Involvement of Aquaporins in Colorectal Carcinogenesis

Involvement of aquaporins in colorectal carcinogenesis

Chulso Moon*,1,2, Jean-Charles Soria3, Se Jin Jang4, Juna Lee1, Mohammad Obaidul Hoque1, Mathilde Sibony5, Barry Trink1, Yoon Soo Chang4, David Sidransky1,2 and Li Mao4

1The Head and Neck Cancer Research Division, Department of Otolaryngology, The Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA; 2Department of Oncology and Sidney Kimmel Comprehensive Cancer Center, The Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA; 3Gustave Roussy Institute, Division of Cancer Medicine, 39 Rue Camille Desmoulins, Villejuif 94805, France; 4Molecular Biology Laboratory, Department of Thoracic/Head and Neck Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX 77030, USA; 5Pathology Department, Hospital Tenon, AP-HP, 4 rue de la Chine, Paris 75020, France

Aquaporins (AQPs) are important in controlling water distinct tissue distribution (King and Agre, 1996). In permeability. As AQP1 is known as a serum-responsive the kidney, lung, eye, and brain, multiple water-channel gene, we hypothesized that AQP expression may be homologues are expressed, providing a network for involved in the development of human cancer. By reverse water transport in those locations, and it was assumed transcriptase–polymerase chain reaction analysis, expres- that alterations in AQP expression or function can sion of AQPs 1, 3, and 5 was found in seven colon and be rate-limiting for water transport across certain colorectal cancer cell lines. Western blot analysis membranes (King and Agre, 1996; King et al., 1997; confirmed their expression in four of these cell lines. In Agre et al., 1998). Therefore, it was believed that AQPs situ hybridization demonstrated that during colorectal have some valuable role in human pathophysiology and carcinogenesis, the expression of AQPs 1 and 5 was that changes in such specific pathways have the potential induced in early-stage disease (early dysplasia) and to cause disease or to be the basis of treatment (King maintained through the late stages of colon cancer et al., 2000). For example, mutations in AQP2 were development. Expression of AQPs 1 and 5 was maintained identified as a cause of nephrogenic diabetes insipidus even in metastatic lesions in the liver. These findings (Deen et al., 1994). Yet, the significance of AQPs in demonstrate that the expression of several AQPs is found mammalian systems is frequently doubted, mainly in tumor cells and is associated withan early stage of because a human AQP1-knockout and several AQP- colorectal cancer development. These novel observations knockout mouse models did not show expected clinical suggest that multiple AQP expression may be advanta- phenotypes (Preston et al., 1994; Bai et al., 1999). geous to tumorigenesis, which may lead to a better Previous reports have shown that AQP1 is one of the understanding of colorectal carcinogenesis. protein families involved in cell cycle control known as Oncogene (2003) 22, 6699–6703. doi:10.1038/sj.onc.1206762 delayed early-response genes (Moon et al., 1995; Ma and Verkman, 1999). Therefore, we believed that this Keywords: aquaporin; in situ hybridization; RT–PCR; gene may have a potential role in uncontrolled cell colon cancer; carcinogenesis replication, as in cancer (Saadoun et al., 2002). To test our hypothesis, we initially used reverse transcriptase– polymerase chain reaction (RT–PCR) to screen the expression profiles of human AQPs in seven colon and colorectal cancer cell lines, and confirmed our results by Introduction Western blot analysis. Based upon these observations, by performing in situ hybridization on 16 colorectal The molecular basis of membrane water permeability cancer tissue samples, we further studied the expression remained elusive until the recent discovery of aquapor- of AQPs in vivo by using the well-established multistep ins (AQPs), water channel proteins (Preston et al., 1992; colorectal carcinogenesis model (Fearon and Vogelstein, Agre et al., 1995). The fundamental importance of these 1991). Our results provide a clear and novel example of proteins is suggested by their conservation from bacteria multiple AQP expression during human colorectal through plants