(1998) 17, 2143 ± 2148 ã 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc SHORT REPORT Conservation of function and primary structure in the BRCA1-associated RING domain (BARD1)

Teck-Choon Ayi1, Julia Tsou Tsan1, Larn-Yuan Hwang1, Anne M Bowcock2 and Richard Baer1

1Department of Microbiology and 2Department of Pediatrics and the McDermott Center for Human Growth and Development, University of Texas Southwestern Medical Center, 6000 Harry Hines Boulevard, Dallas, Texas 75235, USA;

The BRCA1 encodes a tumor suppressor that has within discrete nuclear foci (the `BRCA1 nuclear dots') been implicated in hereditary forms of breast and ovarian together with HsRad51, a mammalian homolog of the cancer. During S phase of the cell cycle, BRCA1 bacterial recA protein. Since recA and its eukaryotic polypeptides are found in discrete nuclear bodies counterparts have been implicated in DNA repair, it (`BRCA1 nuclear dots') together with HsRad51, a was proposed that BRCA1 normally serves to maintain human homolog of the E. coli recA protein, and BARD1, genomic stability and that the ability of BRCA1 to a protein that interacts with BRCA1 to form a stable suppress tumor formation is a re¯ection of its role in heterodimer. BARD1 is structurally similar to BRCA1 in DNA repair (Scully et al., 1997a,b). that both molecules harbor an amino-terminal RING BARD1 is likely to play a critical role in BRCA1- domain and two carboxy-terminal BRCT domains. Here mediated tumor suppression. BARD1 and BRCA1 we describe the sequence and expression have a similar con®guration of amino acid motifs, with pattern of murine Bard1. A comparison of the mouse and a RING domain at the amino-terminus and two human sequences reveals that the recognizable protein tandem copies of the BRCT domain at the carboxy- motifs of BARD1 are well conserved, including the terminus (Wu et al., 1996). In vivo, these exist RING domain, the three tandem repeats, and, to as a heteromeric protein complex, the formation of a lesser extent, the two BRCT domains. However, the which is completely abolished by tumor-associated remaining sequences of BARD1 display a markedly missense mutations in the RING domain of BRCA1 lower degree of phylogenetic conservation, comparable to (Wu et al., 1996). Since germline carriers of these those reported for BRCA1 and BRCA2. Moreover, mutations have a pronounced susceptibility to breast murine Bard1 retains the ability to associate in vivo with and , the interaction between BARD1 BRCA1, and its expression pattern in adult mice mirrors and BRCA1 may be essential for the tumor suppres- that of Brca1, with elevated levels of RNA transcripts sion activity ascribed to BRCA1. Indeed, rare found in the testes and spleen. These data suggest that alterations of the BARD1 gene have been uncovered BRCA1 and BARD1 have co-evolved to participate in a in primary tumors of breast, ovarian, and uterine common pathway of tumor suppression. tissues, suggesting that BARD1 mutations may themselves be contributing factors in the development Keywords: BRCA1; ; RING domain; of some human carcinomas (Hoa Thai et al., 1998). BRCT domain; tumor suppressor Recently, immunostaining experiments have shown that BARD1, BRCA1 and HsRad51 polypeptides colocalize to the same nuclear bodies ± the `BRCA1 nuclear dots' ± during S phase of the cell cycle (Jin et Almost half the families a‚icted with hereditary breast al., 1997; Scully et al., 1997a,b). Moreover, when S cancer harbor germline mutations of the BRCA1 gene phase cells sustain DNA damage, the BRCA1 nuclear (Hall et al., 1990; Futreal et al., 1994; Miki et al., dots disperse (Scully et al., 1997a; Thomas et al., 1997) 1994). Since neoplastic development in BRCA1 and, in at least some cells, BARD1-, BRCA1- and mutation carriers is accompanied by loss or inactiva- HsRad51-staining reappears in distinct nuclear struc- tion of the wildtype allele, the BRCA1 protein is likely tures that represent sites of damaged, replicating