Genomics applied to the treatment of breast cancer

Breast cancer remains a major health issue in the world with 1.7 million new cases in 2012 worldwide. It is the second cause of death from cancer in western countries. Genomics have started to modify the treatment of breast cancer, and the developments should become more and more significant, especially in the present era of treatment personalization and with the implementation of new technologies. With molecular signatures, genomics enabled a de-escalation of chemotherapy and personalized treatments of localized forms of estrogen-dependent breast cancers. Genomics can also make a real contribution to constitutional genetics, so as to identify mutations in a panel of candidate genes. In this review, we will discuss the contributions of genomics applied to the treatment of breast cancer, whether already validated contributions or possible future applications linked to research data.


INTRODUCTION
Breast cancer is the first cancer in terms of incidence among women, with 1.7 million new cases in 2012 worldwide. It is also the second cause of death from cancer in western countries with 40,000 deaths per year [1,2]. In the last 20 years, breast cancer mortality has continuously decreased as a result of mass screening programs and early diagnosis, but also as a consequence of improved treatment for both localized and metastatic disease [3,4].
In the era of personalized cancer medicine, advances in genomics are essential assets. In this review, we will address current knowledge in genomics applied to the treatment of breast cancers.

HISTOLOGIC AND MOLECULAR CLASSIFICATIONS
There are schematically three main histologic types of breast cancer ( Figure 1): i) estrogen-dependent breast cancers expressing the estradiol receptor (ER) and treated with a panel of drugs that target the estradiol receptor pathway [5]; ii) breast cancers overexpressing the human epidermal growth factor receptor 2 (HER2) oncoprotein and treated with anti-HER2-based chemotherapies, the first anti-HER2 being a therapeutic monoclonal antibody, trastuzumab [6]; and iii) "triple negative" breast cancers which lack the expression of the estradiol receptor, the progesterone receptor, and HER2. There are still no targeted therapies for triple-negative breast cancers, which Review Figure 1: Breast cancers landscape evolution from histologic to molecular classifications. www.oncotarget.com have a high metastatic potential, and consequently a bad prognosis [7].
Since 2000, several molecular subtypes have been identified, including the aforementioned histological subtypes (Figure 1). Peru et al. identified six molecular sub-types: Luminal A, Luminal B, Luminal C, HER2enriched, Basal-like and Normal-like [8][9][10]. The luminal A subtype has the best prognosis and the lowest proliferative potential, including highly estrogendependent breast cancers. Luminal B subtype is also estrogen-dependent, but to a lesser extent, and corresponds to a sub-group of more aggressive cancers with a higher proliferative potential [11]. For the basal-like subtype, 50 to 75% are triple-negative breast cancers. Their aggressiveness is frequently the result of the loss of the retinoblastoma protein 1 (pRb1), of a high rate of tumor protein 53 (TP53) gene mutations and also of mutations in genes involved in DNA-repair mechanisms, such as breast cancer type 1 susceptibility (BRCA1) gene [12].
This early molecular classification was later revised, with the following subtypes: Luminal A and B, HER2positive/ER-positive, HER2-positive/ER-negative, basal p53-altered and basal p53-normal ( Figure 1). In this more recent version, 7% of cancers are of unknown subtype because of the heterogenic expression of markers [13,14]. Since 2005, among ER-negative cancers, several teams have identified a "molecular apocrine" subtype, characterized by an activation of the androgen receptor (AR) pathway and by the expression of AR target genes in 50% of cases, or by HER2 overexpression [15][16][17]. In 2012, a new molecular classification was established which includes Luminal, molecular Apocrine and Basal-like subtypes ( Figure 1) [18]. Different immunohistochemistry markers have been proposed for in situ characterization of molecular apocrine breast cancers, including androgen receptor (AR) and gross cystic disease fluid protein 15 (GCDFP15) [17].
In 2011, Lehman et al. proposed a molecular classification of triple-negative breast cancers which we will discuss later [19].

