Oncogene (2008) 27, 2951–2960 & 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00 www.nature.com/onc ORIGINAL ARTICLE A novel role for the -catabolizing enzyme CYP26A1 in Barrett’s associated adenocarcinoma

C-L Chang1,2, E Hong1, P Lao-Sirieix1 and RC Fitzgerald1

1MRC-Cancer Cell Unit, MRC/Hutchison Research Centre, Cambridge, UK and 2Department of Obstetrics and Gynecology, Mackay Memorial Hospital, Taipei, Taiwan

Vitamin A deficiency is associated with carcinogenesis, Introduction and upregulation of CYP26A1, a major retinoic acid (RA)-catabolizing enzyme, has recently been shown in The incidence of oesophageal adenocarcinoma has in- cancer. We have previously demonstrated alterations of creased rapidly in the western world over the last two RA biosynthesis in Barrett’s oesophagus, the precursor decades (Pohl and Welch, 2005). The 5-year survival rate is lesion to oesophageal adenocarcinoma. The aims of this an abysmal 13% because the majority of patients present study were to determine CYP26A1 expression levels and with locally advanced disease (Medical Research Council functional effects in Barrett’s associated carcinogenesis. Oesophageal Cancer Working (MRCOCW) Group, 2002). Retinoic acid response element reporter cells were used to Since most oesophageal adenocarcinomas arise on the determine RA levels in non-dysplastic and dysplastic background of Barrett’s metaplasia, which gradually Barrett’s cell lines and endoscopic biopsies. CYP26A1 progresses to carcinoma through a metaplasia–dyspla- expression levels, with or without induction by RA and sia–carcinoma sequence, this provides the opportunity lithocholic acid, were determined by quantitative reverse to intervene at an early stage. A greater understanding transcriptase-PCR (RT–PCR) and immunohistochemis- of the pathogenesis of Barrett’s associated cancer is try. CYP26A1 promoter activity was determined by a required to develop effective chemopreventive strategies. luciferase reporter construct. CYP26A1 was stably over- An association between (retinoic acid, RA) expressed in GihTERT cells, which were evaluated for deficiency and carcinogenesis has been known since the gene-expression changes (pathway array and quantitative 1920s when vitamin A-deficient rats were observed to RT–PCR), cellular proliferation (cytometric DNA profile develop squamous metaplasia of the lung mimicking the and colorimetric assay) and invasion (in vitro matrigel changes caused by smoking (Wolbach and Howe, 1925). assay) with or without the CYP inhibitor ketaconazole. Since then, a number of experimental animal models have RA levels decreased progressively with the degree of demonstrated that vitamin A deficiency is associated with dysplasia (Po0.05) and were inversely correlated with increased susceptibility to carcinogenesis (Bertram, 1993; CYP26A1 gene levels and activity (Po0.01). CYP26A1 Avidan et al., 2002). In humans, epidemiological data expression was increased synergistically by RA and have demonstrated a relationship between low levels of lithocholic acid (Po0.05). Overexpression of CYP26A1 serum b-carotene and vitamin A intake and the risk of led to induction of c-Myc, epidermal growth factor various human cancers including cervical pre-cancerous receptor and matrix metalloproteinase 3 as well as lesions (Saffiotti et al., 1967; Wald et al., 1980; Smith and downregulation of tissue inhibitor metalloproteinase 1 Waller, 1991). These effects have been shown to occur and 3. Functional effects of CYP26A1 overexpression through defects in DNA repair (Webster et al., 1996), were increased proliferation (Po0.01) and invasion in increased cell proliferation (Reifen et al., 1998) and a vitro (Po0.01), which were inhibited by ketaconazole. number of genotoxic events (Klamt et al., 2003). Overexpression of CYP26A1 causes intracellular RA Disruption of normal signalling may occur as depletion and drives the cell into a highly proliferative and a result of reduced expression of retinoid receptors, invasive state with induction of other known oncogenes. reduced transcriptional responses of RA target genes Oncogene (2008) 27, 2951–2960; doi:10.1038/sj.onc.1210969; and increased RA metabolism (Freemantle et al., 2003). published online 3 December 2007 A member of the cytochrome P450 family, CYP26A1, has recently been identified as a major RA-catabolizing Keywords: vitamin A; oesophageal adenocarcinoma; enzyme, which converts RA to biologically inactive proliferation; invasion; bile acids hydroxylated retinoid derivatives (Niederreither et al., 2002). CYP26A1 has been shown to be upregulated in human familial adenomatous polyposis adenomas, sporadic colon cancers and primary ovarian cancer Correspondence: Dr RC Fitzgerald, MRC-Cancer Cell Unit, MRC/ (Downie et al., 2005; Shelton et al., 2006). Furthermore, Hutchison Research Centre, Hills Road, Cambridge, Cambs CB2 the pattern of retinoid gene expression, including 2XZ, UK. E-mail: [email protected] CYP26A1, has been shown to correlate with retinoid Received 3 May 2007; revised 11 October 2007; accepted 6 November sensitivity in a panel of breast and lung cancer cell lines 2007; published online 3 December 2007 (Liu et al., 2005). Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2952 Apart from the connection between RA deficiency states and cancer, there are several reasons why CYP26A1 might be implicated in Barrett’s oesophagus and associated carcinogenesis. First, Barrett’s oesopha- gus may be considered as an abnormal re-activation of embryonic patterning since the oesophagus undergoes a columnar to squamous transition at 18 weeks gestation. CYP26A1 is necessary for the establishment of the anterior–posterior axis in embryogenesis (Abu-Abed et al., 1998), and we have recently shown that alterations in RA concentration levels can affect differentiation of human oesophagus ex vivo (Chang et al., 2007). Second, chronic exposure to duodeno-gastro-oesophageal reflux- ate is the only well-established risk factor for Barrett’s carcinogenesis (Winters et al., 1987), and the bile acid lithocholic acid (LCA) has recently been demonstrated to compete efficiently with 9-cis-RA for the (RXR)-binding sites on the promoter region of the CYP26A1 gene (Radominska-Pandya and Chen, 2002). Therefore, the specific aims of this study were to determine the relative levels of RA and CYP26A1 expression in neoplastic compared with non-neoplastic Barrett’s cells in vitro and in vivo, to evaluate the effects of LCA on CYP26A1 expression and to examine the molecular and functional effects of CYP26A1 over- expression using an in vitro model system.

