Breast cancer is the most frequently diagnosed malignancy in women in the Western countries1 where estimated life-time risk of developing the disease is around 10%.2 Both non-genetic and genetic risk factors have been identified. Family history of breast cancer, age, previously diagnosed malignancy, nulliparity, young age at menarche, late menopause, high age at first birth, hormone replacement therapy and radiation exposure are among the currently known risk factors (reviewed in Ref. 3), a positive family history being the strongest. Familial aggregation is seen in 15–20% of breast cancer cases and about one fourth of those belong to families that can be defined as high risk breast cancer families.4 While varying in different populations, approximately half of high risk breast cancer families can be explained by inactivation of 1 of the 2 major breast cancer susceptibility genes, BRCA1 or BRCA2.5, 6, 7
Tumors from BRCA1 mutation carriers are of an overall higher histological grade, are frequently lacking estrogen receptor (ER) and progesterone receptor (PgR) expression and are more often aneuploid than sporadic tumors. Similar tendency has been observed in tumors from BRCA2 mutation carriers although these are in general ER positive.8, 9, 10, 11, 12 Tirkkonen et al. used metaphase comparative genomic hybridization (CGH) to define chromosomal alterations in BRCA1 and BRCA2 tumors.13 The study revealed a high frequency of overall DNA copy number changes in BRCA1 and BRCA2 tumors as compared to sporadic breast tumors, possibly reflecting the role of BRCA gene products in maintaining genome integrity (reviewed in Ref. 14). Furthermore, the specific patterns of chromosomal aberrations in BRCA1 and BRCA2 tumors indicate a relationship between genetic predisposition and tumor progression pathways.
CGH analysis revealed a particularly high prevalence of chromosome 5q deletions in BRCA1 tumors.13 As the resolution of metaphase CGH is limited, a more refined approach is needed to define with more accuracy the regions involved. In the current study we conducted loss of heterozygosity (LOH) analysis using microsatellite markers covering chromosome 5 in BRCA1, BRCA2 and sporadic breast tumors, and used a high-resolution tiling BAC array CGH platform to fine map 5q deletions in BRCA1 tumors. We confirm the very high imbalance frequency at specific chromosome regions and pinpoint regions and candidate genes that are affected by homozygous deletions in BRCA1 tumors.
Material and methods
DNA from 174 paired blood and breast tumor samples was subjected to analysis of LOH at chromosome 5. The sample set included tumors from Finnish, Icelandic and Swedish BRCA1 (n = 42) and BRCA2 (n = 67) mutation carriers (referred to as BRCA1 and BRCA2 tumors) as well as tumors from sporadic cases of breast cancer (sporadic, n = 65). The Swedish samples were collected at the Department of Oncology at the Lund University Hospital (19 BRCA1 and 22 sporadic breast tumors), the Finnish samples were collected at the Departments of Pathology, Obstetrics and Gynecology and Oncology, Helsinki University Hospital (14 BRCA1 and 17 BRCA2 tumors) and the Icelandic samples were collected at the Department of Pathology, Landspitali-University Hospital (9 BRCA1 and 50 BRCA2 tumors and 43 sporadic tumors). The samples were viewed by pathologists from each of the contributing institutions and the majority had a tumor cell content of at least 50%. However, since we did not use microdissected tumor tissue for analysis, the aberrations rates observed in the study could be an underestimate of the real situation due to masking of deletions by DNA from normal cells. Clinico-pathological information was obtained from pathological records at the contributing hospitals. The categorized clinico-pathological variables analyzed were ER status (>10 vs. ≤10 fmol/mg protein), PgR status (>25 vs. ≤25 fmol/mg protein), histological type (ductal or lobular), tumor size (<2 vs. ≥2 cm), age at diagnosis (<50 vs. ≥50 years), percentage of cells in S-phase (≤7% vs. >7%), ploidy (diploid or aneuploid) and grade (grade I vs. II and III). The study has been approved by the respective Ethical Committees in each country.
