Pathway-based signatures predict patient outcome, chemotherapy benefit and synthetic lethal dependencies in invasive lobular breast cancer

Introduction

Invasive Lobular Carcinoma (ILC) is the second most commonly occurring histological subtype of breast cancers with up to 15% diagnosed.  E-cadherin loss has been observed in >85% of ILCs and may be the cause of the lack of cellular cohesion. ER+ve ILCs have worse disease-free survival and overall survival at 10 years compared to ER+ve Invasive breast carcinoma-No specific type (IBC-NSTs).

Genomic studies have identified mutations in the PI3 kinase/AKT pathway, MAP kinase pathway and glutamate receptor signalling in ILCs. Despite the morphological and molecular differences between ILCs and IBC-NSTs treatment options are similar, surgery, radiotherapy, chemotherapy and hormone treatment. In the short-term prognosis is good however, long-term survival (>5 years) is poor, complicated by an increased metastases to bone and GI-tract. In this study the authors identified prognostic biomarkers for ILCs using a pathway-specific approach and tested them against a larger cohort of samples

Main Points

• Seven independent breast cancer cohorts were analysed with the Metabric cohort assigned as the discovery cohort as it contained normal breast samples and long-term outcome data.
• 1398 differentially expressed genes were identified between ILC and normal samples.
• 10 of these genes were associated with poor outcome and 42 genes associated with good outcome in the training cohort.
• Pathway analysis identified 135 enriched pathways including cell cycle, metabolic and immune system related pathways. This was condensed to 29 pathways by removal of redundant pathways.
• Using the Metabric study and the SIMMS algorithm, a univariate hazards model was created for each pathway. This hazards model was tested in 6 independent studies and demonstrated 25 pathways were significantly associated with patient outcome. The tope three pathways were cell cycle, regulation of beta cells and haemostasis.
• The SCAN-B cohort represent the latest clinical modalities with chemotherapy treated and naïve subgroups. 17 of the 25 pathways were still prognostic in the untreated group whilst none were prognostic in the chemotherapy treated subgroup suggesting that these pathway biomarkers could predict high risk patients that could benefit from chemotherapy.
• Using a machine learning model, multiple pathways were aggregated and the resulting multivariate signature trained on the Metabric cohort. Deployment of this signature on 6 independent validation cohorts demonstrated that the signature was able to classify patients into appropriate risk groups with a 10-year survival of 60.4%1 for the high risk group and 78.8% for low risk.
• Deployment of the signature on the SCAN-B cohort (With the most up-to-date treatment strategies were included) displayed robust prognostic ability. Further analysis of tumour subtypes, where 18 patients were either TNBC or HER2-positive, five were predicted to be low risk and 13 high risk. The aggregated signature was also able to predict the benefit of chemotherapy in the high risk patients but not low risk.
• The signature was also deployed on the 1795 IBC-NST and ER+/HER2-negative patients in the SCAN-B cohort. Whilst there was some separation of risk groups in this cohort there was significantly reduced performance compared to the ILC cohorts.
• The cohort described by Guedj et al. was used to investigate metastases free survival using the aggregated pathway signature. The signature was able to accurately predict the likelihood of metastasis. This was confirmed by RNA-seq of a separate cohort of ILCs.
• This signature also demonstrated superior performance in prognostic potential compared to 3 clinically approved predictors MammaPrint, Oncotype DX and PAM50.

Conclusion

Discovery and independent validation of biomarkers specific for Invasive lobular breast cancer remains an unmet need. The authors utilised 7 ILC gene expression studies to identify the prognostic potential of dysregulated pathways. The pathway signature score was able to predict high risk patients that would benefit from chemotherapy and accurately estimate the risk of 10-year metastasis. In comparison to 3 clinically available prognostic biomarker panels assessing 6 independent ILC cohorts the pathway signature had the highest prognostic capacity. Identification of this signature addresses the lack of tests available to distinguish ILCs from other histological tumour types and identify drug discovery and treatment strategies.

EPISTEM SERVICES

Gene Expression Analysis

Epistem is a GCLP-accredited laboratory specialising in assessing biomarker, target discovery and personalised medicine to inform drug development and clinical studies. We regularly employ multi-omic assessment studies which cover a wide spectrum of tissues and cells. We offer microarray, qPCR and NGS services for gene expression, whole exome and epigenetic analysis in all species.

Oncology

Epistem offers several pre-clinical human cancer models (including breast cancer) to assess the efficacy of cancer therapies as single agents or in combination with chemo- and/or local radio-therapy. To aid in the assessment we have a Cytation 5 live-cell imaging system and a 15 colour Novocyte flow cytometer, which gives the opportunity to evaluate treatment effects and elucidate the function of specific immune cell types in vitro and ex vivo.

Bioinformatics

All of our models are supported by bioinformatics analysis, including differential expression and pathway analysis. We also offer biomarker signature discovery and analysis using our proprietary software in clinical and preclinical samples.

Plucked Hair Biomarker Platform

Epistem provides a unique plucked hair biomarker platform for targeting intracellular signalling pathways in inflammation, oncology, fibrosis and other therapeutic areas. Plucked hair provides a minimally invasive surrogate tissue to assess epithelial tissue drug-induced changes. Effects on mRNA and protein expression levels can be analysed in both preclinical and clinical samples