FBXW7 influences murine intestinal homeostasis and cancer, targeting Notch, Jun, and DEK for degradation

As Fbxw7-null embryos die in utero (Tetzlaff et al., 2004; Tsunematsu et al., 2004), we generated a conditional transgenic mouse with an Fbxw7 allele, in which exon 5 (encoding the F-box domain) is floxed (Fig. 1 A and Fig. S1). Bitransgenic Villin-Cre⁺Fbxw^fl/fl (Fbxw7^ΔG) mice were then generated to investigate the role of SCF^Fbxw7 E3 ligase activity in the intestine (Fig. S1). Villin-Cre–mediated (el Marjou et al., 2004) Fbxw7 deletion efficiently and specifically occurred in the intestine of Fbxw7^ΔG mice compared with Fbxw^fl/fl (Fig. 1 B). Immunohistochemistry (IHC) analysis of Fbxw7^fl/fl mice showed that nuclear Fbxw7 was widely expressed in the intestinal epithelia, including crypt base cells (CBCs; Fig. 1, C and D, red arrowheads), whereas cytosolic Fbxw7 was hardly detectable (Fig. 1, C and D, black arrowheads). In situ hybridization (ISH) of Fbxw7 in Fbxw7^fl/fl mice showed that Fbxw7 mRNA was strongly expressed in CBCs and transiently amplifying (TA) cells located in crypts (Fig. 1 I, arrowheads), whereas weaker expression of Fbxw7 was detected in differentiated villus cells. Fbxw7 expression was abolished in the Fbxw7^ΔG but not in heterozygous Fbxw7^ΔG/fl mice (Fig. 1, I and J vs. K). Quantitative RT-PCR (qRT-PCR) data also confirmed increased levels of Fbxw7 mRNA expression in the crypt versus villus fraction (Fig. 1, L).

All of the aforementioned analyses demonstrate efficient deletion of Fbxw7 in Fbxw7^ΔG intestinal epithelia but not for Fbxw7^ΔG/fl heterozygous mice when compared with Fbxw7^fl/fl mice (Fig. 1, D vs. F and H, I vs. J and K, and L, right). At 9–10 mo of age, the Fbxw7^ΔG mice showed adenomas and induction of Peyer’s patch formation (mean three to five) in the large bowel (Fig. S2). In addition, smaller polyp-like structures and initiation of crypt budding and fission were found in the small intestine (Fig. S2). These observations indicate that high expression of Fbxw7 in CBCs and TA cells located in the crypts may potentially function in intestinal cell linage commitments, and the loss of Fbxw7 in both alleles could be associated with tumorigenesis.

To study the functional importance of Fbxw7 in intestinal tumorigenesis and any potential synergy between Fbxw7 and APC (adenomatous polyposis coli; Fodde, 2002) in orchestrating human oncosuppression, we generated a double mutant mouse line, Apc^Min/+Fbxw7^ΔG. The Apc^Min/+ mouse is a model of human familial adenomatous polyposis, bearing a nonsense mutation at codon 850 of the APC gene (Fodde, 2002) which, combined with a loss of heterozygosity or promoter hypermethylation, results in ligand-independent activation of the WNT/β-catenin pathway (Kennell and Cadigan, 2009).

Apc^Min/+Fbxw7^ΔG mice exhibited rapid tumor development throughout the intestine with clearly discernible β-catenin^High macroscopic adenomas versus control Apc^Min/+Fbxw7^fl/fl mice at 3 wk of age (Fig. 2, A and B; and Fig. S3, A–F). All Apc^Min/+Fbxw7^ΔG mice developed adenomas (Fig. 2, A and B; and Fig. S3, A–F) without discernible metastases and were culled at 3–5 wk of age (Fig. 2 C; n = 14; P < 0.001). In contrast, single Fbxw7 deletion compromised survival in mice, with 42% of the Fbxw7^ΔG mice dying after 11–12 mo versus the control Fbxw7^fl/fl (Fig. 2 C; n = 14). With a 21 ± 3 wk of maximal mean lifespan (Fig. 2 C; n = 14), both control Apc^Min/+Fbxw7^fl/fl and Apc^Min/+Fbxw7^+/+ yield a very few small-sized polyps without macroscopic adenomas at 3 wk of age (Fig. S4, A–D). The low frequency of β-catenin^High adenomas from Apc^Min/+Fbxw7^fl/fl and Apc^Min/+Fbxw7^+/+ (Fig. S4, A–D) was quantified versus Apc^Min/+Fbxw7^ΔG mice (Fig. 2 D; P ≤ 0.001). Epithelial β-catenin expression remained unchanged at 1 and 3 d postpartum in the Apc^Min/+Fbxw7^ΔG versus the control Apc^Min/+Fbxw7^fl/fl mice (Fig. S4, G and H vs. E and F).

Thus, the result suggests that gut-specific Fbxw7 inactivation in the Apc^Min/+ (Apc^Min/+Fbxw7^ΔG) mice induces tumorigenesis and compromises lifespan and survival. In addition, loss of Fbxw7 in the context of APC deficiency may accelerate extensive intestinal tumorigenesis in mice without an immediate Wnt/β-catenin activation, at 3 d postpartum but with substantial β-catenin accumulation in adenomas at 3 wk of age. This observation suggests that the accumulation of β-catenin may require additional activators after longer loss of Fbxw7 or APC to cause progressive development of intestinal adenomas (Phelps et al., 2009).

To investigate the mechanisms by which mutated Fbxw7 may lead to cancer development, our subsequent analysis was focused on young Fbxw7^ΔG mice that showed signs of altered intestinal epithelial cell (EC) composition accompanied by an induction of apoptosis in cells located in crypts positive for active caspase-3 at position 4–5 (Fig. S5, A and B, left vs. right). There was an increase of apoptotic cells by >55% in the small intestine and 30% in the large bowel in the crypts of Fbxw7^ΔG mice when compared with floxed mice (Fig. 3 A; data are mean percentage ± SEM; n = 3; ≥200 crypts/mouse and for each genotype).

However, Fbxw7^ΔG intestines exhibited a significant loss of goblet cells (Fig. S5 C, left vs. right), along with significantly fewer Paneth and enteroendocrine cells in the Fbxw7^ΔG small intestine (Fig. S5 D, left vs. right; and not depicted). Absorptive enterocytes remained unchanged. To further characterize which epithelial lineages may be involved in tumorigenesis, intestinal sections were stained with the goblet cell marker Mucin 2 (Allen et al., 1998), which showed a remarkable reduction in Muc2 secretory cells in Fbxw7^ΔG mice (Fig. S5 E, left vs. right).

