is a classic tumor suppressor gene in that loss of the wild-type allele (loss of heterozygosity, LOH) is required for tumorigenesis in germline mutation carriers. In sporadic breast tumors, allelic loss of BRCA1
is common 
. Knudson's model would have predicted that in at least a proportion of these cases, BRCA1
is inactivated by a somatically acquired mutation. Contradicting this hypothesis is the observation that somatic BRCA1 mutations are exceedingly rare in sporadic carcinomas. However, BRCA1
message and protein are often decreased in sporadic breast and ovarian cancers 
. In some cases, BRCA1 is down-regulated by aberrant methylation. Methylation of the BRCA1 promoter occurs in 11–14% 
of sporadic breast cancers and in 5–31% 
of ovarian cancers and is often associated with LOH 
. However, other mechanisms responsible for loss of BRCA1 in sporadic disease remain to be determined.
Alu-mediated aberrant homologous recombination contributes to loss of BRCA1 in a significant proportion of inherited disease. Importantly, this mechanism may explain some of the allelic loss of BRCA1 observed in sporadic disease. Thus we have taken a genomewide approach to identify genes that suppress Alu-mediated recombination in yeast with the knowledge that this screen would fail to reveal any human genes involved in suppressing Alu-mediated recombination for which no human homologs are present in yeast. A functional screen of the complete set of yeast deletion strains in the BY4742 background identified twelve strains with Alu-mediated recombination rates greater than 1.5 fold that of the wild-type strain (). To confirm that the 5-FOAR phenotype of the newly identified suppressors of Alu-mediated recombination was the result of deletion of the given open reading frames (ORFs), their counterparts in the BY4741 background were analyzed. Only 4 of the 12 strains were validated in BY4741, raising the possibility that the increased rate of Alu-mediated recombination in the other 8 strains was not related to deletion of the indicated gene. Of the 4 validated strains, mutation rates for arg3, tsa1, and rrm3 were significant when compared to the respective wild-type rate. While the mutation rate between oma1 in BY4741 and BY4742 was consistent, suggesting that this gene has a role in suppressing Alu-mediated recombination, the corresponding p-values were not significant when compared to the respective wild-type strain.
Deletion of TSA1, a thioredoxin peroxidase, has been shown to increase the rate of both spontaneous mutation as well as gross chromosomal rearrangement (GCR) 
. The relative rate of tsa1
-permissive GCR is similar to that for Alu
-mediated recombination (7 and 3.34 
, and this study, respectively). Together these results indicate the importance of this gene in preventing a broad spectrum of types of genomic instability. The human homologs of TSA1 are the four member of the peroxiredoxin (PRDX) family of antioxidant enzymes which reduce hydrogen peroxide and alkyl hydroperoxides. Homozygote Prdx1−/−
mice knockouts develop hemolytic anemia and several malignant cancers including epithelial and mesenchymal tumors such as hepatocellular carcinoma, fibrosarcoma, osteosarcoma, islet cell adenomas, and adenocarcinomas of the lung and breast 
. Heterozygote Prdx1+/−
mice also show increased frequency of hemolytic anemia and malignant cancer.
In contrast, ARG3, an ornithine carbamoyltransferase involved in the biosynthesis of arginine 
and its human homolog OTC (ornithine transcarbamylase) have not been priorly identified as having a role in maintaining genomic stability. In addition, this protein has not been associated with cancer.
The final human homolog validated in this screen, RRM3, was first identified as a suppressor of recombination in ribosomal DNA (rDNA) 
. S. cerevisiae
RRM3 and its paralog, PIF1, belong to the super family IB of 5′-to-3′ directed DNA helicases. The Pif1 family helicases are defined by seven highly conserved helicase signature motifs, three motifs that are shared with E. coli
RecD and in eukaryotes, and a highly conserved 21-residue Pif1 family signature sequence located between motifs II and III 
. Mouse and human PIF1 proteins immuno-precipitate with telomerase activity and TERT, the catalytic subunit of telomerase 
. There is also some data suggesting that, like yeast Pif1, human PIF1 may inhibit telomerase activity in vivo
and in vitro
To determine whether variation in the human homologs of yeast genes that suppress Alu-mediated recombination modify the effect of BRCA1 in mutation carriers, we determined the sequence of these genes in a series of BRCA1 mutation carriers who had breast cancer and/or ovarian cancer at particularly early ages. Of the eleven variants identified in BRCA1 mutation carriers, two were identified in controls, and eight of the remaining mutations were not predicted to be damaging by internal alignment programs. Thus, the variants identified in these candidate genes do not appear to contribute to particularly early onset of disease in BRCA1 mutations carriers.
Numerous genomewide studies have been conducted analyzing LOH in sporadic breast cancers to reveal foci of potential tumor suppressor genes. Recent studies have employed higher resolution array-based CGH (aCGH) showing the enormous complexity of breast cancer genomes. These studies have consistently reported the same large regions of loss (8p, 9p, 13q, 16q) 
; the number and identity of tumor suppressor genes that contribute to sporadic breast cancer remains largely unknown. Interestingly, high resolution mapping of regions of losses with frequencies of >30% included 1p32.1-p31.1, which contains OMA1
, 1p36.33-p34.2, which is very close to PRDX1
, 10q25.3-qtel, which contains PRDX3
, 15q21.3-q24.3, which contains PIF1
, and 11q14.3-qtel, which includes ATM
. In the present study, LOH analysis of the human homolog suppressors of Alu
-mediated recombination in 25 grade 3 invasive ductal carcinomas revealed LOH at 26 of 181 loci (14%), among informative cases. LOH frequencies among the chromosomal regions varied from 0% to 30% ( and Figure S2
). For all genes displaying LOH, the retained allele was sequenced to identify inactivating mutations; however, none were identified in this series.
