Microalterations of Inherently Unstable Genomic Regions in Rat Mammary Carcinomas as Revealed by Long Oligonucleotide Array-Based Comparative Genomic Hybridization Free

The presence of copy number variants in normal genomes poses a challenge to identify small genuine somatic copy number changes in high-resolution cancer genome profiling studies due to the use of unpaired reference DNA. Another problem is the well-known rearrangements of immunoglobulin and T-cell receptor genes in lymphocytes (a commonly used reference), which may misdirect the researcher to a locus with no relevance in tumorigenesis. We here show real gains of the IgG heavy chain V gene region in carcinogen-induced rat mammary tumor samples after normalization to paired mammary gland, a tissue without lymphocyte infiltration. We further show that the segmental duplication region encompassing the IgG heavy chain V genes is a copy number variant between the susceptible (SS) and the resistant (BN) to mammary tumor development inbred rat strains. Our data suggest that the already inherently unstable genomic region is a convenient target for additional structural rearrangements (gains) at the somatic level when exposed to a carcinogen (7,12-dimethylbenz[a]anthracene), which subsequently seem to benefit tumor development in the mammary gland of the susceptible strain. Thus, the selection of an appropriate reference DNA enabled us to identify immunoglobulin genes as novel cancer targets playing a role in mammary tumor development. We conclude that control DNA in array-based comparative genomic hybridization experiments should be selected with care, and DNA from pooled spleen (contains immature lymphocytes and is used as reference in animal studies) or blood may not be the ideal control in the study of primary tumors. [Cancer Res 2009;69(12):5159–67]

Chromosomal comparative genomic hybridization (CGH) has been used to map DNA copy number changes in human breast cancer, comparing whole tumor DNA to normal reference genomes (1). Changes in copy numbers of genes such as ERBB2 and c-MYC have been extensively documented in human breast cancer (2, 3) and have led to the successful development of therapeutics against amplified and overexpressed genes, for example, the use of the drug trastuzumab against ERBB2. Although CGH analysis to metaphase chromosomes has been very useful in identifying DNA copy number changes, it can only provide limited resolution, 5 to 10 Mb (and 1-2 Mb of amplified sequences). This limitation has been circumvented by the array-based CGH (aCGH) strategy (4) first performed with BACs, yielding a significantly higher resolution (200-250 kb). Several studies reported profiles of genomic DNA copy numbers using the novel aCGH technology (5–7), and fine mapping of numerous specific genetic alterations was identified defining breast cancer subtypes at the genomic level and highlighting promising therapeutic targets. Today, an even higher-resolution CGH method is available (8, 9), which is based on the use of oligonucleotide probes mapped to the genome, capable of accurately reporting single-copy number changes.

With the advent of the progressively higher resolution aCGH techniques, the identification of instability in genomes from healthy individuals became feasible, a phenomenon defined as copy number variants (CNV). CNVs are submicroscopic DNA segments >1 kb and are present at a variable copy number in comparison with a reference genome (10–13). They were found to affect up to 12% of the human genome and can also influence susceptibility to diseases (12), including breast cancer (14). The existence of structural variants in “normal” genomes, and the fact that they influence gene expression through gene dosage or position effect, adds another layer of complexity to the genome and consequently to tumor genomic profiling studies, as will be outlined below.

With the recent development of rat genetics and genomics (15), rat models have become increasingly important for the study of human diseases, including breast cancer (16, 17). Rat and human mammary carcinomas share similar developmental and histopathologic features and, unlike mouse carcinomas, show hormonal responsiveness and do not have a known viral etiology (18–21). The rat is therefore a model of choice for the recapitulation of human breast cancer etiology and gene discovery.

