Cdx1 Inhibits the Proliferation of Human Colon Cancer Cells by Reducing Cyclin D1 Gene Expression

Cdx1 inhibits the proliferation of human colon cancer cells by reducing cyclin D1 gene expression

John Lynch*,1, Matthew Keller1, Rong-Jun Guo1, Donald Yang1 and Peter Traber1,2

1Division of Gastroenterology, Department of Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA

The transcription factor Cdx1 regulates intestine-specific renewal that occurs continuously throughout the small gene expression and enterocyte differentiation. It has been intestine and colon (Chung, 2000; Lynch and Hoops, hypothesized to play a role in regulating intestinal cell 2002). proliferation; however, the mechanism for this effect Normally, surface cells are lost to the intestinal lumen remains elusive. In a prior study, we demonstrated that due to programmed cell death (apoptosis) or mechanical Cdx1 expression reduced the proliferation of a nontrans- exfoliation and must be replaced. Renewal depends on a formed intestinal cell line. This study tests the hypothesis regulated process of proliferation of stem cells and that Cdx1 expression inhibits colon cancer cell prolifera- differentiation of daughter cells. Proliferation occurs tion by reducing cyclin D1 gene expression. Cdx1 within a zone or compartment of the crypts of the small expression markedly reduced cancer cell proliferation intestine and colon. Daughter cells continue proliferat- and DNA synthesis and induced an accumulation of cells ing as they migrate up along the crypt–villus axis. At the in G0/G1. A transcriptionally inactive Cdx1 mutant could upper third of the crypt, proliferation ceases and the not elicit this effect, suggesting that it required Cdx1 cells undergo a rapid maturation. The transition from transcriptional activity. Cdx1 expression increased the rapidly proliferating cells to nondividing, mature hypophosphorylation of the retinoblastoma (pRb) and enterocytes is swift. Some of the mechanisms regulating p130 proteins. Reductions in G1 cyclin-dependant kinase these events have been identified, including the Wnt and (cdk) activity accompanied this effect. Cyclin D1 mRNA Notch signaling pathways (Korinek et al., 1998; Jensen and protein levels were diminished by Cdx1 expression. et al., 2000) and certain transcription factors such as the Restoration of cyclin D1 expression reversed the G0/G1 Drosophila Caudal-related factors Cdx1 and Cdx2. block and induced pRb hyperphosphorylation. Lastly, The homeodomain proteins Cdx1 and Cdx2 are well- Cdx1 expression did not alter cyclin D1 mRNA stability recognized regulators of intestine-specific gene expres- but did reduce cyclin D1 promoter activity, suggesting sion and enterocyte differentiation (Suh et al., 1994; Suh that Cdx1 acts to diminish cyclin D1 gene transcription. and Traber, 1996; Taylor et al., 1997; Freund et al., We conclude that Cdx1 reduces the proliferation of 1998; Soubeyran et al., 1999). Cdx1 and Cdx2 are human colon cancer cells by reducing cyclin D1 gene important transcriptional activators, promoting the transcription. intestine-specific expression of a number of genes Oncogene (2003) 22, 6395–6407. doi:10.1038/sj.onc.1206770 including sucrase isomaltase, colon carbonic anhydrase 1, and intestinal phospholipase A/lyso-phospholipase Keywords: homeobox; cyclin D1; antiproliferative; (IPAL) (Suh et al., 1994; Taylor et al., 1997; Drummond colon cancer et al., 1998). Cdx1 and Cdx2 also promote differentiation of intestinal cells (Suh and Traber, 1996; Soubeyran et al., 1999). Additionally, observations utilizing a variety of experimental methods suggest that these factors can regulate intestinal cell proliferation. There Introduction is consensus among investigators that Cdx2 expression reduces intestinal epithelial proliferation (Suh and Colon cancer is the second leading cause of cancer death Traber, 1996; Freund et al., 1998). However, the in the US, responsible for nearly 60,000 deaths each findings with respect to Cdx1’s role in control of year. A large body of evidence supports the hypothesis proliferation have been conflicted. that most human colon cancers arise from adenomas. Cdx1 is expressed in the intestinal crypts and Wnt/b- Adenomas arise when somatic mutations are acquired, catenin signaling is required for Cdx1 expression which disrupt the highly ordered process of epithelial (Subramanian et al., 1998; Lickert et al., 2000). These observations seem to support the premise that Cdx1 promotes proliferation. However, Cdx2, which reduces *Correspondence: JP Lynch, Division of Gastroenterology /660CRB, proliferation, is also expressed in intestinal crypts (Rings 415 Curie Blvd., Philadelphia, PA 19104, USA; E-mail: lynchj@mail. et al., 2001). Additionally, dysregulation of the Wnt/b- med.upenn.edu 2Current address: Baylor college of Medicine, One Baylor Plaza, catenin signaling pathway is nearly universally present Houston, TX 77030, USA in human colon cancers (Chung, 2000; Lynch and Received 3 December 2002; revised 30 April 2003; accepted 7 May 2003 Hoops, 2002); yet Cdx1 expression is preferentially Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6396 silenced with the progression to colon cancer in humans expression vector Adeno-CDX1 as described previously (Silberg et al., 1997; Suh et al., 2002). Together, these (Lynch et al., 2000). pAdCMVlink without a cDNA insert findings cast doubt on the premise that Cdx1 promotes was used to generate the control Adeno-NULL vector. proliferation. Infectious titers were determined by an adenoviral plaque In vitro studies using cell culture models have yielded assay. The human cyclin D1 adenovirus (Ad-cyclin D1; Albrecht and Hansen, 1999) was provided by J Albrecht conflicting observations as well. Cdx1 expression in 3T3 (University of Minnesota). It was expanded and titered in the cells results in soft-agar growth and tumor formation in same manner as the pAd-Cdx1 and pAd-Null viruses. one study (Maulbecker and Gruss, 1993) but not in Adenovirus infections were performed as described (Lynch another study (Lynch et al., 2000). Stable constitutive et al., 2000). DLD and HCT116 cells were initially plated Cdx1 expression in human colon cancer cell lines has sparsely, 14.3 Â 104 per 100 mm plate or 5 Â 103/well in a 24- been reported to have no effect on proliferation (Mallo well culture dish. The adenovirus concentration yielding et al., 1998), or reduce tumor growth synergistically with 490% of cells expressing Cdx1 was determined experimentally Cdx2 in vivo (Mallo et al., 1998). Antisense Cdx1 for each cell line using the antibody CPSP (1 : 200) (Silberg et al., 1997), a rabbit anti-Cdx1 antibody. An adenoviral expression reduced proliferation of human colon cancer 6 CaCo2 cells (Lorentz et al., 1997). Lastly, Cdx1 concentration of 1.68 Â 10 infectious particles/ml was determined to be optimal for DLD1 cells, and 3.75 Â 106 infectious expression in rat IEC6 cells increased cell proliferation particles/ml for HCT116 cells, when cells were initially plated in one study (Soubeyran et al., 1999) but not in another at sparse concentrations. These concentrations were used in study (Lynch et al., 2000). One factor complicating the the subsequent experiments. Based on these values, a calcula- interpretation of these studies is the use of pooled cells tion of the multiplicity of infection (MOI) would yield 67 for constitutively expressing CDX1, which were selected for DLD1 cells, and 150for HCT116 cells. These MOI’s are growth in spite of CDX1 gene expression. Cells with artificially inflated by the relatively low cell density we used in reduced proliferation would therefore be overgrown and order to avoid confounding contributions from contact not detected. It is thus not a sensitive method for inhibition. detecting antiproliferative effects. At 48 h postadenoviral infection, the cells were processed for In our present study with human colon cancer cells, the various end points. For cell counts, the cells were trypsinized at several points postinfection, pelleted by cen- we analyse in detail the molecular basis of Cdx1- trifugation, resuspended in PBS, and counted on a hemecyt- mediated inhibition of proliferation. Transient expres- ometer. Plates of DLD1 cells (100 mm) were infected and sion of Cdx1 in human colon cancer DLD1 and whole-cell protein, nuclear proteins, or total RNA were HCT116 cells resulted in a marked reduction of isolated at 48 h. EMSA for Cdx DNA-binding activity and proliferation, with an accumulation of cells in the G0/ immunoblots for Cdx1 and Cdx2 protein were performed G1 phase of the cell cycle. Importantly, this effect was exactly as described (Lynch et al., 2000). specific, as an N-terminally truncated Cdx1 protein was unable to elicit this effect. Similar to our findings in Cell culture and transfections IEC6 cells, the retinoblastoma protein (pRb) and the related p130protein were predominantly hypopho- DLD1 and HCT116 cells were obtained from the ATCC sphorylated. Unlike our prior studies, this was asso- maintained under standard conditions. Cdx1 truncation constructs were synthesized by PCR using ciated primarily with reductions in G1 cdk6 and cdk2 oligos to amplify Cdx1 truncations from a murine cDNA kinase activities. Cyclin D1 mRNA and protein levels template. These amplified products were sequenced to confirm were diminished as before, but we now confirm that this fidelity of the amplification and subcloned into the mammalian decrease is critical for the Cdx1 antiproliferative effect, expression vector pRC/CMV (Promega). N-terminus deletion as restoration of cyclin D1 activity reversed the Cdx1- primers were: 50-TTGGATCCACCATGAAGACCCGAAC- mediated G0/G1 arrest. Additionally, we demonstrate CAAGG-30 and 50-TTTGATCACTAGGGTAGAAACTCC- that Cdx1 expression does not alter cyclin D1 mRNA TCCTT-30. The C-terminus deletion primers were: 50-TTGG- stability, but does reduce transcriptional activity from a ATCCACCATGTACGTGGGCTATGTGC-30 and 50-TTT- 0 human cyclin D1 promoter. Taken together, we GATCACTACTTCTTGTTTACTTTGCGC-3 . pRC/CMV- conclude that Cdx1 expression in human colon cancer Cdx1 was constructed as described (Taylor et al., 1997). HCT116 cells were transfected using FuGENE6 (Roche). cells reduces proliferation by decreasing cyclin D1 gene Luciferase assays were carried out using the Dual Luciferase transcription. These findings suggest that Cdx1 may Assay Kit and the pRL-CMV vector (both by Promega) as play a role in coordinating cell-cycle exit with the transfection control. Sparsely plated HCT116 cells were induction of differentiation that occurs as cells ascend transfected with pMACSKk.II (Miltenyi) and the appropriate from the intestinal and colonic crypts, and loss of CDX1 pRC/CMV expression vectors in a ratio of 1 : 3. At 48 h, cells expression in human colon cancers may contribute to were washed, trypsinized, and dead cells removed using the their increased proliferation. Dead Cell Removal Kit (Miltenyi), followed by a single enrichment for H-2Kk-expressing cells using the MACSelect Kk.II kit (Miltenyi). Isolated cells were fixed in PBS/70% EtOH at À201C overnight. On the following day, cells were Materials and methods washed in PBS, stained with propidium iodide (50 mg/ml) and RNAse A (360 mg/ml) for 30min at 37 1C, then with just Production and expression of adenovirus vectors and adenoviral propidium iodide (50 mg/ml) in PBS for at least 1 h at 41C. infections DNA content was quantified using a FACSCAN (Becton A mouse Cdx1 cDNA was subcloned into the basic adenoviral Dickson). DLD1 cells were transfected by electroporation transfer vector (pAdCMVlink) and used to produce the exactly as described (Krasinski et al., 2001) using a Pulse