We next examined the effects of Fbxw7 depletion on intestinal cell division using the proliferation marker Ki67. Mutant mice showed a marked increase in proliferating cells from all intestinal crypts (Fig. 3 B; data are mean ± SEM; n = 3; ≥100 crypts per mouse and for each genotype; and Fig. S5 F, left vs. right).

Given these findings, we sought to define the molecular links between Fbxw7 loss and consequent changes in intestinal epithelium. The level of intestinal β-catenin staining was unchanged in either Fbxw7^ΔG or the control Fbxw7^fl/fl mice at 5–6 wk (Fig. S5 G, left vs. right). Next, we analyzed several oncogenic proteins of known Fbxw7 targets by IHC. Staining for Notch 1 (Fig. S5 H, left vs. right) and Notch 4 (Fig. S5 I, left vs. right) showed clearly different expression pattern in the Fbxw7^ΔG CBCs and in Fbxw7^fl/fl, respectively. c-Jun was detected in the CBCs and at the apex of the villi (Fig. S6, A and B) and accumulated within the TA cells in Fbxw7^ΔG versus Fbxw7^fl/fl (Fig. S5 J, left vs. right). These observations indicate that depletion of Fbxw7 may cause Notch and c-Jun activation in progenitor cells. This view is supported by Western blotting, which also confirmed that the expression levels of Notch 1, Notch 4, and phosphorylated c-Jun (p–c-Jun) are increased in Fbxw7^ΔG versus Fbxw7^fl/fl intestine, whereas cyclin E and p–c-Myc are less affected (Fig. 3 C).

Consistent with our observed changes in secretory cell differentiation, the Paneth cell marker crypt-1 (cryptidin-1), the enteroendocrine cell marker chgA (chromogranin A), and the goblet cell marker muc2 (mucin 2) were reduced in Fbxw7^ΔG intestines (Fig. 3 D). To evaluate the role of Notch and c-Jun signal activation on the differentiation and proliferation of intestinal ECs (Fig. S5, A–E), we measured the relative mRNA expression of several Notch/Jun downstream targets by qRT-PCR (Table S1). In general, intestinal expression levels of Hes-1 (hairy and enhancer of split 1), c-jun, and cd44 were increased, whereas the level of Hes-5 was not significantly affected (Fig. 3 E). Expression of secretory progenitor markers Math-1 (mouse atonal homologue 1) and ngn-3 (neurogenin-3) were also significantly repressed (Fig. 3 D).

Hes-1 expression is transcriptionally activated by Notch and is considered the canonical Notch downstream target gene (Jarriault et al., 1995). Hes-1 deletion led to excessive numbers of goblet, enteroendocrine, and Paneth cells. Hes-1^−/− mice showed no changes in intestinal proliferation; the up-regulation of Hes-1 resulted in the repression of ngn-3 and Math-1 (Yang et al., 2001), providing further evidence of the effect of Notch overactivation on the differentiation of Fbxw7^ΔG intestinal secretory precursor cells. Mice lacking Hes-1, Math-1, and ngn-3 show a reduction of goblet, Paneth, and enteroendocrine cells, which may indicate a common progenitor cell for these secretory cell lineages (Jensen et al., 2000; Yang et al., 2001; Jenny et al., 2002).

In the large bowel, CD44 protein expression is normally restricted to the crypt compartment, and expression of cd44 from Fbxw7^ΔG mice was significantly increased (Fig. 3 E and Fig. S5 K, left vs. right). CD44 is a c-Jun target, and it is induced in intestinal tumors (Nateri et al., 2005; Van der Flier et al., 2007). However, induction of CD44 may change the proliferative status of the intestinal progenitor cells when Fbxw7 is inactivated (Fig. S5 F, left vs. right).

Because β-catenin may not been involved in the early events of intestinal cancer initiation in our mutant mice (Fig. S4, E–H; and Fig. S5 G, left vs. right), we examined whether other cell regulators can be involved at these earlier stages. Proteins either absent in control Fbxw7^fl/fl or differentially expressed in Fbxw7^ΔG intestines were identified from cytosolic (Fig. 4, A vs. C) and nuclear (Fig. 4, E vs. G) protein extracts using two-dimensional matrix-assisted laser desorption/ionization (MALDI)–mass spectrometry (MS) assays (Kim et al., 2007). Subsequent amino acid sequencing and bioinformatic analyses identified several proteins known to be involved in carcinogenesis. Among others, we identified different isoforms of TPM α and β mouse proteins with high (Fig. 4 D, yellow circles) and low (Fig. 4, B and D, blue circles) molecular weights (Fig. 4, B vs. D and F vs. H) and also a nuclear phosphoprotein DEK (Fig. 4 H, red circle), all of which were subsequently verified by Western blotting and IHC on Fbxw7^fl/fl and Fbxw7^ΔG intestines.

TPM is comprised of tissue-specific isoforms, including skeletal muscle, smooth muscle, fibroblast, and epithelial isoforms that range 32–45 kD (Gunning et al., 2005; Helfman et al., 2008). TPM staining of Fbxw7^fl/fl intestine showed a cytoplasmic pattern in the smooth muscle cells and vesicular staining in apical ECs (Fig. 4 I) in 6-wk-old mice. In contrast, less apical epithelial staining but stronger staining of the smooth muscle cells was found in the Fbxw7^ΔG intestine (Fig. 4 J). Consistent with this, Western blots of TPM on villus-enriched fractions showed a remarkable reduction in the level of epithelial isoforms (Fig. 4 M, left), whereas TPM on whole intestine (epithelia and muscles) showed an increase in the higher molecular weight isoform (Fig. 4 M, right). The finding shows a remarkable transition of TPM protein isoforms in the Fbxw7^ΔG compared with control Fbxw7^fl/fl intestine.

We next examined which Fbxw7 isoform regulates the E3 ligase activity toward TPM degradation and found that the level of overexpressed TPM1-α protein (Houle et al., 2007) was not affected by overexpression of the FBXW7 isoforms in human CRC HCT116 cells (Fig. S7 A). These data suggest that Fbxw7 may not directly influence the TPM protein level but may indirectly affect TPM alternative splicing (Gooding and Smith, 2008).

