The BRCA1 Gene and Its Protein Product:
Characterization, Therapeutic and
Diagnostic Implications
Sean V. Tavtigian, PhD; Alun Thomas, PhD; Thomas S. Frank, MD; and Mark H. Skolnick, PhD

Sean V. Tavtigian, PhD, is director of cancer research; Alun Thomas, PhD, is director of statistics and informatics; and Thomas S. Frank, MD, is vice president of education and medical director, of Myriad Generics, Inc, Salt Lake City.

Breast cancer is the most common cancer among American women; by age 70, an American woman’s lifetime risk of developing breast cancer is about 10%.¹ However, it has long been apparent that some families show autosomal dominant inheritance of increased risk for breast or breast and ovarian cancer. In 1990, Mary-Claire King and coworkers obtained genetic linkage for early-onset breast and ovarian cancer to chromosome 17q21² the gene predicted to reside at that location was named Breast Cancer 1 (BRCA 1). Molecular cloning of the BRCA I gene by a consortium led by Myriad Genetics in 1994,³ followed by cloning of the second breast cancer predisposition gene, BRCA2, in 1995^4,5 has opened new avenues toward understanding both the genetics and the underlying biochemistry of familial breast and ovarian cancer.

Cancer is a multifactorial disease involving interaction of environmental, hormonal, and dietary risks in addition to genetic predispositions. However, progression of a single cell from a normal to a neoplastic state always involves a series of genetic changes that alter either the regulation or the function of a variety of different genes. Such genes may play roles in a number of overlapping physiologic processes, including genome maintenance, cell cycle control, apoptosis, contact inhibition, invasion and metastasis, or angiogenesis. These cancer genes are often classified in 2 main categories, oncogenes and tumor suppressor genes. The distinction between these 2 categories is that tumor progression is promoted by over expression or gain of function in oncogenes but by nonexpression or loss of function in tumor suppressor genes. BRCA1 falls into the second group; it is a tumor suppressor gene.

The cellular functions of a number of oncogenes and tumor suppressor genes are quite well understood. For example, over-expression or gain-of-function mutations in the proto-oncogene HER2/neu, a member of the epidermal growth factor receptor family, constitutively activate a signaling pathway that promotes progression through the G1 phase of the cell cycle. Similarly, the normal function of the protein encoded by the proto-oncogene CYCLIN D1 is to form an enzymatically active complex with G1-specific cyclin-dependent kinases (CDKs). Overexpression or gain-of-function mutations in CYCLIN D1 promote cell cycle progression from G1 to S phase of the cell cycle. At the cellular level, oncogenes act in an autosomal-dominant fashion; 1 abnormal copy of 1 allele of a protooncogene is sufficient to promote tumor progression.

In contrast, the normal function of the protein encoded by the tumor suppressor gene p16 is to inhibit function of the cyclin D-CDK complexes. Thus, loss of p16 function, which occurs in roughly 50% of all human tumors, has a phenotype similar to overexpression of CYCLIN DI—it promotes cell cycle progression from G1 to S phase. The p53 tumor suppressor gene regulates cellular response to DNA damage and other physiologic stresses by playing roles in cell cycle checkpoint signaling and apoptosis. Loss of p53 function, which also occurs in roughly 50% of all human tumors, allows cells to proliferate despite genome damage that would normally either trigger cell cycle arrest, and hence allow repair, or trigger apoptosis.⁶ At the cellular level, tumor suppressor genes function in an autosomal-recessive fashion. In a single cell, loss of function of both alleles of a tumor suppressor gene is usually required to promote tumor progression.

Mutations of oncogenes or tumor suppressor genes can occur as somatic tumor initiation events, tumor progression events, or germ line cancer predispositions. Whereas somatic mutations are responsible for most p53 and p16 lesions in human tumors, rare germ line mutations of p53 cause LiFraumeni syndrome,⁷ and germ line mutations of p16 cause predisposition to melanoma.⁸ Germ line mutations of tumor suppressor genes manifest their effects in an autosomal-dominant fashion, because each cell in a mutation carrier starts with only 1 normal copy of the tumor suppressor gene. Within individual cells, somatic mutation of the remaining normal allele occurs frequently enough that the carrier experiences a markedly elevated lifetime cancer risk. To date, most highly penetrant cancer predispositions are thought to be caused by germ line mutations in tumor suppressor genes. However, the same phenomenon can occur with germ line mutations in oncogenes. For instance, rare germ line mutations in the RET tyrosine kinase predispose for endocrine neoplasias.^9,10