DNA extraction and PCR
DNA extraction from formalin-fixed paraffin embedded and fresh frozen tumor tissue was performed according to Smith et al.15 DNA was extracted from blood samples according to Miller et al.16 Fine mapping of chromosome 5 was performed using 26 microsatellite markers (MWG-Biotech AG) (Table I). PCR was performed in 15 μl reactions with 25 ng DNA, 2 mM MgCl2, 0.16 mM nucleotides (Amersham Pharmacia Biotech Inc), 0.32 μM each primer and 0.3 U Taq DNA polymerase together with the enclosed 10× buffer (MBI Fermentas). Reactions were performed as follows: 30 sec at 94°C, 45 sec at 55°C and 45 sec at 72°C for 35 cycles. The reactions were preceded with denaturation at 94°C for 3 min and followed by 10 min at 72°C.
Table I. The Frequency of Loss of Heterozygosity on Chromosome 5 in Tumors from Sporadic Patients and Patients with a Germ-Line Mutation in BRCA1 or BRCA2 and the Association between Loss of Heterozygosity at Respective Markers and Clinico-Pathological Variables
Clinico-pathological variables tested were estrogen receptor status, progesteron receptor status, histological type of tumor, tumor size, age at diagnosis, percentage of cells in S-phase, ploidy and tumor grade.
Age (0.02), grade (0.03)
ER (0.01), grade (0.04)
S-phase (0.03), grade (0.04)
S-phase (0.02), ER (0.01), PgR (0.02)
S-phase (0.04), PgR (0.02), grade (0.03)
S-phase (0.02), aneuploidy (0.005), grade (0.05)
Ductal tumor type (0.02)
S-phase (0.01), size (0.04)
The PCR products were separated on a 6.5% polyacrylamide gel (8 M urea) in 1× TBE buffer, and blotted to Hybond-N+ nylon transfer membrane (Amersham Pharmacia Biotech, UK). The DNA bands were visualized using ECL (Amersham Pharmacia Biotech, UK) according to the manufacturers protocol with the modification of employing HEPES buffer (40 mM K-HEPES, pH 7.2 with 1 mM CoCl2), instead of cacodylate, for tailing oligomers with terminal transferase for probe preparation.17
A Chi-square test with Yates correction was used to compare LOH frequencies between the 3 groups of breast tumors, as well as to evaluate the relationship between LOH and different clinico-pathological variables. Processing of the data was performed using SPSS 9.0. The level of significance was set at 0.05.
Microarrays with complete genome coverage were produced from the 32K BAC clone library (CHORI BACPAC Resources, http://bacpac.chori.org/genomicRearrays.php), which includes ∼2,500 clones from chromosome 5. DOP-PCR products were obtained from BAC DNA template and purified using filter based 96-wells plates (PALL), dried and resuspended in 50% DMSO to a concentration of 500–1,000 ng/μl. Arrays were printed on UltraGAPS slides (Corning) using a MicroGrid II spotter (BioRobotics, Cambridge).18 Genomic DNA was extracted from fresh frozen tumor tissue, according to published protocols.19 For all tumor samples, 2 μg of genomic DNA was labeled according to published protocols using a random labeling kit (Invitrogen Life Technologies). Test DNA and male reference DNA was differentially labeled, pooled, mixed with human COT-1 DNA, dried down and resuspended in a formamide-based buffer. The reactions were applied to arrays and allowed to hybridize under cover slips for 48–72 hr at 37°C. Slides were washed according to published protocols19 and scanned using an Agilent Microarray scanner (Agilent Technologies). Identification of individual spots on scanned arrays was performed with Gene Pix Pro 4.0 (Axon Instruments), and the quantified data matrix was loaded into Bio Array Software Environment BASE. Background-correction of Cy3 and Cy5 intensities was calculated using the median-feature and median-local background intensities provided in the quantified data matrix. Within arrays, intensity ratios for individual probes were calculated as background corrected intensity of tumor sample divided by background corrected intensity of reference sample. A signal to noise filter (SNR) of ≥5 for the sample channel and ≥5 for the reference channel was applied to the data and spots that failed to pass these criteria were excluded from further analysis and regarded as missing values. The filtered data was, for each array separately, centralized to a median ratio of unity excluding X and Y chromosome clones. All filtering, normalization and analysis was performed in BASE.20 Subsequently, a moving average of 200 kbp was applied and a BASE implementation of CGH Plotter was used to determine deletion/amplicon boundaries.21 Noise constant was set to 15 and amplification/deletion limits was set to ±0.2.
