Background
Triple-negative breast cancer (TNBC) represents 12 to 17% of primary breast cancer and is the most aggressive and deadly breast cancer subtype [
1]. Furthermore, heterogeneity and lack of targeted therapies represent the two main issues for precision treatment of TNBC patients. Molecular subtyping and identification of therapeutic pathways are therefore required to optimize medical care of these patients.
Recent works based on different approaches identified various numbers of TNBC clusters [
2,
3]. Six molecular clusters were found in two in silico studies. In the first study, basal-like 1, basal-like 2, immunomodulatory, mesenchymal, mesenchymal stem-like, and luminal androgen receptor clusters were described [
3]. In the second study, clusters were named immunity 1, immunity 2, proliferation/DNA damage, androgen receptor-like, matrix/invasion 1, and matrix/invasion 2 [
4]. TNBC status was based on bimodal filtering of estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) gene expressions in the first study, and bimodal filtering of estrogen receptor 1 and Erb-B2 receptor tyrosine kinase 2 genes, and median value of progesterone receptor gene in the second study. In addition, four clusters (luminal androgen receptor, mesenchymal, basal-like immunosuppressed, and basal-like immune-activated) and three clusters (molecular apocrine, basal-like-enriched with low immune response and high M2/M1 macrophages ratio, and basal-like-enriched with high immune response and low M2/M1 macrophage ratio) were found in two internal immunohistochemistry (IHC)-typed TNBC studies, respectively [
5,
6]. Although different number of clusters was found, three clusters seem to be present in each of these works: molecular apocrine and two basal-like-enriched clusters with opposite immune response (pro-tumorigenic and anti-tumorigenic). Today, TNBC subtyping still needs to be refined.
In the present study, an unsupervised analysis was conducted on an internal training cohort composed of 238 TNBC tumors. Fuzzy clustering was followed by functional annotation of the clusters, and we intensively focused on exploring the nature of immune response between the two basal-like-enriched clusters. Together, the results identified three TNBC clusters: one molecular apocrine (C1) and two basal-like-enriched, of which one with pro-tumorigenic immune response (immune suppressive), high neurogenesis activity and high biological aggressiveness (C2), and the other with adaptive immune response associated with complete B cell differentiation and immune checkpoint upregulation (C3).
Materials and methods
Patients
Internal cohort was composed of 238 TNBC patients. One hundred seven patients were part of a previous study, and 131 other TNBC patients were enrolled in the UNICANCER PACS08 (NCT00630032) adjuvant multicenter trial described elsewhere [
6,
7]. This last study was approved by a French Ethic Committee (CPP Ouest V, CHU Pontchaillou, Rennes, France; reference number: 07/09-626).
An external cohort composed of publicly available TNBC patients with available tumor genomic data was built. To avoid cross-platform normalization issues, we exclusively looked for Affymetrix® genomic datasets in repositories such as Gene Expression Omnibus (GEO) and ArrayExpress, selecting those with a medium to large sample size [
8,
9]. Non-TNBC data from these studies were also retrieved (Additional file
1).
Tumor tissues
All tumor tissue samples were surgically collected and processed in two parts by a pathologist. The first part was fixed in 10% neutral buffered formalin for standard histological analysis and IHC. The second part was immediately dissected, snap-frozen in liquid nitrogen, and stored until RNA extraction.
Total RNA was prepared following standard protocols then treated with DNase I using the RNeasy column purification system (Qiagen, France). Assessment of RNA quality, integrity, and purity was done through a Bionalyser 2100 (Agilent Technologies, Palo Alto, CA). RNA samples were considered for further analysis only if they had distinct 28 S and 18 S ribosomal peaks.
Gene expression profiling
Gene expression analysis was performed using Affymetrix® Human Genome U133 Plus 2.0 Arrays (Affymetrix®, Santa Clara, CA) measuring over 54,000 transcripts representing over 20,000 genes. cRNA synthesis, labelling as well as chip hybridization, washing, and image scanning were performed according to the manufacturer’s protocol. Affymetrix® na35 probe set annotation was used.
Internal and external data pre-processing
For both internal and external cohorts, raw data were MAS5-normalized in the Affymetrix® Expression Console (v1.3.1) and then log2-transformed. Genes were then median-centered and scaled in each cohort separately. All microarrays complied with quality criteria. Microarray and patient clinical data have been deposited in the GEO under the GSE83937 accession number. Five publicly available datasets were pooled for a total of 257 TNBC (Additional file
1).
Unsupervised analysis
To organize data into groups with the same underlying molecular characteristics, we performed clustering analysis, based on the 5% most variable probe sets (n = 1843; intersection of the 2 sets of most variable probes), by means of fuzzy clustering method (6).
Cluster functional annotation
To annotate the clusters, we used clinicopathologic characteristics, 54 gene-expression signatures (GES), Gene Ontology enrichment analysis (GOEA), and non-TNBC data. Depending on the nature of data, the following methods helped to explore differences among the clusters: one-way analysis of variance (ANOVA) with Tukey post hoc test, Fisher’s exact test, Cox regression model, Kaplan-Meier curves with log-rank test, and Pearson’s correlation coefficient.
