Introduction
The incidence of breast cancer continues to increase globally. According to the most recent Globocan estimates (2012), breast cancer ranks as the fifth highest cause of mortality from cancer overall, and is the most frequent cause of cancer mortalities in women in developing countries (1). In Africa, there is a substantial variation in the estimated breast cancer incidence across different regions, which may be attributed to the exposure to different environmental factors (2). South Africa (SA) presents the highest incidence of breast cancer, affecting primarily young women (≤50 years old), who are diagnosed at a mean age of 10 years younger compared with women in Western countries. However, the mortality rates in SA remain high (~17%) among female populations of all ages, potentially due to the late stage of the disease at diagnosis (1). The high rates of incidence and mortality are due to multiple and complex social, cultural and economic factors that directly impact breast cancer diagnosis and response to treatment. The health care system in SA and other Sub-Saharan countries, in general, face inadequate health personnel, offer poor health insurance programs and limited facilities for cancer screening and treatment, and are compounded by prohibitive costs for breast biopsies and pathological exams (3). Other factors involve lifestyle-associated risks including smoking, alcohol consumption and high fat diets (4). In addition, several studies have demonstrated that in the Sub-Saharan region, including SA, there is an over-representation of breast cancer subtypes that denote poorer prognosis, most notably among pre-menopausal women (5-9). Although the standard classification of breast cancer, based on three surrogate immunohistochemistry (IHC) markers, the estrogen (ER), progesterone (PR) and human epidermal growth factor receptor 2 (HER2) receptors, is not routinely performed in the majority of SA cancer centers, but studies have demonstrated the prevalence of hormonal negative tumor subtypes, including a high rate of triple negative breast cancer (TNBC) and HER2-positive tumor subtypes. These two subtypes are known to present a clinically aggressive phenotype, including resistance to treatment and high rates of disease recurrence (7-9). Although the determinants for the high prevalence of these unfavorable tumor subtypes is unclear, several studies implicate intrinsic genetic differences, such as the presence of founder mutations in breast cancer susceptibility genes, including the BRCA1 gene (10-12) or unknown genomic variants identified by genomic wide association studies (13-16).
The extensive characterization of the molecular profiling in breast cancer with the use of multiomics platforms, have allowed for extraordinary progress in the understanding of the biological factors and their clinical impact on each breast cancer subtype (17,18). In African countries, however, they remain poorly characterized, mainly due to the unavailability and the high costs of such platforms in addition to the poor annotation of patients’ clinical data (19). Among the African countries, SA has the highest number of published genomic research studies (assessed between 2004-2013), however only 6.1% of these studies are associated with cancer (20). To the best of our knowledge, few breast cancer studies have characterized patient genomic profiles and/or specific molecular alterations in their populations (14,21-25). As a consequence, there is limited knowledge on their tumor biology and the corresponding lack of clinically relevant biomarkers that would be particularly beneficial for the prognosis and treatment of patients with breast cancer representative of the diverse racial groups of SA.
The integrity of the genome is crucial for tumor suppression. Genomic instability, one of the hallmarks of cancer, can be evidenced by the presence of DNA copy number changes (26,27). The presence of gains and losses in specific cytobands of chromosomes is a clear evidence of loss of the genomic integrity control cancers (28). The recurrence of these specific alterations, which may affect the function of cancer driver genes mapped in these regions, including oncogenes and tumor suppressors, occurs during tumor progression and may also facilitate this process (29).
In the present study, the main aim was to determine the patterns of the genome-wide copy number alterations (CNAs) and their potential impact on functional cellular pathways targeted by cancer driver genes in patients with breast cancer from the Western Cape region of SA. This study further assessed the correlations between the genomic results with the patients clinicopathological data. In addition, considering the extensive genetic diversity of the SA population, composed of groups originating from Europe, Asia and various African countries, the present study also aimed to determine whether the main copy number patterns observed were associated with patient’s ethnicity.