Results

Decreased retinoid levels in neoplastic relative to non-dysplastic Barrett’s cells Figure 1 Retinoic acid concentration in a Barrett’s oesophagus First, we determined the relative levels of RA in cells and tissues. Cell lines representative of Barrett’s carcinogenic sequence (a) or Barrett’s metaplasia (BE) and adenocarcinoma (Ca) Barrett’s carcinogenesis. Barrett’s telomerase-immorta- patients (b) were cultured for 24 and 18 h respectively with 100 nM lized cell lines were treated with 10À7 M RA over 24 h and retinal in serum-free medium. The resultant medium (100 ml) from then examined for RA levels using retinoic acid response cell or biopsy culture was then added to a 96-well plate containing element (RARE) reporter cells. The results demon- 3T3 RARE-SEAP reporter cells and incubated for 24 h. SEAP concentration in the medium was determined by Great EscApe strated that non-dysplastic Barrett’s cell lines, BAR detection kit. Relative RA levels were calculated by comparison À11 À11 (7.16 Â 10 ±2.20 Â 10 M) and QhTERT (2.91 Â with a standard curve using serial RA concentrations and then 10À11±1.38 Â 10À11 M), had higher retinoid levels, com- adjusted for the protein content of the cells. In panel a, for each cell pared to dysplastic cell lines, ChTERT (1.51 Â 10À11± line, three biological replicates (each with three technical replicates) 0.09 Â 10À11 M), GohTERT (6.73 Â 10À12±4.29 Â were performed. For panel b, the level of RA was determined in triplicate in each biopsy sample (***Po0.0001 and **Po0.01). 10À12 M) and GihTERT (2.40 Â 10À12±1.53 Â 10À12 M, Po0.0001; Figure 1a), and transformed normal oeso- À12± À12 phagus cell lines Het1a (2.43 Â 10 1.28 Â 10 M) the expression levels of the main RA-producing enzyme À12± À12 and NES (1.60 Â 10 0.32 Â 10 M). We confirmed RALDH2 in Barrett’s cell lines and tissues did not the in vivo relevance of these changes using oesophageal reveal any differences (data not shown). Therefore, we tissues from patients with non-dysplastic Barrett’s and next examined the expression levels of the main adenocarcinoma. Specifically, the mean retinoid level in catabolizing enzyme, CYP26A1. The constitutive ex- Barrett’s oesophageal biopsies (5.92 Â 10À11±1.28 Â À11 pression of CYP26A1 in all the cell lines is very low, but 10 M) was increased 13-fold compared to normal detectable by real-time PCR. We demonstrated that À12± À12 oesophagus (1.94 Â 10 0.73 Â 10 M) and 5.5-fold CYP26A1 mRNA transcript levels were induced by RA compared to adenocarcinoma biopsies (1.04 Â 10À11± À11 in the GihTERT cell line in a dose-dependent manner 0.9 Â 10 M, P ¼ 0.033; Figure 1b). À5 À10 (10 –10 M all-trans retinoic acid (ATRA)), (data not shown) and therefore, 10À8 M was chosen for subsequent Increased CYP26A1 gene expression in Barrett’s experiments. Across the panel of cell lines, RA led to a carcinogenesis significant induction of this gene, which was most Cellular retinoid activity is modulated via the balance marked in the dysplastic cell lines (GohTERT, ChTERT between production and metabolism. Measurement of and GihTERT) compared to the non-dysplastic

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2953 Barrett’s cell lines, BAR and QhTERT (Po0.01; staining was also seen to increase in the epithelial Figure 2a). Measurement of the promoter activity, using compartment throughout the metaplasia–dysplasia– a luciferase reporter vector containing a 354-bp prox- adenocarcinoma sequence (Figure 3b). Together, these imal 50 sequence of CYP26A1, revealed an increased in vivo and in vitro data suggest that CYP26A1 gene reporter activity in dysplastic cell lines consistent with expression is upregulated during the carcinogenic the expression levels (Po0.01; Figure 2b). In patient sequence of Barrett’s oesophagus. biopsies, real-time PCR data revealed a significant upregulation of CYP26A1 in adenocarcinoma com- Lithocholic acid and RA synergistically enhance the pared with non-dysplastic Barrett’s tissues (P 0.01; o expression of CYP26A1 Figure 2c). These results were obtained in the absence of Since duodeno-gastro-oesophageal reflux is causally any exogenous ATRA and reflect basal transcript levels. implicated in Barrett’s carcinogenesis and since LCA In the light of this in vivo data, we next determined the competes with RA for the RXR-binding site, we level of protein expression in tissue sections comprising examined the effect of the secondary bile acid LCA on each stage of the metaplasia–dysplasia–adenocarcino- CYP26A1 gene expression. LCA (15 mM) alone signifi- ma. A western blot was performed with affinity-purified cantly increased its expression in GihTERT cells anti-CYP26A1 antibody alone and following pre- (P 0.05). ATRA and LCA in combination resulted in incubation with CYP26A1 peptide to ensure antibody o a two- and threefold enhancement of CYP26A1 gene specificity (Figure 3a). The degree of immunostaining expression for GihTERT at 10À7 M of RA and Goh- for Barrett’s tissues without dysplasia (Figures 3b, c, 4a TERT respectively, compared to addition of ATRA and b) and with low-grade dysplasia (Figures 3b, c, 4c alone (P 0.05; Figures 5a and b). and d) was generally low. Where staining was present, it o was confined to the deep glands (black arrow) and the lamina propria (red arrow). In contrast, there was Overexpression of CYP26A1 in Barrett’s cells enhances strong immunostaining in samples containing high- cell proliferation and survival grade dysplasia (Figures 3b, 4e and g) and adenocarci- To investigate the functional role of CYP26A1 over- noma (Figures 3b, 4h and i) with a shift towards expression, we used GihTERT since its expression levels increased expression in the epithelial compartment and were most profoundly induced by RA in these cells and reduced expression in the lamina propria (Figure 3c). their transfection efficiency is good. Two stable cell lines The staining was localized primarily to dysplastic were established, using a vector for pCEP4-CYP26A1 epithelial cells within the glands (Figure 4g) compared and a pCEP4 empty vector, (Figure 6a). To examine the with the surface (Figure 4f). The intensity of CYP26A1 molecular mechanisms by which CYP26A1 might be