DEK (Fig. 4 H), a nuclear phospho-protooncogene protein, is implicated in carcinogenesis and up-regulated in a variety of aggressive human tumors (Waldmann et al., 2004; Carro et al., 2006). IHC analysis demonstrated intense staining of DEK in crypt cells from Fbxw7^ΔG intestine (Figs. 4, K vs. L, arrowheads), and Western blotting also confirmed an increased level of DEK in Fbxw7^ΔG intestine (Fig. 4 N). We measured relative DEK expression with real-time qRT-PCR (Table S1) and found that DEK mRNA expression was normal (Fig. 4 O). These data suggest that loss of SCF^Fbxw7 E3 ligase activity can result in the accumulation of DEK protein level.

β-Catenin levels in the epithelia of Apc^Min/+Fbxw7^ΔG and control Apc^Min/+Fbxw7^fl/fl mice at early age (3 d postpartum) do not show an apparent difference (Fig. S4, F vs. H); indeed, IHC analysis detected transformation of TPM isoform composition from the epithelial isoform in Apc^Min/+Fbxw7^fl/fl mice to the muscle isoform in Apc^Min/+Fbxw7^ΔG mice 3 d postpartum (Fig. 5, A vs. B, arrowheads). Results also showed a high level of DEK in the intestines of Apc^Min/+Fbxw7^ΔG mice 3 d postpartum (Fig. 5, C vs. D). Highly elevated DEK levels were also observed in adenomas of adult double mutant mice (unpublished data). Therefore, we investigated whether DEK and TPM play a role in the regulation of cell growth.

Comparable IHC analysis on wild-type FBXW7 (FBXW7^wt) and FBXW7 mutated (FBXW7^mut) human colorectal tumors tissues revealed that TPM and DEK expressions in ECs are correlated with FBXW7 mutations in human colon cancer (Fig. 5, E–N; and Tables S2 and S3). TPM expression was elevated in smooth muscle (stromal) cells from FBXW7^mut tumors when compared with FBXW7^wt tumors (Fig. 5, H vs. L). Quantification data, consistent with our mouse model observation, show that TPM expression in epithelia decreased with FBXW7 mutation (Table S3; P = 0.0024), whereas its expression was elevated in stroma compared with FBXW7^wt tumors (Table S3; P = 0.0106; and Table S2). DEK staining was more intense in FBXW7^mut ECs (Fig. 5, J vs. N), revealing its high levels of expression in FBXW7^mut tumor tissues (Table S3; P = 0.0010).

Notably, lower expression of TPM expression is correlated with a high level of DEK expression in ECs (Fig. 5, H vs. J and L vs. N). Collectively, these data show that accumulation of DEK and loss of epithelial TPM may contribute to the oncogenicity of FBXW7 mutation in both human CRC and in the Apc^Min/+Fbxw7^ΔG mouse intestine, which is indicative of possible roles for DEK and TPM in colorectal tumorigenesis (Prasad et al., 1993; Bharadwaj and Prasad, 2002; Mlakar et al., 2009; Wise-Draper et al., 2009).

To test whether there was a functional relationship between CRC cell proliferation and their FBXW7 status, single HCT116 cells lacking or expressing FBXW7 (Mao et al., 2004a; Mao et al., 2008), HCT116^−/−, HCT116^+/−, and HCT116^+/+ cells, were cultured for colony forming assays (Fig. S8 A). Both CRC HCT116^+/− and HCT116^−/− cell lines, harboring deletion of FBXW7 caused an increase in size and number of cell colonies (Fig. 5 O; 0.0001 < P < 0.05). In separate tissue culture studies, we investigated the role that TPM may play in CRC cells. Expression of TPM1-α is widely distributed in both smooth muscle cells and the cytoskeleton of ECs (Gunning et al., 2005; Helfman et al., 2008). Therefore, the effects of TPM expression on proliferation, cell migration, and wound healing were then tested in both HCT116^+/+ and HCT116^−/− cells. Both cells were thus transfected with FLAG–TPM1-α and compared with pcDNA backbone vector (control)–transfected cells (Fig. S8 B). We observed that ectopic expression of TPM1-α increased cell number in G0/G1 phase and decreased the number of cells in S phase in both HCT116^−/− and HCT116^+/+ cells (Fig. 5 P; 0.0001 < P < 0.001). A wound healing assay using the aforementioned transfection protocol in both cells lines (HCT116^−/− and HCT116^+/+) was performed, in which overexpression of TPM1-α repressed wound healing (Fig. 5 Q; 0.0001 < P < 0.05; and Fig. S8 C). Transfection of FLAG–TPM1-α significantly reduced HCT116^+/+ (P < 0.001) and HCT116^−/− cell motility in trans-well migration assay (P = 0.043). Western blotting analyses suggest that ectopic expression of TPM1-α in both HCT116^−/− and HCT116^+/+ cells led to an increased level of E-cadherin (Fig. S8 B), which is suggestive of wound healing and cell migration repression (Fig. 5, Q and R). We also found that c-Jun protein levels were decreased in cells transfected with TPM1-α (Fig. S8 B), which is a plausible functional correlation to reduced cell proliferation (Fig. 5 P).

We then tested whether DEK induces CRC cell division. HCT116 cells were transfected with the GFP-DEK expression vector (HCT116^GFP-DEK) and DEK-specific duplex small interfering RNA (siRNA; HCT116^DEK−), and cell cycles were compared with cells transfected with sequence scrambled control siRNA (control; HCT116^ssc). Two different siRNA duplexes were used, and down-regulation of DEK was similar (Fig. S8 D). DEK knockdown significantly increased cell number in G0/G1 phase (P < 0.001) and decreased the number of cells in S phase (Fig. 5 S, gray bars). In contrast, ectopic expression of GFP-DEK increased the number of cells forming the S-phase plateau (HCT116^GFP-DEK vs. HCT116^ssc; P < 0.001) as well as the G2/M-phase peak (HCT116^GFP-DEK vs. HCT116^ssc; P < 0.05; Fig. 5 S, gray bars), thus suggesting that DEK might play a role in inducing cell division.

To elucidate whether nuclear accumulation of oncogenic β-catenin in the Apc^Min/+ mice and oncogenic DEK in the Fbxw7^ΔG mice cooperate in carcinogenesis as in the Apc^Min/+Fbxw7^ΔG mice, we examined β-catenin levels in human CRC HCT116^−/− cells, which harbor deletion of FBXW7 and a β-catenin phosphorylation site mutation. Both β-catenin and DEK protein levels were increased in HCT116^−/− cells (Fig. 5 T), suggesting that HCT116^−/− cells may mimic the molecular defects of adenomas developed by Apc^Min/+Fbxw7^ΔG mice. Fbxw7 and β-catenin coimmunoprecipitation (co-IP) experiments did not show a direct interaction between Fbxw7 and β-catenin proteins overexpressed in HEK293T cells (unpublished data). Thus, SCF^Fbxw7 E3 ligase activity may not directly influence the level of β-catenin. Therefore, collectively, these observations from the double mutant mice and human CRC tissues and cells support the conclusion that DEK accumulation and TPM suppression could be necessary events for the transformed phenotype.