LOH on chromosome 5 and association with clinico-pathological variables
LOH mapping was performed using 26 microsatellite markers evenly distributed over chromosome 5. The LOH frequency at individual markers varied from 19 to 82% in BRCA1 tumors, 11–44% in BRCA2 tumors and 7–43% in sporadic tumors (Fig. 1). Of 36 BRCA1 tumors that were informative on at least 10 markers distributed along the chromosome, 6 (17%) showed LOH and 3 (8%) showed retention of heterozygosity (ROH) at all markers, suggesting whole chromosome loss and absence of deletions, respectively. In 65 informative BRCA2 tumors, the corresponding figures were 1 (2% LOH) and 13 (20% ROH), and in 58 sporadic tumors, 0 (no LOH) and 22 (38% ROH), respectively. The LOH frequency was significantly higher in BRCA1 tumors as compared to sporadic tumors for 19 markers, in BRCA1 vs. BRCA2 tumors for 10 markers and in BRCA2 cp. sporadic tumors for 1 marker. This difference in LOH frequency was not due to overall higher grade of the BRCA1 tumors since the pattern was still seen within grade III samples only. Looking at markers with significantly higher frequency of LOH in BRCA1 tumors than in the other groups, e.g. D5S2006, D5S620 and D5S671, for D5S2006, the LOH frequency for BRCA1 tumors was 82%, for BRCA2 tumors 44% and for sporadic tumors 19% while the respective figures for grade III tumors were 100%, 57% and 14%. For D5S620, the corresponding figures were 71%, 34% and 21% vs. 57%, 46% and 33% and for D5S671, 57%, 42% and 11% vs. 64%, 36% and 0%.
To map potential sites of tumor suppressor genes, the smallest regions of overlapping LOH between tumors (LOH targets) were defined. The region between a marker showing LOH and the next adjacent negative marker was defined as deleted, meaning that in some cases, the overlap was at the intersection between 2 adjacent markers. Figure 2 depicts the LOH results for tumors with partial and interstitial LOH patterns. LOH in BRCA1 tumors were in general more extensive and in many cases affecting the whole chromosomal arm. However, based on the LOH pattern in BRCA1 tumors, 2 limited LOH targets were identified, 1 located at 5q35.3 and the other at 5p15.3. Particularly high frequency of LOH (82%) was found at 5q35.3 indicating a target region telomeric of D5S211 (Fig. 2a). LOH at the 5q35.3 region was also seen in BRCA2 and sporadic tumors (Fig. 2b and 2c). On the basis of the fact that a number of tumors manifest LOH at only 1 of the 2 markers at 5q35.3 (D5S2030 and D5S2006) it can be assumed that these 2 markers define the boundaries of an LOH target, including a 2.5 Mb region. Results from the corresponding region at 5p15.3 indicate an LOH target between markers D5S392 and D5S406 (4.7 Mb) in BRCA1 tumors (Fig. 2a), as well as in BRCA2 and sporadic tumors (Fig. 2b and 2c).
Mapping of additional LOH targets, based solely upon the patterns in BRCA1 tumors, revealed rather large regions (Fig. 2a). However, by taking the LOH pattern of all 3 tumor groups into consideration, 9 additional limited target regions were defined; D5S432-1997 (5p15.2-p15.1), D5S648-455 (5p14.3-p13.2), D5S398-2858 (5q11.2-q12.3), D52858-1977 (5q12.3-q13.3), D5S618-644 (5q14.3-q15), D5S404-393 (5q23.1-q31.1), D5S2011-636 (5q31.3-q33.1), D5S636-412 (5q33.1-q33.3) and D5S671-211 (5q34-q35.2) (Fig. 2).
Since the LOH targets appear to be common for all 3 groups of breast tumors (although varying in frequency) it was decided to pool data for association analysis. The results from comparison of LOH at individual markers to various clinico-pathological variables are shown in Table I. Significant correlations (p ≤ 0.05) were found to negative ER and PgR status, high tumor grade, large tumors, high percentage of cells in S-phase, aneuploidy, ductal type of tumors and, in general, young age at diagnosis. An interesting exception was the association between high age at diagnosis and LOH at markers D5S455 (5p13.2) and D5S398 (5q11.2), which is mainly contributed by tumors from BRCA mutation carriers. At 5p13.2, 71% of all tumors showing LOH were diagnosed after the age of 50 (versus 50% expected), the figure for BRCA tumors only was 62% (versus 38% expected) and for sporadic tumors 88% (versus 69% expected). For 5q11.2, the figures were 73% (versus 50% expected) for all 3 groups of tumors together, 75% (versus 38% expected) for BRCA tumors only and 71% (versus 69% expected) for sporadic tumors.