Figure 2 Expression of CYP26A1 in Barrett’s cell lines and tissues and CYP26A1 promoter activity. (a) Relative mRNA expression of CYP26A1 in cell lines representative of the Barrett’s carcinoma sequence cultured with or without 100 nM ATRA. (b) CYP26A1 promoter activity following RA stimulation in cells that were transiently co-transfected with CYP354-Luc and Renilla, and treated with 100 nM RA 24 h after transfection. The luciferase activity was measured 24 h later and the data are given as a fold increase compared with untreated cells. Each experiment was done in triplicate, **Po0.01 for dysplastic compared with non-dysplastic lines. (c) Baseline transcript levels in Barrett’s adenocarcinoma (Ca) compared with Barrett’s oesophagus (BE) biopsies. The gene expression, relative to GAPDH, was determined by real-time PCR. The data are expressed on a logarithmic scale. **Po0.01. ATRA, all-trans RA; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; RA, retinoic acid.

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2954

Figure 3 Expression of CYP261 in Barrett’s tissue. (a) Validation of rabbit polyclonal anti-CYP26A1 antibody on GihTERT cells by western blotting. Membrane was probed with affinity-purified anti-CYP26A1 antibody alone (lane 1) and anti-CYP26A1 antibody pre- incubated with CYP26A1 peptide (lane 2). b-Actin was used as a loading control. Cumulative immunohistochemistry score for the distribution and intensity of CYP26A1 expression in BE (Barrett’s oesophagus), LGD (low-grade dysplasia), HGD and AC tissues in the epithelium (b) and lamina propria (c), **Po0.01 and *Po0.05 for BE vs AC tissues.

implicated in carcinogenesis, we analysed the cell- (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium signalling gene expression pattern of cells overexpressing bromide) assay (P ¼ 0.002, Figure 7a). This difference CYP26A1 compared to that of control cells (Figure 6b). was abolished by adding 2 mM ketoconazole, a pan- The results suggested a gene expression pattern favour- cytochrome P450 inhibitor (P ¼ 0.008; Figure 7a). Flow ing cell survival with antiapoptosis signals (decreased cytometric analysis provided further evidence for Bax and increased BCL2, BCL2L1 and BIRC1) and increased proliferation since GihTERT/pCEP4- decreased cell death signals (TNF, FAS, FASLG). In CYP26A1 cells exhibited a higher percentage of cells addition, differentiation of cell cycle regulators, cyclin- in the S and G2M phase (S ¼ 17.1%, G2M ¼ 24.24%) dependent kinase inhibitors (CDKN1A, CDKN1B, compared to control GihTERT/pCEP4 cells CDKN1C, CDKN2A) and retinoid-binding proteins (S ¼ 10.19%, G2M ¼ 18.1%). In addition, GihTERT/ (RBPs) were also repressed. Oncogenes, in particular pCEP4-CYP26A1 cells were more invasive in vitro than epidermal growth factor receptor (EGFR) and c-Myc, control cells (Po0.001), and this invasive behaviour was were upregulated (Figure 6b), and this was confirmed by inhibited with ketoconazole (Po0.01; Figure 7b). real-time PCR (Figure 6c). In addition to the global analysis of genes involved in cell signalling, we also performed real-time PCR for matrix metalloproteinases (MMPs) and tissue inhibitor metalloproteinases Discussion (TIMPs) in view of their known role in invasion. Out of the genes examined (MMP-1, -2, -3, -7, -9, -10, -14 We have demonstrated decreased RA levels and and TIMP-1, -2, -3), the MMP-3 gene was found to be increased expression of CYP26A1 in Barrett’s associated significantly upregulated (P ¼ 0.003), and TIMP-1 and dysplasia and adenocarcinoma in vitro and in vivo. LCA, -2 were downregulated (P ¼ 0.005 and 0.01, respectively) a component of refluxate, and RA synergistically in GihTERT/pCEP4-CYP26A1 cells, relative to enhance the expression of CYP26A1. Overexpression GihTERT/pCEP4 control cells (Figure 6d). of CYP26A1 led to upregulation of MMP-3, TIMP-1, -2 In view of the gene expression patterns that favoured and oncogenes EGFR and c-Myc and altered the cell survival and invasion, we performed functional biological behaviour of Barrett’s GihTERT cells leading assays for these parameters. These CYP26A1-over- to enhanced proliferation and invasion. expressing cells showed enhanced cell growth, compared CYP26A1 was recently cloned and is believed to be to empty vector-transfected cells in an in vitro MTT responsible for the 4-hydroxylation of ATRA, which in