DEK is involved in gene transcription and RNA splice site recognition (Soares et al., 2006). Posttranslational modifications play an important role in regulating DEK’s binding to DNA and chromatin proteins (Alexiadis et al., 2000; Kappes et al., 2004), and the interaction of DEK with U2AF-RNA requires phosphorylation (Soares et al., 2006). DEK is highly phosphorylated by multiple kinases (Kappes et al., 2004; Soares et al., 2006), and sequence comparison reveals that DEK contains phosphorylation motifs similar to the CDC4-phospho-degron described in other Fbxw7 target proteins (Welcker and Clurman, 2008). In particular, DEK protein harbors four GSK-3β phosphorylation clusters (Fig. S7 B), two at the N terminus (Thr¹⁵ and Thr⁶⁷) and the other two at the C terminus (S¹⁶⁹ and S²²⁷; Kappes et al., 2004). How DEK protooncogene function and its expression levels are regulated is largely unclear.

In examining which FBXW7 isoform regulates the E3 ligase activity toward DEK degradation, it is apparent that overexpression of the FBXW7 isoforms in a human cell line HEK293T reveals that DEK, in the presence of GSK-3β, is targeted by FBXW7-α and less so by FBXW7-γ and FBXW7-β (Fig. 6, A [lane 1 vs. lanes 2 and 3] and C [lane 2 vs. lane 1]). Overexpression of Fbxw7-α in HEK293T did not alter DEK mRNA levels (unpublished data). Moreover, GFP-DEK^Δ100 (100-aa deletion mutation at N terminus of DEK protein) co-overexpressed with FBXW7-α was unchanged (Fig. S7 C). Indeed, phosphorylation occurs at the N terminus of DEK protein upon in vitro translation (IVT) of DEK(1–100) in rabbit reticulocyte (RT) lysates, which was observed with higher electrophoretic mobility and treatment of RT lysates containing p-DEK with λ-phosphatase, resulting in an unphosphorylated, faster-migrating form (Fig. 6 B, top). Consistent with this, previous data also confirmed N-terminal DEK phosphorylation by in vitro kinase assays in which DEK was translated in vitro in RT lysates (Soares et al., 2006).

Mutations at either Thr¹⁵ or Thr⁶⁷ also did not stabilize DEK (unpublished data). In contrast, both phosphorylation clusters of DEK mutant protein (DEK^T15,67A or DEK^2A; Fig. 6 B, bottom) are similar in the presence or absence of FBXW7-α (Fig. 6 C, lanes 4 and 5), suggesting that both N-terminal phosphorylation clusters at Thr¹⁵ and Thr⁶⁷ contribute to recognition of DEK by SCF^FBXW7-α for degradation. Blocking phosphorylation and inhibition of GSK-3β expression using LiCl or proteasome inhibitor–stabilized p-DEK (Fig. 6 C, lanes 3 and 6 vs. lane 2). p-DEK was stable in HEK293T cells (Soares et al., 2006), but coexpression of FBXW7 decreased the steady-state levels of DEK and resulted in reduced protein half-life (Fig. 6 D; Nateri et al., 2004). In contrast, the stability of DEK^2A mutant protein was not affected in the presence of Fbxw7 (Fig. 6 D). These data suggest that the E3 ligase SCF^Fbxw7-α degrades p-DEK in a GSK-3β–dependent manner. Consistent with this, HCT116^−/− cells restored the stability of ectopic expression of DEK protein (Fig. 6 E). Therefore, DEK protein level may be directly modulated by SCF^Fbxw7-α E3 ligase activity.

IP of FLAG epitope–tagged FBXW7-α–GFP revealed its interaction with p-DEK–GFP protein (Fig. 6 F, lane 3) in HEK293T cells. Also, IP of in vitro translated FLAG-tagged Fbxw7 showed a direct interaction with the DEK protein (Fig. 6 G). We next examined whether FBXW7 possesses E3 ligase activity toward p-DEK by assessing DEK ubiquitination in HEK293T cells (Wu et al., 2007). DEK was efficiently ubiquitinated (Fig. 6 H, lane 3 vs. lanes 1 and 2), and therefore SCF^FBXW7 has properties consistent with a function as an E3 ubiquitin ligase that specifically targets p-DEK.

Recent experiments have demonstrated that p-DEK influences alternative splicing of RNA (Soares et al., 2006). We examined whether deregulated p-DEK changes alternative RNA splicing of TPM (Fig. 4, I vs. J and M; Gooding and Smith, 2008). We used a double reporter system for determining splicing efficiency (Fig. 7 A; Nasim and Eperon, 2006). A short splicing region containing TPM target intron and its splicing sites (ss) was introduced between two reporters encoding β-galactosidase (β-gal) and luciferase (Luc), such that the upstream β-gal is expressed regardless of splicing (Fig. 7 A). However, the Luc will be expressed unless the multiple stop-codons (reds) residing in the splicing region are removed (Fig. 7 A). Therefore, the proportion of transcript variants can be evaluated by the relative ratio of Luc to β-gal expression.

We show that endogenous p-DEK (Fig. 7 B, lane 1) but not unphosphorylated DEK (Fig. 7 B, lane 5), significantly enhanced the TPM RNA splicing efficiency. A similar observation was made when DEK was overexpressed in the presence and absence of GSK-3β (Fig. 7 B, lane 3 vs. lane 7). Therefore, the phosphorylation of DEK by GSK-3β is a crucial step to mediate Tpm RNA splicing. Intriguingly, p-DEK targeted by FBXW7 significantly reduced the splicing efficiency (Fig. 7 B, lanes 2 and 4 vs. lanes 1 and 3). In addition, the assessment of splicing efficiency by the Luc/β-gal ratio was also independently confirmed by RT-PCR (Fig. 7 C). Consistent with our mouse model (Fig. 4 M), enhanced DEK in HCT116^−/− cells is sufficient to facilitate TPM shifts to a higher molecular weight (Fig. 7 D, arrowhead). DEK depletion in HCT116^−/− using the DEK-specific duplex siRNA significantly reduced TPM RNA splicing efficiency (Fig. 7 E, left) and also reduced the TPM shifts to the higher molecular weight (Fig. 7 E, right). These data suggest that TPM splicing is directly regulated by GSK-3β/DEK, which in turn is subjected to SCF^FBXW7 E3 ligase activity.

