High-Throughput Loss of Heterozygosity Mapping in 26 Commonly Deleted Regions in Breast Cancer¹

Recent advances in the field of human genetics, especially completion of the Human Genome Project, have produced tremendous amounts of data that will hasten the identification of genes involved in cancer development. For example, the identification and localization of over 20,000 microsatellite markers (1) can be used to perform fine mapping in chromosomal regions known to harbor one or more TSGs.³ In conjunction with the increased map resolution provided by these microsatellites, the development of powerful bioinformatic tools such as the UCSC Human Genome Browser⁴ with exquisite mapping of known and predicted genes provides the ability to identify putative TSGs within minutes in silico, saving years of investigative work in the understanding of cancer genetics. In addition to the biological knowledge generated from the Human Genome Project, technological advances such as CAE have dramatically increased the speed of genomic studies, and recent studies have shown that the use of CAE results in increased sensitivity in the detection of chromosomal loss over traditional slab gels (2). Thus, CAE provides a platform on which genome-wide, high-throughput LOH studies can be performed quickly and accurately.

In addition to these genomic advances, the practice of LM has revolutionized cancer genetics. LOH studies have traditionally been hampered by the heterogeneous mix of cell types found within a tumor, with the DNA from normal cells obscuring the extent of chromosomal loss in tumor cells. LM, which employs a short focused pulse from an infrared (IR) laser to isolate selected cells, can produce a homogeneous cell population from histological tissue sections. The use of LM has shown an increase in the ability to detect chromosomal loss; in breast tumor samples in particular, a 2-fold increase in detection of LOH has been observed (3). Numerous studies have identified regions of chromosomal loss associated with breast cancer in whole tissue or manually microdissected tumor samples, and recently, several investigators have used breast tumor DNA isolated by LM (3, 4, 5); however, there are no reports in which the enhanced technical precision of LM has been combined with the power and speed of CAE. To effectively combine these technologies in large-scale LOH studies, a set of microsatellite markers must (a) fall within a known or candidate TSG region, (b) not be prohibitively expensive, (c) have high heterozygosity, and (d) amplify DNA from laser-microdissected tumor samples that is often of very low quantity and poor quality. Whereas many breast cancer studies have used particular microsatellite markers to detect LOH, a consensus set of markers to be used in a breast cancer-specific LOH study is not currently available. Here, we present a panel of 52 microsatellite markers that correspond to the 26 regions most commonly deleted in breast cancer (6). This panel has the potential to serve as an effective and efficient tool for recognizing clinically relevant patterns of chromosomal loss and for identifying regions likely to harbor breast cancer-related TSGs.

DNA samples were acquired from archival paraffin-embedded tumor samples examined by one pathologist to ensure consistency of diagnosis. Pure tumor cell populations were isolated using the ASLMD laser microdissection system according to standard procedures (Leica Microsystems, Wetzlar, Germany) or using the PixCell II laser capture apparatus (Arcturus Engineering, Mountain View, CA) with a modified protocol (7). For each sample, referent DNA was extracted from disease-free tissues such as nipple or negative axillary lymph nodes using commercially available kits (Qiagen, Valencia, CA). Accurate DNA quantification was not possible due to very low yields following LM, rather, fluorescence-labeled microsatellite primers (Invitrogen Corp., Carlsbad, CA) were used to amplify small aliquots of each DNA sample using either a modified PCR stepdown protocol (8) or a hot-start protocol that includes 1 cycle of 95°C for 5 min and 40 cycles of 95°C for 1 min, X°C (X°C = 50°C to 62°C) for 2.5 min, and 72°C for 1 min. PCR products were then purified using Sephadex resin and run on a MegaBACE-1000 capillary electrophoresis apparatus (Amersham Biosciences, Piscataway, NJ) using standard conditions. Genotypes were assigned using Genetic Profiler version 1.5 (Amersham Biosciences), and allelic ratios were calculated using the ration (T1/T2)/(N1/N2), where T1 and T2 correspond to the less intense and more intense peak heights of the tumor samples, and N1 and N2 correspond to the lower and higher intensity peak heights of the referent samples (2).

The 26 CDRs were identified by integrating genetic data from 143 studies describing LOH in breast cancer (6). Because each study used unique sets and types of polymorphic markers, such as RFLPs and microsatellites, the data were integrated into a single genetic map. Those regions showing loss in at least four independent studies were defined as a CDR (Fig. 1). Because each CDR was defined cytogenetically and contained a number of polymorphic markers, we developed a consensus set of markers with the greatest likelihood of detecting chromosomal loss in a large number of breast tumor samples that could be integrated among and between laboratories. The sequence of each cytogenetically defined CDR (6) was examined using the UCSC Human Genome Browser (9).⁴ All microsatellite markers and their precise map location within each region were identified. The size of the CDRs, many of which spanned multiple cytogenetic bands, ranged from 27.2 Mb on chromosome 8q24 to 2.4 Mb on 11q13.1. Over 200 Mb of sequence was examined, and over 1000 microsatellites within or surrounding each CDR were identified; the average marker density was 1 microsatellite every 350 kb, with high and low marker densities on chromosomes 17p13.1 and 6q16, respectively. A subset of markers from the ends of each region was then selected to assess loss across each CDR with a minimum number of genotypes. Each subset was then examined for informativeness; markers with heterozygosity < 0.60 were excluded. Finally, only those markers already commercially available as fluorescence-labeled primer sets were considered in order to keep the cost of large-scale genotyping to a minimum. Eighty-five marker sets were thus chosen for further analysis.