DNA to function as a tumor suppressor (Smith et al., 1992). (Scully et al., 1997a). Taken together, these observa- Although the biochemical properties of BRCA1 are tions suggest that BARD1 and BRCA1 are function- poorly understood, recent studies have shown that its ally-related proteins that, in association with HsRad51, subcellular distribution is cell cycle-dependent (Chen et participate in the cellular response to DNA damage. al., 1995, 1996; Scully et al., 1996, 1997a,b; Thomas et Recent studies have uncovered striking parallels al., 1996; Ru€ner and Verma, 1997; Thakur et al., between the properties of BRCA1 and those of 1997; Wilson et al., 1997). BRCA1 polypeptides are BRCA2, another tumor suppressor that is also di€usely distributed in the nuclei of resting cells and inactivated in many cases of hereditary breast cancer G1 proliferating cells. However, when proliferating (Wooster et al., 1994, 1995; Tavtigian et al., 1996). cells enter S phase, these polypeptides accumulate Although the proteins encoded by these do not share appreciable , their expression patterns during cell cycle progression are almost indistinguishable (Rajan et al., 1996, 1997; Spillman and Bowcock, 1996; Vaughn et al., 1996; Bertwistle et Correspondence: R Baer al., 1997; Wang et al., 1997). Furthermore, BRCA2 Received 2 March 1998; revised 11 May 1998; accepted 11 May 1998 polypeptides have been shown to associate in vivo with The BRCA1-associated RING domain (BARD1) protein T-C Ayi et al 2144 HsRad51 (Mizuta et al., 1997; Sharan et al., 1997; An interesting feature of BRCA1 and BRCA2 is Wong et al., 1997; Chen et al., 1998; Katagiri et al., their poor degree of evolutionary conservation. For 1998), the same protein that localizes to the BRCA1- example, the mouse and human counterparts of these staining nuclear dots of S phase cells (Scully et al., proteins share less than 60% amino acid identity, a 1997a,b). Although the subcellular distribution of level of phylogenetic conservation that is remarkably BRCA2 polypeptides has not been evaluated by direct low ± especially when compared with those of other immunostaining, Western analysis of subcellular known tumor suppressors (Abel et al., 1995; Bennett fractions shows that BRCA2 is predominantly a et al., 1995; Lane et al., 1995; Sharan et al., 1995; nuclear protein (Bertwistle et al., 1997). This, together Szabo et al., 1996; Connor et al., 1997b; McAllister et with its ability to interact with HsRad51, suggests that al., 1997; Sharan and Bradley, 1997). Moreover, yeast BRCA2 may also reside in the same nuclear structures orthologs of the BRCA1 and BRCA2 genes are not as BRCA1, HsRad51 and BARD1. found within the complete nucleotide sequence of the

Figure 1 An alignment of the amino acid sequences of mouse and human BARD1. The potential initiator methionines (residues 1 and 20 of murine Bard1) are marked with vertical arrows. Sequences corresponding to the RING motif (murine residues 40 ± 84), the three tandem ankyrin repeats (residues 415 ± 513), and the two tandem BRCT domains (residues 593 ± 765) are highlighted. The Q564H mutation of human BARD1 is also indicated (Hoa Thai et al., 1998). The sequences were aligned by introducing gaps to maximize amino acid sequence identity; the values for amino acid identity were then calculated by considering each gap as a single mismatch. The composite sequence of BARD1 cDNA has been deposited in the GenBank database (accession no. AF057157) The BRCA1-associated RING domain (BARD1) protein T-C Ayi et al 2145 S. cerevisiae genome. Yet, BRCA1 and BRCA2 domains of BARD1; namely, the amino-terminal associate in vivo with HsRad51 (Mizuta et al., 1997; RING motif (86.7% identity), the three internal Scully et al., 1997a,b; Sharan et al., 1997; Wong et al., ankyrin repeats (90.1%), and the two carboxy- 1997; Chen et al., 1998; Katagiri et al., 1998), a highly terminal BRCT domains (79.8%). conserved protein with obvious counterparts in not To evaluate