Upon its discovery, analysis of the sequence of BRCA1 did not provide any strong clues to its tumor suppressor function. The BRCA1 transcription unit contains 24 exons and spans 81,000 bases of genomic DNA.^3,11 The first exon is noncoding; translation initiates within the second exon. The canonical splice variant of the transcript encodes a protein of 1863 amino acids. Recognizable sequence motifs within this form of BRCA1 are a RING finger motif near the N-terminus,³ 2 potential nuclear localization signals within exon 11,¹² and a tandem repeat of a sequence element located near the C-terminus termed the BRCT domain¹³; this last sequence element was not defined until some 20 months after BRCA1 was found. As a whole, the protein is relatively hydrophilic, and the C-terminal segment of the protein is relatively acidic. Thus, at the time of its discovery, there were only 2 clues to BRCA1 function: its RING finger domain, which can function either as a nucleic acid-binding or a protein-binding domain, and its potential nuclear localization signals, which suggest that the encoded protein might reside in the nucleus. Further obfuscating matters was the characterization of another splice variant, Brca1-A11b, which lacks the majority of exon 11, including the nuclear localization signals.¹⁴ Sequencing of the BRCA1 orthologs of a number of nonhuman mammals has revealed the surprising observation that its sequence has been relatively poorly conserved during mammalian evolution. For instance, the 57% amino acid sequence identity observed between the human and mouse orthologs places the gene in only the 5th percentile for sequence conservation between these two species.¹⁵ However, the RING finger domain, the nuclear localization signals, and the BRCT repeats are all present in the mouse sequence, suggesting that they are functionally important.

In mice, the expression pattern of Brca1 has been thoroughly analyzed during embryonic development, during spermiogenesis, and in breast tissue during maturation, pregnancy, and regression.^16,17 The data show that Brca1 expression is generally associated with cellular proliferation during development, but the highest levels of Brca1 expression occur in proliferating tissues undergoing terminal differentiation. Examples of such tissue include breast terminal end buds as they differentiate into ductal epithelium and spermatocytes during and after meiosis.

Homozygous disruption of Brca1 is lethal during early embryonic development in mice, but the exact phenotype of the knockout depends on where in the gene the lesion was created. In knockouts specifically targeted to exon 11, Brca1 -I- embryos are slightly underdeveloped by embryonic day 8.5 and typically die by embryonic day 10.5. Even so, at least subsets of embryonic cells are still proliferating at this point.^18,19 In a knockout designed to abrogate function of all known splice forms, the phenotype is more severe; the embryos fail gastrulation, primary cultures of the embryonic cells fail to proliferate, and the tumor suppressor p21 transcript is strongly overexpressed.²⁰ The underlying phenotype seems to be marked failure of cell proliferation in the developing embryo.

Because overexpression of p2l, which, like p 16, blocks cell cycle progression by inhibiting the kinase activity of cyclin-CDK complexes, could explain part of that phenotype, this observation has been pursued further. Both Brca1 : p21 and Brca1 : p53 double knockouts were made and analyzed.^21,22 Although both double knockouts are still embryonically lethal, the embryos usually gastrulate and then achieve variable levels of organogenesis. In fact, at embryonic day 9.5, some of the Brca1 : p21 double knockouts are morphologically almost indistinguishable from wild-type. While these data suggest a regulatory link between BRCA1 and p21, they also pose a dilemma: because it is a tumor suppressor, loss of BRCA1 should promote tumorigenesis; however, in mouse embryos, loss of Brca1 instead blocks cellular proliferation. It is also important to note that, while these studies suggest a regulatory link between BRCA1 and p21, they do not reveal whether activation of p21 is a direct or a relatively indirect consequence of BRCA1 ablation,

Eventually, both the phenotype of Brca1 knockout mice and the tumor suppressor function of the BRCA1 protein must be understood in terms of the protein’s role in the biochemical pathway(s) in which it acts. In an initial exploration, we and others have been using the yeast 2-hybrid system to identify genes that encode proteins that interact with various fragments of BRCA1 ^23-25 Three segments of BRCA1 have received particular attention: the RING finger domain, the segment of exon 11 that contains the 2 potential nuclear localization signals, and the C-terminal region containing the 2 BRCT domains.

Two-hybrid screens using the RING finger domain of BRCA1 as bait identified a novel interacting protein named BRCA1associated RING domain (BARD1). Interestingly, BARD 1 is structurally similar to BRCA1; it contains a RING finger near its N-terminus, 3 ankyrin repeats, a feature not shared with BRCA1, within its central segment, and 2 BRCT domains at its C-terminus.²³ Deleterious point mutations within the RING finger of BRCA1 disrupt the interaction between BRCA1 and BARD1 in vitro, providing some evidence that this protein-protein interaction is important to the tumor suppressor activity of BRCA1.

