Intestinal Cancer in Patients with a Germline Mutation in the Down-Regulated in Adenoma (DRA) Gene

SHORT REPORT Intestinal cancer in patients with a germline mutation in the down-regulated in adenoma (DRA) gene

Akseli Hemminki1, Pia HoÈ glund1, Eero Pukkala2, Reijo Salovaara1,3, Heikki JaÈ rvinen4, Reijo Norio5 and Lauri A Aaltonen1

Departments of 1Medical Genetics, and 3Pathology, Haartman Institute, PO Box 21 (Haartmaninkatu 3), 00014 University of Helsinki, Finland; 2Finnish Cancer Registry, Liisankatu 21 B, 00171 Helsinki, Finland; 4Second Department of Surgery, Helsinki University Central Hospital, Haartmaninkatu 4, 00250 Helsinki, Finland; 5Department of Medical Genetics, the Family Federation of Finland, PO Box 849 (Kalevankatu 16), 00101 Helsinki, Finland

A recent study has revealed that germline mutations of and in dedierentiated states of mucosa, such as in the down-regulated in adenoma (DRA) gene are a likely embryonic tissues or in neoplasia, DRA expression is cause of a recessive intestinal absorption defect, not detected (Schweinfest et al., 1993; Silberg et al., congenital chloride diarrhea. This ®nding was in 1995; HoÈ glund et al., 1996). Although the association accordance with previous works showing that DRA between lost DRA expression and neoplasia is encodes a sodium independent transporter for sulfate apparent, it is not known if this loss has a role in and oxalate. Although DRA was originally reported as a tumorigenesis, or if it is merely a consequence of the candidate tumor suppressor, these studies have ques- dedierentiated status of the epithelium. tioned the relevance of DRA in cancer. To evaluate Congenital chloride diarrhea (CLD) is an autosomal whether further studies on the role of DRA in recessive disorder characterized by voluminous watery tumorigenesis are still of interest, we examined whether stools containing a high concentration of chloride individuals carrying germline DRA mutations have an (McKusick, 1994). Other features include hydramnios excess of intestinal cancer. Cancer status of 229 and premature birth. Heterozygous carriers are members of 36 Finnish congenital chloride diarrhea symptomless. The majority of Finnish CLD patients families (44 homozygous patients, 70 heterozygous originate from the Eastern part of Finland (see map in parents, and 115 grandparents at 50% risk of being a HoÈ glund et al., 1995). The population of this area is DRA mutation carrier) was checked at the Finnish approximately 1.35 million people. Strong evidence was Cancer Registry and the risk of intestinal cancer was recently presented indicating that CLD is caused by found slightly elevated (standardized incidence ratio 3.4, germline mutations in the DRA gene. A three base pair 95% con®dence interval 1.4 ± 7.0, P50.05). While this in-frame deletion (DV317) and a C307W change were result does not unambiguously demonstrate an increased found in all Finnish CLD chromosomes studied intestinal cancer risk in DRA mutation carriers, it should (n=64). However, the C307W seemed like a poly- promote further studies to determine the possible role of morphism because of its high population frequency. DRA in cancer. Furthermore, it was not disease-causing when present without DV317 (HoÈ glund et al., 1996). Keywords: DRA; CLD; cancer; neoplasia; tumor The association between digestive tract cancer and suppressor homozygous defects in another anion transporter, the cystic ®brosis transmembrane conductance regulator gene, CFTR, has been recently documented (Neglia et al., 1995). Some of the molecular mechanisms that Four years ago, the down-regulated in adenoma might associate gastrointestinal cancer predisposition (DRA) gene was cloned (Schweinfest et al., 1993) and and defects in CFTR have been recently revealed consecutively mapped to 7q22-q31.1 (Taguchi et al., (Abraham et al., 1996). To determine whether the 1994). DRA has a great deal of structural similarity germline defect in DRA has relevance in intestinal with the diastrophic dysplasia sulphate transporter cancer, we utilized two unique data sets that were gene, which led investigators to presume a functional