to mammals (King and Agre, 1996; carcinogenesis. Maurel, 1997; Calamita et al., 1998). Thus far, 10 mammalian AQPs have been identified, each with a Results *Correspondence: C Moon, The Head and Neck Cancer Research Division, Department of Otolaryngology, The Johns Hopkins School of Medicine, 818 Ross Research Building, 720 Rutland Ave., Sequence-specific antisense and sense primers were Baltimore, MD 21205-2196, USA; E-mail: [email protected] designed for AQPs 1, 2, 3, 5, and 8. Seven colon and Received 3 February 2003; revised 29 April 2003; accepted 6 May 2003 colorectal cancer cell lines were screened with RT–PCR. Aquaporins and colorectal carcinogenesis C Moon et al 6700 Of the five AQPs examined, AQPs 1, 3, and 5were expression of AQPs 1, 3, and 5. AQP1 was strongly expressed in all the seven cell lines (Figure 1a). AQPs 2 expressed in colonic adenoma with almost no staining in and 8 were not expressed in any of these cell lines (data surrounding normal colonic mucosa (Figure 2a). In not shown). These data indicate that, although AQP3 is addition, it was consistently expressed in primary colon the only known AQP to be expressed in normal colonic cancer (Figure 2b) and in metastatic lesion in the liver surface epithelium, different AQPs seem to be expressed (Figure 2c). AQP5was also expressed in early adenoma, simultaneously in tumor cells. Western blot analysis late adenoma (Figure 2d, e), and in adenocarcinoma demonstrated that AQPs 1, 3, and 5were expressed (Figure 2f) with almost no staining in surrounding concurrently in all four of the tested cancer cell lines normal colonic mucosa (Figures 2d, e). The expression (Figure 1b), confirming our RT–PCR results. NIH3T3 of AQP3 in colonic surface epithelium (reviewed by Ma cells transfected with AQP1 and AQP3 expression and Verkman, 1999) and expression of AQP1 in the constructs were used as positive controls for the Western vascular endothelium were previously described (Nielsen blots of AQPs 1 and 3, respectively. 293, the kidney cell et al., 1997; King et al., 2000). The germinal centers of line, was used as a positive control for AQP5. In the tonsils have also been shown to express AQPs 1 and addition, there was a relatively lower expression of 5(Moon et al., 2003, manuscript in review). As internal AQP3 in the HCT116 p53 À/À clone compared to the controls for in situ hybridization, slides of the germinal HCT116 p53 þ / þ clone, consistent with a previous centers of the tonsils were probed with antisense report that p53 promoted the expression of AQP3 riboprobes of AQP1 (data not shown) and AQP5 (Zheng and Chen, 2001). (Figure 2g) and sense riboprobes of AQP1 (data not To confirm these findings in vivo and to evaluate the shown) and AQP5(Figure 2h). The vascular endothe- pattern of AQP induction during carcinogenesis, we lium was stained with antisense riboprobes of AQP1 performed in situ hybridization with AQP1, AQP3, and (Figure 2i). The expression of AQP3 was maintained in AQP5riboprobes generated by cloning the RT–PCR colon cancer and in different stages of preneoplastic products. We then tested the expression of these AQPs lesions (data not shown). These findings strongly suggest in colon cancer and in different stages of preneoplastic that at least two different AQPs are induced in the early lesions (Fearon and Vogelstein, 1991). From five colon stage of colorectal carcinogenesis. cancer patients 16 tissue samples were studied, including one patient with metastasis to the liver. Of these samples, 12 consisted of adenocarcinomas and the Discussion surrounding normal colonic tissue, while four were metastatic lesions from the liver. Of the five patients, This report provides a novel example of the involvement staged surgically resected series were available for three of water channel proteins in human colon cancer. Thus patients. All 12 colon cancer samples showed strong far, AQP3 is the only AQP known to be expressed in

a - DLD1 SW480 W489 HT20 H455 HCT116 p53-/ Control HCT116 p53+/+ AQP1 (217 bp) AQP3 (885 bp for Control through SW480, 131 bp for W489 through H455) AQP5 (247 bp) GAPDH (782 bp) b Negative Positive HCT116 HCT116 control control p53 +/+ p53 -/- DLD1 SW480 AQP1 (28 kD)

AQP3 (33 kD)

AQP5 (27 kD)

ß-actin (42 kD)