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2955 et al., 2004); however, the molecular mechanism and functional consequences of these alterations have not been investigated. Here we focused on the CYP26A1 enzyme; however, it is also possible that reduced RA levels are due to aberrant expression of cellular retinal- binding protein or lecithin:retinal acyltransferase. These two proteins mediate the esterification of retinal to retinylester, the hepatic storage form of retinal. The integrity of the other signalling components should be investigated further; however, our data suggest that CYP26A1 overexpression is at least partially responsible for the relative RA deficiency in adenocarcinoma compared to Barrett’s metaplasia. Expression of CYP26A1 may be inducible by RA or RAR, RXR agonists. Two putative RARE sites exist on the proximal (À132) and distal (À2090) 50 region of this gene, which have been shown to work cooperatively (Loudig et al., 2005). Analysis suggests that there are potential DR5 RARE- or RXR-binding sites noted around the 2K 50 region. LCA has recently been demonstrated to compete efficiently with 9-cis-RA for the RXR retinoid-binding site. Here we have shown that LCA can activate RA signalling in reporter cells and induce CYP26A1 expression when acting synergistically with RA. Since only modest CYP26A1 induction was observed with LCA alone, the enhancement might occur through the RARE. Hence, because there is repetitive exposure of the Barrett’s mucosa to refluxate containing bile acids in vivo, it is possible that CYP26A1 expression may be induced even under relatively low concentrations of RA. It is also possible that inflammation may play a role, since there are also nuclear factor B-binding sites in the promoter of CYP26A1. Further work is required to determine how CYP26A1 gene is regulated in Barrett’s metaplasia and associated dysplasia. Overexpression of CYP26A1 also results in an increased proliferation index in keeping with data showing that this gene confers a growth advantage Figure 4 Representative immunostaining patterns for CYP26A1. (Osanai and Petkovich, 2005). In view of the multiple CYP26A1 staining of non-dysplastic Barrett’s (BE, A and B) signalling pathways, which might affect cell prolifera- ( Â 200) and low-grade dysplastic epithelia (C and D)(Â 200) are tion, we used an array approach to identify candidates. mostly negative except some weak, focal staining within deep The pattern of gene expression identified included glands (arrow). There are also some positive mesenchymal cells (panel C, arrow). In a high-grade lesion (E)(Â 100), the upper non- apoptosis-regulating genes, repression of tumour sup- dysplastic Barrett’s epithelium is negative (F)(Â 200), compared to pressor genes, de-regulation of cell cycle and activation frank staining in high-grade dysplastic glands (G)(Â 400). Invasive of growth factor (EGFR) and c-Myc genes. These cancerous cells show strong CYP26A1 staining (H and I)(Â 200). alterations are similar to the molecular events reported in Barrett’s carcinogenesis (reviewed in Bax et al., 2005). Increased retinoid activity has been linked to an turn diminishes RA signalling activity. In a mouse abrogation of EGFR signalling (Sah et al., 2002). In model with mutated RALDH2 and CYP26, the hydro- addition, c-Myc expression was also reduced in retinoid- xylated derivates of CYP26 were shown to be biologi- treated human leukaemia cells (Yung, 2004). Therefore, cally inactive (Niederreither et al., 2002). We have it is perhaps not surprising that CYP26A1 overexpres- recently shown that manipulation of RA concentrations sion increases expression of these oncogenes. has profound effects on differentiation status ex vivo The in vitro matrigel assay demonstrated that (Chang et al., 2007). However, this analysis was dysplastic Barrett’s GihTERT cells overexpressing the confined to non-dysplastic Barrett’s oesophagus. Very CYP26A1 gene had increased capacity for invasion, little work has been performed in Barrett’s associated which was associated with the upregulation of MMP-3 adenocarcinoma to determine the availability of RA and and downregulation of TIMP-1 and -3. RA has been whether its signalling is intact. Altered expression of documented to affect invasion by modulating the retinoic acid receptors (RAR) a and g as well as all three TIMPs or MMPs in many instances (Schoenermark RXR has been described (Lord et al., 2001; Brabender et al., 1999; Bigg et al., 2000). The conclusions drawn

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2956 et al., 1992; Van Heusden et al., 2002). In addition, RA metabolism-blocking agents lead to a reduction of tumour size in mouse xenograft models (Patel et al., 2004). In conclusion, we have demonstrated increased expression of CYP26A1 in Barrett’s carcinogenesis and investigated the functional significance of this gene in oesophageal adenocarcinoma including proliferation and invasion. These data suggest that the potential benefit of RA metabolism-blocking agents in cancer therapy deserves further study.

Materials and methods

All the work in this study was performed on human material (biopsy samples or cell lines).