In the present investigation, we undertook the high resolution (385K) NimbleGen long oligonucleotide aCGH in a novel rat model of mammary cancer. We previously established the SS/JrHsdMcwi (Dahl salt-sensitive hypertensive rats) and BN/NHsdMcwi inbred rat strains to be highly susceptible and completely resistant, respectively, to 7,12-dimethylbenz[a]anthracene (DMBA; a prototypical polycyclic aromatic hydrocarbon)-induced mammary cancers (17). The polycyclic aromatic hydrocarbons are known to be an important class of environmental chemical carcinogens and have been implicated in human breast carcinogenesis (22, 23). The goal was to identify structural genomic alterations that occur somatically during chemical carcinogenesis in the susceptible (SS) inbred strain. Analysis of individual DMBA-induced mammary carcinomas from SS inbred rats using a control of pooled spleen DNA from untreated SS rats revealed several tumor aberrations colocalizing with germ-line CNVs that we recently mapped between different inbred rat strains (24). These results suggested that the detected aberrations may not be tumor specific and also suggested the possibility of CNVs existing between animals within an inbred rat strain.

The finding prompted us to reanalyze a subset of the tumors with aCGH using paired spleen DNA (e.g., spleen DNA from the same animal in which the tumor developed) as control. The generated tumor data were this time more homogeneous, enabling the identification of a gained genomic region in the majority of the tumors that overlapped with an intrachromosomal segmental duplication. Segmental duplications are defined as having ≥90% sequence identity, are ≥5 kb in length, and are known to be very dynamic evolutionarily (25, 26). This specific region encompassed the IgG heavy chain V (Ighv) genes and is located on rat chromosome (Rattus norvegicus, RNO) 6 (138.5-140 Mb). The aCGH results were further validated using real-time quantitative PCR (qPCR) with paired mammary gland DNA as reference. Thus, the selection of an appropriate reference DNA enabled us to identify immunoglobulin genes as novel cancer targets playing a role in mammary tumor development.

Animals, treatment, and tissue samples. Virgin female SS and BN rats were obtained from the Department of Physiology, Medical College of Wisconsin, and belonged to the 49th and 32nd inbred generations, respectively. All rats were received at weaning and were housed three animals to a cage on a 12 h light/12 h dark cycle at 22°C room temperature and 55% relative humidity. The animals were provided MCW/Teklad diet #3075S and water ad libitum. All experimental procedures were conducted following approval of the Institutional Animal Care and Use Committee. At age 47 to 57 days, 17 SS virgin females received a single 65 mg/kg DMBA [200 mg DMBA (Acros Organics) dissolved in 10 mL sesame oil (Sigma-Aldrich)], dose by oral gavage. Controls were 6 SS rats, treated with an equivalent volume of sesame oil, without DMBA.

Beginning 3 weeks after gavage, the rats were palpated weekly for mammary tumors until the 15th week. The location of each palpable tumor was noted. The time between the date of DMBA treatment and the date of detection of the first tumor was recorded and defined as latency. At week 15, all rats were sacrificed by CO₂ asphyxiation and the number and location of the mammary tumors were recorded at necropsy. A portion of each tumor was fixed in 10% buffered formalin, processed, and embedded. Paraffin sections (5 μm thick) were stained with H&E. Pathologic diagnoses of the mammary lesions were classified by a pathologist as described (21). The remaining mammary tumor tissues were frozen in liquid nitrogen and stored at −80°C. Spleens and mammary gland tissues from 6 of the 17 DMBA-treated SS animals and mammary gland from one untreated BN animal were frozen in liquid nitrogen and stored at −80°C. Spleens were also harvested from the 6 SS rats (controls) treated only with sesame oil and frozen and stored as described above.

Isolation of DNA. DNA was isolated from 20 to 30 mg mammary tumor and mammary gland tissue or from 10 to 12 mg spleen using DNeasy spin columns (Qiagen). Each DNA sample was eluted from the column with two sequential aliquots of 10 mmol/L Tris, 0.5 mmol/L EDTA (pH 9.0). The concentration and purity of each sample were determined by measuring absorbance at 260 and 280 nanometers and by running aliquots of DNA samples on 1% agarose gel.