available to us. First, all identi®ed patients (n=44), similarity between these two genes (HaÈ stbacka et al., parents (n=70) and grandparents (n=115) of Finnish 1994). Structural similarity with a rat liver sulfate CLD families were checked for cancer at the Finnish transporter led other researchers to study the function Cancer Registry. Second, 728 patients with colorectal of DRA, and the gene was shown to encode a sodium neoplasia, 485 of them from the high CLD frequency independent transporter for sulfate and oxalate (Silberg area, were studied for the presence of the DV317 et al., 1995). The expression of DRA is limited to the mutation. 416 control individuals were also analysed mature mucosal cells of the small and large intestine for this mutation. (Schweinfest et al., 1993; Silberg et al., 1995; HoÈ glund As a result of the centralized health care system of et al., 1996) and to the prostate (HoÈ glund et al., 1996), Finland, diagnostic hallmarks, and well-studied genealogy of the disease (Norio et al., 1971) the ascertainment of CLD is believed to be almost Correspondence: LA Aaltonen complete. Altogether 36 CLD families were known Received 16 April 1997; revised 1 September 1997; accepted 1 to us. In 27 families the three base pair deletion September 1997 DV317 has been demonstrated to cause the disease Mutation in DRA and intestinal cancer AHemminkiet al 682 (HoÈ glund et al., 1996), from the rest of the families no Table 1 Cancer in CLD familes DNA samples were available. In all families, the Number patients were diagnosed by characteristic clinical of cases Average features, and by demonstrating high fecal chloride (parents/ age at content. Previous genealogical studies have identi®ed Cancer location Cancer histology grandparents)diagnosis parents and grandparents of the patients (HoÈ glund et Intestine 7a (3/4) 58 al., 1995; Norio et al., 1971). Since the disease is Rectum adenocarcinoma 3 (2/1) 47 Sigmoid colon adenocarcinoma 2 (0/2) 69 recessive, all parents are obligate heterozygous Ascending colon adenocarcinoma 1 (0/1) 72 carriers, and 50% of the grandparents carry one Ileum adenocarcinoma 1 (1/0) 53 defective allele. One mother of a CLD patient was Lung 8 (1/7) 67 exluded from this study because paternal uniparental microcellular carcinoma 4b (0/4) 69 adenocarcinoma 2 (1/1) 60 isodisomy of chromosome 7 caused the disease in the unknown 2 (0/2) 70 aected child (HoÈ glund et al., 1994). Data was Other 18 (1/17) 68 available from 44 patients, 70 parents and 115 Stomach adenocarcinoma 2 (0/2) 74 grandparents. One parent, and 33 grandparents were Prostrate adenocarcinoma 2 (0/2) 70 lost during follow up, usually because they had died Endometrium adenocarcinoma 1 (0/1) 62 Ovary cystadenocarcinoma 1 (1/0) 42 before personal identi®cation numbers necessary for Breast ductal carcinoma 1 (0/1) 79 computer linking were introduced in 1967. The Kidney renal cell carcinoma 1 (0/1) 66 Finnish Cancer Registry has legal status, is popula- Thyroid papillary carcinoma 1 (0/1) 73 tion based, and has almost 100% coverage since 1953 Eye (choroid epithelium) melanoma 1 (0/1) 57 (KylloÈ nen et al., 1987; Teppo et al., 1994). Follow-up Bladder unknown 1 (0/1) 75 for cancer started at the birth of the ®rst CLD patient Skin basocellular carcinoma 7 (0/7) 69 in each family (for the grandparents follow-up started at the birth of the CLD patient's parent), or on the Total 33 (5/28) 66 1st of January 1967, whichever was later, and ended Results from the Cancer Registry follow-up: cancers found in CLD at death or 31st of December 1993, whichever patients' parents and grandparents. aNumber of cases compared with occurred ®rst. The numbers of observed cases and the age and sex adjusted expected value was statistically signi®cant with the Mantel-Haenzel chi-square test, P50.05. bNumber of cases person-years at risk were counted, by 5 year age compared with the age and sex adjusted expected value was groups, separately for the three calendar periods statistically signi®cant with the Mantel-Haenzel chi-square test, (1967 ± 1975, 1976 ± 1984 and 1985 ± 1993). The ex- P50.01 pected numbers of cases for