A new mechanism of SOX9 action to regulate PKCα expression in the intestine epithelium

Interfering with the molecular mechanisms regulating the expression and/or activity of the effectors of proliferation-to-differentiation transition (PDT) is an attractive approach to inhibit tumor-cell growth and/or reorientate tumor cells towards a post-mitotic differentiated state.

SOX transcription factors, which contain a high-mobility group (HMG) domain, have been clearly evidenced as important regulators of the PDT during development and adult tissue renewal (Episkopou, 2005; Kanai et al., 2005; Shimizu et al., 2007; Zorn and Wells, 2007). In particular, findings from our laboratory indicate an involvement of SOX9 in the control of proliferation, differentiation and cell fate of the intestinal epithelium (Bastide et al., 2007). However, the molecular mechanisms involved in these biological processes still remain to be elucidated. Up to now, CEACAM1 is the only SOX9 direct target identified in the intestinal epithelium whose gene expression is upregulated through the binding of the SOX9 HMG domain on a sequence similar to the in vitro characterized consensus sequence [A/T][A/T]CAA[A/T]G (Kamachi et al., 2000; Zalzali et al., 2008). The expression of genes encoding, for example, the CDX2 cell-differentiation marker, the carcinoembryonic antigen (CEA) or the tight-junction protein claudin 7 is inversely decreased in response to an overexpression of SOX9, but it was concluded that these repressions might be indirect, i.e. mediated through the prior transcriptional activation of as-yet-unidentified repressor genes (Blache et al., 2004; Darido et al., 2008; Jay et al., 2005). Indeed, SOX9 was always found to be a transcriptional activator in a number of physiological situations (De Santa Barbara et al., 1998; Lee et al., 2004; Ng et al., 1997) and no SOX9 DNA-binding sequence could be identified within the promoter of these genes.

In vitro and in vivo data indicate that variations in the amount of the tumor-promoting phorbol-ester receptor protein kinase C (PKC) also drastically influence the PDT and might have an important impact on intestinal tumorigenesis. Indeed, overexpression of both PKCβ2 and PKCϵ stimulates the proliferation of colon epithelial cells (Murray et al., 1999; Perletti et al., 1996) and a high PKCβ2 expression level is considered as an early promoting event in colon carcinogenesis (Gokmen-Polar et al., 2001). By contrast, overexpression of PKCα induces cell-growth arrest and differentiation of intestinal epithelial cells, whereas opposite effects are observed with antisense oligonucleotides encoding PKCα (Scaglione-Sewell et al., 1998). In vivo, cell-growth inhibition and differentiation that occur when intestinal cells migrate from the proliferative crypt towards the differentiated and functional villus correlate with an increase in PKCα expression (Saxon et al., 1994), and knockout of the gene encoding PKCα is associated with the acquisition of the proliferative phenotype of intestinal epithelial cells (Oster and Leitges, 2006). Finally, PKCα is usually decreased in colorectal cancers (CRCs), suggesting a protective role of this enzyme from intestinal tumorigenesis (Kahl-Rainer et al., 1994; Suga et al., 1998).

Leontieva and Black identified two distinct pathways that are able to regulate PKCα contents in intestinal epithelial cells, both of which occur at the post-translational level but with no obvious link with the PDT (Leontieva and Black, 2004). The present study establishes that transcription of the gene encoding PKCα can be repressed both in vitro and in vivo by SOX9 in proliferating intestinal epithelial cells and that this repression is mediated through a DNA-binding-independent transcriptional mechanism.