Gene-expression signatures
Fifty-four GES were selected for functional annotation of breast cancer tumors. Twelve GES were used for breast cancer molecular subtyping: 4-TNBC, androgen receptor (AR), basal-like, tumor identity card (CIT), claudin-CD24, claudin-low, ER, ER-negative, ERBB2, molecular apocrine, PAM50, and TNBCtype. Eleven were used for immune response dissection: B-cell, Cell type Identification By Estimating Relative Subsets Of known RNA Transcripts (CIBERSORT) [
n = 22] (
http://cibersort.stanford.edu/), cytolytic activity (CYT), interferon (type I IFN), interleukin-8 (IL-8), M2/M1 macrophage ratios (M2/M1, M2/M1 [Becker]), MHC-1, MHC-2, STAT1, and T cell [
10]). Three relate to microenvironment cells: epithelial cells, fibroblasts, and neurons (
http://xcell.ucsf.edu/). Three were linked to metabolism evaluation: adipocytes, glycolysis, and iron (IRGS); and 22 to critical biological pathways in cancer: AKT, β-catenin, chromosomal instability (CIN), E2F3, EGFR, HOXA, mitochondrial oxidative phosphorylation (MITO/OXPHOS), MYC, p53, PIK3CA, perineural invasion (PNI), prolactin (PRL), proliferation, PTEN loss, RAS, reactive stroma, SRC, Stroma-CD10, TGFβ, VEGF, wound response, and YAP1-WWTR1. Finally, three prognostic GES were also used: 38-GES, van’t Veer 70-GES, and genomic grade index (GGI). Complete GES list, methods, and references are described in Additional file
2.
GOEA
Functional annotation of each cluster through GO biological process analysis was performed using the ToppGene web tool [
11]. Two methods were used to select genes differentially expressed across the clusters. SAM method was performed to obtain lists of genes with significantly different expression between clusters (one versus one and one versus the others): genes for which all corresponding probe sets had a
q value of 0% were retained. In addition to SAM method, expression of the 5% most variable probe sets was represented on a heatmap, with patients ordered according to the clusters; hierarchical clustering (centered Pearson correlation distance, Ward’s method) was performed on the probe sets in order to visually detect groups of genes with high expression patterns corresponding to specific cluster(s) of patients. Visually identified overexpressed gene sets in heatmap were designated by “H” followed by a number, which referred to the number of the corresponding cluster (1, 2, or 3). If sets of genes contained different subsets, this number was followed by “a,” “b,” or “c.”
Histological evaluations, tissue microarrays, and immunohistochemistry
Eighty-seven cases of archival formalin-fixed, paraffin-embedded tissues of the internal cohort were evaluated on tissue microarrays, containing a median of 3 replicate 0.6 mm cores per case. Tissue microarrays (TMA) construction was described elsewhere [
6]. Furthermore, 42 full sections of these surgical specimens were available for histological evaluation. Immunohistochemistry was performed using the Ventana BenchMark Ultra platform (Ventana Medical Systems, Tucson, AZ). Details of the antigen retrieval technique and dilution of primary antibodies (CD20, CD21, CD138, MECA79, UCHL1/PGP9.5, and S100) are described in Additional file
3.
Tumor-infiltrating lymphocytes (TILs), especially CD138-positive plasma cells and CD20-positive B lymphocytes, were assessed according to recommendations of an international working group [
12,
13]. The presence of tertiary lymphoid structure (TLS) and lymphoid clusters, with and without germinal centers, respectively, was evaluated in the 42 specimens with available full sections, due to their typical localization in the surrounding area of the tumors. Furthermore, the number of CD21-positive follicular dendritic network was assessed by counting positive structures by IHC on a representative full section per tumor [
14]. The presence of high endothelial venules (HEV) was assessed on full sections by counting the number of MECA79-positive vessels in five 400× magnification hotspots per tumor [
15]. Cases were classified as nerve fibers positive versus nerve fibers negative using UCHL1/PGP9.5 (neuronal marker) and S100 (pan-specific Schwann cell marker). Pathologists were blinded for TNBC cluster assignment.
Non-TNBC external data testing
Non-TNBC external data (
n = 894) were used to refine non-basal-like TNBC cluster (Additional file
1).
Statistical analysis
We considered a two-sided
P value of less than 5% to be statistically significant. For SAM method, 0%
q values were retained. All statistical analyses and figures’ generations were performed using R [
16] and the packages affy 1.50.0, amap 0.8.14, cluster 2.0.4, citbcmst 1.0.4, fpc 2.1.10, and samr 2.0.
Discussion
The findings of this study strongly strengthen the fact that TNBC can be divided into three subtypes with potential therapeutic implications. To the best of our knowledge, we identified for the first time links between neurogenesis, tertiary lymphoid structures, plasma cells, B lymphocytes, and triple-negative breast cancer subtypes (C2 and C3).