NimbleGen RN34_WG aCGH. DNA from each of the 19 mammary tumors (developed in 17 DMBA-treated SS rats) was cohybridized to the 385K NimbleGen rat genome tiling path arrays together with a DNA pool consisting of 6 spleens harvested from SS female rats not treated with DMBA. In addition, rehybridization of 6 of the 19 mammary tumors was done, this time using paired spleen tissue as control. The whole-genome oligonucleotide microarray provided by NimbleGen contains 385,000 isothermal probes, 45- to 75-mer, which spans the rat genome at a median probe spacing of 5.3 kb. The oligonucleotide design, array fabrication, DNA labeling, CGH experiments, data normalization, and log₂ (Cy3/Cy5) copy number ratio calculations were done by NimbleGen according to recommended and published procedures (9). The SignalMap software (which is based on the circular binary segmentation algorithm; ref. 27) scores log₂ ratios above 0.20 to 0.25 as DNA copy number gains and below −0.25 to −0.5 as losses. Homozygous deletions are scored whenever log₂ is below −1.0. NimbleGen supplies the predicted segmentation for the unaveraged data (defined as UA window, see Table 1, all 385,000 probes present) as well as three different window sizes. The window sizes correspond to 10 (defined as the 50 kb window in Table 1), 20 (defined as the 100 kb window in Table 1), and 50 (defined as the 250 kb window in Table 1) times median probe spacing on the design.

Detection of homozygous deletion of DNA regions including the zinc finger protein 533, semaphorin 3E, and fragile histidine triad genes. To confirm the homozygous deletions detected by aCGH of the RNO3, 4 and 15 subchromosomal regions in rat mammary cancers, rat genomic sequences were obtained from the University of California-Santa Cruz genome browser⁷

and primer pairs (Table 2) were designed with the Primer3 program.⁸ The PCR product sequences were verified by DNA sequencing (as described; ref. 28) and shown to exhibit 94% to 99% identity to the corresponding sequences in the draft of the rat genome. For the analysis of deletions, PCR amplification of the genes/markers was carried out using template DNA from rat mammary tumors. Paired spleen DNA for animal 593 was included as a positive control, whereas a sample consisting of pooled spleen DNA from 6 SS animals not treated with DMBA was used as positive control for tumors 263 RE (RE = right side of the rat mammary gland milk line; gland E), 334 RB (RB = right side; gland B), 211 RA (RA = right side; gland A), and 211 LB (LB = left side; gland E), as paired control tissue was not available for these particular animals. Distilled H₂O was used as a negative control in all cases. PCR was carried out in a volume of 50 μL containing 100 ng DNA, 0.4 μmol/L of each primer, 2.5 units Platinum Taq DNA polymerase (Invitrogen), 0.2 mmol/L deoxynucleotide triphosphates, 1 to 2 mmol/L MgCl₂ and 1× PCR buffer (supplied by the manufacturer together with the Taq DNA polymerase). The amplification mix was preincubated at 94°C for 3 min followed by 30 cycles consisting of denaturing at 94°C for 30 s, annealing at 57°C for 30 s, and extension at 72°C for 1 min with a final elongation step for 7 min at 72°C. The PCR products from individual experiments were separated on 1% agarose gels and the gels were read under UV light after staining with ethidium bromide. Whenever a negative result was obtained (no visible band), the reaction was repeated for verification.

