Tumor spatial transcriptomics and potential mechanisms of intrinsic resistance to HER2-directed therapy.

Abstract

e12602 Background: Neoadjuvant dual HER2 blockade + chemotherapy (NAC) is standard of care for early-stage HER2+ breast cancer (BC), leading to pathological complete response (pCR) associated with high cure rates in ~50-60% of patients (pts). In contrast, non-pCR pts have significant relapse risk. Mechanisms of intrinsic resistance to HER2-targeted therapy remain poorly understood. Methods: We conducted a phase II trial of NAC with dual HER2 blockade (12 weeks trastuzumab + HER2 tyrosine kinase inhibitor varlitinib + paclitaxel), followed by surgery, in stage I-III HER2+ BC pts. Tumor biopsies were taken at baseline (BL) and at surgery post-NAC. We analyzed 24 paraffin-embedded tumor samples from 8 pCR (BL) and 8 non-pCR (BL, post-NAC) pts with spatial transcriptomics profiling (10X Genomics) to evaluate the multidimensional organization and contribution of cell components to treatment response. Pre-processing of the raw matrix was performed using Seurat (v5.0.3), with normalization conducted via the SCTransform function. Cell composition was deconvoluted using Cell2location (v0.1.3). Copy number alterations were inferred using SCEVAN (v1.0.1), while pathway activities were estimated with PROGENy (v3.20). Cell-to-cell and cell-to-pathway colocation analyses were carried out using mistyR (v1.10.0). Results: At BL, non-pCR pts showed significantly lower tumor HER2 expression, more heterogeneity in HER2 expression and ERBB2 gene amplification (amp), and greater dissociation between ERBB2 amp and HER2 expression vs pCR pts. Significant heterogeneity in ERBB2 amp persisted in non-pCR post-NAC tumors. Non-pCR pts showed significant suppression of immune-related pathways vs pCR pts at BL. Deconvolution analysis showed non-pCR pts to exhibit marked lack of immune cells, including T / NKT cells / plasma blasts at BL and post-NAC as well as reduced B cells post-NAC vs BL tumors of pCR pts. Spatial analysis revealed differences in colocalization of immune cells: BL tumors in pCR pts showed aggregation of memory B cells with other immune cells to form immune hotspots. In contrast, both BL and post-NAC tumors in non-pCR pts showed memory B cells to colocalize with cycling T and myeloid cells with less immune hotspots. Pathway analysis revealed significant differences in immune cell interactions with signaling pathways in pCR vs non-pCR pts. At BL, immune cells were associated with JAK-STAT and TNFα pathway activation in pCR pts, but were strongly correlated with TRAIL in non-pCR pts. Post-NAC, immune cells in non-pCR pts were significantly linked to JAK-STAT and TGFβ signaling. Conclusions: Spatial transcriptomics analyses reveal tumor and immune microenvironment heterogeneity between pCR and non-pCR pts to neoadjuvant HER2-targeted therapy. This heterogeneity may explain intrinsic resistance and could unravel potential therapeutic targets in resistant HER2+ breast cancer.