An array CGH platform, comprising >2,500 BAC clones mapped to chromosome 5 was used, providing a complete coverage and average resolution of 50 kbp. Results from 10 BRCA1 tumors are illustrated in Figure 3b. All tumors showed deletions in at least 1 region on 5q and the overall frequency was high, which confirmed the high LOH frequency seen in the BRCA1 tumors (Fig. 1). Three regions (5q13.2-q23.3, 5q31.3-q33.2 and 5q33.2-q35.1) were found to be deleted in >80% of samples. These regions are large and vary in size from 11.6–59 Mbp, each comprising 30–60 genes, including several candidate tumor suppressor genes such as XRCC4, MSH3, RAD50, RASA1, APC and PPP2R2B. In addition to these heterozygous deletions, two distinct homozygous 5q deletions were detected, spanning from 700 kbp to 1.5 Mbp (Fig. 4a). Tumor Ca13714 harbored a homozygous deletion of ∼1.5 Mbp at 5q12.3-q13.1, which includes 2 known genes, LY64 and PIK3R1. The second homozygous deletion was located at 5q12.1 and also comprises 2 genes, PART1 and DEPDC1B, as previously reported.18 The homozygous deletion in tumor Ca13714 at 5q12.3-q13.1 was verified by PCR using a STS marker that mapped to the putative homozygous deletion. A marker for a fragment mapping to a non-deleted region was used as a reference (Fig. 4b). According to the LOH mapping shown in Figure 2, these 2 homozygous deletions lie within 2 different LOH target regions. The one at 5q12.1 lies within the region defined by D5S398 and D5S2858 and the one at 5q12.3-q13.1 within the region defined by D5S2858 and D5S1977. It should be noted that the array CGH analysis revealed additional homozygous deletions located on other chromosomes in 2 BRCA1 tumors.
Here we report results of a comprehensive study of chromosome 5 LOH in breast tumors from 3 groups of patients; BRCA1 germ line mutation carriers, BRCA2 mutation carriers and unselected cases, comprising mainly patients with sporadic disease. A very high frequency of LOH was seen at several chromosomal regions in BRCA1 tumors, confirming our previous observations of recurrent DNA copy number loss at 5q using CGH analysis.13 LOH was observed at the same regions in BRCA2 and sporadic tumors, albeit at lower occurrence, although one of the loci (5q31.3) contained frequent LOH in all the 3 tumor groups. Based on the patterns of LOH in all tumors, 11 distinct chromosome 5 regions were mapped as LOH targets. Seven of these regions have not previously been reported in breast cancer; 5p15.3, 5p15.2-p15.1, 5p14.3-p13.2, 5q14.3-q15, 5q23.1-q31.1, 5q31.3-q33.1 and 5q33.1-q33.3. Moreover, we used high-resolution array CGH to map DNA copy gains and losses at chromosome 5 in 10 BRCA1 tumors, which demonstrated the high prevalence (>80%) of deletions at regions overlapping the LOH targets, namely at 5q13.2-q23.3, 5q31.3-q33.2 and 5q33.2-q35.1. In addition, 2 novel homozygous deletions were detected comprising restricted regions of <1.5 Mbp at 5q12.3-q13.1 and 5q12.1, respectively.