A significant proportion of high-risk breast cancer families are not explained by mutations in known genes, indicating that still unidentified genes may explain cancer risk in these families. To determine if variation in the human homologs of the yeast mutator candidate genes contributes to increased breast cancer risk in high-risk families, a cohort of Ashkenazi Jewish probands was sequenced for each of the human homologs. Three unreported variants in candidate genes were identified in this population ( and Table S8
). Of these, only variant L319P in PIF1 was reported to be damaging by PolyPhen and not tolerated by SIFT. While this variant was not observed in 368 controls nor has it been reported in dbSNP or the 1000 Genomes Project, we identified it in 2 out of 844 breast cancer cases of Ashkenazi Jewish ancestry. The leucine at amino acid 319 is completely conserved within the PIF1 family of DNA helicases and is located within the putative Pif1family signature motif located between motifs II and III 
. Given that it is predicted to be within a helical domain (PredictProtein and PSIPRED), substitution of the five-membered chemical ring of proline from a linearly structured leucine likely disrupts protein structure. However, given the number of Ashkenazi Jewish controls evaluated for L319P we cannot exclude the possibility that this allele may be a rare PIF1 allele limited to this population.
Snow et al., reported that Pif1 (−/−) mice are viable at expected frequencies and displayed no visible abnormalities or increases tumor burden. These results seem to contradict those present here suggesting that loss of PIF1 function may contribute to breast carcinogenesis. However, for many genes it is well known that findings in mouse mutants cannot necessarily be extrapolated to humans. For example, early attempts to develop mouse models of BRCA1-linked breast cancer were unsuccessful (reviewed in 
). Early embryonic lethality precluded tumor development in Brca1 (−/−) mice. Surprisingly, conventional null or hypomorphic Brca1 alleles revealed lack of tumor formation in heterozygous mice. However, homozygous mice with certain hypomorphic Brca1 alleles can survive to adulthood and display an increased susceptibility to a range of tumors, including mammary carcinomas 
. Tumors can also be induced by conditional inactivation of Brca1 in breast epithelial cells through cre/loxP-mediated recombination 
. Inactivation of Brca1 alone in murine ovarian surface epithelium resulted in an increased accumulation of premalignant changes, but no tumor formation 
. Importantly, somatic loss of both Brca1 and p53 resulted in the rapid and efficient formation of highly proliferative, poorly differentiated estrogen receptor-negative mammary tumors that closely mimic human BRCA-mutated breast cancers with basal-like phenotypes 
suggesting that other genetic events contribute to tumorigenesis. Approximately 50% of familial breast cancer remains unresolved- that is disease cannot be explained by loss of function mutations in known breast cancer genes. Thus other genes are worthy of in-depth genomic analysis in unresolved families regardless of their associated mouse phenotype.
To determine if additional PIF1 variants impact breast cancer risk, we determined the complete PIF1 coding sequence in a series of 400 additional high risk breast cancer probands largely of European ancestry. Of the variants identified in this series, S223T, P357L, and R592C are potentially deleterious ( and ). In S. cerevisiae Rrm3
, threonine occupies the position corresponding to human S233 thus it is unlikely that this variant impacts human PIF1 function; however, the amino acid is immediately adjacent to conserved helicase motif I (a nucleotide binding motif also known as “the Walker A box”) and as such could affect ATPase activity. The human proline at position 357 is adjacent to motif III and completely conserved within the family. Thus it is possible the nonconservative amino acid change P357L contributes to protein destabilization. R592C is a relatively conserved amino acid position (R or Q in humans, mice, both yeast Pif helicases, and E. coli
RecD). While this substitution results in an amino acid with a smaller side chain, it is conservative in terms of hydrophilicity. However, in yeast, mutations in this region tend to disrupt helicase activity but not ATP binding or hydrolysis thereby impairing the ability of the protein to couple conformational changes caused by ATPase activity to DNA unwinding 
. Finally, PIF1 variant P109L, which was found in a BRCA1 mutation carrier with particularly early onset breast cancer as well as in 1 of 198 controls, is predicted to be deleterious. Although this proline resides over 100 residues upstream of motif I, it is completely conserved from yeast to humans.
The Pif1 family of 5′ to 3′ DNA helicases is conserved from yeasts to humans. While the Pif1 helicase function is dispensable in S. cerevisiae
and mouse, the S. pombe
Pfh1 is essential in both mitochondria and nuclei 
. The results shown here demonstrate that the pfh1-L430P
allele does not provide the essential activity of Pfh1 (). Since Pfh1-L430P is expressed, its failure to complement is not due to misfolding and degradation of the mutant protein (). Pfh1-L430P does not complement Pfh1 helicase activity in either the nucleus or the mitochondrial (). Therefore, we conclude the lethality of cells expressing Pfh1-L430P is due to loss of helicase function in both the nucleus and mitochondria.
Here we report the systematic analysis of the complete set of yeast gene deletion mutants to identify genes required for preventing Alu-mediated aberrant homologous recombination events, providing a global view of these nonessential genes in maintaining genome stability. We identified both previously known suppressors of chromosomal rearrangements as well as a number of novel genes. We provide genetic and functional evidence that a rare, loss of function variant in the helicase PIF1 may elevate breast cancer risk. Finally, although the primary aim of this research was to identify novel genes involved in genomic rearrangement at the BRCA1 locus, the genes identified in this screen may also contribute to chromosomal rearrangement at other loci. As such, they should be considered as candidate genes capable of facilitating cancer-inducing deletions, duplications, translocations, and splice variations in other tumor types.