Patient samples Biopsy samples were obtained from patients having endoscopy at Addenbrooke’s Hospital NHS Trust, following approval by the local research ethics committee. Barrett’s oesophageal samples were obtained from surveillance endoscopies and oesophageal squamous tissue specimens from diagnostic endoscopy procedures in patients who had an endoscopically and histopathologically normal squamous oesophagus. Paired research samples were taken together with samples for routine pathological examination. The research samples were either snap frozen in liquid nitrogen and stored at À80 1C or stored in holding medium for organ culture and processed within 6 h. For immunohistochemistry, samples of Barrett’s oesophagus with varying degrees of dysplasia and adenocarcinomas as well Figure 5 Synergistic induction of CYP26A1 by RA and LCA. as samples of normal squamous epithelium were selected from Two immortalized dysplastic Barrett’s cell lines GihTERT (a) and the Addenbrooke’s archive. GohTERT (b) were cultured to confluency and stimulated with 15 mM LCA alone, 10À7 and 10À8 M RA alone or a combination of LCA and RA. After 6 h, total RNA was extracted and CYP26A1 Cell lines gene expression was measured by real-time (RT)–PCR (*Po0.05 A panel of hTERT-immortalized Barrett’s cell lines, which and **Po0.01). LCA, lithocholic acid; RA, retinoic acid. were derived from well-characterized patients with different degrees of dysplasia within Barrett’s oesophagus, was used to determine the retinoid activity at different stages of the metaplasia–dysplasia sequence. The hTERT-immortalized from these experiments must be cautious since over- non-dysplastic Barrett’s cell line BAR (gifts from Rhonda F expressing genes of interest can lead to nonspecific Souza, University of Texas Southwestern Medical Center, TX, effects. However, due to the low basal CYP26A1 USA) was grown on a 3T3 sorry feeder layer in KBM-2 transcript levels in cell lines, knockdown experiments (Cambrex, UK) medium and supplemented as previously were not possible. The abrogation of the functional described (Hormi-Carver et al., 2007). In addition, an hTERT- effects of CYP26A1 overexpression by ketoconazole immortalized non-dysplastic cell line, QhTERT (CP-A), and three dysplastic cell lines, ChTERT (CP-B), GohTERT (CP- suggests that the effects are mediated via the cytochrome D) and GihTERT (CP-D) (gifts from P Rabinovich, Uni- P450 system, but in future more specific antagonists, versity of Washington, WA, USA), were maintained in MCDB which are not yet commercially available, would add 153 medium (Sigma, St Louis, MO, USA) and supplemented further weight. as described previously (Palanca-Wessels et al., 2003). The role of RA in cell differentiation, growth and apoptosis has provided the rationale to use RA as a Reporter cells and retinoic acid levels chemoprevention and chemotherapeutic agent in Bar- Cell lines and biopsies in organ culture were maintained in rett’s oesophagus (Hormi-Carver et al., 2007) and other serum-free medium containing 100 nM retinal for 24 h (Sigma). tumour types. Hence, clinical remission in acute RA levels in the medium were determined using a reporter cell promyelocytic leukaemia patients who respond to RA system. For this, Swiss 3T3 cells were transfected with 1.2 mgof therapy correlates with terminal differentiation of pRARE-TA-SEAP (Clontech, Mountain View, CA, USA) and leukaemic blasts. Similarly, in vitro inhibition of 0.2 mg of pIres-puro (Clontech). Puromycin was used to select cells stably expressing the reporter and after 24 h, the CYP26 leads to an increase in differentiation, decrease phosphatase activity was quantified using a Great EscAPe in proliferation, cell adhesion and metastatic potential. SEAP detection kit (BD Bioscience, San Jose, CA, USA). This has led to the suggestion of a possible chemopre- Relative RA content in the medium was calculated by vention role of using specific CYP26 inhibitors (Wouters comparison with a standard curve (serial RA concentrations

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2957

Figure 6 Validation of the membrane array data. The expression of CYP26A1 (green) was confirmed by immunofluorescence in GihTERT cells stably transfected with pCEP4 empty vector or pCEP4-CYP26A1. The nuclei are labelled with DAPI (blue (a)). The expression pattern of cells overexpressing CYP26A1 was analysed using a gene expression array membrane. This assay was performed in duplicate and results for the two repeats are represented in blue and red (b). Validation of EGFR and c-Myc genes (c) as well as MMP and TIMP genes by real-time PCR (d) from RNA extracted from sub-confluent GihTERT cells stably expressing the CYP26A1 gene or control. DC(t) is the relative cycle threshold of target gene to GAPDH determined from amplification curve in real-time PCR. DAPI, 4’,6-diamidino-2-phenylindole; EGFR, epidermal growth factor receptor; GAPDH, glyceraldehyde-3-phosphate dehydrogen- ase; MMP, matrix metalloproteinase; TIMP, tissue inhibitor metalloproteinase.

10À6–10À12 M) and adjusted for protein content according to the ATCCTCAAGCTCATCAGATTTCCC-30, containing the bicinchoninic acid protein assay using a commercially avail- BamH1 site by Vent (NEB, UK). The PCR fragment was able kit (Pierce, Rockford, IL, USA). digested and cloned in-frame into the eukaryotic expression vector pCEP4 (Invitrogen) and confirmation was obtained by sequencing. pCEP4 is an episomal mammalian vector contain- Expression and reporter constructs ing the Epstein–Barr virus replication origin (oriP) and nuclear To clone the full-length CYP26A1, GihTERT cells were antigen (EBNA-1 gene), to permit extrachromosomal replica- treated with 100 nM ATRA and then RNA was extracted using tion in humans. Thus in our experiment, we used pooled Trizol (Invitrogen, Paisley, UK). Reverse transcriptase-PCR transfected cells selected by hygromycin B, rather than single (RT–PCR) was performed using the forward primer 50-CCG clones. The CYP26A1 promoter reporter was constructed in CTCGAGGATGAAGCGCAGGAAATACG-30, containing two steps. PCR cloning of a 2K fragment of the CYP26A1 Xho 1 restriction site and the reverse primer 50-CGGG promoter sequence was performed using Pfu polymerase

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2958 (Sigma-Aldrich, St Louis, MO, USA) for 30 min at room temperature. Staining was performed using the DakoCytoma- tion EnVision þ System (DakoCytomation Ltd, Ely, UK) staining kit according to the manufacturer’s instructions. The primary antibody, a rabbit polyclonal antibody raised against the human C-terminal 18-amino peptide of CYP26A1, which is specific for the A1 isoform (Alpha Diagnostic International, San Antonio, TX, USA), was diluted in tris buffered saline/ Tween 0.05% containing 1% bovine serum albumin and incubated for 1 h at room temperature. A negative control was performed by omission of the primary antibody. Sections were then counter stained with haematoxylin and mounted. Scoring of immuno-positive cells for CYP26A1 was done separately for the epithelial and lamina propria compartments. CYP26A1 staining was scored according to the overall distribution (0 ¼ none, 1 ¼ o25%, 2 ¼ 25–50%, 3 ¼ 50–75% 4 ¼ >75% of cells stained) and intensity of staining (0 ¼ negative, 1 ¼ weak, 2 ¼ moderate, 3 ¼ strong staining).