BARD 1 has been screened for mutations in a set of 50 breast, 58 ovarian, and 60 uterine primary tumors. Three rare mis-sense changes were found; 2 of these were somatic, and 1 was germ line; the germ line missense allele occurred in a patient who had 3 primary tumors. Both the germ line missense allele and 1 of the somatic mis-sense mutations were hemizygous in their respective tumors, suggesting that the wild-type allele had been lost. These findings are consistent with the notion that these 2 lesions are deleterious missense changes that occurred in a tumor suppressor gene.²⁶

Two-hybrid screens using the segment of exon 11 that contains the 2 potential nuclear localization signals identify hSRP1~ (importin-a) as an interacting protein.²⁵ Importin-a is a component of the nuclear receptor complex and as such could play a role in the nuclear localization of BRCA1. In vitro, engineered missense mutations of BRCA1’s candidate nuclear localization signals abrogate binding to importin-a, and expression constructs directing synthesis of BRCA1 proteins bearing these missense mutations demonstrate reduced or ablated nuclear localization.²⁵ Together, these data demonstrate that the nuclear localization signals are functionally important and supporn the hypothesis that the interaction between BRCA1 and importin-a is physiologically relevant.

Analysis of the C-terminal BRCT domains of BRCA1 may prove particularly informative. The sequence motif defining this domain was not identified until after BRCA1 was discovered.¹³ Shortly after the domain was defined, refined sequence search strategies identified BRCT domains in a set of about 40 nonorthologous genes.^27-29 Interestingly, most proteins in this set to which some function has been assigned are involved in DNA repair.

Several proteins that interact with the C-terminal domain of BRCA1 have been identified.²⁴ One of these, C-terminal Interacting Protein (CtIP) (1604B112 in reference 24), was first identified by virtue of its association with an adenovirus Ela protein binding protein, CtBP (Genbank accession #U72066). The interaction between BRCA1 and CtIP was confirmed by demonstrating both that the interaction occurs in yeast cells in both orientations and that BRCA1 can precipitate CtIP in vitro (A.K.C. Wong, unpublished data, 1998). Further analysis of this protein-protein interaction showed that a germ line truncation mutation removing the last 11 amino acids from the carboxy-terminus of BRCA1, 1853ter, abolishes not only the transcription-activation function of BRCA1 but also binding to CtIP. A sequence-based screen of a panel of 89 tumor cell line cDNAs for mutations in the CtIP coding region identified 5 missense variants. In the pancreatic carcinoma cell line BxPC3, a nonconservative lysine-to-glutamic acid change at codon 337 occurs on a background of either loss of heterozygosity or nonexpression of the wild-type allele (A.K.C. Wong, unpublished data, 1998). As with BARD1, the combination of deleterious alleles of BRCA1 that fail to interact with CtIP and a potentially deleterious missense change of CtIP in a tumor cell line is consistent with the notion that this protein-protein interaction is important to the tumor suppressor activity of BRCA1.

Beyond hints to the function of BRCA1 gleaned thus far from exploration of the biochemical pathway in which it acts, immunostaining and immunoprecipitation experiments have yielded a good deal of information about the subcellular localization and phosphorylation status of the BRCA1 protein. The canonical splice variant of BRCA1 encodes a 1863 amino acid nuclear phosphoprotein that migrates at 220 to 230 kd on PAGE gels.^12,14,30-32 While there was initial controversy over its localization and gel mobility, most of the conflicting data resulted from the use of expression constructs that in fact did not encode full-length BRCA1, notably by Jensen and associates,³³ and antibodies that significantly cross-reacted to other proteins.³¹

The phosphorylation status of BRCA1 is cell cycle regulated and peaks during late G1 or at the G1-S boundary.^32,34,35 During S phase of the cell cycle, BRCA I becomes localized to discrete nuclear dots. Exposure of S phase cells to discrete nuclear dots. Exposure of S phase cells to DNA-damage—inducing agents such as y- and UV radiation, hydroxyurea, or miromycin-C results in the disruption of this localization and redistribution of BRCA1 across the nucleus. Following similar treatments, BRCA1 becomes hyperphosphorylated; within the limits of experimental measurement, the nuclear redistribution and hyperphosphorylation occur simultaneously.³⁴’³⁵ One derailed study of the BRCA1 nuclear dots suggests these structures are actually cross-sectional views of cytoplasmic tubules that originate in the perinuclear region and extend into the nucleus.³⁶ Whether this is an accurate description of nuclear ultrastructure or a subtle artifact caused by a combination of fixarion conditions and antibody cross-reactivity remains an open question.