Only a few previous studies have addressed LOH mapping in BRCA1 tumors,22 or in BRCA2.23, 24, 25, 26 Those results have shown that chromosomal regions with prevalent LOH in BRCA1 and BRCA2 tumors are usually affected also in sporadic breast cancer, although at a lower rate. The high frequency of LOH in tumors from the predisposed individuals may be a result of the inherent genomic instability that characterizes cells with BRCA deficiency. Moreover, the specific chromosomal regions affected during tumor development in BRCA1 and BRCA2 mutation carriers may reflect a progression along ‘genetic programs’ determined by the predisposing gene or cell type of tumor origin. However, these ‘programs’ are not unique to BRCA tumors but are also seen in sporadic tumors, maybe just by coincidence, but possibly because of somatic and epigenetic events in BRCA genes.27, 28 Indeed, lack of nuclear BRCA1 protein in sporadic tumors is associated with early age of diagnosis, indicating that a disturbance of the BRCA1 pathways may induce a more aggressive tumor development normally associated with BRCA1 germline mutations.29, 30 Amplification of a recently identified oncogene, EMSY, encoding a suppressor of BRCA2 protein function, was found to be associated with a pathological profile similar to that of BRCA2 mutated tumors.31 Moreover, frequent LOH at specific loci on chromosome 5q (5q11.2, 5q14, 5q21-q32) in sporadic tumors was described as one of the hallmarks of basal-like breast tumors, which share many characteristic histopathological and molecular features with BRCA1 tumors, suggesting that the basal-like tumors might represent a group of sporadic tumors showing a defective BRCA1 pathway.32 One of the many characteristics of basal-like breast tumors as well as BRCA1 tumors is ER negativity. In our study, 10 of 17 (59%) ER-negative and 32 of 48 (67%) ER-positive LOH-informative sporadic tumors showed LOH at one or more of the chromosome 5 markers used. Looking specifically at the regions described by Wang et al.,32 7 ER-negative (7/17, 41%) and 24 ER-positive (24/48, 50%) showed LOH at 5q14 and 5q21-q32 but none of the ER-negative tumors showed LOH at 5q11.2 whereas 16 (16/48, 33%) ER-positive tumors showed LOH in that region. So with respect to the results of Wang et al.,32 this comparison of different ER-classes of sporadic tumors does not suggest specific relation of LOH at the mentioned 5q-loci to basal-like tumors. However, in lack of direct ways to identify basal-like tumors in our study, these data does not rule out the possibility that the pathway of basal-like tumors includes LOH at one or more specific 5q loci.
Previous chromosome 5 LOH studies have been performed mainly on sporadic breast tumors and focused on regions including MSH3 and APC, genes known to be involved in colon cancer. Benachenhou et al33 reported a target region located at 5p13.1-5q12.3, with LOH frequency of 23%. This region overlaps with 2 target regions identified in our study (5q11.2-q12.3 and 5q12.3-q13.3), in which LOH ranges from 21–43% in BRCA2 and sporadic tumors, and up to 61% in BRCA1 tumors. Although MSH3 was suggested as a possible candidate gene, updated genome information (http://genome.ucsc.edu/) indicates a location telomeric to the reported LOH region. Thompson et al.34 reported 28% LOH frequency at 5q21, the putative location of the APC gene, and suggested either of these 2 genes to be the target genes in sporadic breast cancer. Indeed, it has been shown that APC is frequently mutated (18%) and hypermethylated (44%) in sporadic breast tumors,35, 36 and it would be interesting to analyze the role of APC in BRCA1 and BRCA2 tumors.
We demonstrate a particularly high LOH rate at the 5q34-q35.3 region in BRCA1 and BRCA2 tumors, while being considerably lower in sporadic tumors. Nathanson et al have suggested a possible BRCA1 modifier locus at 5q34-q35 that may harbor a gene or genes that increase the age-adjusted penentrance of BRCA1 mutation.37 Interestingly, we found a correlation between LOH at marker D5S2030 at 5q35.3 and early age of diagnosis. Krop et al identified a candidate 5q35 breast tumor suppressor gene, HIN-1 (high in normal 1), the expression of which is significantly decreased in both pre-invasive lesions and carcinomas. No mutations were found in HIN-1, but promotor hypermethylation was reported in >90% of breast cancer cell lines and 74% of primary tumors, which could explain its decreased expression.38HIN-1 analysis in BRCA1 tumors was not conclusive39 and BRCA2 tumors have not been analyzed.