total cancer and for speci®c cancer types were calculated by multiplying the number of person-years in each age group by the In the unselected patients with colorectal neoplasia, corresponding average cancer incidence in the whole DV317 was not found in the low frequency areas Finnish population during the period of observation. (South-Central Finland) in 243 patients (223 carcino- To calculate the standard incidence ratio, the observed mas, 20 adenomas). In the high CLD frequency area 6/ number of cases (see Table 1) was divided by the 485 (1.2%) patients, 2/198 (1.0%) with adenoma and expected number. The statistical signi®cance was 4/287 (1.4%) with carcinoma, were heterozygous for tested by the Mantel ± Haenzel chi-square test, on DV317 (Figure 1). Four (1.0%) of the 416 controls the presumption that the number of observed cases displayed the mutation (dierence not statistically followed a Poisson distribution. signi®cant). The CLD patients were followed for 697 person In three out of the four unselected patients with years, parents 1250 person years, and grandparents DV317, carcinoma DNA was also available. Loss of 2430 person years. At the end of follow-up, there were the wild type allele could not be detected in any of no cases of cancer reported to the Cancer Registry these tumors (data not shown). The same was true with among the CLD patients (homozygous for the genetic four paran embedded tumors from the CLD families defect), all 32 years old or younger. Two of the parents displaying heterozygosity for DV317. (obligate heterozygous carriers of the defect) had The study focused on occurrence of intestinal altogether three primary intestinal cancers (P50.01, cancer, but incidentally as many as eight cases of standardized incidence ratio (SIR) 20.0, 95% con- lung cancer were detected during the cancer follow-up ®dence interval (CI) 4.1 ± 58.5). One had a cancer of (SIR 2.0, CI 0.9 ± 4.0). Four of these eight cases were the rectum at 39 years of age, the other had cancers of small cell carcinomas (0.6 expected, SIR 7.1, CI 1.9 ± the ileum, rectum and lung at the age of 53 years. The 18.1, P50.01), two were adenocarcinomas and two cancer status of the latter patient was known to us were of unknown histology. beforehand. Four of the grandparents (at 50% risk of The role of DRA in colorectal tumorigenesis is carrying the genetic defect) had colorectal cancer (SIR unclear. The initial observation that the gene is down- 2.1, 95% CI 0.6 ± 5.3). Thus seven cases of intestinal regulated in intestinal neoplasia suggested that it might cancer were observed in the CLD families (SIR 3.4, function as a tumor suppressor (Schweinfest et al., 95% CI 1.4 ± 7.0, P50.05). There was no signi®cant 1993). The alternative hypothesis is that the loss of overall increase in the frequency of cancer in the DRA expression in intestinal tumors occurs merely as kindreds (SIR 1.1, 95% CI 0.7 ± 1.6). If the CLD a consequence of dedierentation (HoÈ glund et al., parent whose cancer status (two intestinal tumors) was 1996). Mechanisms connecting DRA and tumor known to us beforehand was excluded from the study, development have been speculated on previously. then the increase in intestinal cancer risk in the CLD Sulfation is necessary for the normal synthesis of families would not remain signi®cant (SIR 2.4, 95% CI proteoglycans. Defective sulfate transport might 0.8 ± 5.6). disturb this process enough to alter the eect of Mutation in DRA and intestinal cancer A Hemminki et al 683 growth hormones, since proteoglycans have an this study. Further follow-up will more accurately important role in their binding and modulation of determine their risk of malignancy. The intestinal activity (Ruoslahti and Yamaguchi, 1991). cancer risk of the CLD parents was higher than that Proteoglycans have been shown to have decreased of the grandparents. Only 50% of the grandparents sulfation in tumors of the liver and kidney (Nakamura have the DRA defect whereas all parents are obligate and Kojima, 1981; Lapis et al., 1990). Furthermore, the carriers. normal sulfomucin secretion of the colonic mucosa is No individuals with the