Intestinal tumor formation in Apc^Min/+Fbxw7^ΔG mice apparently occurs by repopulation of preexisting crypts (unpublished data) with APC mutant cells, as reported for humans (Shih et al., 2001). To determine whether Fbxw7 depletion results in the rapid transformation of intestinal crypt cells into adenomas in Apc^Min/+Fbxw7^ΔG, we examined the proliferation index of mutants versus litter mate control mice at different ages. Results showed that Apc^Min/+Fbxw7^ΔG mice significantly induced proliferation compared with control Apc^Min/+Fbxw7^fl/fl mice at 3 d postpartum (measurement as Ki67⁺ cells/micrometer; P < 0.001; Fig. 8 A). Ki67⁺ cells were also significantly increased in Fbxw7^ΔG crypts compared with Fbxw7^fl/fl at 3 wk old (measurement as Ki67⁺ cells/crypt; P < 0.001; Fig. 8 B). Further molecular characterization on protein extracts from small intestine isolated at 3 d postpartum indicated that in the Apc^Min/+Fbxw7^ΔG, the levels of Notch 1, c-Jun, and DEK were significantly increased (Fig. 8 C), similar to the levels present in the Fbxw7^ΔG intestine (Figs. 3 C and 4 N). CyclinD1 protein levels, a positive proliferation marker, were also increased, and TPM protein expression patterns consequently changed (Fig. 8 C). We concluded that Fbxw7^ΔG crypt cells bearing multiple subsequent deregulations in Notch, Jun, and DEK activation result both in early tumor initiation and proliferation and in an increase of tumor incidence at 2–3 wk of age.

Despite the well-established links between signaling and transcriptional activity, little is known about how transcriptional activators/repressors are regulated once tumorigenesis has been triggered. One mechanism is by limiting the lifetime of an activator/repressor by destroying it via the ubiquitin–proteasome system (Nakayama and Nakayama, 2006; Welcker and Clurman, 2008). Our study reveals a previously unrecognized role for SCF^FBXW7 E3 ligase and its substrates in regulating intestinal epithelial homeostasis and tumorigenesis.

In this study, the Fbxw7^ΔG mice at 9–10 mo of age showed adenomas, and crypt budding and fission were also found in the small intestine (Fig. S2). Crypt budding and fission occur in normal intestinal growth and repair as well as in polyposis. It has been suggested that these events occur when the number of intestinal stem cells in a crypt exceeds a threshold (Brittan and Wright, 2002). However, further analysis is required to validate this concept in our mutant mice. Furthermore, inactivation of both Fbxw7 and APC appears to be synergistically accelerating the malignant progress from adenoma to adenocarcinoma (Fig. 2 and Fig. S3).

Fbxw7^ΔG mice showed activation of Notch, c-Jun, and their downstream targets (Fig. 3 and Fig. S5, G–K). However, our result is consistent with recent reports that activation of the Notch signaling pathways mainly induced Hes-1 but not Hes-5 expression; Notch signaling is implicated in the regulation of cell fate specification predominantly by Hes-1 (Fre et al., 2005; van Es et al., 2005). The induction of Hes-1 but not Hes-5 presumably indicates that these genes could be regulated by additional transcription factors as well as Notch and consequently has overlapping but distinct patterns of expression in the gut. Whereas Hes-1, -6, and -7 transcripts were apparently restricted to the crypt region, Hes-5–expressing cells were also found in the villus epithelium and in the mesenchyme (Schröder and Gossler, 2002). Consistent with previous studies (Schröder and Gossler, 2002; Riccio et al., 2008), Notch 1 was most prominently expressed in ECs of the crypts (Fig. S5 H, left vs. right), where Fbxw7 is also highly expressed (Fig. 1 I), presumably by mediating Notch 1 down-regulation, and consequently Hes-1 was more affected than Hes-5.

We and others have reported that activation of the c-Jun and Notch signaling pathways is implicated in the regulation of cell fate specification and proliferative capacity in the intestine (Fre et al., 2005; Nateri et al., 2005; van Es et al., 2005; Sancho et al., 2009). Our work demonstrates an important role of Fbxw7 in these processes. In addition, Muc2 expression was also reduced in Fbxw7^ΔG intestines. Collectively, our results (Fig. S5, A–F), along with previous work (Fre et al., 2005; Nateri et al., 2005; van Es et al., 2005; Sancho et al., 2009), suggest that these signals maintain the homeostatic control of intestinal epithelium.

Recent studies have demonstrated that alterations in Notch and c-Jun signals could also be involved in Fbxw7^ΔG intestinal tumorigenesis (Nateri et al., 2005; Akiyoshi et al., 2008; Fre et al., 2009; Sancho et al., 2009). However, Muc2 inactivation also causes intestinal tumor formation (Velcich et al., 2002). Alternatively, expression of the cell cycle regulator cyclin E and the oncogene c-Myc were unaffected by loss of intestinal Fbxw7. Cyclin E expression is not always increased in Fbxw7-deficient cells and/or in Fbxw7^−/− embryos (Mao et al., 2004b; Rajagopalan et al., 2004; Tsunematsu et al., 2004). In addition, cyclin E levels were not affected in Fbxw7-deficient T cells (Onoyama et al., 2007). It is therefore possible that Fbxw7 mediates cyclin E degradation in a tissue-dependent manner. Recent studies have also identified Fbxw7-γ as the major player mediating c-Myc degradation (Welcker et al., 2004a; Bonetti et al., 2008), and it may not be affected in our mice because of the low level of endogenous intestinal Fbxw7-γ isoform expression. Accordingly, we showed that nuclear Fbxw7-α is the major isoform present in intestinal epithelia (Fig. S6 D).

The up-regulation of β-catenin in adenomas from double mutant (Apc^Min/+Fbxw7^ΔG) mice at 3 wk of age (Fig. 2 and Fig. S3) and in HCT116^−/− cells indicates a cross talk between Wnt/β-catenin signaling and FBXW7. However, Fbxw7-deleted intestinal epithelia did not lead to a significant accumulation of β-catenin in the intestines of 5/6-wk-old Fbxw7^ΔG mice (Fig. S5 G, left vs. right) and in 3-d-postnatal Apc^Min/+Fbxw7^ΔG mice (Fig. S4, F vs. H). We conclude that β-catenin accumulation may require additional factors after prolonged Fbxw7 or APC loss (Phelps et al., 2009).