the expression pattern of Bard1, only yeast (Rad51) but also in bacteria (RecA) (Karlin steady-state levels of Bard1 transcripts were deter- and Brocchieri, 1996). mined by Northern analysis of polyadenylated RNA Since BARD1 polypeptides are likely to function in derived from di€erent tissues of adult mice. As the same biochemical pathway as BRCA1, we sought illustrated in Figure 2a, Bard1 transcripts of approxi- to compare the evolutionary stability of BARD1 with mately 6.0 and 3.0 kilobases are especially prevalent in that of BRCA1, BRCA2 and HsRad51. Therefore, a RNA from spleen and testes (lanes 3 and 8, lEXlox cDNA library prepared from an 11-day-old respectively), but not from heart, brain, lung, liver, whole mouse embryo (Novagen, Inc.) was screened by skeletal muscle or kidney. The same expression hybridization with radiolabeled fragments of human pattern was obtained upon hybridization with pE18, BARD1 cDNA. Two independent hybridizing clones a cDNA probe containing codons 8 ± 301 of mouse were isolated (mB-12ex and mB-22ex), each of which Brca1 (Bennett et al., 1995); in accord with previous contained sequences encoding a carboxy-terminal reports (Abel et al., 1995; Bennett et al., 1995; Lane et segment of the mouse Bard1 polypeptide (amino acid al., 1995; Zabludo€ et al., 1996; Blackshear et al., residues 405 ± 765 and 87 ± 765, respectively). A cDNA 1998), this probe detects a major polyadenylated RNA fragment encoding the full-length protein (mB-2-2; species of 7.5 kilobases in the spleen and testes of residues 1 ± 765) was then obtained by screening a adult mice (Figure 2b). The abundance of Bard1 and mouse acinar cell cDNA library with murine sequences Brca1 transcripts in spleen and testes is especially from the mB-22ex insert. The alignment in Figure 1 striking given that these samples displayed the lowest shows that the mouse and human orthologs of BARD1 levels of hybridization with the GAPDH control share *70.4% amino acid identity. Higher levels of (Figure 2c; lanes 3 and 8). conservation are found within the known protein Human BARD1 was originally identi®ed on the basis of its ability to associate in vivo with BRCA1 (Wu et al., 1996). A minimal sequence required for the interaction has been localized to residues 26 ± 142, a segment of human BARD1 that includes the RING domain (Wu et al., 1996). To determine whether this property of BARD1 has been maintained during mammalian evolution, a corresponding segment of murine Bard1 (residues 20 ± 134) was tested for interaction with BRCA1 using a mammalian version of the two-hybrid assay (Fields and Song, 1989; Dang

Figure 3 Mammalian two-hybrid analysis of the in vivo interaction of BRCA1 with the mouse and human counterparts of BARD1. Each culture of 293 cells was transiently co- transfected with the G5E1bLUC reporter plasmid and the two indicated expression vectors. The GAL4 expression vector encoded either the `parental' GAL4 DNA-binding domain Figure 2 The expression patterns of Bard1 and Brca1 in selected (pCMV-GAL4; denoted by `+' in the GAL4 column), the tissues of adult mice. A Northern ®lter of polyadenylated RNA GAL4-mBard hybrid (residues 20 ± 134 of murine Bard1), or the from di€erent mouse tissues (2 mg mRNA/lane) was hybridized GAL4-hBARD1 hybrid (residues 26 ± 142 of human BARD1). sequentially with radiolabeled probes of mouse Bard1 (a), mouse The VP16 expression vector encoded either the parental VP16 Brca1 (b) and human GAPDH (c) cDNAs according to the transactivation domain (pVP-FLAG; denoted by `+' in the VP16 manufacturer's instructions (Clontech). The probes were com- column) or a VP16-BRCA1 hybrid containing residues 1 ± 304 of prised of cDNA sequences from mouse Bard1, mouse Brca1 human BRCA1. Duplicate transfections were conducted for each [pE18; Bennett et al., 1995], and human glyceraldehyde-3- combination of expression plasmids and the normalized luciferase phosphate dehydrogenase (GAPDH), respectively. As shown, activities