Materials and methods
Sample population
Formalin-fixed paraffin-embedded (FFPE) breast tumor specimens were obtained from 28 female patients from the archives of the Division of Anatomical Pathology, National Health Laboratory Services, Groote Schuur Hospital (Cape Town, SA). These cases were obtained retrospectively between 2006 and 2010 with ethical approval from the Groote Schuur Hospital (Human Research Ethics Committee protocol no. 454/2010); patient identities were kept anonymous at all stages of the study. The specimens were collected at diagnosis and prior to any cancer treatment. The clinical data pertaining to age, ethnicity, tumor stage and receptor status were obtained for each patient. The mean age of the study population at diagnosis was 48.2±12.9 years, ranging from 22 to 80 years old. All cases were of infiltrating ductal carcinoma and the mean tumor size was 40.6±18.9 mm. Tumor types were classified according to the Tumor Node Metastasis staging system (30) and were of grade 1, 2 and 3 in 16%, 44%, and 40% of the cases, respectively. Stages I, II and III were observed in 7.4%, 22.2%, and 70.4% of the cases, respectively. The cases were evaluated for ER, PR and HER2 receptor status by IHC, as previously described (31,32). ER and HER2 positivity was observed in 57.1% and 53.8% of cases, respectively. PR status was available for only 21.4% of the patients, out of which 33% (2/6) were positive. Three patients were of the TNBC subtype.
Information regarding ethnicity was obtained from self-reported records. The majority of the patients were of the Colored group (57.1%), while 21.4% were White, 17.9% were Black and 3.6% were Asian. Based on the South African census nomenclature and definitions, these categories refer to people with common characteristics in terms of history and descent, which were mostly defined in the Apartheid period (31). The Colored group is composed of highly admixed individuals who present ancestry from Europe and Asia and various indigenous tribes of SA. They form the majority population of the Western Cape region, and genetic studies have demonstrated that they are mainly descendants of the Khoisan tribe (31,32). White refers to people of European descent, mainly from the Netherlands, Germany, France and England. Black refers to individuals of African tribe origin including Zulu, Xhosa and Basotho tribes (31). The samples were not selected based on any of the clinical information and ethnicity. The main inclusion criteria, subsequent to institutional review board local approval, included patients with invasive breast cancer and with known hormonal and ERBB2 status and clinicopathological data. The distribution of the clinicopathological data and ethnic groups in the studied population is presented in Table I.
Array-comparative genomic hybridization (array-CGH) analysis
DNA copy number analysis was performed using an oligonucleotide array-CGH platform (SurePrint G3 Human CGH Microarray 8×60K; Agilent Technologies Inc., Santa Clara, CA, USA), according to the protocol for FFPE samples that were established as previously described (33,34). Prior to DNA isolation, 5-μm sections of FFPE tissue from each case were evaluated by a pathologist from the Division of Anatomical Pathology, University of Cape Town (National Health Laboratory Service, Groote Schuur Hospital, Cape Town), to confirm the tumor histopathology and the presence of breast epithelial cancer cells. These cells were identified and microdissected in tissue sections that were 5 μm thick from consecutive unstained slides to ensure the array-CGH analysis of a pure tumor cell population. DNA was isolated using the standard phenol-chlorophorm method as was previously described (32). Normal (reference) DNA was prepared from the peripheral blood of a pool of multiple healthy female donors, as previously described (35,36). Equal amounts of tumor and reference genomic DNA (300-500 ng) were digested and enzymatically labeled using the SureTag Complete DNA Labeling kit (Agilent Technologies, Inc.) and hybridized to the arrays, according to the manufacture’s protocols. The array data was analyzed using the Feature Extraction v.10.10 and Agilent Cytogenomics v.3.0 software (Agilent Technologies Inc.), (ADM) after aberration detection method-2 algorithm, a threshold of 6.0 and an aberration filter with a minimum number of >3 probes. Copy number gains and losses were defined as previously described (34): Minimum mean absolute log2 ratio [intensity of the Cy5 dye (reference DNA)/intensity of the Cy3 dye (test DNA) value] of ≥0.25 and ≤-0.25, respectively, as per the array-CGH analytics analysis. Amplifications and deletions were determined using values of log2≥2 and log2≤-2, respectively.
Functional enriched pathways
The identification of the genes and microRNAs (miRNAs) mapped in the cytobands that presented CNAs was obtained from the Agilent Cytogenomics v.7.0 (Agilent Technologies, Inc.) interval base reports (based on the analysis parameters described above). DIANA-miRPath v.3 (37) was used to perform the pathway enrichment analysis, based on the Kyoto Encyclopedia of Genes and Genomes database (37). miRNA targets were identified using TarBase v.7.0 (38). Only miRNA/mRNA targets that presented a miRNA Target Gene (miTG) score >0.7 based on the microT-CDS (39) interactions were included.