Although it has already been demonstrated that different TPM isoforms are implicated in neoplastic-specific alterations and that decreased expression of specific nonmuscle TPM isoforms is commonly associated with the transformed phenotype (Helfman et al., 2008), we detected loss of epithelial isoforms of TPM in 3-d-postpartum Apc^Min/+Fbxw7^ΔG mice (Fig. 5, A vs. B, arrowheads). Furthermore, a strong association of epithelial TPM down-regulation in colorectal tumors with FBXW7 mutation (Fig. 5, H vs. L) and our data showing restored expression of TPM1-α in CRC cells (Fig. 5 P) support the conclusion that TPM suppression is required in the sequence leading from FBXW7 depletion to loss of cell growth regulation (Bharadwaj and Prasad, 2002; Varga et al., 2005; Zhu et al., 2007; Zheng et al., 2008; Mlakar et al., 2009). Nevertheless, TPMs exhibit substantial diversity in terms of their size, binding affinity to actin, subcellular localization, and in interactions with other proteins, suggesting that individual TPMs may differ in their biological functions: cell migration and invasion (Miyado et al., 1996; Gunning et al., 2005; Gupton et al., 2005). In parallel to 3-d-postnatal Apc^Min/+Fbxw7^ΔG mice (Fig. 5, D vs. C), and in line with a role for DEK in tumorigenesis (Wise-Draper et al., 2009), we also showed that human DEK is highly expressed in human CRC tissues and is correlated significantly with FBXW7 (Fig. 5, N vs. J; and Table S3) mutation and also mediates cell division (Fig. S5). Beyond the current scope of this paper, further contribution of DEK and TPM functions could be compared with other effects of Fbxw7.

Mechanistically, biochemical analysis showed that DEK degradation is mediated by GSK-3β–SCF^Fbxw7 and that the knockdown or overexpression of Fbxw7 reversibly affects the level of DEK (Figs. 4 and 6). Of note, DEK transcript levels were unchanged, suggesting that DEK is not transcriptionally regulated by activated Notch and/or c-Jun signaling pathways. In addition, we showed that p-DEK changes the alternative RNA splicing of TPM and that the knockdown or overexpression of DEK can be inversely correlated with the levels of different isoforms of TPM (Fig. 7). These findings therefore suggest that aberrant DEK-TPM signaling after loss of the Fbxw7 causes loss of cell growth regulation and consequently, after APC loss and β-catenin accumulation, promote adenoma progression to adenocarcinoma in our double mutant Apc^Min/+Fbxw7^ΔG mice. However, further in vivo analysis would be required to validate this concept by genetic intercrosses and generation of Tpm- and/or Dek-deficient mice. Heterozygous deletion of Fbxw7 in Fbxw7^ΔG/fl mice with no significant abnormality phenotype may phenocopy FBXW7 mutations in human CRC with no loss of a second allele and may function as a haploinsufficient tumor suppressor (Rajagopalan et al., 2004; Kemp et al., 2005). Further analysis, beyond the current scope of this paper, would be required to validate this concept in Apc^Min/+ mice.

We conclude that Fbxw7 mutation causes failed intestinal differentiation and increased proliferation and occurs in the absence of cyclin E and c-Myc accumulation. Fbxw7 mutation results in accumulation of multiple substrates and in impaired degradation of Notch, Jun, and DEK, which cooperatively contribute to carcinogenesis (Fig. S8 F). Consistent with this model, an intact and functional Fbxw7 is thought to restore homeostatic control mechanisms and impair oncogenic activators. Therefore, strategies based on the partial delivery of Fbxw7-α to intestinal tumor cells would attenuate nuclear DEK and β-catenin accumulation and show promise for fighting colon cancer. Moreover, given the reduced lifespan and increased incidence of intestinal tumors and mortality in Apc^Min/+Fbxw7^ΔG mice, these mice could be useful for testing carcinogens and drug screening.

Floxed Fbxw7 (Fbxw7^fl/fl) mice (Fig. S1) were backcrossed a minimum of five times to a C57BL/6 background; Villin-Cre transgenic mice (el Marjou et al., 2004) and Apc^Min/+ mice (Fodde, 2002) were on a C57BL/6 genetic background (Nateri et al., 2005). Villin-Cre⁺Fbxw7^fl/fl (Fbxw7^ΔG) mice were generated by interbreeding mice carrying Fbxw7^fl/fl and the Villin-Cre allele. Apc^Min/+Fbxw7^fl/fl mice were generated by interbreeding mice carrying the floxed alleles of Fbxw7 (Fbxw7^fl/fl) and Apc^Min/+, and then Apc^Min/+Fbxw7^fl/fl were intercrossed with the Villin-Cre⁺Fbxw7^fl/fl (Fbxw7^ΔG). In all experiments, only littermates were used and compared. Mice were maintained in a specific pathogen-free condition. All experiments were performed at the University of Nottingham in accordance with institutional biomedical service unit guidelines.

Intestines were removed, flushed with cold PBS, and processed as “Swiss roll” gut preparations as described previously (Nateri et al., 2005). Tissues were fixed overnight in 10% neutral-buffered formalin, paraffin embedded, and sectioned at 4-µm thickness for hematoxylin and eosin (H&E) staining or IHC with the following primary antibodies: active caspase-3, c-Jun, and p–c-Myc (Cell Signaling Technology); lysozyme, Chromogranin A, and CD44s (Millipore); anti-Ki67 (Dako); Notch 1, Notch 4, and Muc2 (Santa Cruz Biotechnology, Inc.); β-catenin (BD); DEK (Abcam); and Fbxw7 (Thermo Fisher Scientific). Where antigen retrieval was required, slides were microwaved in Na citrate buffer, pH 6, for 12 min. All secondary antibodies were biotinylated (Vector Laboratories), and after incubation with avidin–biotin complex (VectaStain ABC kit; Vector Laboratories), sections were developed with diaminobenzidine. Alkaline phosphatase, periodic acid Schiff (Sigma-Aldrich), and Alcian blue (BDH) were used to identify intestinal cell types according to manufacturers. In adult mice, the Ki67-positive cells per crypt were measured and/or the caspase-3 in 200 full crypts or villi was scored from at least three mice of each genotype. Data for caspase-3 are represented as percentage over control ± SEM. In 3-d-postnatal mice, Ki67⁺ cells were scored from three mice and measured per micrometer, and data are represented in comparison with littermate control ± SEM.