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6397 Controller II (Bio-Rad), and 6 mg of pRC expression vectors, actin:A-4700 (Sigma). Positive controls for p16, p21, and p27 1 mg of GFP-F (Clontech), and 28 mg of Bluescript II SK þ were included to confirm antibody activity and specificity. (Stratagene) as carrier. Media were changed at 24 h, and at After washing the membrane, the blot was incubated with the 48 h the cells were harvested and fixed as above in PBS/70% secondary antibody (antiRabbit/HRP (Amersham), anti- EtOH at À201C. The following day, the cells were washed and mouse/HRP or antigoat/HRP (Santa Cruz)); the blot was stained as above with propidium iodide. DNA content was developed with the ECL-Plus Western blotting kit (Amer- quantified by flow cytometry, with gating on GFP þ DLD1 sham). cells. A human cyclin D1promoter-luciferase reporter construct D1pro-1749 (PvuII) was provided by Dr Anil Rustgi Cyclin-dependent kinase activity analysis (University of Pennsylvania). Cdk4 and cdk6 kinase activity assays were carried out as BrdU and flow cytometry assays described (Phelps and Xiong, 1997; Lynch et al., 2000). Several 100 mm plates of DLD1 cells were infected and cultured and BrdU incorporation was measured as before (Lynch et al., an immunoprecipitation (IP) protein extract prepared at 48 h. 0 2000) using the 5-bromo-2 -deoxyuridine Labeling and Detec- Cleared lysates were then subject to IP with 1 mg of either anti- tion Kit I (Boehringer Mannheim). DLD1 cells were infected cdk4 (sc-601, Santa Cruz) or anti-cdk6 (sc-177, Santa Cruz), or and cultured as described, with BrdU labeling being performed 1 mg of rabbit preimmune IgG (sc-2026, Santa-Cruz), and then over 1 h. The percentages of BrdU þ cells were transformed kinase activity was determined as described (Lynch et al., using the arcsine transformation prior to statistical analysis 2000). with ANOVA and Tukey tests. Some treated cells at 48 h were Cdk2-associated kinase activity was measured in a similar stained for Cdx1 (using polyclonal CPSP) along with BrdU, fashion (Sheaff, 1997; Hu et al., 2001). The same IP extracts and then counterstained by DAPI. The cells were examined by used in the cdk4/6 assays were incubated with either 1 mg of fluorescent microscopy and images obtained using a Photo- anti-cdk2 (sc-163, Santa Cruz) or rabbit preimmune IgG, metrix Coolsnap-CF Black and White CCD camera (Roeper followed by incubation with prewashed Protein A beads and Scientific) mounted on a Nikon E600 fluorescent microscope. then several washes with IP and K2 kinase buffers (K2 kinase DNA content was measured using a rapid propidium iodide- buffer: 25 mm HEPES (pH 7.5), 15 mm MgCl2, 0.5 mm DTT, staining method as described (Lynch et al., 2000). DNA 1mm NaF, 0.1 mm Na3VO4,100mm PMSF, 2 mg/ml leupeptin, content quantitation was carried out using FACScan (Becton 2 mg/ml of aprotinin). To initiate the Cdk2 reaction, 10 mlof Dickson). kinase buffer with 100 mm cold ATP, 5 mCi g-32P-ATP, and 0.5 mg of histone H1 (Sigma H-4524) was then added to the RNAse protection and RNA stability determinations pelleted beads, and the reaction incubated at 301C for 30min Total RNA was isolated using RNAeasy (QIAGEN) from with frequent agitation. The reaction was halted and products were analysed by SDS–PAGE. several 100 mm plates of infected DLD1 cells. For the cyclin D1 mRNA stability studies, 10 mg/ml Actinomycin D (Sigma) was added to the cultures at 48 h postinfection and cultures continued for 4–24 h. Riboprobes for the human cyclin genes Results were synthesized using the Human Cell Cycle Regulator Template Set (hCYC-1, PharMingen) and T7 RNA polymerase, as described by the manufacturer. The probe set was Inhibition of proliferation of human colon cancer cell lines hybridized to 5 mg of total RNA and then subject to RNAse T1 by Cdx1 expression digestion using the buffers and conditions recommended by m CDX1 expression is diminished or lost in the progres- the RPA II kit (Ambion). The products were analysed on a 7 sion from normal colonocyte to colon adenocarcinoma. urea/ 5% acrylamide sequencing gel, dried down and exposed to film and quantified by exposure to a phosphoimager screen To determine if this loss contributed to the increased (STORM 840, Molecular Dynamics). Total RNA (10 or 20 mg) proliferation characteristic of human colon cancer cells, was analysed by Northern blotting for c-Myc, cyclin D1, and we tested two human colon cancer cell lines for growth 36B4 mRNA expression as described (Lynch et al., 2000). effects with the induction of Cdx1 expression: DLD1 Probes for human c-Myc cDNA (kindly provided by Gary and HCT 116. Neither of these cell lines expresses Wu, University of Pennsylvania, USA), cyclin D1 (Hiroshi significant CDX1 protein and both express CDX2 but at Nakagawa, University of Pennsylvania, USA), and ribosomal different levels (data not shown). We utilized a Cdx1- phosphoprotein 36B4 (Francois Boudreau, University of adenoviral expression vector (Adeno-Cdx1) and con- 32 Pennsylvania, USA), were random-primed P-labeled and firmed the production of authentic Cdx1 protein in our used as a probe (Ambion). cells. An adenovirus with an empty expression cassette (Adeno-Null) was used as a control. Extracts from Immunoblot analysis Adeno-Cdx1, Adeno-Null, and control uninfected cells Several 100 mm plates of DLD1 cells were infected and were compared by Western blot and EMSA (data not cultured. At 48 h, a whole-cell extract was prepared as shown), and clearly demonstrated induction of a DNA- described (Lynch et al., 2000) and the products analysed by binding Cdx1 protein by the Adeno-Cdx1 vector. SDS–PAGE electrophoresis and electroblotted. The following We determined initially an optimal adenoviral in- antibodies were used: pRb:G3-245 (Pharminigen); cyclin fectious particle concentration for each cell type such D1:06-137 (Upstate Biotechnology), SC-246 (Santa Cruz), X and DCS-6 (BD Pharmingen); p130:SC-317; p107:SC-250; that 90% of cells in culture are infected and cyclin A:SC-751; cyclin B1:SC-245; cyclin D2:SC-452; cyclin expressing Cdx1 as determined by immunofluorescence D3:SC-182; cyclin E:SC-481; cdk2:SC-163; cdk4:SC-260; (data not shown). In order to avoid confounding cdk6:SC-177; p15:SC-613; p16:SC-1661; p18:SC-865; p19:SC- contributions from contact inhibition on our measures 1063; p21:SC-397; p27:SC-1641; p57:SC-1040; (Santa Cruz); of proliferation, we carried out adenoviral infections at