obtained from each transfection are illustrated. The elevated levels of Brca1 and Bard1 transcripts are found in spleen expression plasmid encoding GAL4-mBard was constructed by and testes (lanes 3 and 8, respectively), but not in heart (lane 1), inserting a 0.35 kilobasepair fragment of mouse Bard1 cDNA into brain (lane 2), lung (lane 4), liver (lane 5), skeletal muscle (lane 6) the pCMV-GAL4 expression vector (Wu et al., 1996). The GAL4- or kidney (lane 7). The mobilities and sizes (in kilobases) of the hBARD1 and VP16-BRCA1 expression plasmids and the electrophoretic markers are indicated on the left of each methods for mammalian two-hybrid analysis have been described autoradiogram (Wu et al., 1996; Tsan et al., 1997) The BRCA1-associated RING domain (BARD1) protein T-C Ayi et al 2146 et al., 1991; Tsan et al., 1997). Therefore, cDNA during mammalian evolution, consistent with the sequences were inserted into pCMV-GAL4 to generate notion that these polypeptides function as components an expression vector encoding GAL4-mBard1, a of the same heteromeric complex. hybrid protein containing the DNA-binding domain If the formation of BRCA1/BARD1 heterodimers is of the yeast GAL4p transcription factor fused to required for BRCA1-mediated tumor suppression, then residues 20 ± 134 of murine Bard1. As illustrated in defects in BARD1 may also promote cancer formation, Figure 3, transfection of 293 embryonal kidney cells especially within tissues that are a€ected by loss of with this vector did not induce expression of a BRCA1, such as breast and ovarian epithelia. Upon GAL4p-responsive reporter plasmid, G5E1bLUC screening a panel of seemingly sporadic breast, ovarian (lane 4). Likewise, G5E1bLUC activation did not and endometrial carcinomas, we recently identi®ed occur upon transfection of these cells with a vector three missense mutations in the BARD1 gene: Q564H, encoding VP16-BR304, a hybrid polypeptide com- V695L and S761N (Hoa Thai et al., 1998). Each of prised of the VP16 transactivation domain fused to these mutations appeared to be tumor-associated since the amino-terminal 304 residues of human BRCA1 it was not detected in more that 300 controls from the (lane 1). As expected, G5E1bLUC was strongly general population and, in at least two of the three induced by co-expression of VP16-BR304 with cases (Q564H and S761N), tumor formation was GAL4-hBARD, a hybrid containing the DNA- accompanied by loss of the wild type BARD1 allele. binding domain of pGAL4 fused to residues 26 ± 142 Although the V695L and S761N substitutions arose of human BARD1 (lane 5) (Wu et al., 1996). somatically during tumor development, Q564H was of Signi®cantly, however, G5E1bLUC induction was germline origin. Interestingly, the Q564H patient had equivalent, if not greater, in cells that co-express three primary carcinomas of the breast, ovary and VP16-BR304 with the GAL4-mBard hybrid (lane 3). endometrium, suggesting that BARD1 lesions may Thus, the ability to interact with BRCA1 is retained confer a predisposition to carcinoma development. It is in both the human and mouse orthologs of BARD1. noteworthy, therefore, that amino acid Q564 resides The same results were obtained when these segments within a sequence of ®ve residues (NTGQR) that is of BARD1 and BRCA1 were tested for interaction in conserved in the human and mouse BARD1 polypep- the yeast two-hybrid assay (data not shown), tides (Figure 1). indicating that the in vivo association of these The subnuclear co-localization of BRCA1, BARD1, proteins does not require additional mammalian HsRad51 and possibly BRCA2, as well as the direct factors. protein-protein interactions between BRCA1 and The mouse Brcal and Brca2 polypeptides each share BARD1 on one hand and BRCA2 and HsRad51 on less than 60% amino acid identity with their human the other, suggest that these polypeptides may be equivalents. In contrast, other known tumor suppres- subunits of the same multi-component protein com- sors maintain