Array comparative genomic hybridization analysis discloses chromosome copy number alterations as indicators of patient outcome in lymph node-negative breast cancer

Breast cancer is one of the commonest malignancies in women, and mainly occurs in middle-aged or older women with an estimated 1.7million cases and 521,900 deaths in 2012 [1]. For the past decades, the treatment of breast cancer has experienced several changes by tissue-based biomarker and comprehensive analysis of gene expression profiles based on cDNA microarrays has revealed distinct intrinsic subtypes of breast cancer [2‐4]. This intrinsic subtype classification was improved [5] and then modified as “surrogate” immunohistochemical subtypes comprising luminal A-like, luminal B-like (HER2-positive and HER2-negative), HER2-overexpressing, and triple-negative subtypes [6]. This immunohistochemical classification is clinically applied for treatment decisions and prediction of patient prognosis [7‐9].

Lymph node metastasis-negative (pN0) invasive breast cancer was reported to be associated with good prognosis, with 20–30% 10-year recurrence rate [10‐12]. The pN0 breast cancers were classified into subclasses of low and high recurrence risk according to immunohistochemical subtype or histopathological parameters [10‐12]. However, immunohistochemical or histopathological parameters are not quantitative, and low inter-observer reproducibility can be problematic [13, 14]. Therefore, the identification of novel quantitative and reproducible prognostic markers of pN0 breast cancers is of major importance.

Breast cancers have been reported to show a number of genomic alterations, including gene copy number and structural alterations [15‐19]. Gene copy number alterations (CNAs) and gene expression profiles play important roles in carcinogenetic pathways and can serve as potential biomarkers for prognostication and treatment decisions. It has been utilized extensively for studying the characteristics of breast cancer. Recently, the CNA profiles of ASB13 and SGCZ genes have revealed significant association with survival outcome in young women with breast cancer [20].

In the present study, we examined the CNA profiles of 800 cancer-related genes in 51 pN0 invasive breast cancers using array-based comparative genomic hybridization (aCGH). pN0 invasive breast cancer has been considered to be at low risk of recurrence for more than 5 years after radical surgery. However, it has been reported that the routine examination of regional lymph nodes may be inadequate for the detection of obscure metastases, and that micrometastases were only identified through labor-intensive multiple sectioning and additional immunostaining. Therefore, other potential biomarkers are needed in the decision-making process for treatment. Based on this background, we attempted to identify gene CNAs that were associated with relapse of cancer and patient death, and we were able to identify several gene alterations that may be useful as prognostic biomarkers.

Although numerous studies have examined clinicopathological correlation with CNA status [15‐19], limited studies have explored the prognostic implication of CNA status in pN0 breast cancer. For instance, the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) cohort data has demonstrated integrative cluster associations with histopathological subtypes, tumor grades, and lymphocyte distributions [25]. However, as no information is available on well characterized clinical data with long term follow-up in the study, they did not directly explore the prognostic implication of CNA status in pN0 breast cancer. In our study, based on the total number of CNAs and on CNAs of specific chromosome loci using array CGH technology, we attempted to identify high- and low-risk types of pN0 breast cancer. At first, we identified that there were no differences in the total number of CNAs or of CNAs in the specific loci between the relapsed and non-relapsed group. Therefore, we performed an unsupervised hierarchical cluster analysis based on the array CGH dataset from all 51 cases examined.

Based on this cluster analysis, these 51 cases were classified into three clusters, groups 1 to 3, which corresponded to immunohisochemical subtypes. Group 1 was mostly composed of cases belonging to the triple-negative subtype, group 2 of luminal B-like and HER2-overexpressing subtypes, and group 3 of luminal-A like cases.

Such strong correspondence between CNA pattern and “intrinsic subtype” classification was shown in previous reports [15, 26, 27]. Hicks et al. proposed three patterns of genome rearrangements in breast cancer, i.e., sawtooth, firestorm, and simplex. They suggested that cases of the firestorm pattern were frequently Grade 2 or 3 and were accompanied by CNAs of chromosomes 6, 8, 11, 17, and 20, and especially by amplifications of CCND1 in 11q and of ERBB2 (HER2) in 17q [26]. In the present study, group 2 was mostly composed of luminal B-like and HER2-overexpressing subtypes, and frequently carried 1q, 8q, 16p, 17q gains, which was similar to the firestorm CNA pattern [27]. Hicks et al. showed that the simplex pattern was frequent in grade 1 tumors and was characterized by 16q deletion and 16p duplication, often coupled with 8p loss and duplications of 1q and 8q [26]. These characteristics of the simplex pattern were very similar with those of the present group 3, characterized by luminal A-like subtype, gains of 1q and 16p, and loss of 16q.

In contrast, Natrajan et al. demonstrated that the sawtooth pattern was characterized by high histological grade and basal-like subtype, and numerous CNAs were commonly detected [26]. These CNAs included 5q loss and gain of 1q21.1, 8q24.21-q24.23, and 12p13.31, which agreed with the results for group 1 tumors.