The initial observation of a nuclear dot localization pattern for BRCA1 in S-phase cells³⁰ was reminiscent of the localization of Rad5l, a protein involved in recombination and DNA repair.^37,38 In follow-up lo-calization studies, Sculley and associates carried our immunocolocalization experiments with fluorescent anti-BRCA1, antiSCP3 (a component of the meiotic synaptonemal complex³⁹), anti-Rad5l, and, as a follow-up to results from yeast 2-hybrid studies, anti-BARD1 antibodies. During prophase of meiosis I, the homologous chromosomes pair to form synaptonemal complexes, the structure in which meiotic recombination occurs. Staining of zygotene spermatocytes in prophase of meiosis I with anti-BRCA1 and anti-SCP3 antibodies revealed the striking result that BRCA1 was specifically associated with unsynapsed chromosomal segments and the junction at which synapsis was occurring. Double staining with anti-BRCA1 and anti-Rad5l revealed that the 2 proteins were tightly colocalized along these unsynapsed segments.⁴⁰ This staining pattern also correlates with the high levels of BRCA1 expression seen in zygotene and pachytene spermatocytes.¹⁷

In MCF7 breast tumor cells, double staining with anti-Brca1 and anti-Rad5l again revealed that the 2 proteins were colocalized during S phase,⁴⁰ this time in the previously described discrete nuclear dots.³⁰ Immunocoprecipitation experiments from lysates of S-phase MCF7 cells confirmed that Rad5 1 and BRCA1 were indeed physically associated with each other in vivo.⁴⁰ However, the association between these 2 proteins may be indirect because extensive efforts to detect a direct interaction between them in either yeast 2-hybrid screens or in vitro coprecipitation experiments have consistently generated negative results (P. Barrel, unpublished data, 1997). Finally, double staining of 5-phase MCF7 cells with anti-BRCA1 and anti-BARD 1 antibodies revealed BARD1, already identified as a BRCA 1-binding protein in the yeast 2-hybrid screens,²³ largely colocalized with BRCA1 in these nuclear dots. Further exploration of the composition, ultrastructure, and function of these structures is doubtless forthcoming.

Consequences of exogenous BRCA1 expression in cell lines

Given that the phenotype of mouse Brca1 knockouts—lethality due to a marked failure of cell proliferation during early embryogenesis—was unexpected for the loss of function of a tumor suppressor, it was important to investigate the consequences of forced expression of BRCA1 in normal and tumor cell lines. Unfortunately, such studies are complicated by three factors:

· Ideally, one would choose to express BRCA1 in a null background. However, cell cultures derived from Brca1-null mouse embryos fail to proliferate,²⁰ and no easily cultured human BRCA1-null cell lines have yet been reported.

· Technical difficulties hamper production of wild-type expression constructs encoding large proteins such as BRCA1. It is also rather laborious to confirm that the expression constructs remain wild-type after transfer into retroviral or adenoviral vectors or transfection into tissue culture cells. Yet, these confirmations are important because culture in mammalian cells may well select for mutations in tumor suppressor expression constructs.

· Overexpression of the canonical 1863 residue BRCA1 polypeptide appears to be toxic to mammalian cells in culture.¹⁴

An early report purported to show that retroviral-mediated expression of BRCA1 reduced the proliferation of breast and ovarian tumor cell lines in tissue culture without having a similar antiproliferative effect on other types of tumor cell lines.⁴¹ More impressively, retrovirus-mediated expression of BRCA1 in these breast and ovarian tumor cell lines also reduced their ability to form xenografts in nude mice.⁴¹ However, with consensus recognition that the canonical BRCA1 transcript primarily directs synthesis of a polypeptide of 220 kd reexamination of this work provides no evidence that either the wild-type or mutant expression constructs directed synthesis of the expected proteins. Consequently, the resultant data are somewhat suspect.

Fused to the Gal4-DNA binding domain, the C-terminal segment of BRCA1 has transcriptional transactivation activity in both yeast and mammalian cells.²⁴’⁴² Interestingly, several germline missense changes derived from breast and ovarian cancer-affected persons (affecteds) abrogate this activity.⁴² Together with its nuclear lo-calization and the presence of a RING finger, which can be a DNA binding domain, the intrinsic transcriptional transactivation activity in BRCA1 suggests that the gene may be a transcription factor. This suggestion is further supported by a recent demonstration that some BRCA 1 protein both copurifies with and coimmunoprecipitates with RNA polymerase II holoenzyme.⁴³

Two studies have used full-length, correct-sequence BRCA1 expression constructs to address this hypothesis. Ouchi and associates⁴⁴ cotransfected cells with BRCA1 and reporter constructs driven by synthetic promoters containing the binding sites of several transcription factors important in cell cycle progression. Of the 5 reporter constructs rested, only 1 was transactivated by BRCA1. This construct contained 2 copies of the p53 DNA binding site. Furthermore, transfections into a cell line expressing a temperature-sensitive p53 protein and transfections into a cell line derived from a p⁵³-null mouse demonstrated that BRCA1-mediated activation of this reporter also required the presence of functional p53.

In a related study by Somasundaram and associates,⁴⁵ cotransfection of BRCA 1 with a p21-promoter driven reporter construct into a series of human tumor cell lines provided evidence that BRCA1 can transactivate the p21-promoter. This regulatory link was investigated more thoroughly in 2 colon tumor cell lines, SW480 and HCT116. In p2l(+) SW480 and p2l(+) HCT116 cells, transfection of BRCA1 seemed to reduce the fraction of cells entering S phase; no such effect was observed in p21 (-) HCT116 cells. Finally, staining of the transfected SW480 cells with anti-p21 antibodies revealed increased levels of endogenous p21 protein following transfection with BRCA1.

