Characterization of basal-like subtype in a Danish consecutive primary breast cancer cohort

Abstract

Background: Transcriptome analysis enables classification of breast tumors into molecular subtypes. BRCA1/2 predisposed patients are more likely to suffer from a basal-like subtype and this group of patients displays a more distinct phenotype and genotype. Hence, in-depth characterization of this separate entity is needed.

Material and methods: Molecular subtyping was performed on a consecutive and unselected series of 1560 tumors from patients with primary breast cancer. Tumors were classified by the 256 gene expression signature (CIT) and associated with basic clinical characteristics and aggregated expression levels in the hallmark gene sets.

Results: Of the 1560 samples, 168 were classified basal-like and 120 patients were screened for BRCA1/2 mutations, resulting in 19 BRCA1/2 carriers, 95 non-carriers and six patients carried variants of unknown significance. The BRCA1/2 carriers were significantly younger and there were no carriers above 60 years of age. The tumors showed a loss in DNA-repair profile, as well as an upregulation in proliferative cancer signaling pathways. A robust molecular signature for identification of the BRCA1/2 - carriers was infeasible in the current cohort. Patients with a basal like breast cancer had the lowest median age and the largest median tumor size. They were almost exclusively diagnosed in disease stage III.

Conclusions: Basal-like subtype is indeed a separate entity compared with other molecular breast cancer subtypes and the clinical course for this patient group should reflect the aggressiveness of this cancer. Taken together, patients being diagnosed with a basal-like breast cancer are in the youngest segment of breast cancer patients and are mainly diagnosed in stage III disease.

Introduction

Breast cancer is the most common malignancy in women and despite considerable advances in early detection, diagnosis, and treatment, breast cancer remains one of the leading causes of cancer death [1]. Breast tumors are heterogeneous and are the first solid malignancy where specific molecular treatment factors were introduced [2–4]. Breast tumors can be classified into at least four intrinsic subtypes. One of the key factors distinguishing the different subtypes is receptor status; Luminal A and Luminal B are estrogen receptor (ER) and progesterone (PR)-positive and receptor tyrosine-protein kinase erbB-2 (ERBB2)-negative. The ERBB2 positive subtype is characterized by the ERBB2 positive samples and finally the Basal-like subtype is predominated by receptor negative samples [5]. Based on signaling pathways, copynumber alterations, histopathological and clinical features, including metastatic sites and relapse free survival, Guedj et al. refined the taxonomy by introducing six stable molecular subtypes assigned with normal-like, luminal A, luminal B, luminal C, molecular apocrine and basal-like [6]. The classifier provided by Guedj at al. (CIT classifier) captures the characteristics of the four major subgroups of breast cancer patients [7] and at the same time enables finer clustering associated with distinct clinical and molecular characteristics, e.g., the molecular apocrine subtype is an ER/PR/HER2-negative but Androgen receptor positive subcluster with a poor prognosis. Moreover the CIT classifier identifies the normal-like samples which share many molecular features with the luminal A subtype, including low proliferation, however the normal-like subgroup exhibits improved prognostic behavior [6]. Furthermore, the CIT classifier is readily available as open-source software that uses microarray expression data as input. Another advantage of running the classifier based on microarray data is their reusability, i.e., proliferation index and receptor status can be computed based on the same microarray data.

The basal-like subtype is the most distinct type of the intrinsic subgroups and has more common molecular features with squamous cell lung cancer than with the luminal A or B subtypes [7,8]. Hence, this supports the hypothesis that basal-like subtype should be considered a separate entity when breast cancer is assessed [9]. The neoplastic cells of basal-like breast cancers generally express genes in common with normal basal or myoepithelial cell profile of the breast [3,10]. Furthermore, basal-like cancers are predominated by the lack of expression of hormone receptors; ER, PR and of ERBB2 – receptors [11]. If diagnosed by immunohistochemistry (IHC), the assigned diagnose is triple-negative breast cancer (TNBC), still the terminologies tend to be used interchangeably, whether it is IHC or molecular signature derived tumor classification. This is unfortunate because the therapeutic responses differ between basal-like tumors and other TNBC that are non-basal-like [12]. All though the basal-like subgroup has shown to be a more homogenous group on the molecular level than the TNBC diagnosed by immunohistochemistry, patients assigned with this subtype have demonstrated some diversity in their outcome [13]. In fact, a recent study proposed a molecular signature for identification of two separate groups within the basal-like subtype, with significant differences in patient survival outcomes [14]. Basal-like breast cancers are known for their high proliferation rate and their aggressive behavior and patients suffering from this molecular subtype have a poor prognosis and a short-term disease free survival (DFS) as well as overall survival (OS) [12,15].

