NK cell-triggered CCL5/IFNg-CXCL9/10 axis underlies the clinical efficacy of HER2-targeted antibodies in primary HER2-positive breast cancer [array]

Tumor-infiltrating (TI)-NK cell numbers predict better than TIL score the efficacy of HER2-targeted antibodies in primary breast cancer patients. To understand the mechanism/s underlying this association, biological processes enriched in NK cell-infiltrated as compared to NK cell-desert HER2-positive breast tumors paired by TIL score were leveraged from transcriptomic data. NK cell-infiltrated tumors were enriched in transcripts regulated by interferons and NF-kB. Among them, levels of CCL5/IFNG-CXCL9/10 positively correlated with the number of TI-NK cells in the original biopsy. Indeed, coordinated expression of CCL5/IFNG-CXCL9/10 transcripts was also evidenced in tumor biopsies from a phase II clinical trial where IFNG levels associated to the achievement of pathological complete response to anti-HER2 antibody treatment (OR 96.3, p=0.01) and correlated with their TI-NK cell score. In in vitro models, anti-HER2 antibody-dependent NK cell activation led to the secretion of CCL5/IFN-ɣ and the subsequent production of IFN-ɣ-dependent CXCL9/10 by bystander breast cancer cells. Ex vivo treatment of breast tumor-derived multicellular cultures with trastuzumab induced the activation of CD16+ TI-NK cells and their conversion into a CD16-CD103+ subpopulation, both endowed with CCL5 and IFN-ɣ producing potential. CD16+ and CD16-CD103+ TI-NK cell subpopulations shared the expression of EOMES, TBX21 and several KIR genes, indicating their lineage relationship; and their proportions positively correlated with total NK cell, CD8+ and tissue-resident T cell frequencies, immune cell subsets with anti-tumor potential. Remarkably, the coordinated induction of CCL5/IFNG-CXCL9/CXCL10 expression, concomitant to the conversion of CD16+ into CD16+/-CD103+ tumor-infiltrating NK cells, was recapitulated in a humanized HER2+ breast cancer in vivo model treated with a combination of anti-HER2 antibodies and in vitro expanded human NK cells, paralleling tumor growth control. Finally, patients achieving good clinical responses to anti-HER2 antibody-based neoadjuvant treatment showed an early and coordinated increase in sera CCL5 and CXCL9/CXCL10 levels which positively correlated with TI-NK cell numbers in the corresponding diagnosis biopsy. Overall, our data point to NK cells as regulators of the tumor microenvironment through the early secretion of CCL5/IFN-ɣ resulting in the production of CXCL9/10 and the recruitment/differentiation of immune infiltrates with anti-tumor potential, ultimately contributing to anti-HER2 antibody clinical efficacy. Gene expression microarray analysis were performed by the MARGenomics core facility at Hospital del Mar Medical Research Institute, Barcelona. Total RNA was isolated (RNeasy® Micro Kit, Qiagen) from 3 sequential, 10 µm thick, FFPE breast carcinoma sections from 6 TI-NK cell positive and 6 TI-NK cell negative biopsies. Tumors were selected from a previously described cohort, based on double CD3 CD56 immunohistochemistry data (Muntasell A CCR, 2019). Amplification, labelling, and hybridization was performed according to protocol GeneChip WT PLUS Reagent kit (P/N 703174 2017) and hybridized to Human Clariom S Array (Thermo Fisher Scientific). For analysis, R programming (Version 3.4.3) Bioconductor and the Comprehensive R Archive Network (CRAN 2017) packages were used. Samples were background corrected, quantile-normalized and summarized to a gene-level using the robust multi-chip average (RMA). An empirical Bayes moderated t-statistics model (LIMMA) adjusted by Estrogen Receptor status was built to detect differentially expressed genes with a p-value<0.05 and a fold-change>1.5. Data represented in heatmaps was scaled by genes using the z-score normalization. Analysis of biological pathways enriched in TI-NK cell positive biopsies and major regulators of enriched genes was performed uploading upregulated DEG into IPA software.