A SOX9-dependent transcriptome (http://www.ebi.ac.uk/, ArrayExpress database, accession number E-MEXP-859) recently performed in our laboratory suggested that doxycycline-inducible FLAG-SOX9 expression was able to decrease the level of mRNA encoding PKCα by more than 50% in the human HT29Cl16E CRC cell line. As shown in Fig. 1A,B, quantitative reverse transcriptase (RT)-PCR and western blot experiments demonstrated that a 2.5-fold increase of FLAG-SOX9 expression in these cells induced a 75% decrease in the amount of mRNA encoding PKCα and a 50% decrease of PKCα protein, thus confirming a SOX9-mediated repression of PKCα transcription. Consistent with this result, the luciferase activity of reporter plasmids containing either the –1571 to +227 or the –227 to +77 minimum promoter of the gene encoding PKCα was inhibited by 40-60% in HT29Cl16E and HCT116 human CRC cell lines that were transiently transfected with SOX9 (Fig. 1C), demonstrating that SOX9-dependent PKCα repression is not restricted to HT29Cl16E cells and can be mediated through the minimum –227 to +77 promoter of the gene encoding PKCα. SOX9 overexpression did not significantly influence PKCα-promoter-dependent luciferase activity in SW480 human CRC cells, but this might be due to the high endogenous SOX9 in these cells, compared with HT29Cl16E and HCT116 cells (Fig. 1D). Inversely, the activities of the –1571 to +227 and –227 to +77 PKCα-promoter–luciferase constructs were increased from 2 to 4.5-fold in both HT29Cl16E and HCT116 cells, and up to 8.5 fold in SW480 cells, in response to antisense SOX9 (Blache et al., 2004; Darido et al., 2008), suggesting that PKCα expression is repressed by endogenous SOX9 in these cells. Consistent with this, high SOX9 levels correlated inversely with low levels of mRNA encoding PKCα and of PKCα protein in these CRC cell lines (Fig. 1D-F).

In normal small intestine, SOX9 is highly expressed and concentrates in cell nuclei of the proliferative crypt compartment, and gradually disappears along the crypt-villus axis during differentiation (Fig. 2A). Inversely, the amount of PKCα is very low within the crypt and progressively increases during differentiation. Similar patterns were observed in the colon epithelium and in HT29Cl16E cells, which exhibit the ability to differentiate into the goblet-cell lineage when maintained as confluent cultures. This spontaneous differentiation is associated with a decrease in SOX9 expression (Jay et al., 2005) and correlates with the disappearance of SOX9 from the nucleus after 23 days in culture (Fig. 2B), and the concomitant increase in PKCα expression (Fig. 2C).

There was clear evidence of SOX9-dependent PKCα repression by the marked increase of PKCα immunostaining from the bottom to the top of the crypt region in both the small intestine and colon from mice that were deficient for SOX9 in the intestinal epithelium (Bastide et al., 2007) (Fig. 2A). The marked PKCα staining of villus cells is intriguing, but might result from the persistence of PKCα in SOX9-deficient crypt cells up to the time of their differentiation. Previous data reported a long half-life for PKCα, from 6 to more than 24 hours (Borner et al., 1988), whereas renewal of intestinal epithelium required 3-5 days. Finally, the western blot presented in Fig. 2D clearly demonstrates the accumulation of PKCα in small-intestinal and colonic epithelium of SOX9 knockout mice, indicating the loss of SOX9-dependent negative regulation of PKCα expression. PKCα is activated along the crypt-to-villus axis of SOX9-deficient intestinal epithelium (as evidenced by the accumulated enzyme at the cell-cell contacts and brush-border plasma membrane), suggesting that this enzyme might exert a biological function at the bottom of the crypt, where it is normally poorly expressed.