Studies have shown a higher prevalence of BRCA1/2 predisposing mutations among patients with TNBC and basal-like subtype [15,16] and patients with either type of these two breast cancer types are frequently BRCA1 carriers [17,18]. Normal functioning BRCA1/2 proteins suppress genome instability by promoting the homologous recombination repair (HRR)-system [19,20]. Homologous recombination repair is essential to avoid DNA double-strand breaks, often caused by UV light and metabolic processes, and is a major error-free DNA-repair pathway [21,22]. Therefore, basal-like subtype, enriched in patients with germline pathogenic variants in BRCA1/2, may reflect deficiency in HRR and loss of DNA-repair mechanism in their molecular profiles in comparison to the remaining subtypes [23].

In a Danish cohort of more than 1500 prospective and consecutive primary breast cancer patients, we aimed to characterize the basal-like cancers according to basic clinical parameters and molecular hallmarks of cancer [24]. Furthermore, we examined if a distinct molecular profile can identify the BRCA1/2 carriers among the basal-like tumors.

Material and methods

Patients and tumor samples

The sample cohort consisted of 1560 female patients diagnosed with breast cancer of stage I–III. The patients were clinically and diagnostically assessed at Rigshospitalet between 2014 and 2017. Fresh tumor specimens, extracted during surgery, were inspected by pathologists and tumor biopsies of around 100 mg were stored in RNALater (Thermo Fisher Scientific, Waltham, MA, USA). The Danish data Protection Agency (jr. no.: 2012-58-0004) and Danish Breast Cancer Group (jr. no.: DBCG-2015-14) approved the study.

Gene expression and subclass analyses

RNA was isolated using the AllPrep DNA/RNA purification kit (Qiagen, Hilden, Germany). The integrity of the RNA was measured using the Agilent RNA 6000 Nano Kit on an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA). RNA was reverse transcribed and used for cRNA synthesis, labeling and hybridization with GeneChip^® Human Genome U133 Plus 2.0 Array (Affymetrix, Santa Clara, CA, USA) according to the manufacturer’s protocol. In short, arrays were washed and stained with phycoerytrin conjugated streptavidin using the Affymetrix Fluidics Station 450, and scanned in the Affymetrix GeneArray 3000 7G scanner to generate fluorescent images. Cell intensity files (.CEL files) were generated in the GeneChip Command Console Software (AGCC; Affymetrix, Santa Clara, CA, USA). The probe level data (.CEL files) were transformed into expression measures using R version 3.4.1 (https://www.R-project.org/). Raw intensity .CEL files were preprocessed by quantile normalization, and probe summaries were extracted via robust multi-array average (RMA). The CIT classifier [6] was applied to the probe expression data, by assigning each sample, a molecular subtype; normal-like, luminal A, luminal B, luminal C, molecular apocrine and basal-like. Gene expression values were deduced from the probe expression values, by taking the median probe expression value, in case more than one probe corresponded to the same gene. Each of the CIT subtypes was given a score on each of the hallmark gene sets from the Molecular Signatures Database (MsigDB). The score corresponded to the mean value of the expressions of the genes contained in each hallmark gene set and the expressions of all samples from each subtype. The average expression of normal samples (from Rigshospitalet) was deducted from the expression values of the rest of the samples before the aggregation; hence the normal samples were used for centering.