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6398 very low cell densities. Actual adenoviral concentrations Cdx1 virus (Figure 1b, c). Moreover, the nuclei of Cdx1- ranged from 1.68 Â 106 (DLD1) to 3.75 Â 106 (HCT116) expressing cells were not different from Adeno-Null- infectious adenoviral particles per ml of cell culture treated and control DLD1 cells. There was no evidence media. These levels were not cytotoxic. Using this for increased apoptotic cell death with these treatments optimal concentration, we tested for effects on cell (Figure 1b). Thus, Cdx1 expression inhibited prolifera- accumulation after infection with either the Adeno- tion by reducing DNA synthesis in human DLD1 cells Cdx1 vector or our control adenovirus without the and not by increasing cell death. Cdx1-cDNA insert (Adeno-Null). Both cell lines re- sponded to Cdx1 expression with a reduction in cell Cdx1 expression in DLD1 cells induces a G0/G1 cell-cycle accumulation (Figure 1a). The effect became more block apparent 3–4 days after infection with the Adeno- Cdx1 vector. We found that DLD1 cells had the most Inhibition of proliferation and DNA synthesis is significant reduction in accumulation with Adeno-Cdx1 typically associated with a cell-cycle arrest. To investi- and were not significantly effected by the control gate this possibility, cells were infected as before, and Adeno-Null vector. For this reason, we focused the DNA content analysed by propidium iodide staining studies that followed on Cdx1 expression in DLD1 cells. and flow cytometry. We did not observe subdiploid The rate of DNA synthesis was reduced by Cdx1 peaks of DNA content in the cytometry profiles, further expression. BrdU incorporation into DNA was de- demonstrating that apoptosis was not occurring (data creased twofold by 48 h after infection with the Adeno- not shown). Adeno-Cdx1 virus infection induced a

a 30 120

DLD1100 HCT116

20 80

10 40 Cell number x 10E4 Cell number x 10E4 20

0 0 0 1 2 3 4 0 1 2 3 4 5 Days post infection Days post infection

Figure 1 Reduction of cell growth by expression of Cdx1 in human colon cancer HCT116 and DLD1 cells. (a) DLD1 and HCT116 cells cultured and infected with Adeno-Null or Adeno-Cdx1. Cell counts were performed for each treatment on days noted after infection. One of three experiments with similar results is presented for each cell type. n ¼ 3, each day. Symbols: open circle ¼ control uninfected; closed diamond ¼ Adeno-Null infected; open square ¼ Adeno-Cdx1 infected. (b) DLD1 cells were infected and cultured as before, but the rate of BrdU incorporation was determined. Treated cells on day 2 were evaluated with a polyclonal antibody for Cdx1 (CPSP) expression in addition to a BrdU monoclonal antibody. Nuclei were counterstained with DAPI and the cells examined by ﬂuorescent microscopy. (c) On days 1, 2, 3, and 4 after infection, the percentage of cells incorporating BrdU was determined. n ¼ 5 for each condition, each day. *indicates signiﬁcant difference from control sham-infected or Adeno-Null-infected cells, Po0.01. Representative data from one of three experiments. Black bars ¼ control uninfected; grey bars ¼ Adeno-Null infected; white bars ¼ Adeno-Cdx1 infected

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6399 substantial accumulation of cells in the G0/G1 phase of Cdx1 expression affects G1/S transition regulatory the cell cycle, when compared with controls (Table 1). proteins Adeno-cdx1 treatment led to a 1.6-fold increase in the The retinoblastoma family of phosphoproteins (pRb, number of cells in the G0/G1 phase of the cell cycle. We p107, and p130) are key regulators of the G1/S found that Cdx1 expression reduced the S phase transition in eukaryotic cells. We next studied the population by twofold, consistent with our results from expression of these proteins in DLD1 cells to better the BrdU analysis. By contrast, Adeno-Null control understand the Cdx1-mediated G0/G1 block. Hypopho- treatment induced a modest reproducible reduction of sphorylated forms of these proteins are associated with cells in G0/G1. Thus, Cdx1 expression in DLD1 cells an inhibition of transcriptional events required for the inhibits proliferation by blocking cell-cycle progression in G0/G1. G1/S transition (Sherr, 2000). We consistently noted that faster migrating, hypophosphorylated pRb and p130forms were induced by Cdx1 expression over many Table 1 Cell-cycle distribution for DLD1 cells expressing Cdx1 experiments (Figure 2a). Total pRb and p130protein G0/G1 S G2/M levels were not quantiﬁed. p107 was not detected (data Control 4274.2% 3972.7% 2073.1% not shown). Adeno-Null 3473.9% 4672.7% 2074.0% The pRb phosphorylation state exquisitely regulated Adeno-Cdx1 6576.6% 2072.8% 1679.4% during G1 by the D and E cyclins, cyclin-dependent kinases (cdks), and cdk inhibitors (CDKI). We next Data from 14 separate ﬂow cytometry experiments with DLD1 cells examined these regulators for Cdx1-induced changes. were compiled (n ¼ 30). Assignment to G0/G1, S, or G2/M was made Except for p57, expression levels of Cip/Kip (p21, p27) and percentages calculated using the ModFit LT program (Verity Software House). The cell-cycle distributions for each treatment were or Ink4 (p15, p16, p18, p19) cdk inhibitors were below averaged, and the mean and standard deviation for each is reported the level of detection by Western blot (data not shown).