RA and lithocholic acid stimulation Fully confluent 3T3 RA reporter cells were cultured in a 96- well format in serum-free medium for 24 h with either 10À8– 10À7 M of ATRA (Sigma-Aldrich, Dorset, UK), or LCA (Sigma) at 5, 10 and 15 mM or a combination of RA and LCA over the same concentration range. Serum-free medium with the same concentration of ethanol as vehicle served as a control. After 1 h of stimulation, the cells were washed with phosphate-buffered saline and cultured in 100 ml serum-free medium for 12 h. Secreted alkaline phos- phatase (SEAP) activity in the medium was quantified using Figure 7 Functional effect of stable CYP26A1 transfection. (a) the SEAP detection kit. For detecting the LCA-enhanced Cellular proliferation determined by the MTT assay in GihTERT/ expression of CYP26A1 gene, two immortalized Barrett’s cell pCEP4 and GihTERT/pCEP4-CYP cells cultured for up to 4 days. lines, GihTERT and GohTERT, were cultured to reach (*Po0.05 between untreated and treated G1hTert/pCEP4-CYP; **Po0.001 between untreated G1hTert/pCEP4-CYP and Gih- confluency and then stimulated with LCA (15 mM) alone, RA À7 À8 TERT/pCEP4). (b) Invasion assay in GihTERT/pCEP4 and (10 and 10 M) or a combination of LCA and RA. After 6 h, GihTERT/pCEP4-CYP cells cultured in Matrigel-coated inserts total RNA was extracted and CYP26A1 gene expression was in a transwell system in 0.5% fetal bovine serum medium, with 5% measured by real-time PCR. fetal bovine serum medium in the bottom of chamber as chemoattractant. After 24 h, the cells migrating through the MTT assay Matrigel filter were fixed and stained, and the number of cells was measured by counting the average in five randomly chosen GihTERT/pCEP and GihTERT/pCEP-CYP cells were cul- 4 1mm2 fields (**Po0.01). MTT, 3-[4,5-dimethylthiazol-2-yl]-2,5- tured in 24-well plates at a density of 5 Â 10 cells per well for diphenyl tetrazolium bromide. up to 4 days. The pCEP-CYP cells were also cultured with 2 mM of Ketoconazole (Sigma-Aldrich). The MTT assay was performed in quadruplicate for each time point according to the manufacturer’s instructions (Sigma-Aldrich, Paisley, UK). (Promega, Southampton, UK) and genomic DNA prepared The absorbance was measured spectrophotometrically at a from GihTERT cells. The forward primer was 50-GGGG 0 wavelength of 570 nm from which the background was TACCCAGAGTTCACTCGGATGTCACG-3 containing subtracted. the Kpn1 site and the reverse primer was 50-TACCATCTGC AAGGTTTCCC-30. The PCR fragment was digested by Kpn1 and Nco1, cloned into pGL3-basic vector (Promega, Madison, Determination of cell cycle distribution 6 WI, USA) and confirmed by sequencing. The vector was then A total of 3 Â 10 cells were grown for 24 h as a monolayer and digested with Kpn1, EcoR1 and Klenow enzymes. The pGL3- then trypsinized, suspended in culture medium, centrifuged at CYP26A1 promoter (354 bp) was obtained by re-ligation and 150 g for 10 min, and fixed in 70% ethanol overnight. The fixed sequencing confirmation was performed. cells were centrifuged again, and the cell pellet was resus- pended in propidium iodide (0.2 mg mlÀ1; Sigma) and incu- bated for 30 min at room temperature. Cells were passed Immunohistochemistry and scoring through a 70 mM cell strainer (BD Biosciences, Oxford, UK). Archival blocks of tissue samples from Barrett’s oesophagus DNA content was analysed by fluorescence-associated cell (n ¼ 10), low-grade dysplasia (n ¼ 10), high grade dysplasia sorting using the BD LSR1 flow cytometer (BD Biosciences). (n ¼ 10) and adenocarcinoma (n ¼ 10) patients were deparaffi- Data were captured using Cell Quest software (BD Bios- nized in xylene and then rehydrated through alcohol solutions ciences) and analysed with ModFitLT software (Verity Soft- and water. Antigen retrieval was performed by pressure ware, Topsham, ME, USA). cooking of samples for 3 min in 0.01 M tri-sodium citrate (VWR International, Poole, England) at pH 6.0. Nonspecific Real-time RT–PCR binding was blocked using 10% normal goat serum (Vector Total RNA from cells or tissues was extracted by using Trizol Labs, Burlingame, CA, USA) and 10% bovine serum albumin (Invitrogen). Two micrograms of RNA was reverse transcribed

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2959 Table 1 Primer sequences Primer sequences

Gene Forward Reverse

CYP26A1 50-TAAATGGATACCAGATTCCCAAGG-30 50-CTTCCTTGTTGGTGAAGATCTCTG-30 GAPDH 50-TCTCCTCTGACTTCAACAGC-30 50-TTGTCATACCAGGAAATGAGC-30 MMP-3 50-TCTGAAAGTCTGGGAAGAGG-30 50-AAGTCTCCATGTTCTCTAACTG-30 TIMP-1 50-AATTCCGACCTCGTCATCAG-30 50-GGAACCCTTTATACATCTTGGTC-30 TIMP-3 50-TGTAACTGCAAGATCAAGTCC-30 50-CAGGGTAACCGAAATTGGAG-30 EGFR 50-CTCCATCCTGGAGAAAGGAG-30 50-GTCTATCATCCAGCACTTGAC-30 c-Myc 50-GATTCTCTGCTCTCCTCGAC-30 50-TTCTTGTTCCTCCTCAGAGTC-30