DV317 mutation of DRA reverted to a lower sulfate content sialomucous type were found outside the CLD high frequency area. following malignant transformation, apparently as an Within the region, the carrier frequency of DV317 was early event (Filipe and Branfoot, 1974; Shamsuddin et 4/416 (1.0%) in healthy control individuals, a result in al., 1981; Yamori et al., 1987). In hepatoma cell lines, accordance with previously reported numbers of 3/252 transport of inorganic anions including sulfate is (1.2%) (HoÈ glund et al., 1996). A similar percentage of greatly reduced (von Dippe and Levy, 1982), and DV317 heterozygotes was found in individuals with hyposulfated proteoglycans have a lowered anity to benign colorectal tumors (2/198, 1.0%). The DV317 adhesion molecules of surrounding cells (Robinson et mutation carriership in patients with colorectal cancer al., 1984). Altered intercellular contacts and transmis- was slightly, but not signi®cantly higher, 1.4% (4/287). sion may contribute to invasion and metastasis These ®gures suggest that the impact of DV317 on characteristics of malignant cell populations. This cancer susceptibility is not prominent, though a low evidence suggests that sulfation mechanisms may play level of contribution is possible. a role in the carcinogenesis of various tumor types. In this study seven patients with colorectal cancer A mechanism linking DRA defects and the risk of were directly demonstrated to have germline DRA intestinal cancer remains to be demonstrated. A similar defects. The data on these patients' lesions was situation is currently seen in the more obvious scrutinized for possible characteristic features. Six of association between digestive tract cancer and defects the lesions (86%) were located in the rectum or in the CFTR gene, encoding a chloride transporter sigmoid colon (Table 2). All three cases of rectal (Neglia et al., 1995). In a recent article the role in cancer in the CLD families occurred at a young age cancer predisposition of diering extracellular ATP (39, 50 and 53 years) (Table 1). levels in dierent organ systems of CFTR homozygous The number of intestinal cancers in CLD families and heterozygous mice was proposed (Abraham et al., was found signi®cantly increased (SIR 3.4, 95% CI 1996). 1.4 ± 7.0, P50.05), but only if a CLD parent whose Loss of the wild type DRA allele was not detected in two intestinal cancers were known to us beforehand the tumors. This is not surprising considering that was included to the data set. The cancer registration CLD patients are homozygous for the defect, and system in Finland is virtually complete, and the thrive without prominent cancer predisposition. In the computerized record linkage procedure precise (Puk- past CLD was often lethal, which explains the young kala, 1992). Therefore, technical aspects are unlikely to age (all below 32 years) of the patients available for cause bias in the results. Almost all CLD kindreds originated from Eastern Finland, whereas the average cancer incidence of the whole Finnish population was utilized to calculate the SIR ®gures of this report. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Eastern Finland is a low frequency area for colorectal cancer (Pukkala et al., 1987). Thus rather under- than overestimation of the true cancer risk SIR in CLD families is likely. An incidental observation that small cell lung 81 bp cancers had occurred frequently in the studied CLD families was made (SIR 7.1, 95% CI 1.9 ± 18.1, 78 bp

Table 2 Patients with DRA germline mutations and colorectal carcinoma b c Figure 1 Screening for DV317. The arrow indicates the 3 base Site of CLD family Histology Dukes pair deletion observed in a normal tissue sample derived from one Patienta Sex Age cancer history of the unselected colorectal cancer patients (lane 13). Each lane 1 F 79 ascending colon none adenoca. B (1 ± 20) represents one patient. DNA samples from 728 unselected 2 F 81 rectum none adenoca. B patients with colorectal neoplasia (510 carcinomas, 218 adeno- 3 M 71 rectum none adenoca. B mas) were available to us. These samples emanated from an 4 M 64 rectum none adenoca. A ongoing project in which the frequency of mismatch repair gene 5 M 39 rectum father of patient adenoca. A±B mutations in Finland is being studied. 