Detection of relative copy gain or loss by real-time qPCR. Real-time qPCR was used to detect relative copy gain and loss of genomic material between the control tissue sample (mammary gland from the same rat) and the tumor samples from the SS females. Mammary gland DNA was further compared between untreated BN and SS females using qPCR. The segmental duplication region at RNO6q32, 138.5 to 140.0 Mb, was selected for verification in all cases. Unique sequence primer pair (LOC691753; see Table 2) specific to the region was designed (as described above) and used in real-time qPCR using the QuantiTect SYBR Green PCR Kit (Qiagen). Reactions were analyzed on a Rotor-Gene 3000 Real-time Thermal Cycling System (Corbett Research). Primers were also designed for Restin-like 2 (Rsnl2) and aryl hydrocarbon receptor (Ahr), which were used as two reference (negative control) genes. These two genes did not show copy number differences between reference and tumor samples in 25 independent aCGH experiments. Primer sequences and their location on RGSC3.4 rat genome assembly are listed in Table 2. Triplicate PCRs were done for each primer pair in 20 μL reactions, which included 150 ng (8.8 μL) DNA, 10 μL of 2× SYBR Green Master Mix, and 0.6 μL forward and reverse primers (10 μmol/L). All reactions were cycled as follows: stage 1, 95°C for 900 s; stage 2, 50 cycles, 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s; and stage 3, melt curve 60°C to 95°C. The SS mammary gland DNA was used for the reference genes (Ahr and Rsnl2) and for the test primer pair to construct standard curves at undiluted, 1:5, 1:25, and 1:125 concentrations. A single point from the standard curve was used as a positive control in each assay to evaluate the test gene in the test DNA. Specificity for all qPCR was verified by melting curve analysis and 1% agarose gel detection of single product and dideoxy sequencing of the PCR products (as described above). The data were analyzed with the Rotor-Gene 3000 software using the cycle threshold for quantification. Relative gain or loss (fold change) for the test gene (when normalized to each of the reference genes), between tumor and control (mammary gland) DNA samples, was calculated using the mathematical model described by Livak and Schmittgen (29). Each experiment was repeated twice, the mean fold change was calculated and normalized, and the SD is given.

aCGH findings in 19 DMBA-induced rat mammary carcinomas. In the present study, 19 DMBA-induced mammary adenocarcinomas, developed in 17 SS inbred rats, were subjected to high-resolution long oligonucleotide aCGH analysis. DNA from each tumor was cohybridized onto NimbleGen arrays using a control DNA pool of spleens from 6 SS inbred females not treated with DMBA. Data from a representative oligonucleotide aCGH analysis from four different mammary tumors are shown in Fig. 1. No gross chromosomal copy number changes were observed in the 19 tumors; however, several smaller DNA regions (≤1.5 Mb in size) were altered (see Table 1). Three of the DNA copy number imbalance regions were homozygous deletions: (a) tumor 593 LC exhibited a 0.295 Mb deletion at RNO3q23 (see Fig. 1B and Table 1), involving a single gene, zinc finger protein 533 (Zfp533); (b) tumors 263 RE and 334 RB each showed homozygous deletion of a 0.148 and 0.140 Mb DNA segment, respectively, at RNO4q12 (see Fig. 1C; only data from tumor 334 RB is illustrated). In both cases, the event led to deletion of the end part of the semaphorin 3E (Sema3e) gene. (c) Tumors 211 RA (see Fig. 1D) and 211 LB, both of which developed in the same animal, harbored homozygous deletions of 0.145 and 0.100 Mb DNA segments, respectively, at RNO15p15-14. The deletion encompassed the fragile histidine triad (Fhit) gene.

Detection of homozygous deletions by PCR. To investigate the deleted loci in more detail, we went on to analyze a total of 5 markers from the RNO3q23 region, 3 from the RNO4q12, and 4 markers from the RNO15p15-p14 region. The position and the order of the markers were determined by BLAST searches in the rat draft DNA sequence (Build 3.4). The resulting map positions are shown in Table 2 and Fig. 1A to D (red arrows) and agree with the current rat draft sequence. When the tumors were subjected to PCR analysis with the markers, there was evidence that all 5 mammary tumors contained homozygous deletions of the markers designed in the deleted region, whereas the regions flanking the deletion were found to be intact (visible PCR product in gel electrophoresis). The PCR analysis confirmed the aCGH results.

aCGH results and germ-line CNVs. The other DNA copy number aberrations detected by aCGH in the 19 tumors, listed in Table 1 under the unpaired control DNA section, colocalized with mapped germ-line CNV regions in the rat (24). One of these changes was particularly clear in tumor 601 LB (see Fig. 1A and Table 1 under the unpaired control DNA section), a gain of a region located at RNO6q32 (138.5-140 Mb). To test if any of the mapped germ-line CNVs were tumor-specific changes, 6 (32%) of the 19 mammary tumors were reanalyzed with aCGH using paired spleen DNA from the same animal in which the tumor developed as control. The animals in which the 6 tumors developed (601 LB, 602 LC, 559 RA, 565 LA, 561 LB, and 567 RB) were 3 littermate pairs: animal 601 and 602 made up the first pair, 559 and 561 the second pair, and 565 and 567 the third pair. Results from the aCGH analysis are summarized in Table 1 under paired control DNA section, and show that, with one exception (RNO6q32), none of the germ-line CNV regions were detected. These results suggest that CNVs can exist between animals within an inbred rat strain. The average number of germ-line CNVs detected per tumor or animal, respectively, was calculated to be ∼2 (34/19 or 34/17, respectively).

