e12576 Background: Triple-negative breast cancer’s (TNBC) dismal prognosis demands more effective therapy selection. Previously described molecular subtypes reveal its heterogeneity but lack spatial insight and are not currently used in clinical practice. Here, we aim to evaluate by immunohistochemistry (IHC) the expression of key biomarkers in TNBC and its correlation with spatial gene expression obtained by spatial transcriptomics (ST). Methods: We studied a retrospective cohort of 92 patients (pts) treated at the Institut Jules Bordet in Belgium with primary surgery for early TNBC. ST (Visium Spatial Gene Expression, 10X Genomics) was previously performed on fresh frozen surgical samples. Using serial sections of these samples, we performed duplex IHC staining for Trop-2/androgen receptor (AR) and HER2/Ki-67. IHC staining was scored using H-score (Trop-2), Allred score (AR), and current guidelines (Ki67/HER2), adding the ultra-low and null HER2 categories. Statistical analyses included descriptive statistics, Spearman correlation, Mann-Whitney test, and Kaplan-Meier with log-rank tests for estimating outcome distribution. Results: Of 92 pts included, 4 (4.3%), 25 (27.2%), 28 (30.4%), and 35 (38%) pts had HER2 2+, 1+, ultra-low, and null, respectively, tumors. Importantly, we observed a significant difference in tumor cell-derived ERBB2 gene expression of HER2 ultra-low versus HER2 null tumors ( p<0.0001), while there was no difference in ERBB2 expression between HER2 ultra-low and HER2 1+ tumors ( p=0.49). HER2 expression was not correlated with prognosis. Trop-2 expression by IHC was high, medium, and low in 67%, 23%, and 10% of pts, respectively. IHC expression correlated well with tumor-derived gene expression (r=0.54), with intrasample heterogeneity observed both in IHC and spatial gene expression. Trop-2 expression was prognostic, pts with high Trop-2 H-score tumors having significantly worse outcome ( p=0.021). In this cohort, 25%, 40.2%, and 34.8% of pts had AR Allred scores of 0-2, 3-5, and 6-8, respectively. Tumor AR gene expression was well correlated with IHC AR expression (67.7%). IHC AR expression was associated with the presence of adipose tissue in the stroma (r=0.23, p=0.025) but not with prognosis. Of note, we report for the first time, to our knowledge, for TNBC, AR expression both in tumor and stroma cells, both at the IHC and gene expression levels. Importantly, we could construct a classifier that identifies luminal androgen receptor tumors, using AR Allred score > 5 and Ki67 ≤ 40%, with a PPV of 93% and an NPV of 96%. Conclusions: ST data has the potential to improve easily accessible IHC classifiers, with the aim of refining patient selection for ADC and anti-AR therapies in TNBC.
Read full abstract