The pioneering molecular classification system for breast carcinoma was developed by Perou et al.¹ in 2000. This group performed cDNA microarray analysis on 65 samples from 42 patients. Therein, hierarchical data revealed distinctly variable gene expression patterns in different tumors, but similar patterns among paired samples of the same tumor. The “intrinsic” gene expression pattern was characteristic of an individual tumor—as opposed to those that varied as a function of tissue sampling. In this study, four molecular subtypes of breast carcinomas were identified: “luminal,” “HER2-enriched”, “basal-like”, and “normal breast-like”. The “luminal group” of carcinomas were largely hormone receptor (1) and express luminal epithelial genes—traits similar to those of normal luminal epithelial cells. The “HER2-enriched group” was mainly composed of breast car cinomas with amplification of the HER2 gene. The “basal-like group” was ER(-), and frequently corresponded to the triple negative breast carcinomas (TNBCs, i.e., ER(-), PR(-), and HER2(-)). This group of tumors was immunoreactive for cytokeratin (CK) 5/6 and CK17, similar to the reactivity pattern observed in myoepithelial cells of the normal breast epithelium (i.e., “basal” cells, hence the term applied to this group). The category of tumors called “normal breast-like” had a gene expression pattern similar to that observed in normal breast tissue. It subsequently became evident that the latter group most likely resulted from contamination of tissue samples with high levels of normal breast tissue, and may not exist at all.

A year later, Sorlie et al.² performed 85 cDNA microar

ray analyses on 78 breast carcinoma samples. The results thereof further refined the intrinsic molecular classifier and proposed the division of ER(+) luminal group of breast carcinomas into “luminal A” and “luminal B” subgroups. Although both groups expressed hormone receptors, the luminal B subgroup had a higher proliferation rate and lesser expression of ER-related genes relative to those observed in luminal A group of tumors. Some luminal B tumors overexpressed the HER2/neu gene. Luminal B tumors responded less to hormonal therapy, and more to chemotherapy, relative to luminal A tumors. In general, luminal B tumors had a poorer prognosis than did luminal A tumors. The “HER2-enriched” subtype was characterized by high expression of ERBB2 and genes in the ERBB2 amplicon at 17q22.24, including GRB7. Among HER2-overexpressed/amplified tumors, a significant proportion (50% or so) were ER(+), and ER(+)/HER2(+) tumors were classified as luminal B sub-group by both intrinsic classification systems. Additional studies also suggested that ER(+) and HER2(+) (luminal B) breast carcinomas had high levels of genes often expressed
by normal luminal epithelial cells, whereas ER(—)/HER2(1) (HER2-enriched) carcinomas had high levels of genes expressed by progenitor and stem cell-like cells.^3,4 These studies suggested that ER(1)/HER2(1) (luminal B) and ER(—)/ HER2(1) (HER2-enriched) breast carcinomas were biologically distinct. These two groups of tumors not only showed distinct patterns of responses to chemotherapy and HER2-targeted therapy but also displayed relatively distinctive time to and sites of relapse. ER(—)/HER2(1) tumors tended to recur in the first 5 years, and recurrence after 5 years was low. In contrast, ER(1)/HER2(1) tumors had lower recurrence rate during the first 5 years, but their tendency for recurrence persisted for 15 or more years. ER(—)/HER2(1) tumors recurred more often in visceral organs, whereas ER(1)/HER2(1) tumors recurred more often in bone. The latter pattern was also observed in ER(1) luminal-type tumors. In addition, ER(1)/HER2(1) tumors had a lower rate of pathologic complete response (pCR) after neoadjuvant chemotherapy with or without trastuzumab than did ER(—)/HER2(1) tumors, but ER(1)/HER2(1) tumors did not seem to have poorer prognosis, compared with ER(—) / HER2(1) tumors.^5,6 The clinical, pathologic, and molecular features of the basic molecular subtypes of breast carcinomas are summarized in Table 45.1.