Similar to the other SOX-family members, SOX9 is expected to regulate transcription in association with other transcription factors by bending DNA after association of its HMG domain with the minor groove of the DNA helix at the consensus response sequence [A/T][A/T]CAA[A/T]G (Kamachi et al., 2000). Although no obvious SOX-consensus-binding site was identified on the PKCα (–227 to +77) promoter, SOX9 interaction with the promoter of the gene encoding PKCα was not necessarily excluded owing to the degeneracy of the SOX-binding site and recent findings about a SOX9-dependent transcription of S100A1 and S100B through a SOX9-binding site distinct from the consensus sequence (Saito et al., 2007). SOX9-dependent chromatin immunoprecipitation, however, was unable to co-precipitate the PKCα (–227 to +77) minimum promoter (data not shown), suggesting no physical interaction of SOX9 with the PKCα promoter. We therefore suspected an indirect SOX9-dependent transcriptional repression as suggested for CDX2, CEA and claudin 7 (Blache et al., 2004; Darido et al., 2008; Jay et al., 2005). This hypothesis did, however, still imply DNA-binding of the SOX9 HMG domain. As shown in Fig. 3A, the campomelic-dysplasia-associated SOX9 W143R mutant, although unable to bind DNA (Meyer et al., 1997), was located in the nucleus of HT29Cl16E cells and was still able to repress both the PKCα –1571 to +227 and –227 to +77 promoters in a luciferase assay, suggesting that direct DNA binding of the SOX9 HMG domain is not required for this repression (Fig. 3A, compare wt and W143R SOX9 constructs). Deleting the C-terminal transactivation domain (DC304-509W143R) did not have any significant effect on SOX9-mediated PKCα repression, whereas a SOX9 C-terminal deletion including the +208 to +303 (208-303) central region (DC208-509 W143R) completely abolished the repression of the PKCα –1571 to +227 and –227 to +77 promoters. Moreover, expression of a construct in which the 208-303 SOX9 region was fused to two SV40 T-antigen nuclear localization signals (NLSs) in tandem for appropriate targeting to the nucleus [SOX9(NLSx2)-208-303] upregulated PKCα –1571 to +227 and –227 to +77 promoter activity and exhibited a dominant-negative effect on SOX9-dependent PKCα repression. Fig. 3B further demonstrates that overexpression of SOX9(NLSx2)-208-303 correlates with increased endogenous PKCα expression. Together, these data indicate that the SOX9 (208-303) region is a crucial element in SOX9-dependent repression of transcription of the gene encoding PKCα. This region is highly conserved among the SOXE members (SOX8, SOX9 and SOX10). In SOX8, it exhibits a transactivation potential (Schepers et al., 2000) and, in SOX10, it is involved in the development of glial-cell lineages (Schreiner et al., 2007). Therefore, another biological function can be attributed to the SOX9 (208-303) region, i.e. its involvement in a new DNA-binding-independent transcriptional repression mechanism.

The DC304-509W143R SOX9 construct, which lacks the transactivation domain, is still able to repress PKCα transcription. Such a construct has previously been useful to show SOX9-dependent inhibition of RUNX2 transcriptional activity through the direct interaction of the SOX9 HMG domain and RUNX2 Runt domain (Zhou et al., 2006). We identified a putative binding site for the AML1 (also known as RUNX1) transcription factor in position –215 to –210 of the promoter of the gene encoding PKCα (Fig. 4A). However, neither an siRNA against AML1 (decreasing the level of AML1 mRNA by up to 60%) nor a mutation of the putative AML1-binding site had any effect on the level of mRNA encoding PKCα or on PKCα (–227 to +77)-dependent luciferase activity, indicating that SOX9-dependent PKCα repression is not mediated through a similar AML1-dependent mechanism (data not shown). The PKCα promoter, however, does exhibit a functional SP1-binding site (Clark et al., 2002) (Fig. 4A), and previous GST-pulldown data from Wissmuller et al. indicate that SP1 is able to interact with the SOXE members SOX8 and SOX10 (Wissmuller et al., 2006). In order to determine whether SOX9 might exert transcriptional repression through SP1, we first investigated the effect of SOX9 on an artificial promoter containing only SP1-binding sites (Sowa et al., 1997). As shown in Fig. 3A, overexpression of wild-type or mutated SOX9 influenced the activity of the pSP1 promoter in a similar manner as observed for the PKCα promoter in HT29Cl16E cells. Moreover, a mutation disrupting the SP1-binding site drastically decreased the activity of the minimum PKCα promoter and abolished its sensitivity to the wild-type and SOX9(NLSx2)208-303 constructs (Fig. 4B). Finally, Fig. 4C demonstrates that SOX9 co-precipitates with SP1 and is thus able to physically interact with SP1 in our cellular model. Wissmuller et al. reported that the SOX8 or SOX10 HMG domain alone was sufficient to co-precipitate SP1 in their GST-pulldown assays, and that the interaction involved a short conserved sequence located within the SOXE HMG-domain C-terminus (Wissmuller et al., 2006). This sequence apparently is able to bind a number of transcription factors and might be involved in the SOX9-SP1 interaction. Our data further indicate that, in HT29Cl16E cells, the SOX9 (208-303) central region is required for SOX9-mediated PKCα repression, possibly by increasing the affinity and/or stabilizing the interaction between SOX9 and SP1 within a transcriptional repressor complex.