Figure 1 Expression of AQPs 1, 3, and 5in seven colon and colorectal cancer cell lines. RT–PCR was performed using sense and antisense primers for AQPs 1, 2, 3, 5, and 8 for seven cancer cell lines (W489, HT20, H455, HCT116 p53 þ / þ , HCT116 p53 À/À, DLD1, SW480). The name of each cell line is marked on the top of each lane, and the size of each PCR product is marked with an arrow. PCR mixtures without templates were used negative controls, and GAPDH was the loading control. Each PCR product was cloned and its sequence conﬁrmed. All seven cell lines demonstrated expression of AQPs 1, 3, and 5(a). Western blot analysis was performed using antibodies for AQPs 1, 3, and 5for the cell lines HCT116 p53 þ / þ , HCT116 p53 À/À, DLD1, and SW480. The name of each cell line is marked on the top of each lane, and the size of the protein is marked with an arrow. NIH3T3 cells transfected with AQP1 and AQP3 expression constructs were used as positive controls for the Western blots of AQPs 1 and 3, respectively. 293, a kidney cell line, was used as a positive control for the Western blot of AQP5. NIH3T3 cells were used as a negative control, and b-actin was the loading control. All four cell lines demonstrated AQPs 1, 3, and 5expression( b)

Oncogene Aquaporins and colorectal carcinogenesis C Moon et al 6701 a d expression to occur exclusively in the carcinoma cell itself. These findings raise many possibilities as to the role of AQPs during colorectal carcinogenesis. It is possible that induction of AQPs is required in the early stage of colorectal cancer development to function as one of the essential driving forces for initiating carcinogenesis. Considering prior promoter study results, expression of b e AQP1 in the early dysplasia of colon adenoma could have been induced by the binding of immediate early- response gene products to AQP1 binding sites of the human AQP1 promoter as a result of the expression of an oncogene like Fos/Jun (Moon et al., 1993, 1997). In addition, the expression of at least three AQPs in colon and colorectal cancer cell lines suggests the possibility that tumor cells may need multiple AQPs to c f gain advantages for replication or survival. During the cell cycle, as cell volume needs to increase rapidly by absorbing water from the outside with a minimal amount of energy, expression of several types of AQPs in the tumor cells is likely to be advantageous compared with normal cells having a single type of AQP expression. Also, tumor cells may require several AQPs for high metabolic turnover or tumor-specific metabolic g h pathways needed for survival. In fact, there is evidence suggesting that AQPs function as gas channels (Na- khoul et al., 1998), and it is possible that by accumulat- ing mutations inside the critical motifs in their channel structures, some of these AQPs can gain novel functions other than of a water channel (King and Agre, 1996; Heymann et al., 1998). i These questions of whether the expression of multiple AQPs is a driving force or only an associated phenomenon in colorectal carcinogenesis can be at least partially answered by transfecting overexpression or antisense expression constructs (Splinter et al., 2002). In fact, through a recent functional study, we have identified preliminary evidence that AQP1 has onco- genic properties. Ectopic expression of full-length Figure 2 In situ hybridization with AQP1 and AQP5probes in the multistep carcinogenesis model of colon cancer development. The cDNA of AQP1 induced many phenotypic changes AQP1 antisense riboprobe clearly stains not only colonic adenoma characteristic of transformation, including cell prolif- (a, arrow) but also the primary colon cancer (b, arrow) and eration enhancing activity and anchorage-independent metastatic lesions in the liver (c, arrow). Likewise, the AQP5 growth (Moon and Hoque, 2003, data not yet pub- antisense riboprobe clearly stains early adenoma with moderately dysplastic cells (d, arrow), late adenoma with severe dysplastic cells lished). (e, arrow), and adenocarcinoma (f, arrow). There is almost no Based upon our observations, we propose that staining in the surrounding normal colonic mucosa when probed changes in expression levels of several AQPs may play with AQP1 and AQP5antisense riboprobes ( a and d, star). As a role in tumorigenesis, and we provide evidence for positive controls, the germinal centers of the tonsils were stained the potential role of AQPs in human pathophy- with antisense riboprobes of AQP1 (data not shown) and AQP5 (g). Sense riboprobes of AQP1 (data not shown) and AQP5( h) siology (Marples, 2000). We speculate that studying were used as negative controls. As an internal control, vascular the expression status or modifying the expression of endothelium was stained with the AQP1 antisense riboprobe (i) these AQPs may lead to the design of novel ways of diagnosing or controlling human cancers. normal colonic surface epithelium (reviewed by Ma and Materials And Methods Verkman, 1999). Therefore, finding expression of multiple AQPs in both colorectal cancer cell lines and tissue Cell lines samples was unexpected. More surprisingly, AQPs 1 and 5were found to be expressed in the early stage of colon Three colorectal cancer cell lines (HCT116 p53 þ / þ , cancer development (mild dysplasia); based upon RT– HCT116 p53 À/À, DLD1) were maintained in McCoy’s PCR data from cell lines, we had expected AQP 5A modified medium with 10% fetal bovine serum

