Developing a Diagnostic Tool to Predict Response to Chemotherapy

Approximately 16,700,000 people are diagnosed with breast cancer (BC) worldwide each year, with 500,000 people dying annually. In the UK alone, each year nearly 62,000 people are diagnosed with BC, with one in eight women developing BC in their lifetime. Breast cancers are currently treated with a combination of chemotherapy, targeted therapy and surgery. However, breast cancers are not identical; each tumour's individual characteristics affect how they respond to treatment. In many cases, treatment options are limited and patients are often given sub-optimal treatments which are associated with burdensome side effects. For instance, despite chemotherapy being offered to about 60-70% of patients with BC, either alone or in combination with other targeted therapies, results from a meta-analysis of 123 randomised trials including more than 100,000 patients has shown that chemotherapy reduces recurrence and mortality in only 20 to 33% of patients. Therefore, 80-67% of patients endured this aggressive chemotherapy treatment and did not benefit; unnecessarily suffering the serious physical and emotional side effects, including a risk of permanent heart damage. Currently, there is no predictive test for selecting which patients will benefit from receiving chemotherapy and which will not.

In clinical practice, the decision to use chemotherapy or not depends on evaluating the risk of recurrence and prognosis by interpreting prognostic clinicopathological features including high cost multi-gene tests such as Oncotype DX (Genomic Health Inc.), Mamma-Print (Agendia), and PAM50 (NanoString). Unfortunately, almost all these molecular approaches share common issues, such as insufficiently high levels of evidence, overfitting of computational models, and high false discovery rates. Furthermore, they might not be available for clinical, logistical or financial reasons. Therefore, there is an urgent need for a cost effective, reliable, sensitive, specific, validated biomarker based approach for optimising chemotherapy treatments in patients with BC.

Recently, the investigators have shown that an accurate prediction can be made by testing the activity of a protein called 'SPerm associated AntiGen 5' (SPAG5) in tumour tissue. In a study, published in Lancet Oncology (2016), the investigators showed that SPAG5 gene amplification, as well as SPAG5 transcript and SPAG5 protein overexpression, were all associated with poor clinical outcome, and were independent predictors for chemotherapy response.

The prognostic and predictive power of SPAG5 outperforms many currently used tests including: the standard cancer proliferation index (Ki67), prognostic clinicopathological factors such as the American Joint Committee on Cancer (AJCC) stage and Nottingham Prognostic Index, as well as other currently available multigene-tests including PAM-50, 96-gene genomic grade index, Genomic Chemo Sensitivity Predictor, the Diagonal Linear Discrimination Analysis of 30-gene signature, and the Adjuvant Online index.

The most immediately useful aspect of our original findings is the potential ability to distinguish those patients with BC who are likely to benefit from standard Anthracycline combination (AC) chemotherapy regimens from those who will not. Therefore our findings have the potential to deliver an accurate predictive biomarker for chemotherapy response in BC which would enable the effective tailoring of treatment to the individual patient. Furthermore, the analysis of SPAG5 expression could underpin the development of novel strategies for more effective management and treatment of the disease.

The work undertaken by the investigators on SPAG5 testing has relied on a commercially available polyclonal antibody (PAb) against SPAG5 marketed by Sigma Aldrich and produced by Atlas Antibodies. This has a number of problems when aiming to develop a SPAG5 based clinical test. Firstly, a PAb is a mixture of antibodies and so it lacks specificity and sensitivity. Also, as a PAb it is extracted directly from an animal, constraining the supply and placing a finite limit on the availability of the antibody, as the animal will die one day ending our ability to provide the test. Thirdly, the PAb is owned by a company and so the cost and availability of any test will be subject to that company's decision. Therefore it is essential to develop a monoclonal antibody (MAb) that is suitable for immunohistochemistry staining of Formalin Fixed Paraffin Embedded (FFPE) tissue.

To overcome these problems the investigators chose to develop a SPAG5 targeting MAb that can be taken through the evaluation, clinical testing and regulatory approval process to become a SPAG5 clinical test. Moreover, once the MAb is developed, the validation of previous results will be carried out on a large number of BC cases to verify the prognostic and predictive powers of the antibody. In addition, the predictive utility for chemo-sensitivity of SPAG5 MAb and PAb will be compared to each other and to rival Immunohistochemistry (IHC) tests such as Ki67 and the rival gene expression tests. A bespoke gene panel will be developed for the nanoString nCounter™ FLEX instrument, featuring genes used in PAM-50, 21-gene recurrence Genomic-Chemo-Sensitivity-Predictor, and the 30-gene-DLDA tests, alongside SPAG5 and Ki67.

In June 2018 the investigators were granted funding to pursue this work by the NIHR Invention for innovation grant program.