Intriguingly, array CGH analysis identified 2 homozygous deletions on 5q in 2 different BRCA1 tumors, possibly pinpointing potential tumor suppressor loci. One of the deletions spans 1.5 Mbp at 5q12.3-q13.1 and include 2 known genes, LY64 and PIK3R1. LY64 (OMIM:602226) is a cell surface molecule and belongs to the family of pathogen receptors, Toll-like receptors. PIK3R1 (p85α regulatory subunit) is part of the PIK-3 pathway,40, 41, 42 which makes it a good candidate tumor suppressor gene in breast cancer, especially in light of the very frequent somatic activating PIK3CA mutations, which stimulates the PIK-3 pathway, and also the known role of PTEN downregulation in breast cancer.43 The second deletion contains 2 known genes, PART1 and DEPDC1B, and spans 700 kbp as previously described.18DEPDC1B (also known as BRCC3) encodes a protein with pleckstrin/G protein interacting and RhoGAP domains, which could be of potential relevance in tumor development. Our results suggest that several putative tumor suppressor genes may reside on 5q and that high-resolution array CGH represents an appropriate method to localize them.
Chromosomal locus 5q31.3 is another region of interest regarding the high frequency of LOH seen for all 3 groups of tumors, and also corroborated by array CGH analysis, which pinpoints a region of 11.6 Mbp including more than 50 genes. No candidate breast cancer genes have been localized to 5q31.3 although this region harbors various potential tumor suppressor genes, some of which are implicated in other types of malignancies such as myelodysplastic syndromes and acute myeloid leukemia. Several genes have been pinpointed as possible candidates, e.g. EGR1,44CSF1R,45LOC51780,46HSPA947 and IRF1.48IRF1 has also been implicated in human gastric cancer.49 Based upon their function, other genes at 5q31 such as the transcription regulator TCF150 and the RAS inhibitor Human sprouty 451 could be implicated in suppression of malignant growth.
LOH status at several regions on chromosome 5 was significantly correlated to factors of poor prognosis. This correlation was in some cases not limited to BRCA1 tumors but contributed to by all 3 groups of tumors. For example at D5S2858, where LOH associated significantly with ER negativity, 58% of all tumors showing LOH and 28% of all tumors not showing LOH were ER negative, and the respective figures for BRCA2 associated and unselected tumors together were 45% and 23%. At D5S644, 68% of all tumors showing LOH and 43% of all tumors not showing LOH were PgR negative, and the respective figures for BRCA2 associated and unselected tumors together were 54% and 41%. For D5S669 the corresponding figures were 84% and 50% vs. 78% and 45% respectively. At D5S2078, 87% of all tumors showing LOH and 45% of all tumors not showing LOH were aneuploid, while the corresponding figures for BRCA2 associated and unselected tumors together were 82% and 45%. For D5S2011 the figures were 81% and 38% vs. 89% and 36% respectively and at D5S410 the figures were 92% and 50% vs.100% and 47% respectively. This suggests that factors such as loss of ER and PgR expression, and increased genetic instability, previously described as characteristic of BRCA1 mutated tumors8, 9, 10, 11, 12 are also of higher prevalence in BRCA2 mutated and sporadic tumors with loss of certain regions of chromosome 5. An unexpected finding was the association between higher age at diagnosis and LOH at markers D5S455 (5p13.2) and D5S398 (5q11.2), mainly due to tumors from BRCA mutation carriers who generally are of young age at diagnosis. In contrary, we saw a correlation between LOH at 5q35.3 and diagnosis before the age of 50, contributed by all 3 groups. This indicates that 5q35.5 harbor one or more regulatory genes that affect progression, e.g. gene(s) important for cell cycle or DNA repair mechanisms. In agreement with results published by Schwendel et al.,52 LOH status at some chromosome 5 regions was associated with high histological grade, both in sporadic and BRCA tumors.
In the association analysis between LOH and clinico-pathological factors, we pooled the data from all 3 groups. The reason for this was that our LOH data indicates that there are common chromosome targets in all 3 groups of tumors. However, it should be noted that such pooling of data might have been at the cost of some tumor subgroup specific associations.
In conclusion, results based on microsatellite typing of allelic imbalance combined with array CGH data on DNA copy number alterations, suggest that chromosome 5q contains a number of regions influencing tumor progression when deleted, especially in BRCA1 tumors. Homozygous deletions may pinpoint genes that should be more carefully followed-up. The results are likely to be helpful in identifying new cancer genes that might have therapeutic and diagnostic importance for breast cancer patients.
We thank the personnel at the Department of Pathology for providing pathological information, Orn Olafsson for his assistance with the statistical analysis and Dr. Carl Blomqvist and research nurse Minna Merikivi at Helsinki University Central Hospital for their kind help.