Human CRC specimens from a total of five cases with and six cases without FBXW7 mutations were obtained on separate slides/sections. FBXW7 mutations without loss of heterozygosity were identified by sequencing analysis. Ethical approval and research and development approval were obtained from the Local Research Ethics Committee and the Trust Research and Development office, respectively, in Oxford. For IHC, sections were dewaxed with xylene, rehydrated through graded alcohol, and then boiled in 0.01 M Na citrate buffer, pH 6, for 10 min for antigen retrieval. Immunostaining was performed as outlined above for DEK and TPM antibodies. Staining was evaluated semiquantitatively using an ordinal score scale of 1 = lowest staining and 4 = extensive intense staining. Staining was evaluated independently by four observers. Statistical analysis was performed using SPSS, and FBXW7 statuses were examined for both DEK and TPM for each tumor using the mean of scores ± SD, and a p-value <0.05 was considered statistically significant.

The mouse Fbxw7 RNA probe was generated from an 800-bp mouse Fbxw7 PCR product (confirmed by sequencing) transcribed with T7 RNA polymerase using the DIG RNA labeling mix (Roche). In brief, 4% PFA-fixed intestinal Swiss roll preparations were sectioned (8 µm), dewaxed, rehydrated, digested with 10 µg/ml proteinase K (37°C for 18 min), and then incubated in 0.2 M HCl and refixed. After prehybridization for 1 h, slides were hybridized overnight at 57°C in buffer containing 50% deionized formamide, 10% dextran sulfate, 1× Denhardt’s solution, 10 mM Tris-HCl, pH 7.6, 600 mM NaCl, 0.25% Na dodecyl sulfate, 1 mM EDTA, 2 mg/ml transfer RNA, and ∼2 µg/ml denatured digoxigenin–labeled cRNA probe. The slides were rinsed in 5× SSC at 57°C for 10 min, washed once in 50% formamide in 2× SSC at 57°C for 30 min, and then once in 2× SSC at 57°C for 30 min and twice in 0.2× SSC for 30 min each at 57°C. After blocking with 1% blocking reagent (Roche), slides were incubated with sheep antidigoxigenin fab antibody (1:1,000; Roche) for 2 h at 37°C. Substrate was developed using NBT/BCIP.

For qRT-PCR analysis, total RNA was isolated from freshly dissected intestines using TRIZOL reagent (Sigma-Aldrich) and purified using the RNeasy mini kit including DNase (QIAGEN) according to the manufacturer’s instructions. cDNA synthesis was performed using Superscript reagents (Invitrogen) according to the manufacturer’s instructions. Quantitative real-time PCR was accomplished with SYBR green incorporation (Platinum Quantitative PCR SuperMix-UDG w/ROX; Invitrogen) using an ABI7900HT (Applied Biosystems), and the data were analyzed using SDS 2.3 software (Applied Biosystems). Results were normalized to those obtained with β-actin, and data are presented as fold induction/repression over control mice. Details of primers used are shown in Table S1. All assays were performed in triplicate at least three times.

Intestinal samples from young (≤5 wk) mutant Fbxw7^ΔG and control Fbxw7^fl/fl mice were initially homogenized with ice-cold PBS, and protein extracts were derived following fractionation cytosolic and nuclear extracts according to the manufacturer’s instructions (BioVision), resuspended in two-dimensional lysis buffer (Kim et al., 2007), and loaded separately onto Immobiline DryStrip gels containing a preformed pH gradient (pH 3–10), and proteins were run on the Protean IEF Cell (Bio-Rad Laboratories) according to the manufacturer’s instructions, separated on a 12% polyacrylamide gel, and then stained with Coomassie blue. Gels were scanned on a calibrated imaging densitometer (GS-800; Bio-Rad Laboratories), and images were analyzed using PDQuest (Bio-Rad Laboratories). All assays were performed at least three times. For an accurate determination of the weight of the novel protein or proteins, MALDI-MS provided by the protein chemistry facility with a Mass-Prep robotic liquid handling system and MALDI TOF mass spectrometer (Waters Corporation) in the School of Biomedical Sciences, University of Nottingham was used. Peak lists were entered into MASCOT PMF (http://www.matrixscience.com/search_form_select.html) and ExPASy (http://www.expasy.org/tools/aldente/) database search engines.

Intestinal epithelia were released from murine small intestine by incubation for 30 min at 4°C in PBS containing 2 mM EDTA as previously described (Sato et al., 2009), and immunoblots were performed as previously described (Nateri et al., 2005).

Kaplan–Meier survival analysis was used for Fbxw7^ΔGApc^Min/+ mice and their transgenic littermate (Fbxw7^fl/flApc^Min/+, Fbxw7^ΔG) mice over a period of 62 wk (435 d; n = 14). The survival curve shows the percentage of mice remaining alive at 1-wk intervals. The x axis indicates times (in weeks) until mice were euthanized because of severe weight loss and anemia. 14 mice for each genotype group were used, and the survival curves were significantly different from each other (P < 0.001).

Human CRC HCT116 and HEK293T cells were maintained in RPMI supplemented with 10% FBS. HCT116 cell lines harboring deletion of FBXW7 were a gift from A. Balmain (University of California, San Francisco, San Francisco, CA) and B. Vogelstein (Ludwig Center, Johns Hopkins University, Baltimore, MD; Mao et al., 2004a, 2008). Transfection of plasmids encoding FBXW7-α, -β, and -γ, DEK, GSK-3β, Tpm1-α (available from GenBank/EMBL/DDBJ under accession no. NM_000366), Luc, β-gal, GFP (control), or siRNA duplexes along with others was performed using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer’s instruction. FLAG–GSK-3β plasmid was a gift from B.W. O’Malley (Baylor College of Medicine, Houston, TX; Wu et al., 2007), human DEK-GFP construct was a gift from J. Valcarcel (Centre de Regulació Genòmica, Barcelona, Spain) and A. Bürkle (University of Konstanz, Konstanz, Germany; Soares et al., 2006), and FLAG–Tpm1-α was a gift from J. Huot (Le Centre de recherche en cancérologie de l’Université Laval, Québec, Québec, Canada; Houle et al., 2007). To measure proliferation, HCT116 cells were transfected with an enhanced GFP expression vector (pC2-EGFP; Takara Bio Inc.) and/or with sequence scrambled controls and siDEK duplexes (Invitrogen) and peGFP-DEK vector. Transfection efficiency judged by enhanced GFP expression was 50–60%. 36 h after transfection, cells were collected by centrifugation, ethanol (70%) fixed, and treated with 50 µg DNase-free RNase/ml for 30 min, and DNA was stained with 50 µg propidium iodide/ml and analyzed by FACS. All assays were performed at least three times.