Figure 2 Changes in G1/S regulatory mRNA and protein levels in DLD1 cells expressing Cdx1. DLD1 cells were cultured and infected as before, and at 48 h whole-cell extracts were prepared. (a) Immunoblot for pRb and p130proteins in whole-cell extracts from DLD1 cells infected with Adeno-Cdx1 (Cdx1), Adeno-Null (Null), or uninfected control (Con). Arrows indicated positions of differentially phosphorylated species. Loading controls are presented in (b). (b) Western immunoblotting of whole-cell extracts was performed for cyclins using the speciﬁc antibodies as noted. a-actin served as a loading control. (c) Northern blot analysis for cyclin D1 mRNA levels. Total RNA was isolated at 48 h from DLD1 cells expressing Cdx1 and control cells. RNA (20 mg) was loaded in each lane. Blot was sequentially probed for cyclin D1 and the loading control ribosomal phosphoprotein 36B4. (d) Total RNA was isolated as for the Northern blot analysis. Levels of cyclins A, B, C, D1, D2, and D3 mRNA, as well as control L32 and GAPDH, were analysed in duplicate by RNAse protection using the RiboQuant multiprobe template. Protected probe bands were quantiﬁed by exposure to a phosphoimager screen (STORM 840, Molecular Dynamics). Results from two experiments were normalized to controls L32 and GAPDH, averaged, and expressed as percent of control levels graphically. Con: control uninfected cells; Null: Adeno-null- infected cells; Cdx1: Adeno-Cdx1-treated cells

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6400 Although p57 was detected, its levels were low and unaltered by Cdx1 expression (data not shown). Of the G1 cyclins, only cyclin D1 levels were markedly reduced by Cdx1 expression (Figure 2b). Cyclin D3 levels were slightly decreased, cyclin D2 was not detected, and cyclin E was unaffected by Cdx1 (Figure 2b and data not shown). Cyclins A and B1 levels were both reduced, as would be expected for cells accumulating in G0/G1. Taken together, the Western blot data suggest that Cdx1 induces a G0/G1 accumulation by reducing pRb phosphorylation through a speciﬁc reduction in cyclin D1 protein levels rather than by induction of a cdk inhibitor. We next performed Northern blot and RNAse protection analysis on total RNA isolated from sham- infected, Adeno-null-infected, or Adeno-Cdx1-treated DLD1 cells for 48 h. By these techniques, Cdx1 expression reduced cyclin D1 mRNA levels to 33% of control levels (range 25–45%) (Figures 2c, d). Cyclins A and B mRNA levels were reduced by about 50% from control levels (Figure 2c, d). After normalization to controls L32 and GAPDH and averaging the results from several experiments, Cyclin D3 mRNA levels were only modestly diminished, decreased by 30%. In summary, Cdx1 expression reduced mRNA levels of cyclins A,B, D1, and D3, but the most profound effect was observed with cyclin D1 mRNA levels.

Time course analysis of cell-cycle arrest and reductions in cyclin D1 mRNA and protein levels To better understand the relation between the Cdx1- induced G0/1 accumulation and falling cyclin D1 mRNA and protein levels, we performed a time course analysis. DLD1 cells were infected as before, and analysed for DNA content by flow cytometry at 12, 24, and 48 h post-adenoviral infection. There was no Figure 3 Time course analysis of cell-cycle arrest and reductions in cyclin D1 mRNA and protein levels. (a) DLD1 cells were significant difference in the percentage of cells in G0/G1 cultured and treated as before, except that they were fixed and in the two treatment groups during the first 24 h. stained with propidium iodide at the indicated time points. DNA However, by 48 h, we do see a large increase in the content was determined by flow cytometry. Data from three percentage of Cdx1-expressing DLD1 cells in G0/G1 separate experiments were compiled (n ¼ 6). Assignment to G0/G1, S, or G2/M was made as before. The %G0/G1 for each treatment (Figure 3a). Additionally, while there was no difference was averaged, and the mean and standard deviation for each is in the %G0/G1 at 24 h, we did observe a small but reported graphically. Black bar: control uninfected; grey bar: reproducible reduction in the %S phase, with a recipro- Adeno-Null treated; white bar: Adeno-Cdx1 treated. (b) Whole-cell cal increase in the %G2/M phase of Cdx1-expressing extracts and total RNA were isolated from Adeno-Cdx1-treated cells at 24 h (data not shown). This agrees with the small cells at 12, 24, and 48 h postinfection, as well as control uninfected (at 48 h) and Adeno-Null-infected cells (48 h). A Western blot reduction in BrdU incorporation noted at 24 h in the analysis for cyclin D1 was performed on 50 mg of extract. The blot Adeno-Cdx1-treated cells (Figure 1). was stripped and reprobed for loading control a-actin, as well as We then isolated both protein and RNA from Adeno- Cdx1. A Northern blot analysis for cyclin D1 mRNA levels was Cdx1-treated cells at the same time points and also carried out on 10 mg of total RNA. The blot was sequentially probed for cyclin D1 and the loading control ribosomal performed immunoblot and Northern blot analysis for phosphoprotein 36B4. The figure is a composite made from a Cdx1 and cyclin D1. Cdx1 protein was weakly detected single experiment. Control: control uninfected cells; Adeno-Null: at 12 h, but significantly elevated by 24 h, and modestly Adeno-Null infected cells; Adeno-Cdx1: Adeno-Cdx1-treated cells increased further by 48 h (Figure 3b). Cyclin D1 protein levels fell dramatically between 24 and 48 h in the Adeno-Cdx1-treated cells. At 12 and 24 h, cyclin D1 reduction occurring between 24 and 48 h (Figure 3b). A protein levels were only minimally diminished and small reduction of 10and 20%in cyclin D1 mRNA comparable to those of Adeno-Null-treated cells at levels was noted at 12 and 24 h, respectively (Figure 3b). 48 h. A similar pattern was noted in cyclin D1 mRNA This compares with the 20% reduction observed in the levels. Normalized cyclin D1 mRNA levels were reduced Adeno-Null-treated cells at 48 h (Figures 2d and 3b). to 40% of control levels by 48 h, with the greatest Taken together, these observations suggest that the

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6401 G0/G1 arrest and reductions in cyclin D1 mRNA and protein occur concurrently and are preceeded by the induction of Cdx1 expression.

