Genomewide Scan for Loss of Heterozygosity and Chromosomal Amplification in Breast Carcinoma Using SNP Arrays

Abstract

In an effort to identify novel genes implicated in breast carcinogenesis, a genomewide scan for loss of heterozygosity (LOH) and copy number changes in paired-DNA samples extracted from normal and tumor tissue of frozen sections from women undergoing surgery for invasive breast cancer was conducted. The Affymetrix 10K SNP array was used to examine genomewide LOH of chromosomal regions. The number of LOH events, number of informative loci, percent heterozygosity, and percent fractional allelic loss (%FAL) were calculated. LOH events were detected in all samples, however, the proportion of LOH ranged from 0.1-57.2%. Elevated LOH events were detected in two samples, with a %FAL of 57.2 and 56.2. Chromosomal regions exceeding a threshold value for a p-value curve based on multiple-testing adjusted permutation methods were identified as significant regions of shared LOH across samples. Regions with significant LOH included: 2p25.3; 2p21; 2p16.1 – 2p15; 2q23.3; and, 16q12.1. Chromosomal region 1q32.1 was identified as a region with significant copy number amplification. Regions of LOH and copy number changes identified from this analysis may provide insights into the underlying processes of and genes involved in breast carcinogenesis. This study demonstrates a feasible methodological approach for the assessment of LOH and copy number changes.

INTRODUCTION

Carcinogenesis is a process characterized by genetic instability [1]. Loss of heterozygosity (LOH) and chromosomal amplification are important mechanisms involved in carcinogenesis [2], including breast carcinogenesis [3, 4]. Chromosomal regions exhibiting LOH may contain tumor suppressor genes. A vast number of studies have conducted LOH analysis in breast cancer [4]. In recent years, the use of high density single nucleotide polymorphism (SNP)-based microarray technology has lead to genomewide investigations of LOH permitting the investigation of all chromosomes simultaneously with denser marker spacing, thus improving the resolution of the analysis. Jänne et al [5] demonstrated the efficiency and reliability of the 10K Mapping SNP array in their study comparing this platform to a previous generation array containing ~1500 SNPs as well as to single-sequence length polymorphism methods. Although SNP allelotyping technology has been applied successfully to other cancer types [6-8], to our knowledge, only one published study examined LOH in breast cancer using an early version of this technology [3]. Wang et al. [3] examined the LOH profile of 34 invasive breast carcinomas using the Affymetrix HuSNP chip, containing 1494 SNP loci with an average of 2.57 cM between each SNP marker. Regions 17p, 17q, 16q, 11q, and 14q were identified as the most common LOH sites.

In this study, we examined LOH events and copy number change of chromosomes using DNA extracted from microdissected normal and tumor tissue of frozen sections among women (n=16) undergoing breast cancer surgery. While studies have demonstrated that the use of DNA extracted from formalin fixed tissue is feasible [9], fresh frozen samples are the most optimal source of DNA for these analyses. We used the Affymetrix 10K Mapping SNP array, containing ~11,500 SNPs, to identify genomewide LOH and amplification of chromosomal regions. Regions identified from this analysis may provide insights into the underlying processes of and genes involved in breast carcinogenesis.

MATERIALS AND METHODS

Study Population

The study population included 16 women undergoing surgery for invasive breast cancer at Columbia University Medical Center, New York. Paired normal and tumor tissue were micro-dissected from fresh frozen sections available from the Avon Foundation supported macromolecule bank of the Herbert Irving Comprehensive Cancer Center and confirmed by histology at the Department of Pathology, Columbia University, New York.

Affymetrix GeneChip Protocol

Genomic DNA was extracted from paired microdissected normal and tumor tissue. DNA samples were normalized at 50 ng/μL concentration using reduced TE buffer (10mM Tris-HCl, pH 8, 0.1 mM EDTA). Quantification was assessed from 1μL of sample using the ND1000 spectrophotometer; quality was assessed in duplicate from 1μL of sample per well using DNA7500 chip by Agilent BioAnalyzer 2100. Major peaks were seen at >3kb size. For each GeneChip, 250 ng of genomic DNA sample was digested with Xba-I restriction enzyme and maintained at 4°C. Adaptor Xba was then ligated to the digested DNA and stored at −20°C until PCR amplification. For amplification, diluted adaptor-ligated DNA was used as template and, universal primer (10μM; Affymetrix) was used along with dNTPs and Amplitaq Gold DNA polymerase (Applied Biosystems). PCR reaction was conducted using MJ Research PTC 100 Peltier thermal cycler using denaturation at 95°C for 3 minutes, followed by 35 cycles at 95°C for 20 seconds, 59°C for 15 seconds, and 72°C for 15 seconds, followed by final extension at 72°C for 7 minutes. PCR product was checked using DNA1000 chip in the Agilent Bioanalyzer 2100.