much higher levels of sequence plex. Recent studies also suggest that this complex may conservation, with mouse/human identities that range function in the cellular response to DNA damage from 78% () to 98% (NF1) (see Table 1 of (Connor et al., 1997a; Scully et al., 1997a,b; Patel et Connor et al., 1997b). The data provided here reveal al., 1998). As noted above, the primary sequences of an amino acid identity of 70.4% between the mouse BRCA1, BARD1 and BRCA2 are poorly conserved in and human orthologs of Bard1. As expected, sequence mammals (mouse/human identities of *57%, 70% and conservation is greater within the known protein 58%, respectively) and orthologs of these polypeptides motifs of BARD1, including the RING domain are conspicuously absent from at least one lower (86.7% identity), the three ankyrin repeats (90.1%) eukaryote: the budding yeast S. cerevisiae. In contrast, and the two BRCT domains (79.8%). Previous studies HsRad51 is stringently conserved in mammals (mouse/ have shown that sequence conservation is also human identity of 98.8%) and functional orthologs of elevated within the RING (98.0% identity) and HsRad51 have been found in all free-living species, BRCT (63.7%) sequences of BRCA1 (Abel et al., including prokaryotes (Karlin and Brocchieri, 1996). In 1995; Bennett et al., 1995; Lane et al., 1995; Sharan et keeping with its ancient roots, HsRad51 is essential for al., 1995; Szabo et al., 1996). DNA double-strand break repair, as well as for normal At ®rst glance it would appear that the primary DNA recombination during both meiosis and mitosis sequence of BARD1 is signi®cantly more conserved (reviewed recently in reference Cox, 1997). It is than that of BRCA1 (70.4% vs 57.0%, respectively). interesting, therefore, to consider the relationship However, a closer inspection reveals that the higher between HsRad51 and the poorly-conserved polypep- level of sequence conservation displayed by BARD1 is tides with which it associates in the BRCA1 nuclear primarily a re¯ection of its smaller size (777 residues of dots of S phase cells. Perhaps BRCA1, BRCA2 and BARD1 vs 1863 residues of BRCA1) and the fact that BARD1 emerged during the course of eukaryotic the de®ned amino acid domains ± such as the RING evolution to serve as additional regulatory factors to and BRCT motifs ± constitute a much greater exert ®ner control over an existing HsRad51 function, proportion of the full-length BARD1 polypeptide. such as DNA double-strand break repair. Alterna- Thus, when the known protein domains are excluded tively, these polypeptides may represent components of (i.e., the RING and BRCT sequences of BRCA1, and a novel pathway in the cellular response to DNA the RING, ankyrin and BRCT sequences of BARD1), damage that is absent in prokaryotes and lower exceptionally low levels of phylogenetic sequence eukaryotes. If so, then HsRad51, with its re®ned homology are observed in both BRCA1 (55.0% amino ability to catalyze DNA strand exchange, may have acid identity) and BARD1 (61.1% identity). These been co-opted to serve ± along with its other functions comparisons suggest that the primary sequences of ± in the BRCA1/2 pathway of tumor suppression. In BRCA1 and BARD1 have diverged at a similar rate any case, additional insights into the origins of The BRCA1-associated RING domain (BARD1) protein T-C Ayi et al 2147 hereditary breast cancer should emerge from biochem- Galvin Swift and Ray MacDonald for providing a cDNA ical studies of BRCA1, BRCA2 and their associated library of the 266-6 acinar cell line. This work was proteins. supported by grants CA76334 (R Baer) and CA60650 (AM Bowcock) from the National Cancer Institute. R Baer was the recipient of a Faculty Research Award from the Acknowledgements American Cancer Society (FRA-421). LY Hwang was We thank L Michelle Bennett for providing pE18, a cDNA supported in part by the Leukemia Association of North clone of mouse BRCA1. We are also very grateful to Central Texas.