485 of these (287 6 M 53 rectum father of patient adenoca. D carcinomas, 198 adenomas) were derived from individuals living 7 M 75 sigmoid grandfather of adenoca. A±B in the high CLD frequency area in Eastern Finland (HoÈ glund et colon patient al., 1995). The mutation analysis was performed using normal tissue DNA as described previously (HoÈ glund et al., 1996). Features of the patients with colorectal cancer and the DRA Brie¯y: genomic primers were used to amplify the deletion region, mutation in their germline. All patients are heterozygous for the and PCR products were electrophoresed in a 6% polyacrylamide mutation. a1 ± 4 are patients with sporadic colorectal carcinomas, 5 ± gel to reveal the presence or absence of the 3 base pair deletion. If 7 represent CLD family members. bAdenoca. is an abbreviation for DV317 was detected, loss of the wild type allele was searched for adenocarcinoma. cDukes refers to the Dukes classi®cation of in the tumor tissue when available. 416 anonymous randomly colorectal cancers according to degree of invasion of surrounding chosen blood donors from the high CLD frequency area served as tissues and metastasis. DNA was not available from all patients controls. Mutation status was determined as shown above reported in Table 1 Mutation in DRA and intestinal cancer AHemminkiet al 684 P50.01). In the absence of DRA expression in the mutations in dierent populations, and by continuing lung tissue (Schweinfest et al., 1993; Hemminki, molecular studies on the possible role of DRA in unpublished), the association between DRA defects tumorigenesis. and small cell lung cancer remains highly speculative, although DRA germline mutations may have some systemic eects. All the CLD patients and relatives were identi®ed during previous works that were not related to cancer Acknowledgements research. This makes a selection bias favouring We thank Dr Albert de la Chapelle for support, Drs Juha individuals with cancer in the CLD cohort unlikely. Kere and Christer Holmberg for sharing unpublished data, One may argue, whether the family of the individual and Dr Anu Moisio for access to control samples. The with two intestinal cancers (known to us before the unselected tumor samples were derived from a study conducted by the Finnish HNPCC Clinical Consortium, present work) should have been excluded from the whose members are: Jukka-Pekka Mecklin, Heikki JaÈ rvi- study. However, this family was identi®ed as a CLD nen, Heikki Ahtola, Matti Eskelinen, Niilo HaÈ rkoÈ nen, kindred more than 25 years ago; before any of the Risto Julkunen, Eero Kangas, Seppo Ojala, Jukka tumors were diagnosed (Norio et al., 1971). As Tulikoura, and Erkki Valkamo. We thank Minna Veini, individuals that were not alive after year 1966 were Marilotta Turunen, Siv Lindroos, Sinikka Lindh, Tuula excluded, some grandparents with cancer might have Lehtinen, and Kirsi Hopponen for technical assistance. remained unidenti®ed. If the relative risk is at the This study was aided by grants from the Academy of highest in young age groups, exclusion of follow-up Finland, the Federation of Finnish Insurance Companies, before 1967 leads to underestimation of the true risk. the Foundation for Paediatric Research, the Ulla Hjelt At present DRA does not ful®ll the de®nition of a Fund, the Finnish Cancer Society, the Sigrid Juselius Foundation, the Finnish Medical Society Duodecim, the tumor suppressor gene (Haber and Harlow, 1997). Ida Montin Foundation, the Jalmari and Rauha Ahokas Further studies are needed to clarify whether DRA Foundation, the 350th Anniversary Foundation of the mutations contribute to hereditary cancer predisposi- University of Helsinki, the National Cancer Institute tion, and to sporadic tumor initiation and progression. (CA67941), and the Helsinki University Central Hospital. These questions can be approached by studying the Part of the study was conducted at the FolkhaÈ lsan Institute cancer risk of CLD families with dierent DRA of Genetics.