One apparent conflict between these 2 BRCA1 transfection studies was that Ouchi’s group found that BRCA1 acts through promoter sequences at, or very close to, p53 binding sites; in contrast, Somasundaram’s team found that deletion past the p53 binding sites of the human p21 promoter did not reduce transactivation by BRCA1. Furthermore, the finding that BRCA1 overexprestion activates the p21 promoter in human tumor cell lines appears to conflict with the activation of p21 expression by ablation of Brca1 revealed by the mouse knockout experiments.

Current analyses of the protein encoded by BRCA1 seem to be consistent with 2 functions: a role in DNA recombination and/or repair and a role in transcriptional regulation. Taken together, the following 4 observarions are strongly suggestive of a role in DNA recombination and/or repair:

· The physical interaction between BRCA 1 and Rad5 1

· Their colocalization in spermarocytes along unsynapsed chromosomal segments and at the junction at which synapsis is occurring

· Their colocalization in S-phase nuclear dot structures

· Phosphorylation of BRCA1 and its con-comitant delocalization across, or relocalization within, the nucleus

Consistent with this proposed role, BRCA1-null tumors show a higher frequency of chromosomal amplifications and deletions than do sporadic tumors.⁴⁶ On the other hand, 4 other observations point to a role for BRCA1 in transcriptional regulation:

· The C-terminus of BRCA1, which is relatively acidic and contains the 2 BRCT domains, can act as a transcriptional activator in vivo.

· The architecture of BRCA1, with a RING finger near its N-terminus and acidic domain bearing transcriptional activation activity at its C-terminus, is suggestive of a transcription factor.

· In transfection experiments, BRCA1 seems to activate expression both from promoter constructs bearing p53 binding sires and from the p21 promoter.

· BRCA1 appears to associate with the RNA polymerase II holoenzyme complex.

The phenotype of the mouse Brca1 knockouts is ambiguous with respect to these 2 possibilities. Failure of DNA repair per se cannot be the cause of death in Brca1-null mouse embryos because Brca1: p21 and Brca1 : p53 double knockouts proceed further into development than simple Brca1 knockouts. However, loss of Brca1 function in the knockout mice could trigger p21 expression, as a consequence either of DNA damage or of an alteration in signal transduction pathways. Furthermore, a role in DNA recombination and/or repair and a role in transcriptional regulation need not be mutually exclusive. Brca1 could execute both functions independently, or it could have a signal transduction function that links DNA repair to transcriptional regulation.

BRCA1 was identified in 1994 by virtue of the detection of 4 independent chain-terminating mutations and 1 inferred regulatory mutation that cosegregated with breast cancer predisposition in 5 Utah breast cancer pedigrees.³ Since that time, numerous research centers around the world have contributed to characterization of the spectrum of mutations that occur in BRCA1, both as germline predispositions and as somatic tumor progression events. An organization called the Breast Cancer Information Core (BIC) has been set up as a repository for information about mutations and polymorphisms in breast cancer susceptibility genes. Accordingly, its web site keeps a thorough compilation of sequence variations observed in BRCA1.

Loss of function of both alleles of a tumor suppressor gene is usually required for expression of the associated tumorigenic phenotype. For some tumor suppressor genes, loss of function of the first allele is usually a somatic event; for others, a deleterious predisposing allele is usually inherited. The biologic correlate of this distinction appears to be that if loss of function of the second allele can contribute directly to clonal expansion of the cell null for that gene, then loss of function of the first allele will usually be a somatic event. On the other hand, if loss of function of the second allele does not contribute directly to clonal expansion but, for instance, accelerates the loss of tumor suppressors or activation of oncogenes that, as a secondary event, contribute to clonal expansion, then loss of function of the first allele will usually be an inherited event.⁴⁷

BRCA1 falls into the second category. BIC compilations of mutation screening data from lymphocyte DNA of breast or ovarian cancer affecteds versus their tumors reveal that in more than 99% of cases where loss of function of both alleles of BRCA1 can be documented in a tumor, 1 of the deleterious alleles is present in lymphocyte DNA. The mouse knockout phenotype, which demonstrates that loss of Brca1 causes a failure of cell proliferation, is also consistent with placement of BRCA1 in the second category of tumor suppressor genes.