Oncogene Aquaporins and colorectal carcinogenesis C Moon et al 6702 (FBS), and the colon cancer cell line SW480 was thermal cycler (Perkin-Elmer Co.). For W489, HT20, maintained in Leibovitz’s L-15medium with 10% and H455, we performed 40 cycles using the following FBS. Other colon cancer cell lines (W489, HT20, conditions: 10 s at 941C, 50 s at 631C, and 50 s at 721C. H455) were maintained in Dulbecco’s modified Eagle’s b-Actin mRNA amplification was performed on the medium with 5% FBS. cDNA as a positive control for reaction efficiency and was performed for 25cycles. For HCT116 p53 þ / þ , Tissues HCT116 p53À/À, DLD1, and SW480, the following conditions were used: AQP1, 35cycles for 1 min at 95 1C, In all, 16 premalignant and malignant specimens of 1 min at 641C, and 2 min at 721C; AQP3, 33 cycles for 4 colon epithelium were archived tissue samples of 5sat951C, 45s at 60 1 C, and 1 min and 35s at 72 1C; for surgically resected lesions from five patients treated at AQP5, 35 cycles for 1 min at 951C, 1 min at 611C, and Hospital Tenon, Paris, France. In one patient, meta- 1 min at 721C; and for GAPDH, 24 cycles for 1 min at static lesions in the liver were also available. Of these, 12 951C, 1 min at 601C, and 1 min and 30 s at 721C. All of samples were comprised of adenocarcinomas and the the PCR products were introduced into the pCRsII- normal surrounding colonic tissue, while four were TOPO plasmid vector (Invitrogen) using the TOPO TA metastatic lesions from the liver. Of the five patients Cloning System and each AQP gene-specific sequence studied, a complete staged series consisting of normal was confirmed. Water was used as a negative control. colon epithelium, mild and moderate dysplasia (early adenoma), severe dysplasia (late adenoma), and invasive Western blot analysis adenocarcinoma tissues were available for three patients. Phosphate-buffered saline (PBS)-washed cells were lysed by sonication in ice-cold RIPA buffer (50 mm Tris-HCl, m m m RT–PCR analysis of AQPS expression in human tumor pH 7.4, 150 m NaCl, 1 m PMSF, 1 m EDTA, 1% cell lines Triton X-100, 1% sodium deoxycholate, and 0.1% SDS). Protein concentrations were determined by the For the cell lines W489, HT20, and H455, total RNA Bio-Rad DC protein assay (Bio-Rad). Equal amounts of extracted using an RNeasy total RNA kit (Qiagen) from protein (50 mg) were subjected to 14% SDS–polyacryla- cell lines was reverse transcribed by random primer and mide gel electrophoresis, and the proteins were trans- Superscript II reverse transcriptase (Gibco/BRL). For ferred onto nitrocellulose membranes. After 1 h the cell lines HCT116 p53 þ / þ , HCT116 p53 À/À, blocking with 5% nonfat dry milk in 1 Â PBS at room DLD1, and SW480, total RNA was extracted using temperature, the membranes were probed with rabbit Trizol reagent (Invitrogen) and reverse transcribed using anti-AQP1 antibody (Chemicon International), goat random primers and M-MLV reverse transcriptase anti-AQP3 antibody, or goat anti-AQP5antibody (Invitrogen). The resulting cDNA was subjected to (Santa Cruz Biotechnology), all at 1 : 100 dilutions in PCR with exon spanning primers of AQP1, AQP2, 1 Â PBS for 1 h at room temperature. After washing AQP3, AQP5, and AQP8. The AQP1, AQP2, and three times for 5min each in 1 Â PBS, the membranes AQP5products were amplified using sequences span- were hybridized for 1 h at room temperature with ning part of exons 1 and 2, AQP3 product was from part 1 : 5000 dilutions of the secondary antibody (HRP- of exons 2 and 3, and AQP8 product was from part of conjugated anti-rabbit antibody (Amersham) and HRP- exon 2, all of exon 3 and part of exon 4. The following conjugated anti-goat antibody (Santa Cruz Biotechnol- are the primer sequences and the sizes of each ogy)). To ensure equal loading and transfer of proteins, PCR: AQP1, 50-CGCAGAGTGTGGGCCACATCA the filters were probed with b-actin (Sigma-Aldrich) at a and 50-CCCGAGTTCACACCATCAGCC, amplifying 1 : 5000 dilution. Detection was performed using the a 217-bp product; AQP2, 50-TGGCACGGTGGTA- enhanced chemiluminescence system (Amersham). CAGGCTCT and 50-GCCGTCGTGCTGTTGCTGAG, NIH3T3 cells transfected with AQP1 and AQP3 amplifying a 227-bp product; AQP3, 50-GCTGTCAC- expression constructs were used as positive controls TCTGGGCATCCTG and 50-GCGTCTGTGCCAGG- for the Western blots of AQPs 1 and 3, respectively. GTGTAG, amplifying a 131-bp product; AQP5, H293, a normal kidney cell line, was used as a positive 50-CGTTTGGCCTGGCCATAGGCA and 50-TGGC- control for the Western blot of AQP5. NIH3T3 cells CCTGCGTTGTGTTGTTG, amplifying a 247-bp pro- were used as a negative control. duct; AQP8, 50-CGTGATTGCCACGCTGGGGAA 0 and 5 -CACCTCAGGTCCAAAAGCACG, amplifying Probe construction for in situ hybridization a 423-bp product; and GAPDH, 50-GAAATCCCAT- CACCATCTTCCAGG and 50-CATGTGGGCCAT- Human AQP1 complementary DNA samples were GAGGTCCACCAC, amplifying a 782-bp product. obtained by PCR using the same primers as those used For HCT116 p53 þ / þ , HCT116 p53 À/À, DLD1, to obtain the RT–PCR product in Figure 1a. Human and SW480, different primers were used for AQP3. The AQP5complementary DNA was obtained by PCR primer sequences are as follows: 50-GCCGGATCCGC- using sense primer, 50-CGTTTGGCCTGGCCATAGG- CGCCGCCATGGGTCGACAGAAG and 50-GCCT- CA, and antisense primer, 50-CGATTCATGAC- CTAGACTATCAGATCTGCTCCTTGTGC, amplify- CACCGCAGGG. The cDNA specificity was ing an 885-bp product. PCR was performed in a DNA confirmed by gel sequencing. The cDNA was introduced