PCR product was purified and concentrated with QIAquick PCR Purification kit (Qiagen). PCR product was quantified using ND1000. Subsequently, 20 μg of PCR product in 45 μL volume was used for fragmentation. PCR product quality was checked on DNA1000 chip with Agilent Bioanalyzer 2100. Typical electropherograms of 12 purified PCR products are shown as overlay in Figure 1. Fragmentation was done using fragmentation reagent (Affymetrix). For QC purpose, 1 μL of 1:10 dilution of fragmented DNA sample was put on DNA1000 chip read by Agilent 2100 Bioanalyzer. The major peak of the electropherogram was seen at 15 – 50 bp size as shown in Figure 2. The fragmented PCR products were labeled with GeneChip DNA labeling reagent (Affymetrix) by adding 19.4 μL of labeling master mix with 50.6 μL of fragmented DNA to make 70 μL reaction volume in 0.2 mL PCR tube. The sample was incubated at 37°C for 2 hours, followed by heat deactivation at 95°C for 15 minutes, and immediately transferred on ice. This 70 μL of labeled DNA was mixed with 190 μL of hybe cocktail in 1.5 ml Eppendorf tube. The samples were denatured at 95°C for 10 min, then the tubes were transferred to ice for exactly 10 sec. The target DNA was placed on heating block at 48°C for 2 min after which 200 μL of the denatured hybridization sample was injected into the probe array. Hybridization was done at 48°C for 16 – 18 hrs at 60 rpm. Thereafter, staining and washing was carried out followed by scanning with high-resolution Affymetrix GeneChip scanner 3000.

Figure 1.

Electropherogram of 12 purified PCR products are overlaid against DNA marker (in yellow) showing uniform pattern of the major amplicons ranging between 700 and 1000 bp size.

Figure 2.

Electropherogram of 12 fragmented PCR products are overlaid against DNA marker (in yellow) showing uniform pattern of fragmentation: the fragments range between throughout the samples ranging between 15 and 50 bp size.

Statistical Analysis

The number of LOH events, number of informative loci, percent heterozygosity, and percent fractional allelic loss were calculated. SNP calls of paired normal and tumor samples were combined to make LOH calls as described by Lin et al. [10]. Briefly, LOH events were identified when the normal sample at a particular marker was heterozygous (AB) and the tumor sample at the same marker was either A or B. The fractional allelic loss (FAL) was calculated as the number of LOH events per number of informative loci. Regions of shared LOH across samples were identified by a p value curve based on multiple-testing adjusted permutation methods as described by Lin et al. [10]. The significance curve was used to identify significant shared LOH regions across samples exceeding the threshold value 0.25 [10, 11]. Additionally, we also performed copy number analysis of DNA. Arrays were normalized for probe signal intensity and a signal value was computed for each SNP. The raw copy number for each SNP was computed as (Signal/(mean signal of normal samples at this SNP)* 2). A Hidden Markov Model (HMM) was used to infer copy numbers as has been described by Zhao et al. [12]. All statistical analyses were performed using the dChipSNP module [10] of dChip software [13].

RESULTS

Sixteen female breast cancer cases were included in this study. The demographic and clinical characteristics of the study population are shown in Table 1. The majority of women were white with an average age at diagnosis of 58.4 years. The majority of tumors were ER positive or PR positive, with 10 (71.4%) positive for both. Approximately 43% of tumors were HER2/neu positive. Most tumors were grade 2 or 3.

Table 1.

Characteristics of study participants.