A wide variety of deleterious alleles of BRCA1 has been reported. A promoter deletion was responsible for the inferred regulatory mutation detected in the very first BRCA1 mutation report.^3,48 Other large deletions, often created by unequal crossing over between Alu elements located within the introns of BRCA1, delete internal exons of the gene.^49,50 Small insertion or deletion mutations often create either frameshifts in the coding sequence (Figure 1) or splice errors, and nucleotide substitutions can create nonsense and missense mutations as well as splice errors (BIC data).³

Rather than clustering into mutational hotspots, BRCA1 mutations are distributed relatively smoothly across the gene. In the case of chain-terminating mutations, this datum only suggests that retention of function of sequences encoded by N-terminal segments of the gene does not enhance tumorigenesis; in other words, mutations that ablate C-terminal segments of the gene do not display any sort of dominant negative phenotype. In fact, tissue culture experiments provide some evidence that expression constructs encoding only the C-terminal segment of the protein might confer a dominant negative phenotype." However, mutations creating analogous BRCA1 alleles are not likely to occur in vivo.

For approximately the first year after BRCA1 was identified, we periodically updated a graph of the number of distinct mutations detected in BRCA1 versus the number of "independent" families analyzed (Figure 2). Extrapolation from the data suggested that the number of defined mutations, both nucleotide substitutions and insertion/deletion alterations, would continue to grow; therefore, a BRCA1 predisposition diagnostic rest would have to efficiently detect both novel insertion/ deletion mutations and novel nucleotide substitutions in the coding and proximal intronic sequences of the gene. Considerarions of sensitivity, specificity, automatability, and robustness led us to select and then implement polymerase chain reaction (PCR) from genomic DNA followed by double-stranded sequencing of those PCR products as the format for our BRCA1/ BRCA2 predisposition diagnostic test.²⁴

How frequent are the various classes of BRCA1 mutation in breast cancer affecteds? Shattuck-Eidens and associates⁵² used the diagnostic sequencing system to screen 798 women with breast cancer and/or ovarian cancer for mutations in BRCA1. The patients were also selected for some family history of breast or ovarian cancer or young onset of breast cancer.

Occurrences of mutations were correlated with the information on the patient’s personal cancer history, the history of breast and ovarian cancer in the patient’s family, and whether the patient was of Ashkenazi descent. The last factor was examined because of a known high prevalence of 2 deleterious mutations, l85delAG and 5382insC, among this ethnic group.

Deleterious mutations were detected in 102 women (12.8%). These consisted of 48 distinct mutations, of which 24 (50%) were unknown prior to the study. The mutations were scattered throughout the gene, and in contrast to a previous study,⁵³ no correlation was found between the position of a mutation and the cancer phenotype. By far the most common mutations found were l85delAG (seen 27 times) and 5382insC (seen 17 times). They account for all the deleterious mutations among the 71 Ashkenazi patients, seen 17 and 7 times, respectively, but were also the most common mutations in the non-Ashkenazi sample. Twenty-five polymorphisms, 2 of which were previously unreported, were observed in the sample. There were also 27 variants of unknown biologic significance, only 3 of which were previously known. These findings emphasize the potential pitfalls of screening methods that target a small set of known mutations, a particular class of mutation, or a subsection of the gene in the general population.

Perhaps the most difficult aspect of sequence-based BRCA1 predisposition diagnostic testing is interpretation of sequence variants other than chain-terminating mutations. These fall primarily into 2 categories: potential splice alterations and rare missense polymorphisms. From our data about allele frequencies of common polymorphisms in BRCA1, we estimate that about 60% of people carry at least 1 polymorphism within the expressed exons of the gene that can be used to distinguish between transcripts originating from each of their 2 alleles.⁵² For individuals who carry rare nucleotide substitutions near a splice junction or an insertion/deletion polymorphism in intron sequences proximal to a splice junction, the exonic polymorphisms can often be used to test for proper/improper splicing of both alleles by comparing the genotype of a genomic DNA PCR product that contains the polymorphism to the genotype of a cDNA PCR product that crosses the splice junction in question and should contain the polymorphism. There are 3 fundamental approaches to assessing the significance of the rare missense polymorphisms: family studies, analysis of allele frequencies, and functional analyses.

· Family studies. If a rare missense polymorphism segregates with the affecteds in a family, and if the presence of a nonexpressing allele has been excluded, then one could essentially use the polymorphism as a generic marker and measure the association between the polymorphism and breast cancer in the family. However, many families simply do not have an appropriate pedigree structure to carry our such an analysis. Furthermore, there is good reason to think that some predisposing missense mutations will have lower penetrance than terminator mutations, further complicating potential family studies. This analysis would have more power if the rare missense polymorphisms were examined as a class. If, across all the families in which they are studied, they show disproportionate segregation with the affecteds, then some fraction of them are predisposing mutations. Still, because of the unknown penetrance of any given allele, that fraction could be somewhat difficult to accurately calculate.