Oncogene Aquaporins and colorectal carcinogenesis C Moon et al 6703 into the plasmid pCRsII-TOPO, and this construction and then treated with RNase A (40 mg/ml) and RNase was used as a template to generate sense and antisense T1 (10 U/ml) in 10 mm Tris-HCl, pH 8.0, 1 mm EDTA, probes. During transcription, nonradioactive labeling of 0.5 m NaCl for 30 min at 371C. Then sections were single-strand RNA probes was performed using digox- incubated with agitation for 2 h with 2 Â SSC containing igenin-UTP (DIG RNA labeling kit, Boehringer Man- 0.05% Triton X-100 and 2% normal sheep serum at nheim, Germany). The probes were divided into aliquots room temperature. The sections were rinsed in buffer 1 with RNAse inhibitor and stored at À801C until use. (0.1 m maleic acid and 0.15 m NaCl, pH 7.5) for 5 min at room temperature and then incubated with 2% normal In situ hybridization of tissue sections with AQPS sheep serum, 0.3% Triton X-100 in buffer 1 for 30 min riboprobes also at room temperature. Slides were then incubated overnight at 41C with antidigoxigenin antibody (1 : 500). For each case tested, paraffin-embedded tissue sections 4 After two rinses in buffer 1, slides were rinsed briefly in (4 mm) were cut onto silane-coated slides (Sigma). The buffer 3 (100 mm Tris-HCl, 100 mm NaCl, and 50 mm sections were deparaffinized in xylene, rehydrated in MgCl2, pH 9.5). Alkaline phosphatase was detected gradually decreasing concentrations of ethanol, and using 5-bromo-4-chloro-3-indolyl phosphate and nitro- treated with 0.2 n HCl. Sections were then treated with blue tetrazolium chloride. Slides were then rinsed in proteinase K for 15min at 37 1C. Sections were washed buffer 3 (10 mm Tris-HCl, and 1 mm EDTA, pH 8) and thrice with 1 Â PBS, postfixed in 4% paraformaldehyde mounted with aqua mounting medium (Fischer). Sec- for 5min at room temperature, and rerinsed with tions pretreated with RNase to prevent hybridization or 1 Â PBS. Then sections were acetylated in 0.25% acetic incubated with the digoxigenin-labeled sense probe in anhydride and 0.1 m triethanolamine for 10 min. The the same conditions were used as negative controls. sections were dehydrated in gradually increasing concentrations of ethanol and air-dried prior to hybridization. Sections were prehybridized for 1 h at 421Cin Acknowledgements hybridization buffer (20 Â SSC, 50% deionized We thank Dr Li K Su, who provided total RNA for the three formamide, 2.5mg of predenatured salmon sperm colon cell lines used in this study. The study was supported in part by NIH/NCI Grant P50 CA96784-01 (to CM), American DNA, 1 g of dextran sulfate, 2% 100 Â Denhart’s Cancer Society Grant RPG-98-054 (to LM), Fondation de solution, 2% DTT, and 4 mg of yeast tRNA). Hybridi- France, AP-HP and Lilly Fondation Grant (to J-CS), Cancer zation was performed at 421C for 4 h in hybridization Center Grant P30 CA 16620 (to MD Anderson Cancer buffer containing 400 ng/ml of the probe. The tissue was Center), and Tobacco Research Fund from State of Texas twice washed in 2 Â SSC for 5min at room temperature (to MD Anderson Cancer Center).