Cdx1 expression reduces G1/S cyclin-dependent kinase activity The G1 cdk kinase activity appears diminished based on the pRb western. This could be due to reduced cdk protein levels or inhibited kinase activity. Western immunoblotting and in vitro kinase assays were performed to differentiate between these possibilities. Cdk4, cdk6, and cdk2 levels and activity were measured on the same extracts prepared from sham-infected, Adeno-null-infected, or Adeno-Cdx1-treated DLD1 cells. Cdk4 protein levels were modestly reduced by Cdx1 (Figure 4a). Additionally, Cdk4-associated activity was weak; however, we did observe a 40–50% reduction in cdk4 kinase activity with Cdx1 expression (Figure 4a, c). This reduced activity may derive from both reduced cdk4 protein and reduced cyclin D1 levels. Cdk6 kinase activity was considerably greater than that associated with cdk4. We observed a more substantial 60–75% reduction in cdk6 kinase activity with Cdx1 expression Moreover, there was no change noted in cdk6 protein levels (Figure 4b, c). This suggests

Figure 5 Cdx1 expression reduces cdk2-associated pRb kinase activity in DLD1 cells. (a) The DLD1 IP-extracts were examined for cdk2-associated kinase activity using purified histone H1 as substrate. An immunoblot for cdk2 and a-actin was performed to confirm equal levels. A representative of three experiments is presented. (b) The 32P-labeled products were quantified and normalized as before, and the average and a standard deviation calculated (n ¼ 6)

that the reduction was likely due to diminished cyclin D1 expression alone. By contrast, Adeno-Null treatment only slightly reduced cdk6 kinase activity by an average of 20%. Cdx1 expression reduced cdk 2 kinase activity by 70% (Figure 5). This reduction was not due to loss of cdk2 protein (Figure 5), but more likely due to diminished cyclin A and cyclin B levels as well as some redistribution of low levels of cdk inhibitors (p21, p27, or p57) from cyclin D1-cdk4/6 complexes to cdk2 (Sherr, 2000). Figure 4 Cdx1 expression reduces cdk4/cdk6-associated pRb kinase activity in DLD1 cells. (a) DLD1 cells were cultured and treated as before, and IP-extracts isolated at 48 h. Duplicate 100 mg Cyclin D1 expression ‘rescues’ Cdx1 G0/G1 arrest samples of each extract were subject to IP with an antibody for cdk4 or with a control rabbit IgG. The cdk4-associated kinase The data indicate that diminished cyclin D1 levels may activity present in the IP products was assayed. An immunoblot for be the primary mechanism by which Cdx1 inhibits the cdk4 and a-actin was performed to confirm cdk4 protein levels. Con: control uninfected cells; Null: Adeno-null-infected cells; proliferation of DLD1 cells. If this is the case, Cdx1: Adeno-Cdx1-treated cells. (b) Similar study as in (a) except restoration of cyclin D1 expression should reverse duplicate samples of each extract were subject to IP with an Cdx1-mediated inhibition of proliferation. We tested antibody specific for cdk6 or with a control rabbit IgG. (c) The 32P- this hypothesis by coinfecting cells with Adeno-Cdx1 labeled products were quantified by a phosphoimager. The 32P incorporation values were normalized across three experiments and and an adenovirus containing a human cyclin D1 cDNA then averaged and a standard deviation calculated (n ¼ 6). Black (Ad-cyclin D1). A small amount of the Ad-cyclin D1 bars: Cdk 4 kinase activity. White bars: Cdk6 kinase activity virus was sufficient to reverse the G0/G1 accumulation

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6402 2.0 vectors. The Cdx1-CD protein was not detected because the epitope recognized by the polyclonal antibody was in the deleted C-terminus. However, the Cdx1-CD expression vector was nearly as strong as the Cdx1-Wt 1.5 protein in transactivation of the Cdx-responsive sucrase- isomaltase promoter (Figure 7c). The Cdx1-ND vector was unable to activate this reporter, as was expected given that the N-terminus contains a transcription-

Fold G0/G1 Change activating domain (Taylor et al., 1997). Alternatively, 1.0 the Cdx1-ND vector was able to act as a dominant- negative inhibitor of endogenous CDX2 transcriptional activity in DLD1 cells (data not shown). MOI-Cdx1 ––75 60 60 These constructs were then tested for their ability to MOI-Null 75 60 – 15 – induce antiproliferative effects in colon cancer cells. As MOI-D1 – 15 ––15 transient transfection methods will be significantly less efficient at expressing Cdx1 than the adenoviral method, we needed to identify and enrich for transfected cells in pRb * our study population. We first utilized the MACSelect < Transfected Cell Selection Kit, however, only the HCT116 cells would express the truncated MHC H- k Figure 6 Cyclin D1 expression reverses the Cdx1 growth 2k marker used in the selection. HCT116 cells were inhibition. An adenovirus containing a human cyclin D1 cDNA transiently transfected with each of the constructs, along (Ad-D1) was used to express cyclin D1 in DLD1 cells. The amount with pMACS kk.II (Miltenyi), the marker for transfec- of each virus added to the mix is listed in the table beneath the tion. After 48 h, an enriched population of transfected graph and above the western; the final adenovirus MOI was 75 for each treatment group. At 48 h postinfection, cells were prepared for cells was isolated using the MACSelect system. These DNA content analysis and flow cytometry or whole-cell extracts cells were fixed and stained for propidium iodide, and prepared for Western blots. Results of DNA content analysis are flow cytometry was utilized to quantify the DNA presented graphically as the fold change in G0/G1 compared with content in the enriched pool. Cdx1-Wt expression led Adeno-Null-treated cells (MOI ¼ 75). Four separate experiments were performed (n ¼ 12). The data were transformed using an to a small but statistically significant increase in the arcsine transformation, then evaluated for statistical significance accumulation of cells in G0/G1 when compared with the with ANOVA and Tukey rank mean tests. indicates significant empty vector control and Cdx1-ND mutant (data not difference from Adeno-Null-infected control and Adeno-Null/Ad- shown). The Cdx1-CD mutant differed only from the D1-treated cells, Pp0.025. A representative Western blot for pRb Cdx1-ND construct. Neither were significantly different is presented. *, hyperphosphorylated pRb; o, hypophosphorylated pRb from the control transfection. Therefore, only Cdx1-wt could elicit an accumulation of HCT116 cells in G0/G1. We utilized electroporation as an alternate approach to study the truncation constructs in DLD1 cells. induced by Cdx1 at 48 h (Figure 6). Most significantly, Cotransfection of a farnesylated-GFP expression con- cyclin D1 coexpression also reversed the induction of struct could serve as both a marker of and control for hypophosphorylated pRb (Figure 6). Adeno-Null mod- transfection, and then DNA content could be measured estly reduced the Cdx1-mediated G0/G1 accumulation, in the GFP þ cells by flow cytometry. Additionally, we but did not affect pRb phosphorylation levels (Figure 6). used electroporation to transfect transiently DLD1 cells, Therefore, cyclin D1 expression is sufficient to abrogate as this was the only method yielding transfection Cdx1’s antiproliferative effects. efficiencies high enough for the flow-cytometry method. DLD1 electroporation efficiencies ranged between 30 The Cdx1 G0/G1 arrest is a specific activity of the Cdx1 and 70%. At 48 h, the cells were fixed and stained with protein propidium iodide. Flow cytometry was used to quantify DNA content in GFP þ cells. Cdx1-wt and Cdx1-CD To address the issue of the specificity of Cdx1’s were equally effective at inducing a G0/G1 accumula- antiproliferative effects, we tested truncated Cdx1 tion of cells (Figure 8). On average, there was a 1.7-fold constructs. We generated Cdx1 clones in which the N- increase in the number of cells in the G0/G1 phase. This terminus (Cdx1141À268) or the C-terminus (Cdx1 induction equaled that observed by the adeno-Cdx1 Cdx11À223) was deleted (Figure 7a). These constructs construct in earlier experiments (Table 1). The Cdx1- were then subcloned into an expression vector and ND construct behaved indistinguishably from the empty tested for activity (Figure 7b, c). Transient transfection control pRC/CMV construct with respect to G0/G1 of these four expression vectors (pRC/CMV, the wild- accumulation (Figure 8). In summary, the N-terminus type pRC/CMV-Cdx1-Wt, and the truncations pRC/ but not the C-terminus of Cdx1 is required to inhibit CMV-Cdx1-ND and pRC/CMV-Cdx1-CD) demon- proliferation in HCT116 and DLD1 cells. This indicates strated high levels of Cdx1-Wt and Cdx1-ND protein that the inhibition of proliferation is a specific effect of expression (Figure 7b). As DLD1 cells transfect poorly Cdx1 and correlates with the ability of the protein to using most techniques, we used HCT116 cells to test the function as a transcription factor. Furthermore, the