Overall, our data show that IHC-typed TNBC regroup three different molecular subtypes of tumors, which would necessitate different and appropriate therapies. C1 is clearly a molecular apocrine cluster, which displays luminal, PIK3CA-mutated, and HER2E hallmarks. Use of non-TNBC data to refine C1 cluster annotation permitted to take into account relativity, which can weaken TNBC subtyping interpretation. When taking into account all breast cancers (non-TNBC and TNBC), C1 has to be considered as an intermediate group, from a biological aggressiveness point of view, between non-TNBC and basal-like enriched clusters.
C2 and C3 displayed basal-like hallmarks. However, these two basal-like enriched clusters showed a major biological discrepancy relative to immune response, which was characterized by a decreasing anti-tumorigenic immune gradient from C3 to C2 and a decreasing pro-tumorigenic gradient from C2 to C3. Furthermore, high neurogenesis activity was found for C2 tumors. In addition to immune response, the immune system is also known to play a pivotal role in tissue repair and regeneration. It is astonishing to notice that these two roles seem to be illustrated in C3 and C2, respectively.
We will not extensively develop the list of treatments, including immunotherapies, which could be proposed for TNBC patients, as recent review articles have been published on this topic [
35]. We will select and discuss a few points, which concern the potential therapeutic ramifications of our findings and the treatment that might be used in function of cluster functional annotation (Additional file
29).
C1 tumors (molecular apocrine) could potentially be treated by antiandrogens or, better, by an association of PI3K inhibitors and antiandrogens, which demonstrated a synergistic anti-tumor effect [
17]. Despite the fact that TNBC are defined by HER2 IHC negativity,
ERBB2 GES displayed high expression of
ERBB2 pathway in C1 tumors. The value of targeting the
ERBB2 pathway in C1 tumors warrants further investigations.
Immunotherapies, which aim to combat immunosuppression or stimulate adaptive immune response, could potentially be proposed for C2 (basal-like pro-tumorigenic immune response) and C3 (basal-like adaptive immune response) patients, respectively [
36]. TLS evaluation and targeting represent a promising approach for the design of immune-based strategies which aimed at stimulating C3 patient immune response [
37]. Therapeutic regimens should associate different immune checkpoint inhibitors, to enhance global efficacy and cytotoxicity of chemotherapy, because immunomodulators are not directly cytotoxic against tumor cells.
Immune checkpoints screening showed that upregulation of these markers characterized C3 compared to C2. We do not know if this immune response limitation is responsible for the global inefficacy of anti-tumor immune attack. Whatever the solution is, we can hypothesize that immune checkpoint inhibition should reinforce immune response against tumor cells in C3 [
38]. However,
VTCN1 (
B7-H4) displayed the only C2 > C3 profile. This gene codes for a B7 immunoregulatory protein, which exerts an immunosuppressive effect through inhibition of T cell activation, proliferation, and clonal expansion, and is considered as a protumorigenic factor [
39]. In patients with ovarian carcinoma and glioma, macrophages expressing
VTCN1 have been directly linked to inhibition of T cell immune response [
40,
41]. Therefore,
VTNC1 might actively participate in C2 immunosuppressive phenotype.
Tumor-associated macrophages are crucial actors of tumor fate and therefore represent important and promising immunotherapeutic targets [
42‐
45]. Consequently, numerous macrophage-directed therapeutic approaches are under investigation and should take into account M2/M1 macrophage status according to TNBC subtype.
C2 patients could potentially be considered for anti-neurogenic therapies as recent studies have unraveled the role of neurogenesis in cancer progression and anti-neurogenic therapies are emerging in oncology [
46‐
48]. In prostate, gastric, skin and pancreatic cancers, it has been shown that the infiltration of new nerves in the tumor microenvironment is necessary to primary tumor growth and metastasis [
49‐
52]. The release of neurotransmitters by nerve endings results in the stimulation of corresponding receptors in both stromal and cancer cells, leading to increased tumor growth and dissemination. In particular, the release of noradrenaline by sympathetic nerves induces an angiogenic switch via the stimulation of beta adrenergic receptors in endothelial cells [
53]. In breast cancer, neurogenesis has been reported in the tumor microenvironment, in particular from nerves of sympathetic origin, and the density of nerve infiltration is associated with cancer aggressiveness [
54]. Neurogenesis was found across all breast cancer subtypes, including TNBC. Our present study shows that neurogenesis-related gene expression level is more specifically increased in the C2 cluster of TNBC, suggesting that anti-neurogenic therapies, such as those targeting the neurotrophic tyrosine kinase receptor 1 (NTRK1) could be relevant in C2 TNBC [
46].
Acknowledgements
We wish to thank the Comité Féminin 49 and its President, Mrs. Marie-Christine Laffineur, for their generous funding. PACS08 tissues used in this work were provided by R&D UNICANCER tumor bank located in Léon Bérard Cancer Center in Lyon. We thank the transcriptomic team (Magali Devic, Elise Douillard, Emilie Moranton, and Nathalie Roi). The authors would also like to thank, for fruitful discussions and suggestions, Pascale Jeannin, Yves Delneste (macrophages), and Catherine Pellat-Deceunynck (plasma cells).