Correlation of CNAs with clinicopathological data and ethnicity
Spearman bivariate correlation analysis was performed to determine the association among the number of CNAs and the most frequent affected cytobands with patient clinicopathological data (age <50 or ≥50 years, mean tumor size + and - standard deviation, tumor grades 1-3, tumor stages I-III, ER and HER2 expression status) and ethnicity. Data is presented as Spearman’s correlation coefficient (r), and P<0.05 was considered to indicate a statistically significant difference. Statistical analysis was performed using SPSS Statistics 20 (IBM Corp., Armonk, NY, USA).
Results
DNA copy number changes
Array-CGH analysis was performed for all cases. The CNAs ranged from 1 to 25, with a mean number of 8.68±6.18 CNAs per case. Twelve cases presented CNAs above or equal the mean and 16 cases below. The combined genomic profile of the cases analyzed are presented in Fig. 1. The most frequent affected chromosome cytobands were Xp22.3 and 6p21-p25 (57.14% of the cases), 19p13.3-p13.11 (35.7%), followed by 2p25.3-p24.3, 4p16.3-p15.3, 8q11.1-q24.3 and 16 p13.3-p11.2 (32.14%) (Table I). Chromosomes 2p and 4p were mostly affected by losses and deletions of copy number, whereas chromosomes 8q, 16p and Xp by gains and amplifications, and chromosome 6p by gains and amplifications (mainly 6p22.2-p21) and losses/deletions (mainly 6p25-p24).
A higher number of CNAs (CNAs ≥8.68) were significantly correlated with the cases that present with CNAs in the 6p (r=0.458; P<0.05), 16p (r=0.486; P<0.01) and 19p (r=0.486; P<0.01) cytobands. CNAs on 2p were significantly correlated with CNAs on 4p (r=0.580; P<0.001); CNAs on 4p were significantly correlated with CNAs on 6p (r=0.442; P<0.05) and Xp (r=0.596; P<0.001); CNAs on 6p were significantly correlated with CNAs on 19p (r=0.697; P<0.001) and Xp (r=0.563; P<0.01) and CNAs on 16p were significantly correlated with CNAs on 19p (r=0.542; P<0.01).
The identification of the genes mapped in these cytobands, using the Agilent Cytogenomics interval base reports, revealed a total number of 2,719 genes and 93 miRNAs (Table II). To investigate the function of these specific genes and miRNAs that may be affected by the presence of CNAs in these cytobands, pathway enrichment analysis (DIANA-miRPath v.3) was used. Among the top 15 pathways identified, based on P-value, were pathways associated with prion diseases, fatty acid biosynthesis and metabolism, adherens junction, extracellular matrix (ECM)-receptor interaction, hippo and tumor protein p53 (TP53) signaling pathways (Fig. 2; Table III). Among these top pathways, the ones that were affected by the largest number of miRNAs were the adherens junctions (14 miRNAs), fatty acid metabolism and the viral carcinogenesis pathway (12 miRNAs each) (Table III).
Using miRNA target prediction analysis (using TarBase v.7.0), 512 targets predicted to be regulated by these miRNAs were identified (miTG score >0.7- microT-CDS interactions). A number of common miRNA targets were observed within these 15 pathways, including the cancer driver genes CCND1 in 46.7% of the pathways, CDKN1A, MAPK1 and MDM2 in 40%, SMAD2 in 33.3% and CCND2, IGF1R, KRAS, SMAD3 and TP53 genes in 26.7% (data not shown).
The integration of these miRNAs gene targets with the 2,719 genes that were mapped in the cytobands with CNAs, as generated by the Cytogenomic interval base reports (Table III), revealed a number of 65 common genes (Fig. 3). From these common genes, that may be potentially affected by copy number and miRNA expression regulation, 35.4% were mapped at the 6p25.3-p21.1 cytoband, 18.5% at 8q11.1-q24.3, 17% at 16p13.3-p11.2 and 19p13.3-p13.11, 7.7% at Xp22.33-p21.1, 3% at 4p16.3-p15.2 and 1.5% at 2p25.3-p24.3. The majority of these 65 common genes were involved in viral carcinogenesis (33.8%) and cell cycle pathways (16.9%) (Table IV). Among these genes are known cancer driver genes, including CDKN1A, MAPKs, MYC and VEGFA.