Cell migration was measured using 24-well plates with polycarbonate membrane trans-well inserts with an 8-µm pore size (Corning). 2.5 × 10⁴ cells were seeded, and cell migration was assessed after 24 h by fixing the cells attached to the lower surface in 10% methanol for 30 min. Cells were then stained with methylene blue and manually counted. Cell wound healing assays were performed in 6-well dishes. Cells were grown to >90% confluence and then starved in low serum media (0.1% serum in RPMI) for 24 h. A sterile 200-µl pipette tip was used to scratch three separate wounds. Healing across the wound line was assessed at 24 h. Images were analyzed using the ImageJ analysis program (National Institutes of Health), and the area of the wound was measured. For the wound healing assays, total pixel numbers with an intensity value of 0 were analyzed. All assays were performed at least three times in triplicate. Colony forming assay was performed using single cell suspensions, and 50 live cells (stained with Trypan blue) were plated in a 24-well precoated poly-l-lysine plate. After 2 d of incubation, single GFP^+ve cells and, after 7 d, colonies were identified for fluorescent microscopy experiments. Colonies were counted after 2 wk, and images were captured. Images were analyzed using the ImageJ analysis program, and the area of the colony was measured as outlined above. The experiments were repeated in triplicate and on two separate occasions.

Mice were sacrificed by cervical dislocation, and the intestines were removed and flushed extensively with cold PBS. The first 3 cm of the small intestine and large bowel were taken and homogenized in RIPA lysis buffer (Sigma-Aldrich) supplemented with protease inhibitor cocktail (Sigma-Aldrich). Immunoblots were performed as previously described (Nateri et al., 2005). Equal amount of total protein was denatured in the SDS sample loading buffer, separated on 8–16% SDS–polyacrylamide gradient gels, and transferred to polyvinylidene fluoride membranes, which were immunoblotted with various antibodies as indicated. Antibodies to polyclonal anti-Fbxw7 (Thermo Fisher Scientific and Invitrogen); cyclin E, c-Jun, p–c-JunS63, and p–c-MycT58/S62 (Cell Signaling Technology); DEK and β-actin (Abcam); and Notch 1 and Notch 4 (Santa Cruz Biotechnology, Inc.) were used. Antibodies to monoclonal GFP were obtained from Invitrogen, and antibodies to FLAG and hemagglutinin (HA) were obtained from Sigma-Aldrich. Western blot analysis for HCT116 and HEK293T cells was performed as previously described (Nateri et al., 2004). Cells were treated with LiCl (Sigma-Aldrich) and MG132 proteasome inhibitor (EMD) when required. The experiments were repeated on at least three separate occasions.

Here we used 100 µl of anti-FLAG–conjugated agarose beads (Sigma-Aldrich) to immunoprecipitate Fbxw7 from whole cell extracts of the HEK293T cells using RIPA buffer (150 mM NaCl) with protease inhibitors. After IP, the beads were washed thoroughly with RIPA buffer adjusted to 500 mM. Immunoprecipitated proteins were eluted using 2× SDS loading buffer and then boiled at 95°C for 4 min. Denatured proteins were subsequently separated on 10% SDS PAGE and immunoblotted against anti-GFP and anti-HA antibodies as required after transferring to polyvinylidene fluoride membranes as previously described (Wu et al., 2007). The experiments were repeated on at least two separate occasions.

HEK293T cells were transfected with DEK wild-type (DEK^wt) and/or DEK phosphorylation mutant (DEK^T15,67A or DEK^2A) with Fbxw7-expressing plasmids. 30 h after transfection, cells were radioactively labeled for 3 h with 0.2 mCi [³⁵S]methionine and [³⁵S]cysteine/ml in a labeling medium lacking methionine and cysteine (ICN), followed by chase in complete medium with 1% FCS serum. Cell extracts and IP were performed with rabbit polyclonal antibody to GFP agarose-conjugated beads (Abcam) as previously described (Nateri et al., 2004).

Activity of Luc and β-Gal proteins was measured using a Dual-Light reporter system (Applied Biosystems) according to the manufacturer’s instruction. The reporter assays were performed in triplicates using a multifunction microplate reader (FLUOstar OPTIMA; BMG Labtech) after transient transfection of the bicistronic reporter along with plasmids encoding DEK, FBXW7-α, GSK-3β, and GFP (control) into HCT116 cells. For statistical analysis, the ratio of Luc activity to β-Gal was obtained from three independent experiments. Mean ± SD (n = 3; P < 0.001) values are shown. RT-PCR analysis of total RNA isolated from transfected cells represents an independent experiment but parallel to the reporter assay. DNA bands were derived from spliced and unspliced RNA.

The significance of differences between mean and median was determined using the Student’s t test and the Mann-Whitney U test, as appropriate. Significance testing was performed using SPSS version 15. Mean ± SD (n = 3; *, P < 0.05; **, P < 0.01; ***, P < 0.001) values are shown.

Fig. S1 shows the generation and characterization of intestinal (gut)-specific Fbxw7 knockout (Fbxw7^ΔG) mice. Fig. S2 shows that intestinal targeted inactivation of Fbxw7 leads to colonic and intestinal polyposis. Fig. S3 shows polyps and adenomas in Apc^Min/+Fbxw7^ΔG mice at an early age (3 wk old). Fig. S4 shows that a very small number and size of intestinal polyps/adenomas appeared from 3-wk-old Apc^Min/+ mice with β-catenin^High staining. Fig. S5 shows Fbxw7^ΔG lineage potential and elevation of Fbxw7 substrate expression upon depletion of Fbxw7 in mouse intestine. Fig. S6 shows IHC for c-Jun and FBXW7-α, -β, and -γ protein isoforms. Fig. S7 shows Western blot analysis of TPM protein expression upon overexpression of FBXW7s and phosphorylation of human DEK protein by GSK-3β. Fig. S8 shows cell growth regulation by human DEK and TPM1-α proteins. Table S1 lists primers used for genotyping and qRT-PCR. Table S2 lists TPM and DEK expression scores in sections from human colorectal tumor tissues with and without FBXW7 mutations. Table S3 shows the score quantification of TPM and DEK IHC performed in FBXW7^wt and FBXW7^mut human colorectal tumors.

We thank B. Vogelstein for providing HCT116 knockout cells, B.W. O’Malley, J. Valcarcel, and A. Bürkle for vectors, M. Ilyas, A. Grabowska, and C. Anbalagan for reading the manuscript, and E. Nye and P. Clarke for technical advice.

This study was supported by Cancer Research UK, research projects A9275 and A10247, and the University of Nottingham New Research Fund to A.S. Nateri.

No financial conflicts of interest exist.