In summary, the present study indicates that PKCα is a new SOX9 target gene in intestinal epithelium and reveals a new DNA-binding-independent mechanism for SOX9. Because SOX9 is upregulated by the Wnt-APC pathway in intestinal epithelial cells (Blache et al., 2004), our results provide an explanation for the correlation that is observed in both normal and malignant intestinal epithelium between activated Wnt-APC signaling, upregulation of SOX9 (Cardoso et al., 2006) and reduced expression of PKCα (Oster and Leitges, 2006). PKCα overexpression has been associated with cell-growth arrest and differentiation (Scaglione-Sewell et al., 1998), and PKCα knockout has been associated with tumor formation in the intestine epithelium (Oster and Leitges, 2006). However, in SOX9-deficient intestine overexpressing PKCα, neither reduced proliferation nor increased differentiation was observed; instead, hyperproliferation, no paneth cells and fewer goblet cells were found (Bastide et al., 2007). This result then raises the issue of the functional role of SOX9-dependent PKCα repression in colorectal cancer. Besides PKCα repression, SOX9 also exerts a negative-feedback loop on the Wnt-APC pathway (Bastide et al., 2007) and, therefore, it is possible that cell-growth inhibition due to PKCα overexpression is not sufficient to overcome the increase in proliferation caused by upregulation of the Wnt-APC pathway in intestine that is deficient for SOX9. Previous data from our laboratory indicate that SOX9-dependent inhibition of the Wnt-APC pathway involves the classic mechanism of SOX9 action, i.e. the binding of SOX9 to DNA (Bastide et al., 2007). This inhibition might result from an upregulation of groucho-related inhibitors of the β-catenin–Tcf complex (Bastide et al., 2007) and of CEACAM1, a direct SOX9 target (Jin et al., 2008; Leung et al., 2008; Zalzali et al., 2008). Indeed, the W143R SOX9 mutant, which is unable to bind DNA, is unable to inhibit the Wnt-APC pathway, although it is able to repress PKCα expression, as shown in the present study. Because the DNA-binding-dependent activity of endogenous SOX9 is weak in colon cancer cells (Darido et al., 2008), it may be postulated that the SOX9-dependent negative-feedback loop on the Wnt-APC pathway is at least partially lost in CRC cells. By contrast, PKCα repression, which does not require SOX9 DNA-binding, is maintained, thus favoring proliferation, inhibiting differentiation and potentially facilitating tumor progression, particularly in the context of a constitutively activated Wnt-APC pathway. Elucidating the causes of the decrease of SOX9 DNA-binding-dependent activity in CRC cells is now an attractive issue to address, and might provide a new approach to influence the balance between proliferation and differentiation of CRC cells.

HT29Cl16E, HCT116, SW480 and the SOX9-inducible HT29Cl16E cells were cultured and induced as previously described (Jay et al., 2005).