Andre et al. classified breast cancer cases into three groups, i.e., non-negative matrix factorization (NMF) classes I to III, according to data of CGH and cluster analysis [15]. A total of 65% of NMF class I cases were triple negative, and 6p gain, 5q loss, and 15q loss were reported to be common. NMF class II included most of the HER2-overexpressing subtype, and was characterized by 17q gain corresponding to HER2 amplification. In NMF class III, 73% were ER-positive, 97% were HER2-negative, and 1q gain and 16q loss were common.

With regard to patient outcome, the OS of group 1 was significantly worse than that of group 2. Unfortunately, we could not identify any significant clinical/biological association with outcome in the non-relapsed groups, but among the relapsed cases, group 1 patients showed shorter OS and RFS than did group 2. Furthermore, we showed that 5q loss, 12p gain, and absence of 16p13.3 gain were correlated with worse patient outcomes in all cases and/or subsets of recurrent cases in pN0 breast cancer patients.

5q15 loss was commonly detected in group 1, and was correlated with shorter OS in all 51 cases and shorter OS and DFS in the 19 relapsed cases. Curtis et al. classified 2000 breast cancer samples into 10 integrated clusters (IntClust 10) composed mainly of a basal-like subtype that frequently exhibited 5q loss, 8q gain, 10p gain, and 12p gain [16]. These characteristic CNAs in the IntCrust 10 were partly compatible with those in the present group 1.

Curtis et al. also indicated that 5q harbored genes encoding numerous signaling molecules or transcription factors, cell division genes, and 5q deletions that can modulate the coordinate transcriptional control of genomic and chromosomal instability and cell cycle regulation within cancer cells [16].

One of target genes at the 5q15 locus was nuclear receptor subfamily 2, Group F, Member 1 (NR2F1) [28]. NR2F1 provokes a reduction in chemokine CXCL12 expression and enhancement of CXCR4 expression, and it stimulates breast cancer cell migration [29]. On the other hand, NR2F1 also functions as a dormancy gene, so suppression of this gene results in growth of ER-positive MCF-7 cells in vivo [29, 30]. Based on these observations, enhanced migration and release from the dormant state in tumor cells by the loss of NR2F1 may be associated with earlier relapse [28].

The 12p13.31 gain was selectively detected in group 1, whereas 16p13.3 gain was only detected in groups 2 and 3. The 12p13 gain was shown to be common in the CNA in basal-like subtype [27]. In contrast, 16p gain was shown to be frequent in luminal subtype, and it was correlated with a good prognosis [16].

One of the target genes at the 12p13.31 locus is TNFRSF1A. This gene encodes a receptor of tumor necrosis factor-α (TNF-α), and their ligand-receptor interaction is conditional for the presentation of cellular growth, invasion, and metastasis. The development of primary cancers and metastases were inhibited in TNFRSF1A-deficient mice [31]. TNFRSF1A expression is reportedly associated with poor prognosis in diffuse large B-cell lymphoma [32]. In addition, an RNA sequence analysis showed that the breast cancer-associated fusion transcript SCNN1A-TNFRSF1A may play a role in the development of breast cancer [33]. Furthermore, Egusquiaguirre et al. have demonstrated that elevated TNFRSF1A levels may predict a subset of breast tumours that are sensitive to STAT3 transcriptional inhibitors [34].

One of the target genes at the 16p13.3 locus is ABCA3, which encodes an ATP-binding cassette (ABC) transporter or a family of transmembrane proteins that can transport a wide variety of substrates across biological membranes in an energy-dependent manner [35]. Negative ABCA3 cytoplasmic immunoreaction or decreased ABCA3 expression was significantly associated with lymph node involvement and worse clinical outcome [36].

All relapsed cases showing 5q15 loss or 12p13.31 gain experienced relapses within 25 months after tumor resection. Therefore, 12p13.31 gain and 5q15 loss were considered markers of pN0 breast cancers showing highly aggressive clinical behavior, with most recurrences being triple-negative breast cancer (TNBC).

Limitations of this study include its small scale and its retrospective case-control design. The number of genes mounted on the array was 800. Nonetheless, we were able to show specific CNAs that were correlated with worse or better patient prognosis using array CGH technology in invasive pN0 breast cancers. The data acquired were compatible with the results of stricter studies using current single nucleotide polymorphism (SNP) arrays, and are considered applicable for the identification of high-risk pN0 breast cancer, after the results are confirmed for larger scale cohorts.

In conclusion, a CGH array analysis of pN0 invasive breast cancers sorted the cases into three distinct clusters according to an unsupervised hierarchical cluster analysis. We were not able to identify CNAs specific to the non-relapsed group that could be applicable for avoiding unnecessary adjuvant chemotherapy. However, we did identify several specific CNAs as prognostic markers, i.e., 5q15 (NR2F1) loss, 12p13.31 (TNFRSF1A) gain, and absence of 16q13.3 gain, in all pN0 cases and in the high-risk group of pN0 cases that showed relapse. These specific CNAs could be potential candidates for prognostic biomarkers in pN0 invasive breast cancer for monitoring occult metastases which cannot be detected by present histologic examination of the lymph nodes.