HER2 heterogeneity and treatment response-associated profiles in HER2-positive breast cancer in the NCT02326974 clinical trial

Abstract

BACKGROUNDHER2-targeting therapies have great efficacy in HER2-positive breast cancer, but resistance, in part due to HER2 heterogeneity (HET), is a significant clinical challenge. We previously described that in a phase II neoadjuvant trastuzumab emtansine (T-DM1) and pertuzumab (P) clinical trial in early-stage HER2-positive breast cancer, none of the patients with HER2-HET tumors had pathologic complete response (pCR).METHODSTo investigate cellular and molecular differences among tumors according to HER2 heterogeneity and pCR, we performed RNA sequencing and ERBB2 FISH of 285 pretreatment and posttreatment tumors from 129 patients in this T-DM1+P neoadjuvant trial. A subset of cases was also subject to NanoString spatial digital profiling.RESULTSPretreatment tumors from patients with pCR had the highest level of ERBB2 mRNA and ERBB signaling. HER2 heterogeneity was associated with no pCR, basal-like features, and low ERBB2 expression yet high ERBB signaling sustained by activation of downstream pathway components. Residual tumors showed decreased HER2 protein levels and ERBB2 copy number heterogeneity and increased PI3K pathway enrichment and luminal features. HET tumors showed minimal treatment-induced transcriptomic changes compared with non-HET tumors. Immune infiltration correlated with pCR and HER2-HET status.CONCLUSIONResistance mechanisms in HET and non-HET tumors are distinct. HER2-targeting antibodies have limited efficacy in HET tumors. Our results support the stratification of patients based on HET status and the use of agents that target downstream components of the ERBB signaling pathway in patients with HET tumors.TRIAL REGISTRATIONClinicalTrials.gov NCT02326974.FUNDINGThis study was funded by Roche and the National Cancer Institute.

Keywords: Breast cancer; Oncology.

Figure 1. Transcriptional differences based on HER2 heterogeneity and response to treatment.

(A) Schematic outline of the clinical trial and sample collection. (B) PCA plot depicting transcriptome variation of 242 pretreatment samples colored on the basis of HER2 heterogeneity and pCR status. (C) Box plots depicting ERBB2 mRNA expression in the indicated groups at patient level using the mean expression of 2 biopsies from the same patient. (D) Scatterplot showing the correlation of RCB score and ERBB2 expression in tumors assessed at patient level. (E) Volcano plots illustrating differentially expressed genes (DEGs) between pretreatment samples from the indicated comparisons. Top DEGs are indicated. (F) Box plots showing enrichment scores of 4 different ERBB/PI3K pathway signatures in HER2 non-HET/pCR, non-HET/no pCR, and HET/no pCR patients. Mean scores of 2 pretreatment biopsies from the same patient were used. (G) Heatmap depicting relative expression of KEGG ERBB signaling signature genes in HER2-HET and non-HET samples without pCR normalized to pCR samples. Genes are ordered from commonly different in the 2 groups compared with pCR cases to uniquely different in the HET and non-HET groups. P values were calculated based on 2-tailed Mann-Whitney U test (C and F) and Pearson’s correlation (D).

Figure 2. Intrinsic subtype correlation to HER2 heterogeneity and response.

(A) Box plots depicting PAM50 subtype probability scores in the indicated groups using the mean scores of 2 biopsies from the same patient. (B) Heatmap showing relative expression of PAM50 genes in HET and non-HET samples without pCR normalized to pCR samples. (C) Representative immunofluorescence image of HER2 and cytokeratin 5 (CK5) in a HER2-HET tumor. Immunostaining was performed once. Scale bar: 100 μm. (D) Scatterplot representing correlation of HER2 and CK5 signal intensity in 4,367 cancer cells quantified from 8 tumors across all 3 subgroups. (E) Stacked bar plot depicting the percentage of CK5^hi and CK5^lo cancer cells within HER2^hi and HER2^lo cancer cell subpopulations. High and low were defined by median of all cells quantified. P values were calculated based on 2-tailed Mann-Whitney U test (A), Pearson’s correlation (D), and Fisher’s exact test (E).

Figure 3. Treatment-induced changes in HER2 heterogeneity.