References

Abraham EH, Vos P, Kahn J, Grubman SA, Jeerson DM, Norio R, Perheentupa J, Launiala K and Hallman N. (1971). Ding I and Okunie P. (1996). Nature Med., 2, 593 ± 596. Clin. Genet., 2, 182 ± 192. Filipe MI and Branfoot AC. (1974). Cancer, 34, 282 ± 290. Pukkala E, Gustavsson N and Teppo L. (1987). Atlas of Haber D and Harlow E. (1997). Nature Genet., 16, 320 ± 322. cancer incidence in Finland 1953 ± 82. Cancer society of HaÈ stbackaJ,delaChapelleA,MahtaniMM,ClinesG, Finland publication No. 37: Helsinki. Reeve-Daly M-P, Daly M, Hamilton BA, Kusumi K, Pukkala E. (1992). Use of Record Linkage in Small-area Trivedi B, Weaver A, Coloma A, Lovett M, Buckler A, Studies: Geographical and Environmental Epidemiology. Kaitila I and Lander ES. (1994). Cell, 78, 1073 ± 1087. Elliott P, Guzick J, English D, Stern R (eds.). Oxford HoÈ glund P, Haila S, Socha J, Tomaszewski L, Saarialho- university press: Oxford, pp. 125 ± 131. Kere U, Karjalainen-Lindsberg ML, Airola K, Holmberg Robinson J, Viti M and HoÈ oÈ k M. (1984). J. Cell. Biol., 98, C, de la Chapelle A and Kere J. (1996). Nature Genet., 14, 946 ± 953. 316 ± 319. Ruoslahti E and Yamaguchi Y. (1991). Cell, 64, 867 ± 869. HoÈ glund P, Holmberg C, de la Chapelle A and Kere J. Schweinfest CW, Henderson KW, Suster S, Kondoh N and (1994). Am. J. Hum. Genet., 55, 747 ± 752. Papas TS. (1993). Proc. Natl. Acad. Sci. USA, 90, 4166 ± HoÈ glund P, Sistonen P, Norio R, Holmberg C, Dimberg A, 4170. Gustavson KH, de la Chapelle A and Kere J. (1995). Am. Shamsuddin AKM, Weiss L, Phelps PC and Trump BF. J. Hum. Genet., 57, 95 ± 102. (1981). J. Natl. Cancer Inst., 66, 413 ± 419. KylloÈ nen LEJ, Teppo L and Lehtonen M. (1987). Ann. Chir. Silberg DG, Wang W, Moseley RH and Traber PG. (1995). Gynaecol., 76, 185 ± 190. J. Biol. Chem., 270, 11897 ± 11902. Lapis K, Kavalsky I, Jeney A, Pogany G, Molnar G, Taguchi T, Testa JR, Papas TS and Schweinfest C. (1994). Repassy D, Szecseny A and Karacsonyi S. (1990). Tokai J. Genomics, 20, 146 ± 147. Exp. Clin. Med., 15, 155 ± 165. Teppo L, Pukkala E and Lehtonen M. (1994). Acta Oncol., McKusick VA. (1994). Mendelian Inheritance in Man: a 33, 365 ± 369. catalog of human genes and genetic disorders, 11th ed. von Dippe P and Levy D. (1982). J. Biol. Chem., 257, 4381 ± Johns Hopkins University Press: Baltimore and London. 4385. Nakamura N and Kojima J. (1981). Cancer Res., 41, 278 ± Yamori T, Kimura H, Stewart K, Ota DM, Cleary KR and 283. Irimura T. (1987). Cancer Res., 47, 2741 ± 2747. Neglia JP, FitzSimmons SC, Maisonneuve P, SchoÈ ni MH, SchoÈ ni-Aolter F, Corey M, Lowenfelts AB and the Cystic ®brosis and cancer study group. (1995). N. Engl. J. Med., 332, 494 ± 499.