References

AbelKJ,XuJ,YinGY,LyonsRH,MeislerMHandWeber Patel KJ, Yu VPCC, Lee H, Corcoran A, Thistlewaite FC, BL. (1995). Hum. Mol. Genet., 4, 2265 ± 2273. Evans MJ, Colledge WH, Friedman LS, Ponder BAJ and Bennett ML, Haugen-Strano A, Cochran C, Brownlee HA, Venkitaraman AR. (1998). Mol. Cell, 1, 347 ± 357. Fiedorek FTJ and Wiseman RW. (1995). Genomics, 29, Rajan JV, Marquis ST, Gardner HP and Chodosh LA. 576 ± 581. (1997). Developmental Biol., 184, 385 ± 401. Bertwistle D, Swift S, Marston NJ, Jackson LE, Crossland S, Rajan JV, Wang M, Marquis ST and Chodosh LA. (1996). Crompton MR, Marshall CJ and Ashworth A. (1997). Proc. Natl. Acad. Sci. USA, 93, 13078 ± 13083. Cancer Res., 57, 5485 ± 5488. Ru€ner H and Verma IM. (1997). Proc. Natl. Acad. Sci. Blackshear PE, Goldsworthy SM, Foley JF, McAllister KA, USA, 94, 7138 ± 7143. Bennett LM, Collins NK, Bunch DO, Brown P, Wiseman Scully R, Chen J, Ochs RL, Keegan K, Hoekstra M, RW and Davis BJ. (1998). Oncogene, 16, 61 ± 68. Feunteun J and Livingston DM. (1997a). Cell, 90, 425 ± Chen P-L, Chen C-F, Chen Y, Xiao J, Sharp ZD and Lee 435. W-H. (1998). Proc.Natl.Acad.Sci.USA,95, 5287 ± 5292. Scully R, Chen J, Plug A, Xiao Y, Weaver D, Feunteun J, Chen Y, Chen C-F, Riley DJ, Allred DC, Chen P-L, Ho€ Ashley T and Livingston DM. (1997b). Cell, 88, 265 ± 275. DV, Osborne CK and Lee W-H. (1995). Science, 270, Scully R, Ganesan S, Brown M, Caprio JAD, Cannistra SA, 789 ± 791. Feunteun J, Schnitt S and Livingston DM. (1996). Science, Chen Y, Chen P-L, Riley DJ, Lee W-H, Allred DC and 272, 123 ± 125. Osborne CK. (1996). Science, 272, 125 ± 126. Sharan SK and Bradley A. (1997). Genomics, 40, 234 ± 241. Connor F, Bertwistle D, Mee PJ, Ross GM, Swift S, Sharan SK, Morimatsu M, Albrecht U, Lim D-S, Regel E, Grigorieva E, Tybulewicz VLJ and Ashworth A. Dinh C, Sands A, Eichele G, Hasty P and Bradley A. (1997a). Nature Genet., 17, 423 ± 430. (1997). Nature, 386, 804 ± 810. Connor F, Smith A, Wooster R, Stratton M, Dixon A, Sharan SK, Wims M and Bradley A. (1995). Hum. Mol. Campbell E, Tait T-M, Freeman T and Ashworth A. Genet., 4, 2275 ± 2278. (1997b). Hum. Mol. Genet., 6, 291 ± 300. Smith SA, Easton DF, Evans DG and Ponder BA. (1992). Cox MM. (1997). Proc. Natl. Acad. Sci. USA, 94, 11764 ± Nature Genet., 2, 128 ± 131. 11766. Spillman MA and Bowcock AM. (1996). Oncogene, 13, Dang CV, Barrett J, Villa-Garcia M, Resar LMS, Kato GJ 1639 ± 1645. and Fearon ER. (1991). Mol. Cell. Biol., 11, 954 ± 962. Szabo CI, Wagner LA, Francisco LV, Roach JC, Argonza R, Fields S and Song O-k. (1989). Nature, 340, 245 ± 246. King M-C and Ostrander EA. (1996). Hum. Mol. Genet., Futreal PA, Liu Q, Shattuck-Eidens D, Cochran C, 5, 1289 ± 1298. Harshman K, Tavtigian S, Bennett LM, Haugen-Strano Tavtigian SV, Simard J, Rommens J, Couch F, Shattuck- A, Swensen J, Miki Y et al. (1994). Science, 266, 120 ± 122. Eidens D, Neuhausen S, Merajver S, Thorlacius S, Ot K, Hall JM, Lee MK, Newman B, Morrow JE, Anderson LA, Stoppa-Lyonnet D et al. (1996). Nature Genet., 12, 333 ± Huey B and King M-C. (1990). Science, 250, 1684 ± 1689. 