Deconvolution of global gene expression profiles identifies differentially expressed immune-related gene profiles on aromatase inhibitor-resistant estrogen receptor positive tumours


Session type:


Milana Bergamino Sirvén1, Gabriele Morani, Joel Parker, Gene Schuster, Elena López-Knowles, Judith Bliss, Mitch Dowsett, Maggie Cheang
1Institute of Cancer Research (ICR)



The standard of care for post-menopausal women with estrogen receptor positive (ER+) breast cancers includes aromatase inhibitors (AI) over 5-10 year-period. However, approximately 25% of early ER+BC relapses.  ER+BC is thought to be less immunogenic than other subtypes, but recent studies have shown that targetable immune-checkpoint components expression in Luminal tumours are associated with poorer prognosis. In this study, we aim to identify biological characteristics to predict AI resistance.


POETIC trial was a phase III, randomized 2:1 study for 4486 ER+BC post-menopausal patients to determine whether peri-operative AI followed by standard adjuvant therapy improved outcome compared with standard adjuvant therapy. Pre and post-2-weeks of perioperative AI RNA-later-samples from 118 ER+HER2- BC patients were analysed. Global gene expression profiles were obtained using Illumina-HumanBeadChips (Gellert et al. Nat.Commun 2016). Module-scores (N=649) covering different cancer-pathways and immune-related genes were calculated for each sample.  Proliferation rate was estimated as percentage of cancer cells staining for Ki67 and categorised into High (≥ 10%) and Low (<10%) at baseline (B) and 2wk.


In POETIC trial, patients with Ki672w remaining “High” after 2wks of AI have substantially poorer prognosis than those with Ki672w at a low level. We categorised tumours into classes according to Ki67 at B and at 2wk: 28 HighB-High2wk, 77 HighB-Low2wk, 13 LowB-Low2wk and 0 LowB-High2wk.  There were 107 modules differentially expressed across these classes with a false discovery rate <0.5%. Unsupervised hierarchical clustering of samples with these module-scores revealed that 29 immune-related modules covering immune-tolerance, IFN-gamma biology and gene signatures associated with response to checkpoint inhibitors such as durvalumab, were significantly upregulated in HighB-High2wk tumours. These modules include some genes involving immune-tolerance previously associated with resistance to endocrine therapy. As expected, downregulation of proliferation-related modules was observed in all tumours following 2wks of AI.


Several immune-related gene modules were upregulated in BC tumours with early resistance to AI. Our results suggest that high immune-tolerance could be associated to early resistance to AI and warrant further investigation on the use of immune-checkpoint component inhibition in this setting.


Impact statement

A potential role of immunotherapy in ER+BC-resistant to AI, providing rationale for immune-checkpoint inhibition in this setting.