Investigation of copy number gains in an area of segmental duplication and genomic variability. Interestingly, the only DNA copy number gain that persisted in the paired aCGH experiment was located at RNO6q32 (138.5-140 Mb). As this region encompassed the Ighv genes, the question of whether the change was tumor specific still remained, because the paired control sample consisted of DNA from spleen, a lymphoid organ containing immature B lymphocytes with rearranged segments of the light and heavy immunoglobulin chains for antibody diversity (30). We therefore performed genomic qPCR analysis using another paired control tissue without lymphocyte infiltration; each tumor DNA was compared with corresponding mammary gland DNA. Thus, genomic qPCR was employed to evaluate the detected DNA copy number gain at RNO6q32 (138.5-140 Mb) in the 6 tumors (601 LB, 602 LC, 559 RA, 565 LA, 561 LB, and 567 RB), one of which was a negative control (tumor 561 LB) showing no gain in the aCGH analysis. The test gene, Ighv locus (LOC691753), residing at RNO6q32 (139.5 Mb; see Table 2 and Fig. 1A) was normalized to two different reference genes (Rsnl2 and Ahr). Each assay was done twice, and the mean fold change was calculated (and the SD was given) and transformed to a log₂ scale. The obtained results were similar using either of the two reference genes. Normalized data from the Rsnl2 gene are shown in Fig. 2.

The Ighv locus was clearly gained in 4 tumors (601 LB, 602 LC, 559 RA, and 565 LA) when compared with the paired mammary gland (see Fig. 2). Thus, our qPCR data show gain of the RNO6q32 region in 4 tumors (601 LB, 602 LC, 559 RA, and 565 LA), which agrees with the both paired and unpaired aCGH analyses, except in the tumor 567 RB. According to the aCGH analysis, the RNO6q32 region was also gained in tumor 567 RB (see Table 1 under both unpaired and paired control sections). However, the qPCR analysis showed no gain of the region in this tumor (see Fig. 2) when compared with paired mammary gland, suggesting that the observed gain in the aCGH analysis rather reflected losses of the region in the control DNA (spleen) and was thus a false-positive tumor aberration. As anticipated, tumor 561 LB (the negative control) that did not harbor a gain of the region in the aCGH analysis (see Table 1) was also negative in the qPCR analysis.

To investigate the possibility of the segmental duplication regionalso being variable in copy number between rat strains, we compared mammary gland DNA from an untreated BN female with mammary gland DNA from an untreated SS female. As before, the Ighv locus was compared with both Rsnl2 and Ahr reference genes, with similar results obtained using either of the reference genes. The results showed a significant decrease in copy numbers of the segmental duplication region containing the Ighv genes in the resistant BN strain compared with the susceptible SS strain (log₂ ratio = −2.12; normalized data from the Rsnl2 gene are shown in Fig. 2). The data show that the segmental duplication region encompassing the immunoglobulin genes also is polymorphic between the two strains and is thus here mapped as a CNV.

