Molecular stratification within triple-negative breast cancer subtypes

Abstract

Triple-negative breast cancer (TNBC) has been subdivided into six distinct subgroups: basal-like 1 (BL1), basal-like 2 (BL2), mesenchymal (M), mesenchymal stem-like (MSL), immunomodulatory (IM), and luminal androgen receptor (LAR). We recently identified a subgroup of TNBC with loss of the tumor suppressor PTEN and five specific microRNAs that exhibits exceedingly poor clinical outcome and contains TP53 mutation, RB1 loss and high MYC and WNT signalling. Here, show that these PTEN-low/miRNA-low lesions cluster with BL1 TNBC. These tumors exhibited high RhoA signalling and were significantly stratified on the basis of PTEN-low/RhoA-signalling-high with hazard ratios (HRs) of 8.2 (P = 0.0009) and 4.87 (P = 0.033) in training and test cohorts, respectively. For BL2 TNBC, we identified AKT1 copy gain/high mRNA expression as surrogate for poor prognosis (HR = 3.9; P = 0.02 and HR = 6.1; P = 0.0032). In IM, programmed cell death 1 (PD1) was elevated and predictive of poor prognosis (HR = 5.3; P = 0.01 and HR = 3.5; P < 0.004). Additional alterations, albeit without prognostic power, characterized each subtype including high E2F2 and TGFβ signalling and CXCL8 expression in BL2, high IFNα and IFNγ signalling and CTLA4 expression in IM, and high EGFR signalling in MSL, and may be targeted for therapy. This study identified PTEN-low/RhoA-signalling-high, and high AKT1 and PD1 expression as potent prognostications for BL1, BL2 and IM subtypes with survival differences of over 14, 2.75 and 10.5 years, respectively. This intrinsic heterogeneity could be exploited to prioritize patients for precision medicine.

Figure 1

Most PTEN-low/miR-low (group-a) tumors cluster with the BL1 TNBC subtype. (A) Classification of 205 TNBCs into six Lehmann TNBC subtypes: basal-like 1 (BL1), basal-like 2 (BL2), mesenchymal (M), mesenchymal stem–like (MSL), immunomodulatory (IM), and luminal androgen receptor (LAR) as well as unspecified tumors (UNS). Also shown are the four TNBC subtypes; PAM50 + Claudin-low; IntClust, PTEN-low/miR-low and the BL1-group “a” subgroup. Pathological data, miRNA expression and classification of PTEN-low/βCatenin-high, PTEN-low/RB1Loss-high, PTEN-low/RBKO-high, and PTEN-low/RhoA-high are indicated. (B,C) Survival of six TNBC subtypes as well as unspecified TNBC tumors (UNS) in a cohort of 205 TNBCs. In C, BL1 is stratified into BL1-group-“a” versus all other BL1 lesions. (D) Survival of BL1-group-“a” versus all other BL1 lesions or all other TNBC subtypes and corresponding hazard ratios (HRs) in the 205 TNBC cohort. (E) Survival of BL1-group-“a” versus all other BL1 lesions or all other TNBC subtypes and corresponding HRs in a cohort of 44 TNBCs. (F) Survival of 72 and 16 BL1 TNBCs in both the training (D) and validation (E) sets. Low PTEN expression in BL1 correlated with poor prognosis in both 72 BL1 tumors from the training set and 16 BL1 tumors from the validation set. Low PTEN expression and high pathway activity of β-Catenin, or loss of RB1 using two different signatures RBKO (RB sig. loss and RB knockout) predicted poor clinical outcome in both cohorts.

Figure 2

Identification of signaling pathways that are altered in each specific TNBC subtypes. (A) Classification of 205 TNBCs into subtypes, and activity of 25 different signaling pathways. (B,C) In both training and validation sets high MYC pathway activity alone stratifies BL1 TNBC into high and low risk groups. (D,E) E2F2 and TGFβ pathways are significantly elevated in BL2. (F,G) IFNα and IFNγ pathways are significantly induced in IM. (H) EGFR activity is significantly higher in MLS compared to all other TNBC subtypes. Only IFNα signalling affected clinical outcome (Supplemental Fig. 1).

Figure 3

Expression levels and prognostic power of RhoA signaling and RB loss in TNBC subtypes. (A) High pathway activities of RhoA and RBKO pathways in PTEN-low/miR-low (group-“a”) vs PTEN-low/miRNA high (group “c”), vs all other PTEN-low TNBC (group “b”). PTEN(+) includes all PTEN-high TNBC samples. RhoA and RBKO pathways are elevated in group-a and exhibit poor clinical outcome in both the training and validation sets. (B) High pathway activities of RhoA and RBKO pathways in TNBC subtypes. Both pathways are elevated in BL1 and predict poor clinical outcome in the training and validation sets. (C) Exceedingly poor survival of BL1 TNBCs with low PTEN expression and high RhoA pathway activity.

Figure 4

Copy number alterations (CNAs) in TNBC subtypes and identification of AKT1 gain and high expression as prognostic markers in BL2. (A) CNA in breast cancer related oncogenes and tumor suppressors. Note CNAs in AKT1 in several BL2 TNBC samples. (B) CNA in AKT1 and surrounding genes on chromosome 14q32.3. (C) AKT1 showed the most frequent high CNA gain that predicts poor outcome in BL2 subtype. (D,E) High mRNA expression of AKT1 predicts poor prognosis in BL2 tumors from both training and validation cohorts.

Figure 5

Stratification of IM TNBCs by expression of the immune checkpoint blockade gene PDCD1 (PD1). (A) Expression of immune checkpoint and inflammation associated genes in TNBC subtypes. (B,C) PDCD1 expression is elevated in IM TNBC subtype and predicts poor prognosis both in the training (205) and validation (190) TNBC cohorts.