Abbreviations: EGFR, epidermal growth factor receptor; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; MMP, matrix metalloproteinase; TIMP, tissue inhibitor metalloproteinase. using a SuperScript II reverse transcriptase kit (Invitrogen) in expression in CYP26A1-expressing GihTERT cells was calcu- 20 ml of total reaction solution, and diluted into 100 ml lated relative to that of GihTERT control cells. The data were in RNAse-free water. Each RNA sample was prepared in not analysed statistically since duplicate values are not sufficient triplicate. The primers used are listed in Table 1. Quantitative to perform a statistical test. However, two putative overexpressed PCR was performed on 2 ml of cDNA with the SYBR Green genes were confirmed by RT–PCR to validate the array data. JumpStart Taq Readymix (Sigma-Aldrich). PCR consisted of 1 1 40 cycles of 94 C denaturation (10 s), 57–59 C annealing (20 s) Matrigel invasion assay and extension (20 s). The cycle threshold C(t) was determined A total of 2 Â 104 stable GihTERT cells expressing CYP26A1 for each sample, and the average C(t) of triplicate samples was or control were plated in triplicate on cell-culture inserts of calculated. The expression of each gene relative to glyceralde- transwell chambers (Corning, Corning, NY, USA) containing hyde-3-phosphate dehydrogenases (GAPDH) was determined 6.5-mm filters (pore size, 8 mm) and coated with 25% Matrigel. as DC(t). We used serial dilutions of pCEP4-CYP26A1 vector The top chamber contained 0.5% fetal bovine serum medium DNA as an internal control and the melting curve was with 5% fetal bovine serum medium in the bottom chamber as constructed for each primer. a chemoattractant. After 24 h, the migrating cells present on the bottom part of the filters were fixed and stained, from Expression array analysis which the percentage of invading cells was determined. RNA from GihTERT/pCEP4-CYP26A1 or GihTERT/pCEP4 stable cells was analysed using a pathway array membrane Statistics (SuperArray, Bethesda, MD, USA). mRNA was isolated using Data are presented as the mean and the standard error of the oligodT-linked magnetic bead-based purification kit (Super- mean. The one-way analysis of variance test was used to Array). A total of 0.3 mg purified mRNA was reverse transcribed identify specific differences within a population followed by to cDNA probe by SuperScript II reverse transcriptase, with the t-test to identify specific differences between selected incorporation of Biotin-16-dUTP (Roche, Nutley, NJ, USA). groups. Analysis was carried out using Prism v3.0 (GraphPad Hybridization was performed following the manufacturer’s Software, CA, USA). A P-value 0.05 was considered protocol and the signal was detected using a chemiluminescent o significant: *P 0.05, **P 0.01 and ***P 0.0001. CDP-Star kit (SuperArray). The membranes were exposed to o o o X-ray films, which were developed, scanned and analysed by GEArray expression analysis software (SuperArray). The experi- Acknowledgements ment was performed in duplicate with independent preparation of mRNA. The data from each membrane were normalized by We thank Guillermo De La Cueva Mendez for his help and comparison to a panel of house-keeping genes present on the advice and Madhumita Das for her help with the invasion membrane (GAPDH, RPL13A, PPIA and b-actin). Gene assays.

References

Abu-Abed SS, Beckett BR, Chiba H, Chithalen JV, Jones G, Metzger D teinases-1 (TIMP-1) gene expression by all-trans retinoic acid in et al. (1998). Mouse P450RAI (CYP26) expression and retinoic acid- combination with basic fibroblast growth factor. Eur J Biochem 267: inducible retinoic acid metabolism in F9 cells are regulated by retinoic 4150–4156. acid receptor gamma and retinoid X receptor alpha. JBiolChem273: Brabender J, Marjoram P, Salonga D, Metzger R, Schneider PM, Park 2409–2415. JM et al. (2004). A multigene expression panel for the molecular Avidan B, Sonnenberg A, Schnell TG, Chejfec G, Metz A, Sontag SJ. diagnosis of Barrett’s esophagus and Barrett’s adenocarcinoma of (2002). Hiatal hernia size, Barrett’s length, and severity of acid reflux the esophagus. Oncogene 23: 4780–4788. are all risk factors for esophageal adenocarcinoma. Am J Gastro- Chang CL, Lao-Sirieix P, Save V, De La Cueva Mendez G, Laskey R, enterol 97: 1930–1936. Fitzgerald RC. (2007). Retinoic acid-induced glandular differentia- Bax DA, Siersema PD, Van Vliet AH, Kuipers EJ, Kusters JG. (2005). tion of the oesophagus. Gut 56: 906–917. Molecular alterations during development of esophageal adenocar- Downie D, McFadyen MCE, Rooney PH, Cruickshank ME, Parkin DE, cinoma. J Surg Oncol 92: 89–98. Miller ID et al. (2005). Profiling cytochrome P450 expression in Bertram JS. (1993). Inhibition of chemically induced neoplastic ovarian cancer: identification of prognostic markers. Clin Cancer Res transformation by carotenoids. Mechanistic studies. Ann NY Acad 11: 7369–7375. Sci 686: 161–175; discussion 175–176. Freemantle SJ, Spinella MJ, Dmitrovsky E. (2003). in cancer Bigg HF, McLeod R, Waters JG, Cawston TE, Clark IM. (2000). therapy and chemoprevention: promise meets resistance. Oncogene Mechanisms of induction of human tissue inhibitor of metallopro- 22: 7305–7315.