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6403

a pRC/CMV 1 154 213 268 pRC/CMV-CDX1(wt) N-terminusHD C-terminus pRC/CMV-CDX1-ND 141 268 (Cdx1141-268 )

pRC/CMV-CDX1-CD 1 223 (Cdx11-223 )

bcSucrase Isomaltase Promoter Activation E WT C CD ND pRC/CMV

pRC/CMV-CDX1(wt)

* pRC/CMV-CDX1-ND

pRC/CMV-CDX1-CD actin 0 10 20 30 SI-(-183 to+54)Luc Fold Activation Figure 7 The N-terminus of Cdx1 is necessary for SI promoter activation. (a) Illustration of the Cdx1 truncation mutants. (b) Immunoblot for Cdx1 and Cdx1-N-terminus truncated proteins after transient transfection. At 48 h, whole-cell protein extracts were prepared subjected to PAGE and immunoblot using a polyclonal antibody (CPSP) recognizing an epitope on Cdx1’s C-terminus. (c) Transcriptional activity of the mutant and full-length Cdx1 constructs, cotransfected with a Cdx responsive reporter, the SI (À183 to þ 54) luciferase, and transfection control (pRL-CMV) in HCT116 cells. Luciferase activity was measured at 48 h and normalized to total protein and renilla luciferase levels absence of a pro-proliferative effect by the dominant- Wt or the Cdx1-ND mutant. We observed a modest negative Cdx1-ND construct in DLD1 cells that 30% reduction of cyclin D1 promoter activity in the full- indicates that the endogenously expressed CDX2 length hD1(-1749) promoter with Cdx1-Wt expression protein had no significant antiproliferative effect on its (Figure 9b). This reduction is observed despite the own. induction of the basal vector (pA3-Luciferase) over four-fold by Cdx1-Wt. A greater reduction of cyclin D1 Cdx1 reduces cyclin D1 gene expression promoter activity by Cdx1 might therefore be masked by this nonspecific vector induction. In summary, Cdx1 To determine if Cdx1 reduces cyclin mRNA levels by expression does not significantly alter cyclins A, B, and increasing its rate of destruction, we carried out mRNA D1 mRNA decay rates, but does reduce cyclin D1 stability studies using the RNA polymerase inhibitor, promoter activity in DLD1 cells. Actinomycin D. We observed a gradual decay of cyclin A, B, and D1 mRNAs over the 24 h period (data not shown). Cdx1 expression did not appear to alter the rate Discussion of mRNA decay. The mRNA present from these three cyclins were quantified, and the values, expressed as The continuous renewal of the intestinal epithelium is a percentage of T0 RNA levels, were plotted against time. highly ordered and regulated process. This process Cyclins A, B, and D1 mRNA decay rates were not becomes increasingly disordered in the progression to significantly altered by Cdx1 expression (Figure 9a). To adenomas and carcinomas in the human colon. The data confirm that the Actinomycin D treatment was fully presented here establish in a robust fashion that inhibited RNA polymerase activity, we examined c-Myc restoration of Cdx1 expression in human colon cancer mRNA levels by Northern blot. c-Myc mRNA expres- cells leads to inhibition of proliferation by a mechanism sion was undetectable by 4 h of Actinomycin D that involves reduction of cyclin D1 gene expression, treatment (data not shown). Of note, c-Myc mRNA with associated reduction in G1 Cdk kinase activity. levels were diminished by Cdx1 expression 3–4-fold We used a variety of independent techniques to when compared with adeno-Null controls (data not express Cdx1 transiently in human colon cancer cells. shown). In a prior study, we used both transient and stable- A full-length human cyclin D1 promoter-luciferase inducible techniques to express Cdx1 in nontransformed reporter construct was assayed in the presence of Cdx1- cells. In each case, we observed that Cdx1 expression led

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6404 factors, including Notch, Hedgehog, Wingless (Wnt), pRC/CMV Hes1, Math1, NeuroD, Forkhead-6, KLF-4, KLF-5, HNF-1, GATA 4, and the Caudal-related genes (Cdx1 pRC/CMV-CDX1(wt) * and Cdx2) (Kaestner et al., 1997; Freund et al., 1998; Korinek et al., 1998; Mutoh et al., 1998; Beck et al., 2000; Jensen et al., 2000; Clatworthy and Subramanian, pRC/CMV-CDX1-ND 2001; Dang et al., 2001; Fang et al., 2001; Sun et al., 2001; Yang et al., 2001; Boudreau et al., 2002). Many of pRC/CMV-CDX1-CD * these signaling pathways and transcription factors are recognized to regulate both differentiation and prolif- 0.8 1.0 1.2 1.4 1.6 1.8 2.0 eration (Korinek et al., 1998; Mutoh et al., 1998; Jensen Fold Change G0/G1 Figure 8 The N-terminus of Cdx1 is necessary for growth inhibition. DLD1 cells were electroporated as described. At 48 h, the cells were isolated, stained with propidium iodide and analysed a 100 for GFP expression and DNA content by ﬂow cytometry. The Cyclin A percentage of transfected cells in G0/G1 was determined by gating 80 on GFP þ cells only. This was then normalized to the transfection control (%GFP þ ). Six separate transfections were performed and 60 the results for each were averaged and the result expressed as fold change when compared to the control pRC/CMV-transfected cells 40 20

0 to reduced proliferation that was not associated with 10 20 30 increased apoptotic cell death. Experiments with Cdx1 100 Cyclin B truncation mutants establish that this is a speciﬁc effect 80 of the Cdx1 transcription factor. Furthermore, these antiproliferative effects likely require Cdx1 transcrip- 60 tional activity, as it was the transcriptionally inactive 40 Cdx1-ND mutant that lost the ability to induce the G0/ mRNA Levels