Characteristic	No. (%)
Sex
Female	16 (100)
Mean age at diagnosis in years (SD)	58.4 (11.0)
Race
White/Non-Hispanic	14 (87.5)
Hispanic	1 (6.25)
Other/Unknown	1 (6.25)
Family history of breast cancer
Yes	7 (43.7)
No	9 (56.3)
ER status
Positive	14 (87.5)
Negative	2 (12.5)
PR status
Positive	10 (62.5)
Negative	6 (37.5)
Stage (TNM)
I	6 (37.5)
II	6 (37.5)
III	4 (25.0)
IV	−
HER2/neu
Positive	6 (42.9)
Negative	8 (57.1)
Grade
1	2 (12.5)
2	7 (43.8)
3	6 (37.5)
Not determined	1 (6.2)

ER, estrogen receptor; PR, progesterone receptor; SD, standard deviation.

The average genotype call rates for tumor and normal samples were 92.8% ± 2.9% and 92.4% ± 7.3%, respectively. On average, 3324 loci were informative corresponding to a heterozygosity of approximately 29.7%.

Table 2 summarizes the LOH findings for the paired samples. LOH events were detected in all samples, however, the proportion of LOH ranged from 0.1-57.2%. This range is consistent with findings from other studies for breast cancer [3, 14]. In particular, elevated LOH events were detected in samples 1 and 10, with a %FAL of 57.2 and 56.2, respectively.

Table 2.

LOH events from normal and tumor DNA pairs

Sample	No. of LOH	No. of Informative Loci	Heterozygosity (%)	FAL (%)
1	1984	3466	30.9	57.2
2	57	3151	28.1	1.8
3	9	3319	29.6	0.3
4	4	3658	32.6	0.1
5	17	3380	30.2	0.5
6	35	3450	30.8	1.0
7	24	3423	30.5	0.7
8	66	2809	25.1	2.3
9	65	3668	32.7	1.8
10	2073	3686	32.9	56.2
11	5	3794	33.9	0.1
12	300	2754	24.6	10.9
13	97	3346	29.9	2.9
14	49	2870	25.6	1.7
15	75	3527	31.5	2.1
16	111	2890	25.8	3.8

No. of mapped loci = 11,205

Heterozygosity (%) = No. of informative loci / No. of mapped loci * 100

Fractional allelic loss (FAL) (%) = No. of LOH / No. of informative loci * 100

Several regions of significant LOH were detected, these included 2p25.3; 2p21; 2p16.1 – 2p15; 2q23.3; and, 16q12.1. All known genes contained within these regions are shown in Table 3. There was no clustering of %FAL by patient characteristics including stage, grade, PR status, ER status, HER2/neu status, family history or race.

Table 3.

Known genes contained within regions of LOH

Cytoband	Accession no.	Gene symbol	Gene description
2p25.3	XM_039762	MYT1L	myelin transcription factor 1 -like
	NM_002936	RNASEH1	ribonuclease H1
	NM_001011	RPS7	ribosomal protein S7
	NM_024027	COLEC11	collectin sub-family member 11
	NM_199232	ALLC	allantoicase
2p21	NM_005400	PRKCE	protein kinase C, epsilon
2p16.1-2p15	NM_006296	VRK2	vaccinia related kinase 2
	NM_018062	FANCL	Fanconi anemia, complementation group L
	NM_018014	BCL11A	B-cell CLL/lymphoma 11A (zinc finger protein)
	NM_022894	PAPOLG	poly(A) polymerase gamma
	NM_002908	REL	v-rel reticuloendotheliosis viral oncogene homolog
	NM_144709	FLJ32312	hypothetical protein FLJ32312
	NM_002618	PEX13	peroxisome biogenesis factor 13
	NM_032506	KIAA1841	KIAA1841
	NM_152392	AHSA2	AHA1, activator of heat shock 90kDa protein ATPase homolog 2
	NM_014709	USP34	ubiquitin specific peptidase 34
2q23.3	XM_371575	PRPF40A	PRP40 pre-mRNA processing factor 40 homolog A
	NM_152522	ARL6IP6	ADP-ribosylation-like factor 6 interacting protein 6
16q12.1	NM_002968	SALL1	sal-like 1

In our analyses of chromosomal copy number variation, the 1q32.1 region was identified as a significant region of copy number amplification. As seen in Figure 3, although several additional chromosomal regions exhibited evidence of copy number variations there changes were less consistently seen across the samples as compared to 1q.