Oncogene Role of CYP26A1 in Barrett’s carcinogenesis C-L Chang et al 2960 Group. MRCOCW (2002). Surgical resection with or without Reifen R, Nyska A, Koperstein L, Zusman I. (1998). Intestinal and preoperative chemotherapy in oesophageal cancer: a randomised hepatic cell kinetics and mucous changes in vitamin-A-deficient rats. controlled trial. Lancet 359: 1727–1733. Int J Mol Med 1: 579–582. Hormi-Carver K, Feagins LA, Spechler SJ, Souza RF. (2007). All Saffiotti U, Montesano R, Sellakumar AR, Borg SA. (1967). trans-retinoic acid induces apoptosis via p38 and caspase pathways Experimental cancer of the lung. Inhibition by vitamin A of the in metaplastic Barrett’s cells. Am J Physiol Gastrointest Liver Physiol induction of tracheobronchial squamous metaplasia and squamous 292: G18–G27. cell tumors. Cancer 20: 857–864. Klamt F, Dal-Pizzol F, Roehrs R, de Oliveira RB, Dalmolin R, Sah JF, Eckert RL, Chandraratna RA, Rorke EA. (2002). Retinoids Henriques JA et al. (2003). Genotoxicity, recombinogenicity and suppress epidermal growth factor-associated cell proliferation by cellular preneoplastic transformation induced by vitamin A supple- inhibiting epidermal growth factor receptor-dependent ERK1/2 mentation. Mutat Res 539: 117–125. activation. J Biol Chem 277: 9728–9735. Liu T, Bohlken A, Kuljaca S, Lee M, Nguyen T, Smith S et al. (2005). Schoenermark MP, Mitchell TI, Rutter JL, Reczek PR, Brinckerhoff CE. The retinoid anticancer signal: mechanisms of target gene regula- (1999). Retinoid-mediated suppression of tumor invasion and matrix tion. Br J Cancer 93: 310–318. metalloproteinase synthesis. AnnNYAcadSci878: 466–486. Lord RV, Tsai PI, Danenberg KD, Peters JH, Demeester TR, Shelton DN, Sandoval IT, Eisinger A, Chidester S, Ratnayake A, Tsao-Wei DD et al. (2001). -alpha messenger Ireland CM et al. (2006). Up-regulation of CYP26A1 in adenoma- RNA expression is increased and retinoic acid receptor-gamma tous polyposis coli-deficient vertebrates via a WNT-dependent expression is decreased in Barrett’s intestinal metaplasia, dysplasia, mechanism: implications for intestinal cell differentiation and colon adenocarcinoma sequence. Surgery 129: 267–276. tumor development. Cancer Res 66: 7571–7577. Loudig O, Maclean GA, Dore NL, Luu L, Petkovich M. (2005). Smith AH, Waller KD. (1991). Serum beta-carotene in persons Transcriptional co-operativity between distant retinoic acid response with cancer and their immediate families. Am J Epidemiol 133: elements in regulation of Cyp26A1 inducibility. Biochem J 392: 241–248. 661–671. Niederreither K, Abu-Abed S, Schuhbaur B, Petkovich M, Chambon P, Van Heusden J, Van Ginckel R, Bruwiere H, Moelans P, Janssen B, Dolle P. (2002). Genetic evidence that oxidative derivatives of retinoic Floren W et al. (2002). Inhibition of all-TRANS-retinoic acid acid are not involved in retinoid signaling during mouse development. metabolism by R116010 induces antitumour activity. Br J Cancer Nat Genet 31: 84–88. 86: 605–611. Osanai M, Petkovich M. (2005). Expression of the retinoic acid- Wald N, Idle M, Boreham J, Bailey A. (1980). Low serum-vitamin-A metabolizing enzyme CYP26A1 limits programmed cell death. Mol and subsequent risk of cancer. Preliminary results of a prospective Pharmacol 67: 1808–1817. study. Lancet 2: 813–815. Palanca-Wessels MC, Klingelhutz A, Reid BJ, Norwood TH, Opheim KE, Webster RP, Gawde MD, Bhattacharya RK. (1996). Effect of different Paulson TG et al. (2003). Extended lifespan of Barrett’s esophagus vitamin A status on carcinogen-induced DNA damage and repair epithelium transduced with the human telomerase catalytic subunit: a enzymes in rats. In Vivo 10: 113–118. useful in vitro model. Carcinogenesis 24: 1183–1190. Winters CJ, Spurling TJ, Chobanian SJ, Curtis DJ, Esposito RL, Patel JB, Huynh CK, Handratta VD, Gediya LK, Brodie AM, Hacker JF et al. (1987). Barrett’s esophagus. A prevalent, occult Goloubeva OG et al. (2004). Novel retinoic acid metabolism complication of gastroesophageal reflux disease. Gastroenterology blocking agents endowed with multiple biological activities are 92: 118–124. efficient growth inhibitors of human breast and prostate cancer cells Wolbach SB, Howe PR. (1925). Tissue changes following deprivation in vitro and a human breast tumor xenograft in nude mice. J Med of fat-soluble A vitamin. J Exp Med 42: 753–777. Chem 47: 6716–6729. Wouters W, van Dun J, Dillen A, Coene M-C, Cools W, De Coster R. Pohl H, Welch HG. (2005). The role of overdiagnosis and (1992). Effects of , a new antitumoral compound, on reclassification in the marked increase of esophageal adenocarcino- retinoic acid-induced inhibition of cell growth and on retinoic acid ma incidence. J Natl Cancer Inst 97: 142–146. metabolism in MCF-7 human breast cancer cells. Cancer Res 52: Radominska-Pandya A, Chen G. (2002). Photoaffinity labeling of 2841–2846. human (RXRbeta) with 9-cis-retinoic acid: Yung BY. (2004). c-Myc-mediated expression of nucleophosmin/B23 identification of phytanic acid, , and litho- decreases during retinoic acid-induced differentiation of human cholic acid as ligands for RXRbeta. Biochemistry 41: 4883–4890. leukemia HL-60 cells. FEBS Lett 578: 211–216.

Oncogene