It should be noted that the subtypes derived by the CIT classifier correlate to the ones from the PAM50 classifier, with few differences; CIT normal-like subgroup does not correspond to normal breast tissue but exhibits similar expression profiles to the luminal A subgroup. The CIT normal-like samples would classify as PAM50-luminal A. The CIT luminal C and molecular apocrine subtypes include samples with overexpressed HER2; however, the two subtypes represent patients with distinct ER status (positive and negative, respectively). CIT luminal A and B largely overlap with the PAM50 luminal A and B subtypes.

Blood sample and germline mutation screening

Genomic DNA was isolated using the ReliaPrep Large Volume HT gDNA Isolation Kit (Promega, Madison, WI, USA) and a Tecan Freedom EVO HSM2.0 Workstation according to the manufacturer’s instructions. Mutation screening was done by the breast cancer-predisposing gene-panel as previously described [25]. Sequencing was performed on a MiSeq (Illumina, San Diego, CA, USA) to an average depth of at least 100×. Sequencing data were analyzed using Sequence Pilot (JSI Medical Systems, Ettenheim, Germany), where variants are called if the non-reference base frequency was above 25%. Variants are numbered according to the following GenBank accession numbers: NM_007294 (BRCA1) and NM_000059 (BRCA2) using the guidelines from the Human Genome Variation Society (www.hgvs.org/mutnomen). All class 3–5 variants were verified by Sanger sequencing on an ABI 3730 DNA Analyzer using DNA purified from a second blood sample.

Statistical analysis

The data processing was performed using the R software (https://www.R-project.org/) and packages from the Bioconductor project [26]. Cell intensity files were processed with the R package simpleaffy (http://bioinformatics.picr.man.ac.uk/simpleaffy/). Molecular subtypes were predicted using package citbcmst (https://CRAN.R-project.org/package=citbcmst). Linear models were fit with R package limma [27]. The principal component analysis (PCA) visualizations and the heatmaps were performed in the Qlucore Omics Explorer™ software (Qlucore AB, Lund, Sweden). The rest of the graphics were done using the R package ggplot2 [28].

Results

Subtype distribution

The patient cohort consisted of 1560 consecutive primary breast cancers in stage I–III. The distribution of subclasses were derived from the 256 gene signature (referred to as the CIT classifier further on), and resulted in 161 normal-like (normL), 777 luminal A (lumA), 284 luminal B (lumB), 93 luminal C (lumC), 77 molecular apocrine subtype (mApo) and 168 basal-like (basL) samples. The intrinsic subtypes were depicted in the PCA plots of gene expression values, where the differentiation of the basal-like group compared to the rest of the groups was evident (Figure 1(a)). The deviation appeared even before reducing the gene set, from all the genes contained in the Affymetrix platform, to the 256 gene signature, which is illustrated in Figure 1(b). The heterogeneous luminal types were harder to distinct, which is also clearly shown in both Figure 1(a,b).

Figure 1. The three most variable components of the principal component analysis (PCA) of the gene expression profiles. (a) All genes of the Affymetrix platform are considered and (b) only genes from the 256 gene signature (CIT) are considered. The distinction between the basal-like and the more heterogeneous luminal subtypes is clearly illustrated by the PCA plot, even before selecting for the CIT gene signature.

Clinical characteristics

The 168 samples classified as basal-like subtype included some ER and HER2 positive samples based on IHC, substantiating the fact that basal-like subtype are not identical to TNBC. In particular, among the 168 patients; 10 had HER2 positive status and 158 had negative status, 56 were ER positive with a cutoff of 1% and 112 were negative, 31 were ER positive with a cutoff of 9% and 137 were negative. Patients with basal-like breast cancer were significantly younger at the time of diagnosis compared to other patient groups (see Figure 2(a)). Tumor size did not vary significantly across the subtypes (see Figure 2(b)). Apart from normal-like, age at diagnosis was inversely correlated to tumor size e.g., the younger the age at diagnosis, the larger the tumor size. We correlated clinical stage at the time of diagnosis to the six individual molecular subclasses, and the relative comparison showed common trends. Hence, normal-like and luminal A breast cancers were predominantly diagnosed at stage II, luminal B and C as well as molecular apocrine subtype were more often diagnosed at stage III rather than stage II. However, patients with basal-like cancers were almost exclusively diagnosed with stage III disease (see Figure 2(c)).

