP), sodium orthovanadate (Na3VO4), leupeptin, aprotinin, β-glycerophosphate and okadaic acid). Western blotting analysis was performed by means of standard techniques. Blots were incubated with antibodies detecting β-catenin (1/2,500; BD Transduction), phospho-β-catenin (S45; 1/750; Cell Signaling Technology), phospho-β-catenin (S33/S37/T41; 1/750, Cell Signaling Technology), phospho-β-catenin (T41/S45; 1/750; Cell Signaling Technology), CKIα (C-19; 1/1,000; Santa Cruz Biotechnology), CKIε (1/250; Santa Cruz Biotechnology), CKIδ (1/4000; Icos), HSP90α (1/5,000; Calbiochem), tubulin (1/5,000; Sigma), MDMX (clone 82; 1/1,000; Sigma), mouse p53 (CM5; 1/1,000; Novocastra), BAX (1/200; Santa Cruz Biotechnology), p21 (F-5; 1/200; Santa Cruz Biotechnology), MKP1 (1/750; Santa Cruz), p19ARF (1/100; Upstate), human p53 (DO-1/1801 hybridoma mix; 1/10), phospho-Jun (S63; 1/1,000; Cell Signaling Technology), phospho-histone H2A.X (S139; 1/1,000; Millipore), PP2A-C (1/1,000; rabbit serum provided by D. Virshup), cyclin D1 (SP4; 1/500; Lab Vision) and PROX1 (1/1,000; R&D Systems). Secondary antibodies were HRP-linked goat anti-mouse, goat anti-rabbit, goat anti-rat and rabbit anti-goat antibodies (all 1/10,000; Jackson). Blots were developed using ECL (GE Healthcare). Total RNA was extracted from cell pellets using TRI Reagent (Sigma) and phenol–chloroform methods. RNA (2 μg) was subjected to reverse transcription using M-MLV RT (Invitrogen), and mRNA expression levels were measured by quantitative real-time PCR using SYBR Green (Invitrogen) in a 7900HT Fast Real-Time PCR system (ABI). Relative quantities of gene transcripts were analysed in qBase 1.3.5 software and normalized to the Ubc and Hprt transcripts. Sequences of PCR primers are as follows: 5′-TAGGCGGAATGAAGATGGAC (forward primer for Axin2); 5′-CTGGTCACCCAACAAGGAGT (reverse primer for Axin2); 5′-ATGCGTCCACCAAGAAGCTGA (forward primer for Bax); 5′-AGCAATCATCCTCTGCAGCTCC (reverse primer for Bax); 5′-CAGTATCTCCCGGACTGAGG (forward primer for Cd44); 5′-GCCAACTTCATTTGGTCCAT (reverse primer for Cd44); 5′-GGTGCGGAAGATCGGATCT (forward primer for Csnk1a1); 5′-TTCACTGCCACTTCCTCGC (reverse primer for Csnk1a1); 5′-TGAGCCCCTAGTGCTGCAT (forward primer for Myc); 5′-AGCCCGACTCCGACCTCTT (reverse primer for Myc); 5′-TTGACTGCCGAGAAGTTGTG (forward primer for cyclin D1); 5′-CCACTTGAGCTTGTTCACCA (reverse primer for cyclin D1); 5′-CACCGACAACTCTGTGAAGC (forward primer for cyclin D2); 5′-TGCTCAATGAAGTCGTGAGG (reverse primer for cyclin D2); 5′-GCTGGCGCTATCTATCCTTG (forward primer for cyclin G1); 5′-GGTCAAATCTCGGCCACTTA (reverse primer for cyclin G1); 5′-GTTAAGCAGTACAGCCCCAAA (forward primer for Hprt); 5′- AGGGCATATCCAACAACAAACTT (reverse primer for Hprt); 5′-ATCTCCACGCCTGACCATGT (forward primer for Ifitm1); 5′-CACCCACCATCTTCCTGTCC (reverse primer for Ifitm1); 5′-CTGCTGCCTGGGCTTCATAG (forward primer for Ifitm3); 5′-GGATGCTGAGGACCAAGGTG (reverse primer for Ifitm3); 5′-TACCTGCCCCTACCCTGATG (forward primer for Ly6a); 5′-AGGAGGGCAGATGGGTAAGC (reverse primer for Ly6a); 5′-TGTGTGAGCTGAGGGAGATG (forward primer for Mdm2); 5′-CACTTACGCCATCGTCAAGA (reverse primer for Mdm2); 5′-TCACTAATGCCAAACAGTCCAA (forward primer for Mmp7); 5′-AAGGCATGACCTAGAGTGTTCC (reverse primer for Mmp7); 5′-GTCACACGACTGGGCGATT (forward primer for p19Arf); 5′-GACTCCATGCTGCTCCAGAT (reverse primer for p19Arf); 5′-TCCACAGCGATATCCAGACA (forward primer for p21); 5′-AGACAACGGCACACTTTGCT (reverse primer for p21); 5′-TGAAACGCCGACCTATCCTTA (forward primer for p53); 5′-GGCACAAACACGAACCTCAAA (reverse primer for p53); 5′-TGGAGAGGGTCAGAAAGCAAA (forward primer for Pls3); 5′-AATCCACAACCGCCAAACTG (reverse primer for Pls3); 5′-AGTTCCTGCTGGGTGCTGTC (forward primer for Plat); 5′-CGGGGACCACCCTGTATGTT (reverse primer for Plat); 5′-ATACCGAGCCCTCAACATGC (forward primer for Prox1); 5′-CGTAACGTGATCTGCGCAAC (reverse primer for Prox1); 5′-CAAGAAGAGCAGCATCGACA (forward primer for Puma); 5′-TAGTTGGGCTCCATTTCTGG (reverse primer for Puma); 5′-GGAGCTCAGCAAGACTCTGG (forward primer for Sox9); 5′-TGTAATCGGGGTGGTCTTTCT (reverse primer for Sox9); 5′-CAGCCGTATATCTTCCCAGACT (forward primer for Ubc); 5′-CTCAGAGGGATGCCAGTAATCTA (reverse primer for Ubc). Csnk1a1Δgutp53Δgut (double-knockout) male mice (n = 2) were used as donors of intestinal villi, with Csnk1a1Δgutp53+/Δgut (n = 2) and Csnk1a1+/Δgutp53Δgut (n = 2) mice as controls. After knockout induction for 12 days and concomitant treatment for 1 week with an antibiotic mix of imipenem, vancomycin and metronidazole (each 0.25 mg ml−1) in the drinking water, the mice were killed, their small intestines flushed twice with PBS and inverted, and their villi scraped from the upper mucosal layer. Villi were centrifuged and resuspended in 1 ml PBS. NOD SCID male mice aged 8 weeks were used as recipients for intestinal villi. The recipient mice were anaesthetized with ketamine/xylazine, the left kidney was exposed, and 200 μl resuspended villi from donors were transplanted under the kidney capsule36. Five mice received double-knockout villi, three received Csnk1a1Δgutp53+/Δgut villi and two received Csnk1a1+/Δgutp53Δgut villi. Transplanted recipient mice were kept in specific-pathogen-free conditions and given the above antibiotic mix in their drinking water. Six weeks after transplantation, the recipients were killed and FFPE sections of the kidneys were analysed. For LOH analysis, laser capture microdissection was carried out to obtain DNA from normal and cancerous tissues, extracted by means of standard protocols. PCR was performed with the following primers: 5′-GTAATTGGACCCGATGAATCG (forward primer), 5′-AAACGCAGCAGTGCAACAAAC (reverse primer), detecting 126 base pairs (bp) and 187 bp wild-type and floxed Csnk1a1 alleles, respectively; 5′-CTAGCTTGGCTGGACGTAAAC (forward primer), 5′-AAACGCAGCAGTGCAACAAAC (reverse primer), detecting 150 bp of Csnk1a1 null allele. Human fibroblasts (IMR-90) at early passage were grown with MEM supplemented with 10% serum. Cells were split after 2 days and left in culture for a total of 8 days, then split again and analysed at sub-confluence for SA-β-gal and γ-H2AX nuclear foci, after a total of 10 days in culture. Csnk1a1+/+, Csnk1a1+/fl and Csnk1a1fl/fl MEFs were collected from embryonic day 13.5 mice and prepared according to standard procedures. Early-passage MEFs were seeded at 1 × 106 cells per 10-cm dish and grown with DMEM supplemented with 10% serum. The floxed Csnk1a1 allele was subjected to Cre-mediated excision by transduction of the cells with 5 × 107 plaque-forming units (PFU) ml−1 of Cre-expressing adenovirus (Ad-Cre) (Gene Transfer Vector Core, University of Iowa). Cells were collected 6 days after transduction with Ad-Cre. Cells (n ≥ 100) were counted by a cytometer, and their duplication rates were calculated. RKO cells were grown in DMEM supplemented with 10% serum. MG132 (20 μM; Calbiochem) was added for 3 h before cell collection. For cycloheximide treatment, RKO cells were transduced with shRNA vector for 7 days, then treated with 2 μg ml−1 doxorubicin for 3 h or left untreated. Cells were then washed, incubated with 40 μg ml−1 cycloheximide for the indicated times, and collected. COLO 205 human colorectal cells were grown in RPMI supplemented with 10% serum. Transfection into COLO 205 cells was done using Amaxa Cell Line Nucleofector Kit T (Lonza), and cells were collected 48 h after transfection. MCF 10A breast epithelial cells were grown in F12/DMEM. SA-β-gal staining for human and mouse fibroblasts was carried out as described previously37. Cells were then washed