Protein signaling and drug target activation signatures to guide therapy prioritization: Therapeutic resistance and sensitivity in the I-SPY 2 Trial

Abstract

Molecular subtyping of breast cancer is based mostly on HR/HER2 and gene expression-based immune, DNA repair deficiency, and luminal signatures. We extend this description via functional protein pathway activation mapping using pre-treatment, quantitative expression data from 139 proteins/phosphoproteins from 736 patients across 8 treatment arms of the I-SPY 2 Trial (ClinicalTrials.gov: NCT01042379). We identify predictive fit-for-purpose, mechanism-of-action-based signatures and individual predictive protein biomarker candidates by evaluating associations with pathologic complete response. Elevated levels of cyclin D1, estrogen receptor alpha, and androgen receptor S650 associate with non-response and are biomarkers for global resistance. We uncover protein/phosphoprotein-based signatures that can be utilized both for molecularly rationalized therapeutic selection and for response prediction. We introduce a dichotomous HER2 activation response predictive signature for stratifying triple-negative breast cancer patients to either HER2 or immune checkpoint therapy response as a model for how protein activation signatures provide a different lens to view the molecular landscape of breast cancer and synergize with transcriptomic-defined signatures.

Keywords: LCM; RPPA; biomarker; breast cancer; clinical trial; drug target; neoadjuvant; phosphoprotein; protein; resistance.

Graphical abstract

Figure 1

I-SPY 2 Trial design, RPPA workflow, and patient distribution (A) I-SPY 2 Trial schematic. (B) Patient number distribution by trial arm and HR/HER2 status in the reverse phase protein array (RPPA) dataset. Ctr, control; N, neratinib; PD1-inh, PD1 inhibitor; TDM1/P, TDM1 + pertuzumab; VC, veliparib + carboplatin. (C) RPPA workflow (image modified from Loebke et al.⁹).

Figure 2

Association of protein/phosphoprotein expression with pCR by arm (A) Dot plot of protein/phosphoprotein analytes (columns) having significant associations with pCR in one or more treatment arm(s) of the I-SPY 2 Trial or across 8 arms (rows); X, data not available. (B) Boxplots of TYK2 Y1054/Y1055 and STAT1 Y701 (top) with JAK2 Y1007 and STAT5 Y694 (bottom) expression by pCR status across all arms. Green, no pCR; orange, pCR. (C) Boxplots of AR S650 (top) and ER total (bottom) by pCR status in the PD1-inh arm. Blue, no-pCR; pink, pCR. (D) Boxplot of cyclin D1 expression within each arm by pCR status. Blue, non-pCR; pink, pCR. Unadjusted p values annotated within each graph; n.s., not significant; Boxes show median and 25th to 75th interquartile range (IQR). Whiskers denote largest/smallest values within 1.5× the IQR.

Figure 3

Protein signaling pathway activation-based characterization of receptor subtypes (A) One-way clustering of analytes with a significant expression difference in at least one subtype pair. Mean intensity values for each endpoint within each subtype were calculated and used for clustering. Mean values from blue to red represent low to high. Bar graph: pCR rate (%) for each subtype across all arms in the study. Blue arrows, PD1/PDL1; green arrow, STAT3 S727; and red arrows, STAT5 Y694 and TYK2 Y1054/Y1055 expression levels. (B) Association dot plot of protein/phosphoprotein analytes (rows) having significant association with pCR in one or more HR/HER2 subtypes and/or across all subtypes. (C) Boxplots for total ERα (left), AR S650 (center), and cyclin D1 total (right) in all patients (upper) and the HR+HER2− subset (lower) demonstrating associations with non-pCR. Blue, no pCR; pink, pCR. Boxes show median and 25th to 75th IQR. Whiskers denote largest/smallest values within 1.5× the IQR.

Figure 4

Druggable targets revealed by protein pathway activation clusters Two-way, unsupervised hierarchical clustering map of protein/phosphoprotein analytes (rows) and 736 patients (columns) comprising the RPPA dataset demonstrating 11 distinct signaling-based clusters. Heatmap color scale: red/white/blue, higher/intermediate/lower levels of expression.

Figure 5

Protein/phosphoprotein signaling activation clusters linkage to sensitivity/resistance (A) One-way hierarchical clustering map of protein/phosphoprotein analyte-mean signaling levels (columns) within each RPPA signaling cluster (rows). Heatmap color scale: red/white/blue, higher/intermediate/lower levels of expression. (B) Sankey plot illustrating relationship of trial arms (left), RPPA signaling clusters (center), and pCR (right).

Figure 6

DRFS association within non-pCR patients by RPPA cluster (A) Hazard ratio (HR) for DRFS for pCR by signaling cluster (box size, power; whiskers, 95% confidence interval (CI). (B) Kaplan-Maier plots of DRFS in the non-pCR patient subset for all 11 RPPA signaling clusters (B.1). (B.2–B.4) Kaplan-Meier plots of DRFS comparing patients achieving pCR across the whole population (blue curves), to non-pCR patients (red curves) in cluster 9 (B.2), cluster 2 (B.3), and cluster 8 (B.4). (C) Dot plot demonstrating association of individual protein/phosphoprotein analytes (columns) with DRFS in each of the 11 RPPA signaling clusters (rows) within non-pCR patients. Analytes included are limited to those with significant or nominal association in at least one RPPA signaling cluster.

Figure 7

HARPS in TNBC (A) Two-way scatterplot of HER2 Y1248 (y axis) and EGFR Y1173 (x axis) of TNBC patients (n = 252) with pre-defined neratinib response cut-point (blue solid line). (B) Response rates observed for TNBC patients whose tumors were HARPS+ or HARPS– across the six treatments arms for TNBC patients. (C) Donut plots of the TN RPS signature distribution in both the HARPS+ and HARPS– cohorts. Numbers indicate individual patient numbers for each signature with overall number shown in the middle of circle.