DNA Repair Deficiency as a Biomarker in Breast Cancer

June 1, 2015

In this interview, we are discussing the use of DNA repair deficiency as a biomarker in breast cancer.

Today we are speaking with Philip Schouten, MSc, a [PhD] graduate student and Sabine Linn, MD, PhD, a medical oncologist specializing in breast cancer--both of the Netherlands Cancer Institute. Both Philip and Dr. Linn conduct research on cancer biomarkers, particularly biomarkers that can be used to guide decision making in treatment.  They recently penned an editorial in the Journal of Clinical Oncology on the challenges of using DNA repair deficiency as a predictive biomarker in breast cancer, particularly for triple-negative disease.

-Interviewed by Anna Azvolinsky, PhD

OncoTherapy Network: Dr. Linn, could you first describe the current agents to treat triple- negative breast cancer?

Dr. Linn: Yes, of course, I think we should talk about the adjuvant setting and metastatic setting. In the adjuvant setting, mainly doxorubicin, cyclophosphamide, and a taxane are used. In the neoadjuvant setting, so before surgery, one might also add carbloplatin to a taxane since two studies have shown that this increases the pathologic complete response remission rate with about 15% to 20%. But there is discussion about whether this should be standard or not. It is not standard yet in the guidelines. For BRCA mutation carriers, there is an ongoing study globally that randomizes patients to 1 year of olaparib or placebo, and olaparib is a PARP inhibitor. And this is after standard chemotherapy and this study is called the OlympiA study. In the metastatic setting, currently there is no standard first-line chemotherapy, and these patients can only be treated with chemotherapy and it's all in the palliative setting so it helps to prolong survival, but all of these patients will eventually die of the disease. Agents that are commonly used is capecitabine, the combination of carboplatin and gemcitabine, especially in platinum-naïve  patients, also taxanes are used, especially when patients did not receive taxanes in the adjuvant setting, and other agents commonly used are vinorelbine and eribulin. And then in the experimental setting, there are many, many agents being studied. I think the most promising agents being studied now are PARP inhibitors, but also immune-checkpoint inhibitors.

OncoTherapy Network: Philip, what are the genetic commonalities among these types of tumors and maybe also the differences for this heterogeneous set of tumors that are neither HER2 nor estrogen receptor-positive?

Philip Schouten: So triple-negative breast cancers are a heterogeneous disease, and can be exemplified by the five subtypes in the gene expression analysis done by Lehmann et al., but for the commonalities there is enrichment for BRCA1 mutations and if you are a BRCA1 mutation carrier and you are diagnosed with breast cancer, then there is a high chance the disease is triple-negative.

OncoTherapy Network: Dr. Linn, could you explain exactly what is DNA repair deficiency and what are the tests to identify whether a tumor has this phenotype?

Dr. Linn: I think Dr. Schouten would like to answer this question as well.

Philip Schouten: So there are multiple types of DNA repair for different kinds of lesions that can happen to the DNA. One of most toxic lesions is a DNA double-stranded break, and these are only repaired error-free by homologous recombination and one of key players in the homologous recombination DNA repair pathway is BRCA1. So, that’s the marker of interest in some studies. To find such a DNA repair deficiency, you could look at multiple measures. So there are single genes that when inactivated, for example, BRCA1 or BRCA2, or other genes in the same pathway, if you inactivate them, the DNA repair cannot be done through homologous recombination and this inactivation can occur due to mutation in the gene or promoter methylation which silences a gene. Then, you can also look at the effects of such DNA repair inactivation has on the genome. And there have been multiple gene expression and copy number array signatures that can identify aberrations that are caused by the inactivation of the DNA repair pathway. So, the signatures that are out there now are measuring telomeric imbalance, it's called an HRD-LOH score, there’s a score that measures large aberrations   over the genome, it’s called the large-scale transition score loss of heterozygosity over the genome and this is developed by Myriad Genetics. There is a chromosome instability score and we made our own BRCA1 and BRCA2-like signature based on copy number data and you can also look at the functional inactivation, and there are several labs in the world that use Rad51 foci formation to identify those tumors that are supposedly deficient in DNA repair. 

OncoTherapy Network: Based on prior trials and other types of research, what do we know about targeting DNA repair deficiency and why is it potentially difficult?

Philip Schouten: So there have been many studies done over the last 15 years trying to identify a biomarker to identify the tumors that have a deficiency in homologous recombination and then second, you want to identify a drug regimen that will target this deficiency. And despite all of these studies, we think that the most difficult part is to interpret many of these studies because the methodological set up doesn’t allow for the proper conclusions. Sometimes the control groups are missing and then basically you don't know if you have a prognostic or predictive marker, and you don't know if you should treat or not treat [based on this marker].

Dr. Linn: I think that in fact, it is not so difficult to target because we know from some retrospective studies that for instance, when you apply drugs that cause DNA double-stranded breaks in BRCA1 mutation carriers for instance, there are indications that works particularly well in this subgroup of patients and when they have this mutation, we think that they have homologous recombination deficiency in their tumors. The issue is that the set up of most studies done so far have not been right. And that is because if you want to properly evaluate a biomarker, you need patients with the biomarker that have been treated with the drug of interest, but also patients with the biomarker that have not been treated of interest and on the other side, you need patients that do not have the biomarker of interest, but are treated with the drug of interest, and also a control group that does not have the biomarker on interest and has been treated with a different treatment. And so you need four groups of patients and what has mainly been done is studies have been published where you see patients with the biomarker of interest that have been treated with the drug of interest and that has been compared with patients that do not have the biomarker of interest and have been treated with the same drug of interest. The problem there is you don’t know whether the biomarker has prognostic value or predictive value, or is a mixture of both. So the problem is that the proper control group that hasn’t been treated with the drug of interest is mostly missing in studies.

OncoTherapy Network: So do you see a way forward to amend this or is it difficult to address this issue?

Dr. Linn: So it is not so difficult. So one way is something that we have been doing with the BRCA1-like class, we have analyzed this biomarker, or putative biomarker, using patients’ material from those who have been treated in the context of a randomized clinical trial, where half of the patients have been treated with drugs that cause a lot of DNA double-stranded breaks while the other half of patients have been treated with standard chemotherapy that causes not so many DNA double-stranded breaks. And when you use this set up, so in a post-hoc analysis of a randomized clinical trial, you can show that the biomarker works or doesn’t work. We found indications that the BRCA1-like classified can identify patients who derive substantial benefit from high-dose or intensified chemotherapy containing alkylating agents that cause a lot of DNA double-stranded DNA breaks. The hazard ratio was 0.2 in three separate studies and it was always the same hazard ratio so it looks pretty strong. Another way to do this is to take a phase II study where you have been treating patients with our without a biomarker of interest with a drug of interest and then look in a database--preferably nationwide--where you find matched controls to the cases in your phase II trial who have or have not the biomarker and have not been treated with the drug of interest. Then it's a matched case control design where you also have patients that have the biomarker of interest but have not been treated with the drug of interest. This is something that can be rather easily accomplished so you don't need a prospective, randomized trial, but could even use studies that have been carried out already. And then the strongest way to do this is to set up a prospective, randomized clinical trial and then depending on whether you believe in your biomarker or not, you can include only patients with the biomarker of interest or you include all patients, and then stratify for the biomarker. And some of these strategies are currently being carried out.

OncoTherapy Network: Thank you both so much for joining us today.

Dr. Linn: You’re welcome!