Determination of the optimal cycling conditions was critical in the development of this mapping panel. All DNA samples were acquired from paraffin-embedded tissue samples that had been archived for up to 10 years. Furthermore, because LM is a time-intensive process, a minimal number of cells were isolated, often resulting in quantities of DNA too low to detect using spectrophotometry. The use of CAE overcomes, in part, the poor quality and quantity of DNA isolated from these samples by greatly reducing the amount of PCR product needed to generate reliable genotypes versus the amount required for detection with either radioisotope- or fluorescence-labeled products run on polyacrylamide slab gels (9, 10). A minimal peak threshold of 1000 relative fluorescence units (rfu) was required for sample analysis, and whereas 7 of 85 (8%) of markers failed to amplify under all PCR conditions tested, an additional 12 of 85 (14%) markers that effectively amplified high-quality genomic DNA failed to amplify the tumor DNA derived from laser microdissection of the archived, paraffin-embedded tumor samples. Of the remaining markers that robustly amplified all DNA samples, two markers were chosen to represent each CDR, producing a final set of 52 markers (Table 1).

To detect chromosomal loss, microsatellite markers with a high probability of amplifying heterozygous loci were used. The effectiveness of this marker set in detecting heterozygosity in patient samples is seen in Fig. 2. Ten referent DNA samples, isolated from disease-free nipple sections, were screened for all 52 markers, and the extent of heterozygosity in these patients ranged from 100% to 50% (Fig. 2,A). More importantly, when markers were analyzed in pairs within a CDR, over 60% of the 260 genotype pairs were informative for both markers in a given region, whereas less than 5% were uninformative at both loci (Fig. 2 B).

To confirm the utility of this mapping panel, DNA from 10 tumor samples ranging from stage 0 to stage III and the corresponding referent samples were screened. Fig. 3 shows the results from a stage IIIA infiltrating ductal carcinoma with positive ER and PR status from a 71-year-old woman with 5 of 7 positive lymph nodes. Allele ratios were calculated using the peak height, rather than peak area, as suggested in previous studies (2, 11). The appropriate normalized ratio for inferring chromosomal loss varies widely in the published literature, with values ranging from 20% to 50%. Because samples were collected using LM, a stringent detection value of >50% loss was used. Results from Fig. 3 show that loss was detected at either one or both markers for 18 of 26 (69%) CDRs. Five of the CDRs showed loss at one microsatellite but were uninformative at the other. Where only one of the two markers showed loss, the critical region for the putative TSG can be further delimited. Every chromosomal region studied showed loss in at least 1 of the 10 samples examined, with 10–30% of these showing loss at both markers.

Chromosomes 16 and 22 have two adjacent CDRs; both pairs of markers tested on each chromosome indicated loss, suggesting large chromosomal deletions. Chromosome 17 is characterized by three CDRs: two located close together on the p arm; and a third located on the q arm. Loss was seen in three of the four markers on 17p and in both markers on 17q. Although this screening set cannot distinguish whether there are two or three separate regions of loss on chromosome 17 in this patient or whether the loss encompasses one large contiguous deletion, these markers are effective in detecting LOH. In cases where only one of the two markers showed loss (chromosome 3p14.1, 5q21.1–21.3, and 17p13.1), LOH was confirmed by using DNA collected from a separate tissue section from the same tumor and carrying out independent amplification and genotyping.

All 10 tumor samples showed loss, ranging from loss at 6 of 26 (23%) regions in a stage I infiltrating ductal carcinoma with positive ER/PR status from a 54-year-old woman to loss of 20 of 26 (77%) regions in a stage IIIA lobular carcinoma with positive ER/PR status from a 76-year-old woman with 4 of 17 positive lymph nodes. This panel set, therefore, successfully combines the technologies of CAE and laser microdissection as well as validates the identification of these 26 regions as important regions in breast cancer.

The ability to use archived tumor samples is critical to breast cancer research because earlier detection methods have greatly decreased the number and size of tumors available for research projects. Because archival pathology collections are most often comprised of formalin-fixed and paraffin-embedded samples, the quality of the nucleic acids and proteins extracted from fixed tissues is poorer than those extracted from flash-frozen specimens. Thus, DNA isolated from archival samples is likely unsuitable for genomic technologies such as the chip-based comparative genomic hybridization and microarray analysis. This study has shown that DNA from archival, paraffin-embedded breast tumor samples can be isolated using LM and effectively genotyped using CAE. This mapping panel can be used as a high-throughput screening tool for identifying genomic patterns of chromosomal loss associated with breast cancer and for identifying and characterizing regions that may harbor TSGs. Using the technologies afforded by CAE, a minimum of 12 microsatellite markers can be run simultaneously in 1 well of a 96-well plate, allowing one tumor sample to be screened for all 52 markers in one 75-min run. With the increased throughput of CAE machines now capable of running 384 samples, large-scale studies of tumor samples can be undertaken, and LOH profiles for each sample can be generated within days. Generation of large amounts of LOH data in different stages and types of breast cancer will allow specific patterns of loss to be identified and perhaps used in a clinical setting as a predictor of progression and outcome of breast cancer. Because these markers have been specifically mapped within the completed sequence of the human genome, and because these markers have all been shown to effectively detect LOH in breast samples, this standardized panel can be used to identify subsets of samples with loss at specific loci, and by using one set of markers, data from different studies can be pooled. Coupled with the extensive identification and mapping of genes available within minutes from web sites such as the UCSC Human Genome Browser,⁴ subsets of samples can be more extensively studied using a fine mapping approach, and the types and number of underlying TSGs can be elucidated. This marker set, therefore, will be useful in studying the development and progression of breast cancer and may identify predictive patterns of loss.

We thank Drs. Vimal Mital, Harold Ashcraft, and John Yerger and the staff of the Windber Medical Center pathology for access to archival breast samples and critical advice.