0 20 G1 arrest. Lastly, and perhaps most importantly, we have been able to demonstrate in several divergent cell 0 10 20 30

lines a conserved mechanism by which Cdx1 induces % of T 100 cell-cycle withdrawal. In both studies, Cdx1 expression Cyclin D1 lead to a G0/G1 accumulation and reduced pRb and 80 p130protein phosphorylation. This was due to reduced 60 cyclin D1 gene expression, which thereby diminished G1 cdk kinase activity. 40 These findings are consistent with current models of 20 intestinal epithelial development. Control of proliferation, differentiation, and lineage determination are not 0 10 20 30 well understood, but are known to depend upon the Time (hours) actions of several signaling pathways and transcription b 200 Figure 9 Reductions in Cyclin D1 mRNA levels are due to changes in Cyclin D1 gene expression. (a) Cyclin D1 mRNA stability is not significantly altered by Cdx1 expression. DLD1 cells were infected and cultured as before. At 48 h, actinomycin D was added to the cultures (final concentration ¼ 10 mg/ml). Total RNA was isolated at specific 150 intervals thereafter. This RNA was subject to RNAse protection analysis. Cyclins A, B, and D1 RPA band intensities were quantified, and the average of the duplicates was calculated and background subtracted. These values were then plotted as %T0 levels against length of Actinomycin D treatment to determine the decay rate of the 100 mRNA. Solid line/open boxes: Control Adeno-Null-infected cells; Dashed line/closed diamonds: Adeno-Cdx1-infected DLD1 cells. (b) Cdx1 expression reduces cyclin D1 promoter activity. Duplicate DLD1 cell transfections were performed as described with human cyclin D1 promoter-luciferase reporter constructs and Cdx1-Wt or Cdx1-ND Relative Luciferase Activity 50 expression constructs. At 48 h, the cells were isolated and luciferase activity determined. Luciferase activity was normalized to renilla luciferase levels (transfection control) and total protein. This activity was expressed as relative luciferase activity. Black bar: Cdx1-ND; grey bar: Cdx1-Wt. One of several experiments is presented -1749 pA3

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6405 et al., 2000; Mariadason et al., 2001; Naishiro et al., served differences. Classically, the study of antiproli- 2001; van de Wetering et al., 2002). In general, factors ferative agents has been carried out with stable-inducible that promote differentiation also reduce proliferation. expression constructs or transient expression methods Given that Cdx1 and Cdx2 have been implicated in including adenoviruses and transfection (Polyak promoting enterocyte and colonocyte differentiation, et al., 1997; Somasundaram et al., 1997; van de Weter- they might be expected to inhibit proliferation as well. ing et al., 2002). Stable-inducible or transient Our observations support this hypothesis. expression of Cdx1 is by design more sensitive for Our findings may help resolve conflicting reports. It the detection of antiproliferative effects than the has been hypothesized that Cdx1 promotes intestinal methods utilized in the other reports–namely, stable epithelial proliferation, because Cdx1 is expressed highly transfection with a constitutive CDX1 expression vector in the crypt compartment and because Cdx1’s expres- and pooling transfected cells rather than isolating sion is induced by oncogenic Wnt/b-catenin and Ras individual clones (Lorentz et al., 1997; Soubeyran signaling (Lorentz et al., 1999; Lickert et al., 2000; et al., 1999, 2001). One potential criticism of our study Soubeyran et al., 2001). However, known proliferation- is that we cannot be certain that expression of Cdx1 and tumor-suppressors like Cdx2 and p16 are expressed leads to the induction of genes that are in vivo in crypt cells (Dai et al., 2000; Rings et al., 2001). transactivated by Cdx2. However, the hypothesis that Furthermore, Cdx1 is expressed at low but detectable there are Cdx1- and Cdx2-specific targets may be an levels in the nonproliferating surface and villous artefact of early studies that suggested Cdx1 and Cdx2 epithelial cells as well (Silberg et al., 1997, 2000). had distinctive, nonoverlapping patterns of intestinal Moreover, Cdx1 expression is generally diminished or expression. It is increasingly apparent that the Cdx1 and actively silenced in human colon cancers (Silberg et al., Cdx2 have complementary and overlapping roles in 1997; Suh et al., 2002) despite the near-universal development (van den Akker et al., 2002). This is activation of Wnt/b-catenin and Ras signaling path- understandable given their structural homology, com- ways in human colorectal carcinomas. This last find- mon DNA sequence requirements, and overlapping ing is inconsistent with Cdx1 acting to promote the distributions of expression. oncogenic Wnt/b-catenin and Ras signaling pathways. It is interesting to speculate how the cell-cycle We would expect Cdx1 levels to be increased if this inhibitory property of Cdx1 may function normally in were the case. intestinal crypt cells as well as in neoplasia. Cdx1 (and Lastly, Cdx1 has been described to be an oncogene Cdx2) may serve as a feedback mechanism to constrain due to its presence in human metaplasias of the the proliferation of the rapidly dividing crypt epithelial esophagus and stomach (Soubeyran et al., 2001), its cells, while at the same time promoting their differentia- reported ability to induce soft-agar growth in 3T3 and tion. In intestina neoplasia, Cdx1 may be preferentially IEC6 cells (Maulbecker and Gruss, 1993; Soubeyran targeted for silencing because of its effects upon cyclin et al., 1999), induce PCNA expression and BrdU D1 levels and activity. Cyclin D1/cdk kinase activity is incorporation (Oh et al., 2002), as well as promote known to be critically important for the proliferation of proliferation in IEC6 and CaCo2 cells (Lorentz et al., neoplastic cells in the presence of wild-type pRb 1997; Soubeyran et al., 1999). However, the expression (Donnellan and Chetty, 1998) and its expression is of Cdx1 in intestinal metaplastia is not evidence for a typically upregulated in a number of human cancers proliferative function but rather an association with an including colon cancer (Donnellan and Chetty, 1998). In intestinal cell phenotype; Cdx2 and sucrase isomaltase colon carcinogenesis, this is an early event and is most have also been associated with intestinal metaplasia (Bai commonly caused by increased transcriptional activity et al., 2002). Additionally, the expression of both Cdx1 due to markedly enhanced Wnt/b-catenin- and Ras- and Cdx2 decreases or is lost during the transition from signaling, rather than gene duplications (Albanese et al., intestinal metaplasia to dysplasia to cancer (Silberg et al., 1995; Arber et al., 1996; Shtutman et al., 1999; Tetsu 1997; Bai et al., 2002), further supporting a role in and McCormick, 1999). It therefore stands to reason constraining rather than augmenting proliferation. that the neoplastic cells that inactivate Cdx1 and Finally, in contrast to the above reports, we found enhance cyclin D1 gene expression would obtain a Cdx1 expression reduced rather than increased prolif- proliferative advantage. eration of cells in culture, and a stable-inducible Cdx1 In summary, the expression of Cdx1 in human colon expression vector could not induce 3T3 cell soft-agar cancer DLD1 cells reduces proliferation by diminishing growth. cyclin D1 gene expression. This results in decreased G1 In considering why our observations differ from other cyclin D1-cdk4/6 kinase activity and hypophosphory- reports, several possibilities emerge, including that the lated pRb and p130proteins. This is a specific effect IEC6 and 3T3 cells in our study could correspond to likely requiring Cdx1 transcriptional activity, as it different subclones, or that murine Cdx1 (used in our requires an intact Cdx1 N-terminus but not the C- study) and human CDX1 (used in other studies) possess terminus. This mechanism may regulate the swift different biological properties. Both possibilities are transition from rapidly proliferating crypt cells to unlikely. In particular, murine and human Cdx1 nondividing, mature enterocytes in the intestinal and proteins are highly conserved, with nearly 85% con- colonic crypts. It may also function to reduce the served amino-acid identity. Rather, it is more plausible proliferation of neoplastic cells and thereby retard that different techniques are responsible for the ob- tumor formation in the colon.

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6406 Abbreviations (J Lynch), Pilot Project awards from the McCabe Fund (to J CDX1, CDX2, human genes; Cdx1, Cdx2, mouse genes or Lynch) and Center for Molecular Studies in Digestive and species not speciﬁed. Liver Disease (to J Lynch), and the Morphology, Cell Culture, and Molecular Biology Cores of the NIH/NIDDK Center for Acknowledgements Molecular Studies in Digestive Diseases at the University of This work was supported by National Institute of Diabetes Pennsylvania (P30-DK50306). This work was also supported and Digestive and Kidney Diseases Grants PO1-DK49210 in part by Grant IRG-78-002-25 from the American Cancer (project to P Traber and S Silberg) and DK02695-02 Society (J Lynch).