Correlation of CNAs with clinicopathological parameters and ethnicity
Prior to the association between the clinicopathological data and ethnicity (Table I) with the copy number data, a significant correlation between tumor size and tumor stage (r=0.532; P<0.01) was observed. Inverse correlations were observed between ER status and tumor grade (ER positivity and lower tumor grade r=-0.664; P<0.001) and HER2 status and tumor size (r=-0.408; P<0.05). No correlation was observed among the other clinicopathological parameters and/ or the diverse ethnic groups.
For the association with the number of CNAs, the cases were divided into two groups; one with the number of CNAs higher and one lower than the mean number of CNAs observed among the cases (≥8.68 and <8.68, respectively). Unpredictably, the cases in the group with a lower number of CNAs were significantly associated with higher tumor size (r=-0.401; P<0.05) and tumor grade (r=-0.516; P<0.01). The association of the most frequent cytobands affected with the clinicopathological parameters above, revealed correlation of CNAs on 4p (r=-0.452; P<0.05) and 6p (r=-0.491; P<0.05) with tumor grade. No significant association was observed among CNAs and/or specific cytobands with any ethnic group.
Discussion
Genome-wide molecular signatures are well-defined and established in breast cancer, characterizing the distinct intrinsic subtypes and their impact on prognosis and clinical outcome (18,19). However, there are limited genomic signatures characterizing breast cancer based on the ancestry of a patient, particularly associated with their tumor somatic profiles. In patients of African descent, the majority of the studies available are based on genetic variants, including single nucleotide polymorphism and copy number variation (CNV) (13,15,16). While they are relevant in determining cancer risk and susceptibility in specific populations, they fail to characterize the tumor genomic signatures and corresponding imbalances in gene dosages and mutations that are prevalent and may affect cancer genes.
CNAs in the tumor genome may result in the gain/amplification of oncogenes and/or loss/deletion of tumor suppressor genes and are major drivers of tumor development (27,40). A genome-wide analysis of CNAs conducted in different types of cancer revealed that a specific tumor type presents 17% of regions with amplifications and 16% of deletions, compared with <0.5% in normal samples (40). The patterns of these somatic tumor alterations, as CNVs, may also vary within ethnic groups (35,41). The majority of the studies assessing these alterations, in addition to their gene and epigenetic expression profiles, are in White European and African American patients with breast cancer (35,40-45). Few studies have been performed in patients from Sub-Saharan countries (24,25). Thus, the consequences of CNAs due to the genetic instability of their tumor types remain largely unknown.
In the present study, genome-wide copy number profiling in patients with breast cancer patients from the Western Cape region of SA was performed. Array-CGH analysis revealed recurrent CNAs in the cases analyzed, with the most frequent affecting the cytobands 2p25.3-p24.3, 4p16.3-p15.3, 6p25.3-p22.3, 6p22.3-p.21.1, 8q11.1-q24.3, 16p13.3-p11.2, 19p13.3-p13.11 and Xp22.3. A number of these CNAs, including gains on 8q and 16p, are commonly reported in breast cancer, and are not necessarily associated with specific tumor subtypes and/or ethnicity (18). A number of reports, however, have demonstrated that patients with TNBC and other BRCA1-associated tumor types, which are frequently identified in women of African descent (46,47), present a significantly higher frequency of focal amplifications on 8q24, where C-MYC is located (48). This amplification was, however, not a distinguishing characteristic associated with ancestry, as it was observed both in patients of European and African-American descent (18).
Interestingly, losses on 2p, 4p and gains on 6p, 8q, 16p and 19p were previously reported in the array-CGH analysis of a set of 28 breast cancer cases from patients from Sub-Saharan Africa (24). Among these alterations, gains on 6p presented a significant higher level of copy number increase when compared to American patients with breast cancer, previously analyzed (49). In a previous study in African-American (AA) and Non-Hispanic White patients (NHW) patients with breast cancer (35), it was observed that in the AA group there was a higher frequency of copy number gains, affecting 6p and 8q, when compared with the NHW group. However, in these cases, a significant higher level of genome-wide CNAs was observed in the AA group, which may justify the observed increased in the copy number in these chromosome regions. In fact, CNAs were the only significantly different variable between AA and NHW patients with TNBC tumor subtypes (P<0.01) when clinicopathological parameters were taken into account (35).