MOLECULAR SIGNATURES AND ESTROGEN-DEPENDENT BREAST CANCERS
For estrogen-dependent breast cancers, the contribution of genomics has been small. However, molecular signatures have substantially modified clinical practice, enabling the therapeutic decision for adjuvant chemotherapy to be redefined for localized breast cancers [20]. Three commercially available tests are currently used: MammaPrint ® (Agilent, The Netherlands) [21,22], OncotypeDx ® [23] and PAM50 (Prediction of Microarray using 50 classifier genes plus 5 reference genes) (Prosigna ® kit) (see Table 1) [24,25]. These tests quantify the expression levels of a limited panel of genes in the primary tumor. Most of them were developed for formalin-fixed paraffin-embedded tissue samples for implementation in daily practice.
These molecular signatures are used to classify patients according to their risk of metastatic relapse, to guide the decision for adjuvant chemotherapy when conventional criteria are insufficient [20]. This is particularly true for ER-positive, HER2-negative breast cancers without lymph node involvement (N0): in this sub-group of patients, adjuvant chemotherapy significantly reduces the risk of metastatic relapse only for high-risk patients [26][27][28] [29]. A meta-analysis of 147 articles concluded that molecular signatures for breast cancer enable 10% of patients at high clinical risk of relapse to be reclassified as low-risk patients, thus reducing the use of chemotherapy, with a favorable cost-efficiency ratio and improved quality of life for non-treated patients [30].
For estrogen-dependent breast cancers, the other contributions of genomics remain in the research field. For example, recent studies have identified the presence of mutations of the estrogen receptor 1 (ESR1) gene, such as the D538G or Y537S/C/N mutations, associated with resistance to anti-estrogens. These mutations change the conformation of the ligand binding site, thereby reducing the affinity of tamoxifen for the estrogen receptor [31,32]. The systematic screening for these mutations is not currently recommended. Further studies are required to demonstrate their possible usefulness in guiding hormonetherapy prescription in daily practice.

GENOMICS APPLIED TO THE TREATMENT OF HER2-OVEREXPRESSING BREAST CANCERS
It is recommended to determine HER2 status in the primary tumor or in metastatic samples using a standardized immunostaining method. For doubtful cases, in situ hybridization methods are currently used to determine the HER2 gene copy number [33,34]. In 2013, an international consensus clearly defined the criteria for HER2 protein overexpression and for HER2 gene amplification [35].
However, these methods entail certain limitations, typically for micro-invasive foci in a primary tumor or a micro-metastatic axillary lymph node [36].
Digital droplet PCR (ddPCR), easier to implement than in situ hybridization methods, seems to be a reliable alternative for the evaluation of the HER2 copy number in breast or gastric cancers [37][38][39]. A major limitation could be intra-tumor heterogeneity. Indeed, primary breast cancers are heterogeneous [28], and this is also the case for HER2 status [40]. Combining laser-microdissection with ddPCR overcomes this limitation, by enabling a precise assessment of the HER2 copy number within a cancer sample. In a recent study, we validated the use of laser microdissection combined with ddPCR to assess HER2 copy number in micro-invasive breast cancers with at least 50 invasive cancer cells. We then applied this methodology to a 45-year-old patient with extensive in situ breast cancer, and no associated micro-invasion except a micro-metastasis found only on one section of the sentinel axillary lymph node. We first laser-microdissected the micro-metastatic foci, and then used ddPCR to demonstrate that HER2 was amplified. This led us to optimize the adjuvant treatment for our patient, and she received trastuzumab-based adjuvant chemotherapy [37].
For HER2-overexpressing breast cancers, another issue remains to be deciphered: is there any clinical benefit to be drawn from the systematic determination of HER2 amplification level within the tumor? Singer et al. demonstrated a correlation between the level of HER2 amplification and the response to anti-HER2 treatments in the neoadjuvant setting [41]. Laser-microdissection of cancer cells from a metastatic tumor biopsy combined with ddPCR could be used to overcome signal dilution by enrichment with tumor cells. So, in a skin metastasis of a HER2-expressing breast cancer, we have demonstrated that HER2 copy number evaluated by ddPCR passed from 6 to 34, without and with prior laser microdissection respectively ( Figure 2).