References

Albanese C, Johnson J, Watanabe G, Eklund N, Vu D, Mallo GV, Soubeyran P, Lissitzky JC, Andre F, Farnarier C, Arnold and Pestell RG. (1995). J. Biol. Chem., 270, Marvaldi J, Dagorn JC and Iovanna JL. (1998). J. Biol. 23589–23597. Chem., 273, 14030–14036. Albrecht JH and Hansen LK. (1999). Cell Growth Differ., 10, Mariadason JM, Bordonaro M, Aslam F, Shi L, Kuraguchi 397–404. M, Velcich A and Augenlicht LH. (2001). Cancer Res., 61, Arber N, Hibshoosh H, Moss SF, Sutter T, Zhang Y, Begg M, 3465–3471. Wang S, Weinstein IB and Holt PR. (1996). Gastroenterol- Maulbecker CC and Gruss P. (1993). Cell Growth and Differ., ogy, 110, 669–674. 4, 431–441. Bai YQ, Yamamoto H, Akiyama Y, Tanaka H, Takizawa T, Mutoh H, Naya FJ, Tsai MJ and Leiter AB. (1998). Genes Koike M, Kenji Yagi O, Saitoh K, Takeshita K, Iwai T and Dev., 12, 820–830. Yuasa Y. (2002). Cancer Lett., 176, 47–55. Naishiro Y, Yamada T, Takaoka AS, Hayashi R, Hasegawa Beck F, Tata F and Chawengsaksophak K. (2000). BioEssays, F, Imai K and Hirohashi S. (2001). Cancer Res., 61, 2751– 22, 431–441. 2758. Boudreau F, Rings EH, van Wering HM, Kim RK, Swain GP, Oh EJ, Park JH, Cho M, Lee WJ, Choi YH and Yoo MA. Krasinski SD, Moffett J, Grand RJ, Suh ER and Traber PG. (2002). Int. J. Oncol., 20, 23–29. (2002). J. Biol. Chem., 277, 31909–31917. Phelps DE and Xiong Y. (1997). Methods Enzymol., 283, Clatworthy JP and Subramanian V. (2001). Mech. Dev., 101, 194–205. 3–9. Polyak K, Xia Y, Zweier JL, Kinzler KW and Vogelstein B. Chung DC. (2000). Gastroenterology, 119, 854–865. (1997). Nature, 389, 300–305. Dai CY, Furth EE, Mick R, Koh J, Takayama T, Niitsu Y and Rings EH, Boudreau F, Taylor JK, Moffett J, Enders GH. (2000). Gastroenterology, 119, 929–942. Suh ER and Traber PG. (2001). Gastroenterology, 121, Dang DT, Mahatan CS, Dang LH, Agboola IA and Yang 1437–1450. VW. (2001). Oncogene, 20, 4884–4890. Sheaff RJ. (1997). Methods Enzymol., 283, 173–193. Donnellan R and Chetty R. (1998). Mol. Pathol., 51, 1–7. Sherr CJ. (2000). Cancer Res., 60, 3689–3695. Drummond FJ, Sowden J, Morrison and Edwards YH. (1998). Shtutman M, Zhurinsky J, Simcha I, Albanese C, D’Amico M, FEBS Lett., 423, 218–222. Pestell R and Ben-Ze’ev A. (1999). Proc. Natl. Acad. Sci. Fang R, Olds LC, Santiago NA and Sibley E. (2001). Am. J. USA, 96, 5522–5527. Physiol. Gastrointest. Liver Physiol., 280, G58–G67. Silberg DG, Furth EE, Taylor JK, Schuck T, Chiou T and Freund JN, Domon-Dell C, Kedinger M and Duluc I. (1998). Traber PG. (1997). Gastroenterology, 113, 478–486. Biochem. Cell. Biol., 76, 957–969. Silberg DG, Swain GP, Suh ER and Traber PG. (2000). Hu B, Mitra J, van den Heuvel S and Enders GH. (2001). Mol. Gastroenterology, 119, 961–971. Cell. Biol., 21, 2755–2766. Somasundaram K, Zhang H, Zeng YX, Houvras Y, Peng Y, Jensen J, Pedersen EE, Galante P, Hald J, Heller RS, Ishibashi Wu GS, Licht JD, Weber BL and El-Deiry WS. (1997). M, Kageyama R, Guillemot F, Serup P and Madsen OD. Nature, 389, 187–190. (2000). Nat. Genet., 24, 36–44. Soubeyran P, Andre F, Lissitzky JC, Mallo GV, Moucadel V, Kaestner KH, Silberg DG, Traber PG and Schutz G. (1997). Roccabianca M, Rechreche H, Marvaldi J, Dikic I, Dagorn Genes and Dev., 11, 1583–1595. JC and Iovanna JL. (1999). Gastroenterology, 117, 1326– Korinek V, Barker N, Moerer P, van Donselaar E, Huls G, 1338. Peters PJ and Clevers H. (1998). Nat. Genet., 19, 379–383. Soubeyran P, Haglund K, Garcia S, Barth BU, Iovanna J and Krasinski SD, Van Wering HM, Tannemaat MR and Grand Dikic I. (2001). Oncogene, 20, 4180–4187. RJ. (2001). Am. J. Physiol. Gastrointest. Liver Physiol., 281, Subramanian V, Meyer B and Evans GS. (1998). Differentia- G69–G84. tion, 64, 11–18. Lickert H, Domon C, Huls G, Wehrle C, Duluc I, Clevers H, Suh E, Chen L, Taylor J and Traber PG. (1994). Mol. Cell. Meyer BI, Freund JN and Kemler R. (2000). Development, Biol., 14, 7340–7351. 127, 3805–3813. Suh E and Traber PG. (1996). Mol. Cell. Biol., 16, Lorentz O, Cadoret A, Duluc I, Capeau J, Gespech C, Cherqui 619–625. G and Freund J-N. (1999). Oncogene, 18(1), 87–92. Suh ER, Ha CS, Rankin EB, Toyota M and Traber PG. Lorentz O, Duluc I, Arcangelis AD, Simon-Assmann P, (2002). J. Biol. Chem., 227, 35795–35800. Kedinger M and Freund JN. (1997). J. Cell. Biol., 139, Sun R, Chen X and Yang VW. (2001). J. Biol. Chem., 276, 1553–1565. 6897–6900. Lynch J, Suh ER, Silberg DG, Rulyak S, Blanchard N and Taylor JK, Boll W, Levy T, Suh E, Siang S, Mantei N and Traber PG. (2000). J. Biol. Chem., 275, 4499–4506. Traber PG. (1997). DNA Cell Biol., 16, 1419–1428. Lynch JP and Hoops TC. (2002). Hematol.–Oncol. Clin. N. Tetsu O and McCormick F. (1999). Nature, 398, Am., 16, 1–36. 422–426.

Oncogene Cdx1 inhibits proliferation by reducing cyclin D1 expression J Lynch et al 6407 van de Wetering M, Sancho E, Verweij C, de Lau W, Oving I, van den Akker E, Forlani S, Chawengsaksophak K, de Graaff Hurlstone A, van der Horn K, Batlle E, Coudreuse D, W, Beck F, Meyer BI and Deschamps J. (2002). Develop- Haramis AP, Tjon-Pon-Fong M, Moerer P, van den Born ment, 129, 2181–2193. M, Soete G, Pals S, Eilers M, Medema R and Clevers H. Yang Q, Bermingham NA, Finegold MJ and Zoghbi HY. (2002). Cell, 111, 241–250. (2001). Science, 294, 2155–2158.

Oncogene