Impact of mutational profile at diagnosis for MDS/CML

Bookmark and Share
Published: 24 Jun 2017
Views: 2990
Rating:
Save
Dr Guillermo Montalban Bravo - MD Anderson Cancer Center, Houston, USA

Dr Montalban-Bravo speaks with ecancer at EHA 2017 about genomic sequencing of over 200 patients at diagnosis, identifying mutational markers which may predict response to hypomethylating agents

He describes that mutations in TET family proteins and p53 were among the most common aberrations, and discusses how treatment pathways can be uncovered through sequencing analysis.

Dr Montalban-Bravo last spoke with ecancer at ASH 2016 about pure erythroid leukaemias.

For more on the clinical utility of sequencing and genome libraries, watch Dr Yu Shyr speak at AACR 2016.

ecancer's filming has been kindly supported by Amgen through the ECMS Foundation. ecancer is editorially independent and there is no influence over content.

The abstract we have submitted and which is an oral presentation is a study where we try to evaluate the impact of the mutational profile of patients with MDS at baseline, at the time of diagnosis, to try and see if any of the features that we can identify by next generation sequencing help predict the responses to HMA therapy, to hypomethylating agents, which, as you know, are the standard therapy for these patients. So we did baseline sequencing using an amplicon-based next generation sequencing evaluating 28 genes which we know are recurrently mutating in myeloid malignancies in these patients. The 222 patients, all of them had been treated, were then treated with hypomethylating agents and our objective was to try and identify specific features or mutational profiles that could predict for response.

We not only evaluated the actual presence or absence of a mutation, we wanted to study whether specific mutational patterns or clonal sizes of mutations, now that we can also use variant allele frequencies of mutations, so how predominant a specific mutation in the whole bone marrow cellularity is, to try and see if maybe there are mutations that are present at a lower threshold do not predict for response and others did. What we found was that among these 222 patients the presence of certain mutations like mutations in ASXL1 and RANX1 and mutations when we looked at actual pathways, not so much as the actual gene, patients who had mutations in genes involving chromatin regulation or kinase signalling had a lower likelihood of achieving a response.

We then wanted to also interrogate whether the variant allele frequencies impacted response. So in order to try and see if there was any specific cut-off of variant allele frequency predicted for response we generated specificity and sensitivity ROC curves and we found that those patients who had a p53 mutation at a variant allele frequency greater than 0.31 were the ones that really responded worse than patients with a mutation in p53 at a lower VAF. So we then wanted to know also the impact on complete response and in CR we found that, again, kinase signalling and mutations in ASXL1 also predicted for a lower likelihood of response.

The next question was not only whether response rates change with specific mutations at baseline but if we can predict differences in the time to response and duration of response. What we found is that, again, p53 mutations at a variant allele frequency of higher than 0.31 as well as the increased number of mutations and NPM1 mutations could predict for a longer or shorter time to response. So patients with NPM1 tended to respond quicker, probably because they tended to have higher blasts and therefore the responses can be evaluated earlier. On the contrary, patients who had a high number of mutations, three or more, tended to have shorter durations of response as well as p53 mutations. So we do see that gene sequencing evaluation at the time of initial diagnosis may help predict how the patients are going to respond and therefore perhaps that can help us think on future treatments like a transplant and trying to see if we can decide which patients should proceed to other therapies earlier on because we expect they’ll respond in less time. Of course, it’s only 222 patients, this needs to be evaluated on larger cohorts but we think it’s an important finding.

I notice among the readouts that the TET mutations seem to be one of the most common. Are there any potential actionable targets or any drugs that would work towards that?

That’s a good question. We know that TET2, there have been previous publications, probably it’s one of the only genes in which we have three or four big papers in which there has been experience showing that perhaps these patients respond better to treatment, particularly those who have a high variant allele frequency of the TET2 do not have ASXL1 mutations. We could not reproduce this on our test cohort, even though we did see that ASXL1 predicted for a shorter response, but technically there’s a biological mechanism behind, a reasoning that would anticipate that patients with TET2 probably should respond better to treatment.

In terms of specific targets we don’t have any TET2 specific inhibitors on clinical trial but hypomethylating agents should technically work well in these patients just because we know they have a hypermethylated phenotype. But there’s no specific TET2 inhibitor such as IDH inhibitors or FLT3 inhibitors that are being developed.

For the less common ones was there any chance of, say, a co-dependence, that if there was maybe one then you could get away with just an at risk phenotype but the two together was the trigger for disease?

Correct. So we wanted to evaluate also whether co-mutation patterns mattered. Of course when you have a cohort of 222 patients there’s a lesser amount of possible co-mutations. So we did find five significant co-mutated genes and we also evaluated using clonal relationship analysis, we wanted to see whether any of the genes that typically co-occurred tended to happen one on a major clone versus the second tended to appear on a minor clone and whether that impacted response. We did not find any differences in terms of, for example, ASXL1 or RANX1 or ASXL1 and ECH2 which tended to co-occur. There were no significant differences whether one of them was to be normally the major or minor clone and the presence of both did not impact response. What we did see that, irrespective of what gene is affected, the presence of multiple mutations did tend to show that patients were going to respond a less amount of time. It did not significantly predict for achieving a response differently but they did respond for a lower amount of time.

Are there plans to match these mutational sequences against, for example, some of the vast gene libraries that other labs have been piling together?

That’s a good point. That would actually be very interesting, trying to look at available data from other studies where we know there’s populations of patients that have been treated with hypomethylating agents to see if we could find the same signals. Actually that’s something that we could potentially design. Of course the main question is trying to come up with multicentre studies or cohorts of patients where we can increase the end to have more potency to determine this kind of analysis.