GENOMICS APPLIED TO THE TREATMENT OF TRIPLE-NEGATIVE BREAST CANCERS
Triple-negative breast cancers remain a therapeutic challenge. Lehmann  BL1 subtype is associated with greater chemo-sensitivity, whereas BL2 and LAR subtypes are more chemo-resistant subtypes [42]. Unfortunately, this classification still has few clinical applications, even in clinical trials dedicated to triple-negative breast cancers. For example, the BL2 subtype, characterized by an activation of the epidermal growth factor pathway, could respond to anti-epidermal growth factor receptor (EGFR) therapies. A phase II study assessed the benefit of cetuximab, a monoclonal anti-EGFR antibody, in the treatment of 173 women with metastatic triple-negative breast cancers. The results were disappointing, with a non-significant survival gain of 2.2 months [43]. For anti-EGFR-based treatments, a study could be dedicated exclusively to BL2 subtypes, ideally with high EGFR copy numbers and a quadruple wild-type status of KRAS, NRAS, BRAF, and PIK3CA genes, as for colon cancers [43,44].
The same arguments can be applied to the LAR subtype, characterized by androgen receptor signaling pathway activation, and corresponding to certain molecular apocrine sub-types [16]. The LAR subtype accounts for 11% of triple-negative breast cancers [19]. In two phase II studies using anti-androgens for the treatment of patients with metastatic triple-negative breast cancers, disease stabilization at 6 months was observed for less than 20% of patients [45,46]. It should be noted that the patients were not pre-selected according to LAR subtype.
In 2012 and 2013, we conducted a pilot study among five women with metastatic triple-negative breast cancers. For the five patients, we performed transcriptomic analyses on metastasis biopsies, and classified their respective metastatic cancers according to Lehmann's classification. We also established individual xenograft models from the same metastasis biopsies. For each patient xenograft model, we tested a panel of drugs or drug combinations, guided by transcriptomic data. One patient was classified BL2, with EGF pathway activation and no mutation of the EGF pathway genes. In the corresponding xenograft, the most effective regimen was a combination of paclitaxel and cetuximab. This regimen was offered to the patient as a third-line resort treatment with almost complete metabolic response [47]. On the basis of transcriptomic analyses and chemosensitivity data obtained from the different xenografts, we personalized the resort treatment for the four other women in our study. In all cases, despite the fact that this resort treatment was third-line or fourthline, the time-to-progression was longer than that observed with previous lines of treatment [48].

GENOMICS AND TUMOR HETEROGENEITY
Tumor heterogeneity is probably insufficiently taken into account in daily clinical practice, particularly for the treatment of metastatic disease. Most molecular analyses are performed on primary tumors, even in metastatic stages. However, metastatic clones can be a minority in the primary tumors they are deriving from [49,50]. For HER2 status assessment, we have previously noted the benefit of combining molecular and tissue analyses, particularly with the contribution of laser-microdissection to overcome the limitation of tumor heterogeneity [37-39, 51, 52].
Molecular analyses on metastases are rare, mainly because of difficulties in obtaining these samples, and despite the fact that radiology-guided biopsies have considerably reduced this limitation [53]. For breast cancer, eleven studies have been dedicated to wholegenome analyses of metastatic biopsies (Table 2). In addition, most of these studies only included small numbers of patients, and the genome analyses were generally performed on tumors that were not lasermicrodissected. One large study included 216 metastatic samples, some of them paired with samples from the corresponding primary breast cancers. They showed that metastatic clones are enriched with certain molecular abnormalities compared to the primary tumors [54].
Our research team is conducting a program on brain metastases. As part of this program, we performed transcriptomic analyses on laser-microdissected metastatic lymph-nodes of 28 women with HER2-overexpressing or triple negative metastatic breast cancers. Supervised analyses compared the transcriptomic profiles of women who developed brain metastases with those who did not. We identified CDKN2A/p16 as a gene associated with the risk of brain metastases and decreased survival [55].