A problem with the high-resolution copy number arrays in cancer genome profiling studies is that germ-line CNVs can be mistaken for somatic alterations or unnecessary rejection of true positives if unmatched reference DNA is used. Another problem is the well-known rearrangements of immunoglobulin and T-cell receptor genes (11, 30) that normally occur in lymphocytes (a commonly used reference in tumor genome profiling studies), which may misdirect the researcher to a locus with no relevance in tumorigenesis. To avoid misinterpretations in such studies, germ-line CNVs (in the cases where unpaired reference is used) and immunoglobulin genes are excluded as cancer regions and thus as potential therapeutic targets. We here show genuine small somatic gains of the immunoglobulin (Ighv) gene region in carcinogen-induced rat mammary tumor samples after normalization to paired reference DNA, that is mammary gland, a tissue without lymphocyte infiltration. We further show that the segmental duplication region containing the Ighv genes is a CNV between the susceptible (SS) and the resistant (BN) to mammary tumor development inbred rat strains. Our data suggest that the already inherently unstable genomic region is a convenient target for additional structural rearrangements (expansion/gains) at the somatic level when exposed to a carcinogen, which subsequently seem to benefit tumor development in the mammary gland of the susceptible strain.

We started off analyzing 19 DMBA-induced rat mammary carcinomas with the long oligonucleotide aCGH technique to evaluate copy number changes at a very high resolution (5.3 kb). The overall generated tumor data were remarkably homogeneous with no gross chromosomal changes, which is in agreement with their primarily diploid genome status (>85%; ref. 31). Thus far, only a limited number of studies have reported DNA copy number profiles of rat mammary carcinomas. Our previous study of E₂-induced rat mammary carcinomas using chromosomal CGH revealed a clear pattern of nonrandom chromosomal involvement (32). Furthermore, mammary cancers induced in rats by the heterocyclic amine 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine exhibited discernable somatic abnormalities (33) although distinct from chromosomal copy number changes observed in the estrogen-induced mammary tumors (32). These studies, including our present, suggest that distinct pathways may act in a carcinogen/estrogen-specific manner and contribute to tumor development.

Although no gross chromosomal aberrations were detected in the SS mammary tumors, several smaller segments (stretching from 100 kb to 1.5 Mb) exhibited DNA copy number imbalance. Thus, three important findings emerged from this study. First, when comparison was made of tumor DNA with a pool of normal spleen DNA from untreated SS animals, several tumor aberrations were found to overlap with germ-line CNVs that we recently catalogued between different inbred rat strains (24). These changes disappeared when we reanalyzed a subgroup (32%) of the tumors with aCGH using available paired spleen DNA as reference. The finding suggested the presence of germ-line CNVs between animals (e.g., between the control and the tumor-bearing animals' germ-line DNA) within the SS inbred strain as a result of de novo CNV genesis. We previously mapped a total of 33 CNVs in animals between inbred ratstrains (the SS and BN strains included) using the same aCGH platform as in the present communication, which is at a considerably higher frequency than what was observed between animals within an inbred strain (on average ∼2 CNVs). Because the latter CNVs (see Table 1) also varied in copy number between different inbred rat strains (24), they may represent hotspot CNVs. Nonetheless, it is clear that the few detected changes between animals within a strain would complicate tumor genomic profiling studies.

Secondly, we found a genomic region of segmental duplication (1.5 Mb in size) encompassing the Ighv genes to be gained in 80% of the reanalyzed tumors, which we further validated by genomic qPCR using paired mammary gland DNA as control. The region is located at RNO6q32 and is homologous to HSA14q32. Because of the known DNA rearrangements of this region in lymphocytes (30), immunoglobulin genes tend to be excluded as potential cancer target regions in human and rodent genomic profiling studies in which blood or spleen is used as control. Indeed, one of the structural changes detected here with aCGH affecting the immunoglobulin region in one of the tumors (567 RB) was a false-positive. Thus, we propose that future tumor genomic profiling studies using high-resolution aCGH should ideally be conducted with paired control DNA from normal matched tissue (paired normal mammary gland for mammary tumor) to be able to include immunoglobulin genes as candidate genes. Furthermore, new facts have come to light that other genes may have DNA copy number variations according to tissue type (somatic CNVs; ref. 34), which supports our recommendation of selecting a paired reference from the same tissue in which the tumor developed to avoid unnecessary complications in tumor genomic profiling studies.

By using paired normal mammary gland as control in the present study, real gains were detected and validated in 80% of the reanalyzed tumors, suggesting the participation of immunoglobulin genes in mammary tumor development. In concordance with our finding, other studies show that nonhematopoietic human carcinomas (including breast cancer) and cancer cell lines do express immunoglobulins and are thought to promote tumor growth by antibody enhancement (35, 36).