· Allele frequencies. What fraction of rare missense polymorphisms are actually predisposing mutations? Right now, there are not sufficient data to address this question. However, if the frequency of this class of polymorphism turns out to be higher in affecteds than in unaffecteds or, after appro-priate statistical adjustment, higher in affecteds without frameshifts than in affecteds with frameshifts, then an estimable fraction of these polymorphisms are predisposing. Of course, looking at these polymorphisms as a class provides little information about any particular missense polymorphism, but, because each such allele is by definition rare, a prohibitively large number of affecteds and unaffecteds would have to be studied in order to accumulate significant data for each.

· Functional analyses. The secondary structure of BRCA1 is likely a series of independently folding globular domains.^3,13 While terminator mutations eliminate all domains occurring downstream of them in the normal protein structure, predisposing missense mutations are likely to have an effect only on the structure of a single domain. So far, 3 genes encoding proteins that interact with BRCA1 have been identified, though that number is likely to rise. Eventually, all of these protein-protein interacting domains will be precisely mapped; at that point, it will become possible to build individual missense polymorphisms into BRCA1 expression constructs and to test whether they interfere with any of the protein-protein interactions that map to their general vicinity. The subset of missense polymorphisms that alter protein-protein interactions is likely to be enriched for predisposing missense mutations. However, this approach will almost certainly miss some predisposing missense mutations, and it is also likely that some of the protein-protein interactions in which BRCA1 engages are irrelevant to its role as a tumor suppressor protein. As other BRCA1 functional as says will likely have analogous drawbacks, it seems likely that a globally informative analysis of rare missense polymorphisms of unknown significance will require a combined generic and biochemical approach.

Given the ability to test for mutations in BRCA1, it becomes possible to estimate the risks associated with carrier status. Two recent studies have examined this issue, 1 focusing on the risks to carriers from families with evidence of linkage to BRCA1,⁵⁴ the other focusing on the risks of carrying 1 of 2 specific frameshift mutations in BRCA1, l85delAG and 5382insC.⁵⁵ The data are summarized in Figure 3. For breast cancer, the 2 studies are in good agreement; by age 60, the estimated cumulative risk of breast cancer for a BRCA1 carrier is about 54%. The 2 studies’ estimates of ovarian cancer risk are more divergent. Easton and associates⁵⁴ estimate a cumulative ovarian cancer risk by age 60 of about 30%; Struewing’s group,⁵⁵ closer to 17%.

Because several factors significantly correlate with the presence of a deleterious BRCA1 mutation, it is also possible to estimate the probability that a breast or ovarian cancer affected individual carries a deleterious BRCA1 allele. The age of first diagnosis of breast cancer is a predictor, with younger onset indicative of presence. In ascending order, the diagnoses most indicative of BRCA1 mutation were unilateral breast cancer, bilateral breast cancer, ovarian cancer, unilateral breast cancer and ovarian cancer, bilateral breast cancer and ovarian cancer. The stronger the family history of breast cancer, and even more so of ovarian cancer and breast and ovarian cancer in the same woman, the higher the probability of carrying a BRCA1 mutation.

An equation that estimates the probability that an individual carries a deleterious allele of BRCA1 was derived from these data by logistic regression analysis.⁵² Representative graphs of these data are presented in Figure 4. For each diagnostic situation, probability of being a carrier is presented for both the affected individual and a first-degree relative (sibling or offspring) of the affected individual. The gray zone in each graph represents estimated probabilities of carrying a deleterious BRCA1 mutation ranging from 0% to 10%, the interval above which current American Society of Clinical Oncology guidelines would recommend cancer predisposition resting.⁵⁶ Our data indicate that this threshold would be met by the following:

· A woman diagnosed with ovarian cancer before the age of 60 years and a relative with ovarian cancer, or an unaffected first-degree relative of that woman if her ovarian cancer occurred before the age of 50 years

· A woman with ovarian cancer before the age of 50 years regardless of family history, or an unaffected first-degree relative of that woman if her ovarian cancer occurred before the age of 40 years

· A woman with breast cancer before the age of 40 years and a relative with ovarian cancer

· A woman with breast cancer before the age of 35 years and 2 relatives with breast cancer

While probability of a positive test is one of the factors that helps determine the clinical utility of resting, the potential influence of results on patient management is a more important factor.

Ultimately, the value of predisposition diagnostic testing for BRCA1 must lie in reduction of morbidity or mortality for carriers of deleterious mutations in BRCA1. Understanding the biochemistry of tumor suppression by BRCA1, and phenomena surrounding the genesis of breast or ovarian tumors in carriers, will provide additional opportunities for enhanced clinical management of women with mutations in these genes.