Figure 3.

Copy number summary plot for all samples displaying the proportion of samples with copy number amplification to at least 3 copies in red and copy number reduction to at least 1 copy in blue.

DISCUSSION

The present study demonstrated the feasibility and utility of using fresh frozen tissue for the examination of LOH using microarray technology. The Affymetrix GeneChip Human Mapping 10K Array Xba 131 was used to perform genomewide LOH and copy number change analyses of 16 breast cancer samples using 11,205 SNPs. Frequent allelic loss was seen on chromosomes 2p, 2q, and 16q. LOH on chromosome 2 has been infrequently reported in the existing literature on allelic loss in breast cancer based on traditional methods of LOH analyses. In this study, significant chromosomal loss was found at 2p. This finding for breast cancer has been reported by few studies previously [15, 16]. Additionally, we see chromosomal loss at 2q. Loss at this region was reported by Piao et al. [17] and in a pooled analysis by Miller et al. [4] who showed that loss in region 2q was strong in spite of comparatively few observations. Osborne and Hamshere [18] also demonstrated loss at the 2q region by incorporating data across studies. One explanation for this infrequent finding is that the spacing of markers in this chromosomal region may have been too sparse in previous studies to detect loss in the region. Additionally, LOH at the 2q region could be associated with a subtype or clinical characteristic that has not been sufficiently prevalent within all studies.

Significant chromosomal loss was also found at 16q. Previous studies indicate that loss in this chromosomal arm is typically among the highest loss rates [14, 19]. Additionally, pooled analyses by Miller et al. [4] and Osborne and Hamshere [18] demonstrated significant LOH in this chromosomal region.

Several of the genes contained within the regions of significant LOH in this study have been previously implicated in carcinogenesis. While this study investigated a small sample and inference from these findings is somewhat limited, several potential candidate genes of interest have been identified which should be further examined including: PRKCE, FANCL, BCL11A, and SALL1.

Chromosomal amplification was detected in region 1q32.1. Amplification of this region has been previously reported for both breast cancer as well as other cancers [20, 21]. Examination of genes contained within region 1q32.1 reveal that it is a very gene-rich region, containing approximately 60 known genes. Region 1q32.1 has been identified as a region of genomic amplification in other studies of carcinogenesis. Corson et al. [20] detected amplification of chromosomal region 1q32.1 in retinoblastoma and observed the overexpression of KIF14, located at 1q32.1, in breast cancer cell lines, primary retinoblastoma, lung cancer cell lines, and medulloblastoma. Additionally, SRGAP2 (also known as FNBP2, KIAA0456, or srGAP3) is also located at chromosomal region 1q32.1. The SRGAP2 gene was observed to be amplified and overexpressed in breast cancer cell lines [22]. Additionally, SRGAP2 was overexpressed in melanoma, germ cell tumors, chondrosarcoma and retinoblastoma [22].

The two samples with elevated FAL included: 1) a female with a stage 3 and grade 3 tumor, negative for both ER and PR status, HER2/neu positive, and no family history of breast cancer and, 2) a female with a stage 1 and grade 1 tumor, positive for both ER and PR status, HER2/neu negative, and a family history of breast cancer. Given the sample size of this study, we were not powered to examine associations of clinical characteristics with LOH or copy number change. However, this is an important aspect that we plan to address in future studies with a larger sample.

CONCLUSIONS

In summary, regions of frequent allelic loss detected in breast cancer in this microarray based genomewide study included: 2p25.3; 2p21; 2p16.1 – 2p15; 2q23.3; and, 16q12.1. Additionally, 1q32.1 was detected as a region of chromosomal amplification. The findings from this study are consistent with previous studies of LOH and copy number change in breast cancer based on more traditional analyses. Regions of LOH and amplification identified from this analysis may provide insights into the underlying processes of and genes involved in breast carcinogenesis. This pilot work also demonstrates the utility and feasibility of microarray SNP chips for identifying novel loci involved in breast cancer.

List of Abbreviations

LOH

loss of heterozygosity

%FAL

percent fractional allelic loss

SNP

single nucleotide polymorphism

HMM

Hidden Markov Model

ER

estrogen receptor

PR

progesterone receptor

SD

standard deviation