Figure 2. (a) Violin and box plots of the age distribution at the time of diagnosis (y-axis) of molecular subtypes (x-axis). (b) Violin and box plots of the tumor size (y-axis) of molecular subtypes (x-axis). (c) Relative frequency distribution chart of disease stage (x-axis) and relative frequency of occurrence (y-axis). The chart is shown separately for each subtype. (d) Violin and box plots of age distribution at the time of diagnosis (y-axis) of BRCA1/2 carriers and non-carriers (x-axis). VUS: variant of unknown significance. (e) Violin and box plots of tumor size (y-axis) of BRCA1/2 carriers (x-axis). (f) Relative frequency distribution chart of disease stage (x-axis) and the relative frequency of occurrence (y-axis). The chart is shown separately for group of patients (carriers and non-carriers).

Hallmark characterization

To unravel the basal-like portrait we examined the signaling pathways from the well-established hallmarks of cancer by assigning a subtype score to each pathway (Figure 3). The heatmap clearly depicts that the basal-like cancers were receptor-negative (see estrogen and androgen response pathways) and shows a clear upregulation of molecular profile related to the proliferative cancer signaling pathways (see E2F targets, mitotic spindle, KRAS and WNT). In addition, the TP53 tumor suppressor pathway was downregulated. Basal-like and molecular apocrine subtypes were clearly distinct by their metabolic and protein secreting pathways. Immune response pathways were upregulated in basal-like similar to molecular apocrine and luminal C. One of the DNA repair pathways was downregulated; however two other DNA repair pathways were upregulated. Thus, this indicates that neoplastic cells from basal-like samples have lost their cellular modeling and developmental features. In conclusion, the molecular hallmarks paint a portrait that the basal-like samples in our cohort were neoplastic, with highly proliferative features and clear immuno-response profiles.

Figure 3. Heatmap of scaled mean expression values of molecular subtypes for each hallmark gene set. Expressions level of genes contained in each hallmark and the expressions of all samples from each subtype are aggregated, by the mean value. The color ranges from bright green for the lowest expression values to bright red for the highest. Heatmap of the scaled mean expression values of patients above 60 years of age (BRCA1/2 non-carriers, assigned α) and BRCA1/2 carriers, assigned β, for each hallmark gene set. Expressions of genes contained in each hallmark and expressions of samples from each group are aggregated, by the mean value. The color ranges from bright green for the lowest expression values to bright red for the highest.

BRCA1/2 predisposed patients

As part of our diagnostic pipeline our basal-like patients were offered genomic screening to detect if they were carriers of a BRCA1 or BRCA2 pathogenic variants. One of the hypotheses in the present study was to identify distinct features between the BRCA1/2 carriers opposed to non-carriers, also diagnosed with a basal-like breast cancer. At the time of data collection for the present study, 120 patients were screened for genetic variants in BRCA1/2 out of a total of 168 breast cancer patients with a basal-like subtype. Forty-eight patients did not enter the screening procedure since a blood samples were not obtained at the day of surgical procedure, which was the rule for inclusion in this clinical prospective study. The results of the screening showed that 19 patients were pre-disposed, 6 had a variant of unknown significance (VUS) and 95 were BRCA1/2 negative. We explored the clinical characteristics of the BRCA1/2 carriers in comparison to non-carriers with a basal-like subtype.

The clinical characteristics of the screened patients show that the BRCA1/2 carriers were significantly younger than the BRCA1/2 non-carriers (see Figure 2(d)). Neither tumor size nor stages were significantly different in BRCA1/2 carriers as compared to non-carriers (see Figure 2(e,f)).