References

Agre P, Bonhivers M and Borgnia MJ. (1998). J. Biol. Maurel C. (1997). Annu. Rev. Plant Physiol. Plant Mol. Biol., Chemother., 273, 14659–14662. 48, 399–429. Agre P, Brown D and Nielsen S. (1995). Curr. Opin. Cell Biol., Moon C, King LS and Agre P. (1997). Am. J. Physiol., 273, 7, 472–483. C1562–C1570. Bai C, Fukuda N, Song Y, Ma T, Matthay MA and Verkman Moon C, Preston GM, Grifﬁn CA, Jabs EW and Agre P. AS. (1999). J. Clin. Invest., 103, 555–561. (1993). J. Biol. Chem., 268, 15772–15778. Calamita G, Kempf B, Bonhivers M, Bishai WR, Bremer E Moon C, Williams JB, Preston GM, Copeland NG, Gilbert and Agre P. (1998). Proc. Natl. Acad. Sci. USA, 95, 3627– DJ, Nathans D, Jenkins NA and Agre P. (1995). Genomics, 3631. 30, 354–357. Deen PMT, Verdijk MAJ, Knoers NVA, Wieringa B, Nakhoul NL, Davis BA, Romero MF and Boron WF. (1998). Monnens LAH, van Os CH and van Oost BA. (1994). Am. J. Physiol., 274, C543–C548. Science, 264, 92–94. Nielsen S, King LS, Christensen BM and Agre P. (1997). Am. Fearon ER and Vogelstein B. (1991). Cell, 61, 759–762. J. Physiol., 273, C1549–C1561. Heymann JB, Agre P and Engel A. (1998). J. Struct. Biol., 121, Preston GM, Carroll TP, Guggino WB and Agre P. (1992). 191–206. Science, 256, 385–387. King LS and Agre P. (1996). Annu. Rev. Physiol., 158, 619– Preston GM, Smith BL, Zeidel ML and Moulds JJ. (1994). 648. Science, 265, 1585–1587. King LS, Nielsen S and Agre P. (1997). Am. J. Physiol., 273, Saadoun S, Papadopoulos MC, Davies DC, Bell BA and C1541–C1548. Krishna S. (2002). Br. J. Cancer, 87, 621–623. King LS, Yasui M and Agre P. (2000). Mol. Med. Today, 6, Splinter PL, Masyuk AI and LaRusso NF. (2002). J. Biol. 60–65. Chem., 278, 6268–6274. Ma T and Verkman AS. (1999). J. Physiol., 517, 317–326. Zheng X and Chen X. (2001). FEBS Lett., 489, 4–7. Marples D. (2000). Lancet, 355, 1571–1572.

Oncogene