The reporter constructs containing the human pPKCα and the pSP1 promoters were previously described (Clark et al., 2002). The putative AML1-binding site in position –215 to –210 was disrupted using the oligonucleotide 5′-GTGTTCCCAGCATTGTCAGGCACTCGCTGCCTCCTCC-3′ and its complementary sequence, as was performed for the murine Ada gene (Schaubach et al., 2006). The FLAG-tagged wild-type and truncated versions of SOX9 exhibiting the W143R mutation are described elsewhere (Jay et al., 2005; Bastide et al., 2007). The SOX9(NLSx2)208-303 construct was obtained with a synthetic insert encoding the FLAG and two SV40 T-antigen NLSs (5′-ATGGACTACAAGGACGACGATGACAAGGGTACCGATCCAAAAAAGAAGAGAAAGGTAGATCCAAAAAAGAAGAGAAAGGTA-3′) subcloned in frame with the SOX9 (208-303) region within the pcDNA3 polylinker. The SOX9 antisense construct has been previously described (Blache et al., 2004).

Assays were performed in triplicate, at least three independent times as in Blache et al. (Blache et al., 2004). Transfection efficiencies were normalized by co-transfecting the phRG-TK standardization vector (Promega). Results are expressed as the ratio of luciferase activity relative to the control pGL3 and pcDNA3 empty plasmids.

Equal amounts of total-protein extracts from cultured cells (30 μg) or mice intestinal epithelium (20 μg) obtained by scraping the inner intestine layer were loaded on a 10% SDS-PAGE and transferred onto PVDF membranes. Primary antibodies were the mouse anti-PKCα (1:1000; Upstate Biotechnology), mouse anti-FLAG (1:500; Sigma), rabbit anti-SOX9 (1:1000) (Bastide et al., 2007), mouse anti-actin (1:5000; Sigma), rabbit anti-SP1 (1:200; Santa Cruz Biotechnology). Secondary antibodies were the IRdyeTM 800- and IRdyeTM 700DX-conjugated anti-mouse and rabbit (1:10,000; Tebu-Bio). Results were analyzed using the OdysseyR infrared imaging system (LI-COR Biosciences).

Immunoprecipitations were performed with magnetic beads (Ademtech), 1 mg SOX9-induced HT29Cl16E whole-cell extracts and 2 μg of either the anti-FLAG or the anti-SP1 antibodies, according to Ademtech recommendations with a mouse anti-IgG (Santa Cruz Biotechnology) or a rabbit serum as controls.

Experiments were performed with the following primer pairs: GAPDH, 5′-GACCACAGTCCATGCCATCACT-3′ and 5′-TCCACCACCCTGTTGCTGTAG-3′; SOX9, 5′-GCCAGGTGCTCAAAGGCTA-3′ and 5′-TCTCGTTCAGAAGTCTCCAGAG-3′; PKCα, 5′-GCTTCCAGTGCCAAGTTTGC-3′ and 5′-GCACCCGGACAAGAAAAAGTAA-3′(Bastide et al., 2007). Results are expressed as the variation of the studied transcript standardized to GAPDH and relative to the control conditions.

Paraffin-embedded intestine sections were obtained as previously described (Bastide et al., 2007). HT29Cl16E cells were plated in 24-well plates (50,000/well) on coverslips. Antibodies were the rabbit anti-SOX9 (1:200), rabbit anti-PKCα (1:400; Sigma) and the Cy3-conjugated secondary anti-rabbit antibody (1:1000; Santa Cruz Biotechnology). Preparations were mounted in Mowiol and observed using an epifluorescent microscope (Zeiss).

HT29Cl16E cells (350,000) were seeded in 60-mm dishes and transfected with 100 nM of an siRNA targeting the AML1 messenger sequence 5′-CCGCCGCUUCACGCCGCCUUC-3′ (Wang et al., 2005). After 48 hours, cells were harvested and mRNA content was analyzed by real-time RT-PCR.

Chromatin immunoprecipitation (ChIP) assays were performed using the Magna ChIP G Kit and a ChIPable HT29 chromatin (Upstate) with either the rabbit anti-SOX9 or rabbit serum as a control. Forward 5′-GTGTTCCCAGCACCGCAAGG-3′ and reverse 5′-GAGAGTCGGGCTGGTGCTG-3′ primers were used to amplify the PKCα –227 to +77 minimum promoter by PCR.