337. Hoa Thai T, Du F, Tsan JT, Jin Y, Phung A, Spillman MA, Thakur S, Zhang HB, Peng Y, Le H, Carroll B, Ward T, Yao Massa HF, Muller CY, Ashfaq R, Mathis JM, Miller DS, J, Farid LM, Couch FJ, Wilson RB and Weber BL. (1997). Trask BJ, Baer R and Bowcock AM. (1998). Hum. Mol. Mol. Cell. Biol., 17, 444 ± 452. Genet., 7, 195 ± 202. Thomas JE, Smith M, Rubinfeld B, Gutowski M, Beckmann Jin Y, Xu XL, Yang M-CW, Wei F, Ayi T-C, Bowcock AM RP and Polakis P. (1996). J. Biol. Chem., 271, 28630 ± and Baer R. (1997). Proc. Natl. Acad. Sci. USA, 94, 28635. 12075 ± 12080. Thomas JE, Smith M, Tonkinson JL, Rubinfeld B and Karlin S and Brocchieri L. (1996). J. Bacteriol., 178, 1881 ± Polakis P. (1997). Cell Growth Di€er., 8, 801 ± 809. 1894. Tsan JT, Wang Z, Jin Y, Hwang L-Y, Bash RO and Baer R. Katagiri T, Saito H, Shinohara A, Ogawa H, Kamada N, (1997). Mammalian cells as hosts for two-hybrid studies of Nakamura Y and Miki Y. (1998). Genes protein-protein interaction. In: The Yeast Two-hybrid Cancer, 21, 217 ± 222. System, PL Bartel and S Fields, (eds.) Oxford: Oxford Lane TF, Deng C, Elson A, Lyu MS, Kozak CA and Leder P. University Press, 217 ± 232. (1995). Genes Dev., 9, 2712 ± 2722. Vaughn JP, Cirisano FD, Huper G, Berchuck A, Futreal PA, McAllister KA, Haugen-Strano A, Hagevik S, Brownlee Marks JR and Iglehart JD. (1996). Cancer Res., 56, 4590 ± HA, Collins NK, Futreal PA, Bennett LM and Wiseman 4594. RW. (1997). Cancer Res., 57, 3121 ± 3125. WangS-C,LinS-H,SuL-KandHungM-C.(1997). Miki Y, Swensen J, Shattuck-Eidens D, Futreal PA, Biochem. Biophys. Res. Comm., 234, 247 ± 251. Harshman K, Tavtigian S, Liu Q, Cochran C, Bennett Wilson CA, Payton MN, Elliot GS, Buaas FW, Cajulis EE, LM, Ding W et al. (1994). Science, 266, 66 ± 71. Grosshans D, Ramos L, Reese DM, Slamon DJ and Mizuta R, LaSalle JM, Cheng H-L, Shinohara A, Ogawa H, Calzone FJ. (1997). Oncogene, 14, 1 ± 16. Copeland N, Jenkins NA, Lalande M and Alt FW. (1997). Wong AKC, Pero R, Ormonde PA, Tavtigian SV and Bartel Proc. Natl. Acad. Sci. USA, 94, 6927 ± 6932. PL. (1997). J. Biol. Chem., 272, 31941 ± 31944. The BRCA1-associated RING domain (BARD1) protein T-C Ayi et al 2148 Wooster R, Bignell G, Lancaster J, Swift S, Seal S, Mangion Wu LC, Wang ZW, Tsan JT, Spillman MA, Phung A, Xu J, Collins N, Gregory S, Gumbs C, Micklem G et al. XL, Yang M-CW, Hwang L-Y, Bowcock AM and Baer R. (1995). Nature, 378, 789 ± 792. (1996). Nature Genet., 14, 430 ± 440. Wooster R, Neuhausen SL, Mangion J, Quirk Y, Ford D, Zabludo€ SD, Wright WW, Harshman K and Wold BJ. Collins N, Nguyen K, Seal S, Tran T, Averill D et al. (1996). Oncogene, 13, 649 ± 653. (1994). Science, 265, 2088 ± 2090.