A direct correlation was observed among the total number of CNAs and the main cytobands observed; a higher number of CNAs were significantly correlated with the cases that presented with CNAs in the 6p, 16p and 19p cytobands. Direct correlations were observed among CNAs on 2p, 4p, 6p, 16p, 16p and Xp. No inverse correlations among the main cytobands affected by CNAs were observed.
The mean number of CNAs and the specific cytobands affected were correlated with clinicopathological parameters and the diverse SA ethnic groups. Surprisingly, a negative correlation was observed between the mean number of CNAs and tumor size and grade. CNAs on the 4p, 6p and Xp cytobands were also negatively correlated with tumor grade. In general, the increase in genomic instability is associated with poor prognostic parameters (50). However, a number of observations have demonstrated that there are clonal subpopulations of tumor types, including tumor types of epithelial origin, with high levels of somatic CNAs that are associated with a potentially better prognosis (and overall survival time) compared with tumor types with intermediate levels of CNAs (50-53). This interesting observation reflects the well-known intra-tumor heterogeneity marked by the presence of coexisting clones with diverse genomic compositions (52,53). This may explain the unexpected results of the present study, as it was observed that there were higher levels of CNAs in the aforementioned cytobands in tumor types with lower tumor grade and smaller size.
Alterations on 8p were significantly correlated with PR status; however, this correlation may be spurious, considering that only 6 cases presented PR status information. CNAs on chromosome 21p were the only ones associated with ethnicity; no other significant differences were observed between the mean number of CNAs and/or the other most commonly affected cytobands (2p, 4p, 6p, 8q, 16p, 19p and Xp) and ethnicity, indicating that these alterations occur in SA patients irrespective of their genetic descent. Furthermore, multivariate analyses revealed that no clinical or pathological variable, including age, tumor size, grade and stage, and ER, PR and ERBB2/HER2 status were significantly correlated with ethnicity of the patients. However, the comparison of the mean number of CNAs of these SA patients to the mean number of CNAs present in the AA and NHW patients of our previous study (35) revealed significant differences between the Black SA and AA patients (P<0.05), but interestingly no differences between the White SA and NHW patients (P>0.05), which may indicate a common European ancestry. However, when the Colored SA group of patients were compared with the AA and NHW groups, significant differences were observed (P<0.0001 and P<0.001, respectively), which may be resultant of the large variation in ethnicity in this particular group of patients (from Europe and Asia and various tribes of SA).
Finally, by performing functional enrichment analysis of genes and miRNA targets mapped in the cytobands most affected by CNAs, it was observed that they affected a number of critical cancer-associated pathways, including fatty acid biosynthesis and metabolism, adherens junctions, ECM-receptor interaction, hippo and TP53 signaling pathways. These pathways include a number of cancer driver genes associated with breast cancer (54), including CCND1, CDKN1A, MAPK1, MDM2, TP53 and SMAD2. A number of the pathways observed in the present study were also identified in the AA patients with breast cancer with TNBC of a previous study (35), in which a direct integration of copy number and miRNA data conducted on the same samples was performed. Remarkably, CNAs are one of the mechanisms that may result in miRNA dysregulation (55-57). In fact, 15/26 miRNAs of the miRNAs that were identified in the AA patients (of the TNBC subtype) (35) were mapped in cytobands that are frequently affected by recurrent CNAs in breast cancer, including the 6p25.3-p21.1, 8q11.1-q24.3, 16p13.3-p11.2 and 19p13.3-p13.11 similar to the present study.
In conclusion, the present study presents the pattern of copy number changes in the genome of SA patients with breast cancer and demonstrates their potential impact on critical cancer pathways associated with their tumor genomes. Additional genomic tumor profiling in SA and other Sub-Saharan patients with breast cancer in independent and larger sample size cohorts are required in order to determine the association of these alterations with their prognosis and clinical outcome, including response to therapy.