with PBS, fixed in 4% paraformaldehyde for 15 min, washed with TBS, incubated for 10 min in 0.25% Triton X-100-containing TBS, washed with TBS, blocked in 3% BSA with 0.1% Triton X-100/TBS for 1 h and incubated for 16 h with anti-phospho-histone H2A.X (S139) antibody (1/1,000; Upstate). Staining was detected as described above. As a positive control for senescence and DDR, cells were treated with 30 Gy ionizing irradiation. pLKO-based (Open Biosystems) and pLL3.7-based lentiviral vectors were used to transduce cells with shRNA-encoding sequences. The lentiviral packaging system was provided by I. Verma38, 39. Human shRNA-encoding sequences that were used are as follows: CSNK1A1 5′-GCAGAATTTGCGATGTACTTA; CSNK1D and CSNK1E 5′-GGGCTTCTCCTATGACTAC; APC 5′-GCAGAGGAAGGTGATATTC; β-catenin 5′-GTGGGTGGTATAGAGGCTC; p19Arf 5′-GTGATGATGATGGGCAACGTT; scrambled control 5′-TCCTAAGGTTAAGTCGCCCTCG; p53 5′-CGGCGCACAGAGGAAGAGAAT; PROX1 5′-TGAGCCAGTTTGATATGGATTTCAAGAGAATCCATATCAAACTGGCTCTTTTTTC; scrambled control 5′-TGGTCGAACGTCTACTGGAATTCAAGAGATTCCAGTAGACGTTCGACCTTTTTTC. Mouse shRNA-encoding sequence that was used: mSAF (encoded by Hnrnpu) 5′-CTGATGAAGTTGAACTCTC. A combination of CMV–p21 and CMV–GFP expression vectors was used for overexpression of p21, at a ratio of 10/1, respectively. CMV empty vector was used as a control. The in vitro Matrigel invasion assay was carried out in 8 μm Transwell dishes. Matrigel (BD Biosciences) was added to each well (20 μg per well) and was left to gel for at least 1 h. MCF 10A cells were seeded at low and high densities on top of the Matrigel and incubated for 20 h at 37 °C. For visualization, invasive cells were fixed in 4% paraformaldehyde and stained with 0.3% crystal violet. COLO 205 cells were seeded on top of the Matrigel and incubated for 48 h at 37 °C. To induce invasion of COLO 205 cells, 12.5 ng ml−1 PMA40 was added to the lower chambers. For visualization, invasive cells were fixed in 4% paraformaldehyde and stained with Hoechst stain. Gene expression was compared between Csnk1a1+/+ (WT), Csnk1a1+/Δgut (Het) and Csnk1a1Δgut (KO) using Mouse Genome 430 2.0 Arrays (Affymetrix). Gene expression was compared between Csnk1a1+/Δgut (Het), Csnk1a1Δgut (KO), Csnk1a1+/Δgutp53Δgut (p53 KO) and Csnk1a1Δgutp53Δgut (Csnk1a1 and p53 DKO) using Mouse Gene 1.0 ST Arrays (Affymetrix). Gene Expression Omnibus accession number for microarray data is GSE24760. RNA was purified from two biological replicates in each group using the TRIzol extraction method (Invitrogen). Labelled cRNA was produced and hybridized according to manufacturer’s instructions at the Center for Genomic Technologies, Hebrew University. Gene intensities were extracted from CEL files with Expression Console 1.1 (Affymetrix), using the RMA-Sketch method. Raw data were normalized using median intensities41. For each pair of biological replicates, the geometric mean and the ratio were calculated. Transcription factor occurrence was calculated using the DiRE program42. Cutoff was defined as the 20 genes with the highest importance score. Download references We thank S. Robine for the Vil1–Cre–ERT2 mice, O. Sansom and B. Romagnolo for intestinal sections of ApcΔgut mice, K. Rajewsky for the pGEM–loxP–Neo–loxP and pCA–NLS–Cre vectors; and E. Horwitz, M. Farago, D. Naor, N. Asherie and D. Knigin for providing expertise and reagents. We are grateful to A. Yaron for critical reading of the manuscript. This work was supported by the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (AMRF), the Israel Science Foundation, the RUBICON EC Network of Excellence, the Israel Cancer Research Fund and Deutsches Krebsforschungszentrum–Ministry of Science and Technology (DKFZ–MOST). Z.W. is supported by a Marie-Curie Intra-European Fellowship.