To extend upon the observation that BRCA1/2 carriers were significantly younger than the BRCA1/2 non-carriers, we sought to identify differentially expressed genes between the two patient groups. An additive linear model with age and pathogenic mutations as the response variables was fit to the gene expression data. The age was divided into groups of over 60 years of age and younger patients with a pathogenic mutation in BRCA1/2. The contrasts of gene expression profiles between these two patient groups were examined and differentially expressed genes for that contrast and the corresponding p and q-values were computed. The histogram of the p-values resembled a uniform distribution. The analysis showed that IMPDH2 gene was the most differentiated (q-value =0.063) and upregulated in the patients >60 years of age and without a pathogenic mutation. To pursue possible differences in molecular signaling pathways between the two clinically distinct subclusters we generated a heatmap based on the hallmark signatures (see Figure 3). The main differences in the BRCA1/2-carriers compared to the non-carriers were a clear loss in DNA-repair mechanism and relative upregulation in KRAS, Hedgehog and WNT_BETA_CATENIN – signaling pathways. Furthermore, the proliferation rate was altered among the two sub-clusters and the heatmaps demonstrate a higher proliferation in the non-carrier-cluster. The ER-response signals were relatively higher in the BRCA1/2-carrier cluster.

Discussion

The prognostic and predictive features associated with a basal-like breast cancer, as well as distinct molecular profiles, indicate that this subtype should be considered a separate entity. In particular, as it has been shown, that patients with basal-like breast cancer have a profound increase in the risk of recurrence and mortality [12]. In this study, we examined the clinical features of consecutive breast cancer patients and established that basal-like subtype is diagnosed in significantly younger patients who are almost exclusively in stage III disease. We further exploited the molecular hallmarks of basal-like samples and confirmed that this subtype is distinguished by its high proliferative signaling pathways. However, we received a complex signal in the DNA repair pathways, although we expected them to be lost since this subtype is known to have a higher proportion of BRCA1/2 carriers.

We exploited the basal-like subtype by exploring the hallmark gene sets of two clinically separate phenotypes and found that the downregulation of DNA – repair mechanisms is more substantiated among the BRCA1/2 carriers. The differences in the hallmark gene sets were not adequate to establish a robust molecular signature for future identification of BRCA1/2 based on their molecular signature alone. Moreover, no significant correlation was established between age at diagnosis, pathogeny of mutations and gene expression profile. Since the comparison of the two basal-like subgroups did not generate a clear signature at the molecular level it may possibly imply that there are other genes, not yet annotated, that predispose to same type of distinct breast cancer type [29]. However, a previous Danish study including >180 samples and close to a third (n = 55) of BRCA1/2 carriers were somewhat closer in building a molecular signature for identification of mutation carriers, although it was concluded that further validation is needed [30]. Additionally, the sample size of our study is not scaled appropriately as the number of patients carrying a pathogenic mutation (n = 19) is largely lower, compared to the number of genes under consideration (n = 20545). Intriguingly though, none of the hereditary basal-like cancers was identified in the patient group above 60 years of age. However, we cannot conclude that patients above a certain age with a basal-like subtype are not at risk of being predisposed. Repeating this comparison in a larger study would be of great interest. Comparing our cohort and number of patients with a pathogenic BRCA1/2 carrier, we found that previous studies included a much higher frequency (up to 60% of BRCA1/2 carriers; data not shown) [30,31]. With a larger patient cohort and more BRCA1/2 carriers in particular, we would expect to predict when there is high probability that the patient is a non-carrier of pathogenic mutation. However, the results are somewhat unique since they reflect a true clinical setting as all tissue samples were obtained and analyzed within the 10 d of the surgical procedure from primary breast cancer patients. Overall, patients with basal-like breast cancer are at great risk of recurrent disease and death. Thus, the clinical course for this patient group should be tailored and reflect the aggressiveness of this cancer. In line with this, upfront identification of BRCA1/2 carriers could lead to stratified treatment with poly(ADP-ribose) polymerase inhibitors (PARP-inhibitors) [32] and result in a greater implementation of precision medicine for this challenging group of patients.

Taken together, we established that basal-like subtype is diagnosed in significantly younger patients who are almost exclusively in stage III disease. Patients with a basal like breast cancer had the lowest median age and the largest median tumor size. The BRCA1/2 carriers were significantly younger and we did not identify any carriers above 60 years of age.

Acknowledgments

We would like to thank Line Offenbach Jacobsen, Søren Petersen and Bettina Marsot Andersen for excellent laboratory assistance.

Disclosure statement

The authors have no conflicts of interest to declare.

Additional information

Funding

The authors received no funding for this study.