(A) Sankey plot illustrating HER2 IHC score changes in pretreatment and posttreatment biopsies from 49 patients. Samples with HER2 heterogeneity shift are highlighted. (B) Representative images of intrapatient paired pretreatment and posttreatment HER2 IHC. Images were taken under ×40 magnification. (C) Sankey plot illustrating HER2 heterogeneity status shift in pretreatment and posttreatment biopsies from 50 patients. (D) Line plots depicting Shannon’s equitability index changes of ERBB2 copy number in paired pretreatment and posttreatment biopsies among all or non-HET/HET patients. Samples with HER2 heterogeneity shift are highlighted. (E) Representative images of intrapatient paired pretreatment and posttreatment ERBB2 FISH. Scale bar: 10 μm. Two-tailed Wilcoxon’s matched-pairs signed rank test (D).

Figure 4. Treatment-induced changes in transcriptomic profiles.

(A) PCA plot depicting transcriptomic variation among pretreatment (n = 81) and posttreatment (n = 43) biopsies. (B) Box plot showing Euclidean distances between pretreatment and posttreatment RNA-Seq samples in non-HET and HET patients. (C) Bar plot showing ranking of samples based on treatment-induced Euclidean distances. Pretreatment HET, HET shift, and PIK3CA/ERBB2 mutation are indicated. (D) Box plot showing transcriptomic Euclidean distances between pretreatment and posttreatment samples in tumors with or without PIK3CA or ERBB2 mutations. (E) Line plots showing enrichment scores of ERBB/PI3K pathway signatures in pretreatment and posttreatment samples of each patient. Mean scores of 2 biopsies were used to represent the pretreatment group of each patient. (F) Volcano plots illustrating DEGs between pretreatment and posttreatment biopsies from non-HET and HET samples. (G) Heatmap of top REACTOME pathways differentially enriched between pretreatment and posttreatment samples from non-HET and HET cases. Red and blue represent increased and decreased pathway enrichment, respectively, in posttreatment samples compared with pretreatment counterparts. Color scale represents magnitude of change. P values were calculated based on 2-tailed Mann-Whitney U test (B and D) and Wilcoxon’s matched-pairs signed-rank test (E).

Figure 5. Preexisting immune microenvironmental differences and treatment response.

(A) Box plots depicting predicted overall stromal and immune scores in the indicated patient groups. Mean scores of the 2 pretreatment biopsies from the same patient were used. (B) Box plots showing predicted abundances of lymphocyte and myeloid cell subsets from the 2 pretreatment biopsies from the same patient in the indicated groups. (C) Heatmap showing relative expression of immune checkpoint genes in samples without pCR normalized to pCR cases. Asterisks indicate statistically significant (P < 0.05) differences. (D) Box plots depicting BCR and TCR richness and diversity index scores in the indicated groups. Average measurements of the 2 pretreatment biopsies from the same patient were used for pretreatment values. P values were calculated based on 2-tailed Mann-Whitney U test in all panels.

Figure 6. Digital spatial profiling of the immune environment.

(A) Representative images of digital spatial profiling (DSP) of tumor and immune regions of interest (ROIs). Scale bars: 50 μm. (B) PCA plot depicting the DSP signal variation from tumor and immune ROIs of the 58 target protein expression included in the DSP panel among 30 pretreatment samples of the indicated patient groups. (C) Heatmaps showing relative expression of 58 proteins by DSP in samples without pCR normalized to pCR samples. Asterisks indicate statistically significant (P < 0.05) differences. Comparisons were separated into tumor and immune ROIs. (D) Box plots showing the normalized expression of HER2 among the 3 groups in pretreatment samples in both tumor and immune ROIs. P values were calculated based on 2-tailed Mann-Whitney U test (C and D).

Figure 7. Treatment-induced changes in the immune environment.

(A) Line plots depicting predicted overall stromal and immune scores among the 4 subgroups based on HER2 heterogeneity and pre-to-posttreatment samples from the same patient. Mean scores of 2 biopsies were used to represent the pretreatment group of each patient. (B) Line plots showing predicted lymphocyte and myeloid cell subset abundances in the indicated pre-to-posttreatment pairs. Average ratio of 2 biopsies was used to represent the pretreatment sample of each patient. (C and D) Heatmaps showing treatment-induced changes in the expression of immune checkpoint genes (C) and 58 proteins used in digital spatial profiling (D). Asterisks indicate statistically significant (P < 0.05) differences. In D, comparisons are separated into tumor and immune ROIs. (E and F) Line plots depicting BCR and TCR richness (E) and diversity index scores (F) in the indicated pre-to-posttreatment pairs. Mean measurements of 2 biopsies were used to represent the pretreatment sample of each patient. P values were calculated based on 2-tailed Wilcoxon’s matched-pairs signed-rank test in all panels.