Parker et al.⁷ narrowed down the classifier gene list to 50 genes on the basis of gene expression microarray data for the intrinsic molecular classification for breast carcinomas. A quantitative reverse transcription-polymerase chain reaction (qRT-PCR) assay for these 50 genes was developed to assign intrinsic subtypes (luminal A, luminal B, HER2-enriched, basal-like, and normal breast-like) for formalin-fixed paraffin-embedded (FFPE) samples. The prognostic value of prediction analysis of microarray (i.e., PAM50, termed as such due to its reproducibility in subtype classification) was later validated on an additional 786 breast carcinomas and shown to be superior to that of routine clinicopathologic factors.⁸ In the National Cancer Institute of Canada Clinical Trials Group MA.12 study, the PAM50 assay also showed predictive value for adjuvant tamoxifen treatment in addition to offering prognostic value.⁹ The MA.12 study was a prospective, randomized trial of premenopausal women with stage I to III breast carcinomas post-adjuvant chemotherapy to evaluate the effect of tamoxifen versus placebo. Intrinsic subtypes assigned by PAM50 showed independent prognostic value in addition to standard clinicopathologic variables. In addition, patients with the luminal A subtype defined by PAM50 showed greater benefit from adjuvant tamoxifen.

In 2011, Lehmann et al.¹⁰ reported their analyses of 587 TNBCs from 14 human breast carcinoma gene expression datasets (n: 2,353, TNBC: 386) as the discovery set, and seven different gene expression datasets (n: 894, TNBC: 201) as the validation set. The group identified six TNBC subtypes: basal-like 1 (BL1), basal-like 2 (BL2), immunomodulatory (IM), mesenchymal (M), mesenchymal stem-like (MSL), and luminal androgen receptor (LAR) subtypes (Fig. 45.1).

These subtypes were reported to display distinct gene expressions and had different clinical outcomes.

The “BL1” subtype showed enrichment of cell cycle-related gene expression, and the “BL2” subtype showed ele vated expression of genes involved in growth factor signaling (including EGF, NGF, MET, and IGF1R). The BL2 subtype was suggestive of basal/myoepithelial differentiation, with high levels of p63 and CD10 expression. Both BL1 and BL2 had a high Ki67 proliferation rate and a higher pCR rate (63%) after neoadjuvant taxane treatment. “M” and “MSL” subtypes had enrichment expression of genes involved in cell motility and cell differentiation pathways. Metaplastic carcinomas (which included matrix-producing, sarcomatoid, and squamous cell carcinomas) shared molecular features with M and MSL subtypes. Compared with the M subtype, the MSL subtype showed lower levels of proliferation genes and shared molecular features of claudin-low subtype (see later).¹⁰

The “IM” subtype had enrichment of genes involved in im mune cell responses and showed substantial overlap with the gene signature for medullary breast carcinoma—a TNBC with a good prognosis.¹¹ The “LAR” subtype expressed luminal CK18 and had elevated expression of genes involved in hormonal pathways, including androgen receptor. The LAR subtype corresponded to the molecular apocrine subtype described previously by Farmer et al.¹² Lehmann et al’s study indicated that each of the molecular subtypes in TNBC has distinct “druggable” targets for potential targeted therapy. This finding could change the current “one size fits all” chemotherapy regimen usually employed for TNBC to individu ally targeted therapy based on specific molecular subtypes. Their study also showed that TNBC does not consist only of basal-type breast carcinoma and that TNBC was a heterogeneous group, which included basal-like (47%), luminal A (17%), luminal B (6%), normal breast-like (12%), HER2 (6%), and unclassified (12%) carcinomas as categorized by the intrinsic classification.

Most of the initial molecular classifications were based on analyses of relatively small cohorts of retrospectively collected breast carcinomas, and did not include low-incidence subtypes. Claudin-low tumor, a recently identified molecular subtype, is characterized by low gene expression of tight-junction proteins claudins 3, 4, and 7 and E-cadherin. Clau din-low tumors have been shown to express low to absent luminal epithelial markers and high epithelial-mesenchymal transition (EMT) markers.³ Tumor-initiating cells/stem-like cells (CD44(+)/CD24(-/low)) sorted from primary breast carcinomas show exclusive features of claudin-low molecular subtype and are resistant to chemotherapy.¹³ The MSL subtype identified from TNBC by Lehmann et al.¹⁰ shared features similar to those of the claudin-low molecular subtype.

Curtis et al.,¹⁴ using integrated copy number alteration (CNA) and gene expression, analyzed over 2,000 breast carcinoma samples and found 10 integrative subgroups with distinct clinical outcomes. The group also identified additional heterogeneity within the intrinsic subtypes defined by gene expression alone. The integrative classification also identified the putative molecular drivers in each subgroup, for example, loss of expression of the PPP2R2A gene, a Bregulatory subunit of the PP2A mitotic exit holoenzyme complex located on chromosomal 8p21 that is highly associated with luminal B breast carcinomas.