The modern paradigm for drug development calls for identification of an enzymatic step that can be inhibited by a pharmaceutical agent, a protein-protein interaction that can be altered by a pharmaceutical agent, or a protein-signaling molecule interaction that can be appropriately mimicked or blocked by a pharmaceutical agent. In the area of anticancer pharmaceuticals, the farnesyl transferase inhibitors now entering clinical trials are probably the best example of a successful application of this paradigm. Development began with recognition that the ras protooncogene, which, in its activated form, constitutively stimulates cell proliferation, must be localized to the cell membrane in order to execute its role in signal transduction. With the discovery that this localization is accomplished by the addition of a lipid moiety, called a farnesyl group, to Ras by the enzyme farnesyl transferase, the stage was set for development of farnesyl transferase inhibitors as anticancer agents.^57,58

Unlike ras proto-oncogenes, BRCA1 encodes a tumor suppressor protein the function of which is absent in the breast or ovarian tumors of BRCA1 carriers. Thus, the question becomes, do we know enough about the biochemical pathways in which the BRCA1 protein functions to pick a point at which those pathways would be susceptible to interference or inhibition in a way that would either block tumor proliferation or stimulate apoptosis? If loss of BRCA1 causes a defect in recombination or repair, the answer is "no" Restoration of repair is a rather unlikely target for small molecule development. If loss of BRCA1 causes a defect in signal transduction, the answer is a more guarded "no," because although the signaling pathways in which BRCA1 might function are not well enough explored to pick an appropriate target, such a target may in fact exist.

The goal of gene therapy is to restore the normal function of a gene in such a way as to achieve a beneficial patient outcome. BRCA1 carriers have only 1 functional copy of the gene in their somatic cells instead of 2. Thus, in asymptomatic carriers, it would seem that if gene therapy could place a second functional copy of the gene in most or all of their mammary ductal epithelial cells and ovarian mesothelial cells, then their breast and ovarian cancer risks would be reduced. However, this goal is beyond the scope of present technology.

The breast and ovarian tumors of BRCA1 carriers lack functional BRCA1 protein. It has been suggested that replacement of BRCA 1 function in breast or ovarian tumors would be an effective strategy,⁴¹ but questions about the likely value of this approach have also been raised.⁵⁹ The primary defect in BRCA1-null tumors may well be a deficit in DNA recombination or repair and attendant genomic instability. Indeed, genomic instability has already been documented in BRCA 1-null tumors⁴⁶ and may well account for the notorious difficulty of culturing those tumors ex vivo. If this is in fact the case, then replacing BRCA1 function in these tumors could help stabilize their genomes, an outcome that would be without therapeutic value. However, if the primary defect in BRCA1-null rumors is a flaw in signal transduction related to the need for DNA repair, then a gene therapy approach might have the beneficial effect of restoring some aspect of growth control or perhaps drive the cells toward apoptosis.

Even before there is a pharmaceutical or a gene therapy intervention targeted direcrly at carriers, we have the opportunity to use this information for medical management of women at increased risk of breast and ovarian cancer. A task force organized by the National Human Genuine Research Institute recently issued guidelines to facilitate early detection of breast cancer in women who carry mutations in BRCA1 or BRCA2.⁶⁰ This task force recommended that such women commence yearly mammography and annual or semiannual clinician-administered breast examination beginning at age 25. However, 2 potential complications need to be weighed: (1) Mammography is less effective in premenopausal and perimenopausal women than in postmenopausal women; a more effective, possibly MRI-based, screening method is needed. Until better detection methods are available, monthly self-examinations and frequent physician-administered examinations offer the best chance of early breast tumor detection. (2) Because BRCA1 plays a role in DNA recombination or repair and definitely responds to UV- and y-irradiarion—induced DNA damage, the possibility remains that carriers may be hypersensitive to ionizing radiation; if this is indeed the case, then the increased likelihood of early detection through mammography may be at the expense of increased cancer risk.

Ovarian cancer, being harder to detect than breast cancer, is associated with greater mortality. From either the excess gonadotropin or the incessant ovulation theories of ovarian carcinogenesis, one might predict that the incidence of ovarian cancer could be lowered by long-term oral contraceptive (OC) use.^61,62 Numerous studies find this to be the case,⁶³ with one study finding a risk reduction of 1100 for each year of OC use.⁶⁴ If reduction of ovarian cancer risk with long-term OC use extends to BRCA1 carriers, early identification of carriers followed by long-term OC treatment could prove enormously beneficial. However, such a treatment must be evaluated with respect to the possible effects on breast cancer risk. In the absence of identified effective chemoprevention for ovarian cancer, increased surveillance or prophylactic oophorectomy after age 35 or when childbearing is completed remains an option.⁶⁰

The final area that needs to be addressed is the treatment of affected carriers. Easton and associates⁵⁴ found that after diagnosis of a first breast tumor, BRCA1 carriers face disproportionately high risks of either a second independent breast tumor or an ovarian tumor. Myriad Genetics’ BRCA1 and BRCA2 test results also strongly support this conclusion (T.S. Frank, unpublished data, 1998). Because treatment options for a second breast tumor are limited by the modality of treatment of a first tumor, affected women carriers and their oncologists may opt for aggressive surgical treatment of a first tumor as a prophylactic measure against subsequent tumors.