GENOMICS AND CONSTITUTIONAL GENETICS OF BREAST CANCER
In the last ten years, a panel of genes has been proposed for the diagnosis of hereditary familial cancers. In the context of hereditary predisposition for breast and/or ovarian cancers, the United States National Comprehensive Cancer Network recommends a panel of nineteen genes and proposes corresponding clinical screening tests (Table 3) [56]. In France, the Genetics and Cancer Group, supported by the French National Cancer Institute, recently updated their recommendation to test a panel of 13 genes accompanied by prevention and screening measures for patients and their families (Tables  4 and 5) [57,58]. With twenty-eight platforms covering the French territory and dedicated to molecular biology, patients at risk for hereditary cancer can have the benefit of recent technologies applied to constitutional genetics. These platforms have implemented high-throughput sequencing tools like the Next Generation Sequencing     [59].
In sporadic breast cancers, NGS sequencing enables four subtypes to be discriminated on the basis of different genetic and epigenetic modifications, with three genes (PIK3CA, TP53, and GATA3) that are modified in more than 10% of patients. Basal-like breast cancers typically harbor mutations in the TP53, RB1, and BRCA1 genes, together with MYC amplifications [60].
Because of the potential therapeutic applications, the identification of inactivating mutations in the BRCA1 gene, a tumor suppressor gene, is important. BRCA1, because of its critical role in DNA repair mechanisms through homologous recombination, is one of the most important genes associated with hereditary breast cancer [61]. More than 75% of BRCA1-mutated breast cancers have a triplenegative phenotype and are classified as basal-like [8]. Constitutional BRCA1 mutations are of high penetrance, occur in 10% of breast cancer patients and in 20% of young women with triple-negative breast cancers [62]. BRCA1 sporadic mutations are also found in 1% of breast cancers, and the promoter can be hypermethylated in 11 to 14% of cases, resulting in BRCA1 gene inactivation [63][64][65].
A deficit in homologous recombination via BRCA1 inactivation has provided the rationale to concomitantly inhibit other DNA repair pathways, particularly the (ADPribose) polymerase (PARP) enzyme pathway. Olaparib, a PARP inhibitor, showed considerable benefit in patients with metastatic BRCA1-mutated breast cancers [66,67]. We, therefore, need to look for BRCA1 germline mutations in patients at risk for hereditary breast cancer. For sporadic breast cancer, particularly for the triple negative sub-type, we also need to identify BRCA1-inactivating mutations in tumors.
In this domain of constitutional genetics, many questions remain unanswered, particularly the translational value of identifying mutations of unknown significance in genes of low to moderate penetrance. The contributions of Genome-Wide Association Studies (GWAS) have not been very great. To date, more than sixty GWAS have been conducted on breast cancer samples. A meta-analysis of these GWAS identified 84 loci of interest possibly associated with an increased risk of breast cancer [68,69]. Numerous low penetrance variants have been identified, without validating their functional significance. One of these variants concerns the oncogene FGFR2, the FGFR2 protein being overexpressed in 5% of breast cancers. This variant corresponds to a single nucleotide polymorphism (SNP) that affects the binding site of FGFR2, thus activating the downstream signaling pathway in a ligandindependent manner [70]. It is necessary to address the potential benefit of targeting FGFR2 for therapeutic purpose. Another SNP, in the 8q24 region, participates in regulating MYC oncogene transcription which is distant from this SNP by more than 300 kb [71]. Most GWAS studies suggest that mutations in low penetrance genes could partially explain genetic predisposition to breast cancer, even though their functional significance remains unclear [72].

CONCLUSIONS
The contribution of genomics applied to the treatment of breast cancer remains moderate. In practice, it is limited to informing adjuvant treatment decisions for early-stage diseases and to HER2-overexpressing breast cancers whatever the stage. However, breast cancers are heterogeneous and complex. Treatments need to be adjusted according to molecular subtypes and guided by the underlying genetic events. Several programs are ongoing to map the complex genetics of breast cancer, using multi-omic approaches such as the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) [73], which will help to take tumor heterogeneity into account more efficiently. Also, the increasingly widespread utilization of NGS will help to decipher the individual molecular complexity of breast cancers. This will rapidly increase the contribution of genomics in shaping breast cancer treatment in the next few years, especially in the present era of personalized treatments.

Author contributions
The following made substantial contributions to conception and design of the paper: Hamdan D, Janin A, and Bousquet G; performed revision and corrections: Leboeuf C, Nguyen TT and Meles S.