Interestingly, the rat genomic region of segmental duplication, containing the Ighv genes, is remarkably large (1.5 Mb in size) and is termed a “duplication block” because of its size (25). Many studies have shown significant association between the location of segmental duplications and regions of chromosomal instability or evolutionary rearrangements (37–40) and are thought to have a role in >25 recurrent genomic disorders (41). Segmental duplications were recently found to be exposed to unequal crossing over during meiosis, generating gametes that either lack or carry a double dose of the critical interval, increasing the susceptibility to disease (42). The expansion of the segmental duplication region in the rat mammary tumors may similarly have been exposed to unequal crossing over during mitosis, and the selective pressure in the presence of a carcinogen may have favored survival and expansion of the cells carrying a double dose of the immunoglobulin genes, promoting tumor development. Although the exact mechanism behind the detected gains is unclear, our data support a role for immunoglobulin genes in chemically induced rat mammary tumorigenesis.

The region of segmental duplication was further found to be significantly increased in copy number in the mammary cancer susceptible SS strain compared with the resistant BN inbred rat strain, suggesting that the region has expanded in the SS lineage during rat evolution, and is thus here mapped as a CNV. Taken together, our data suggest that the already inherently unstable CNV region is evolutionary very dynamic and has a tendency to further expand after carcinogen exposure, promoting tumor development.

Finally, three different genomic regions were found to be homozygously deleted in a subset of the mammary tumors, of which one of the regions encompassed the well-known breast cancer tumor suppressor gene Fhit (located on RNO15p15-14, which is homologous to human chromosome HSA3p14.2), previously reported to be genetically altered early on in breast tumor development (43). We found the Fhit gene to be deleted in 11% of the rat mammary tumors. Interestingly, Fhit knockout mice are highly susceptible to carcinogen (N-nitrosomethylbenzylamine) induction of esophagus and stomach tumors, and Fhit replacement by gene therapy in these mice induces apoptosis and significantly reduces tumor burden (44).

The second gene affected by homozygous deletion was Sema3e, which is located on RNO4q12 and is homologous to HSA7q21. This gene belongs to the semaphorin/collapsin family of molecules and was found deleted in 11% of the mammary tumors. Christensen and colleagues reported that the expression of Sema3e in murine mammary adenocarcinoma cells correlates with the ability to form experimental lung metastasis in vivo (45, 46) and further showed the frequent expression of Sema3e in human cancer cell lines and solid tumors from breast cancer patients, which support the role of this protein in tumor progression and metastasis. Rat mammary tumors induced by chemical carcinogens rarely develop metastases, and the deletion of Sema3e in some tumors may play a role in this characteristic.

The third genomic locus found deleted encompassed the Zfp533 gene located on RNO3q23 (homologous to HSA7q21) and was absent in only 5% of the rat mammary tumors. Zfp533 belongs to the zinc finger family of transcription factors, and to date, there is no available information about this gene, but it may have a tumor suppressor role in the chemically induced rat mammary tumor model.

In conclusion, we show the importance of selecting a correct paired tissue as reference in tumor genome profiling studies to include all genuine small somatic alterations that may in the future serve as therapeutic targets. Our experiment allowed the identification of immunoglobulin genes as novel cancer targets, an event that may have a direct biological and clinical significance. Thus, control DNA in aCGH experiments should be selected with care, and DNA from pooled spleen or blood may not be the ideal control in the study of primary tumors.

No potential conflicts of interest were disclosed.

Grant support: Healthier Wisconsin Research Initiative (S.L. Sugg and H.J. Jacob), Lady Harley Riders Breast Cancer Research Fund (S.L. Sugg), Medical College of Wisconsin Breast Cancer Research Fund (T. Adamovic, S.L. Sugg, and H.J. Jacob), and Scott and Peggy Sampson Memorial Breast Cancer Research Fellowship (T. Adamovic and H.J. Jacob).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.