Treating acute leukaemias of ambiguous lineage

Dr Lindsey Montefiori - Department of Pathology, St. Jude Children's Research Hospital, USA

We are focussed on a group of high risk acute leukaemias that are collectively known as lineage ambiguous acute leukaemias. These are relatively rare but they have very poor prognosis and one of the reasons for this is because there are multiple challenges in terms of accurately diagnosing these leukaemias in patients because they typically express antigens associated with both the myeloid and the lymphoid lineage. So that poses challenges for diagnosing and classifying them and also in choosing the most appropriate therapy.

There is relatively little known about their underlying biology because they are so rare so this study focussed on this collective group of acute leukaemias, taking an approach that’s orthogonal to looking at cell surface markers to classify them. So we looked at the transcriptional profiles of as many lineage ambiguous leukaemias as we could compile, spanning from mixed phenotype acute leukaemia, acute undifferentiated leukaemia, early T-cell precursor acute leukaemia which is considered a subtype of T-ALL but these leukaemias frequently express myeloid and stem cell antigens. So by looking at the gene expression profiles of these leukaemias we aimed to better classify them and understand their underlying biology.

Can you explain the methods used?

As I said, we took a transcriptional approach. So we analysed RNA-Seq data; in this pan-acute leukaemia analysis there were over 2,500 samples. This allowed us to perform various methods to look at the relationships between different leukaemias. So we used [?] projection analysis and hierarchical clustering to examine the relationships among this entire cohort. As we expected, the group of B-cell acute lymphoblastic leukaemias formed very distinct gene expression clusters that corresponded to genomic alterations but this group of the non-B leukaemias comprised of T-ALL, AML, and these lineage ambiguous leukaemias we did observe substructure within this group and, interestingly for us, specifically within the lineage ambiguous leukaemias we identified a new subtype and I say new just because it hasn’t really been described in the literature before and this is likely due to limits on sample size. So in this analysis we identified 63 samples that fell within a very distinct gene expression cluster.

What were the key results?

The first main result is this identification of this subgroup. All of these samples are defined by having structural alterations, genomic structural alterations, that target the BCL11B locus on chromosome 14. We identified this from whole genome sequencing analysis and in a subset RNA-Seq was able to identify this as well. The characteristics of these structural alterations is that this leads to expression of BCL11B from only one allele, so the allele that has this structural alteration. But importantly we found that this group was comprised of multiple diagnostic entities and this ties back to the challenges that I mentioned in the beginning.

So in the clinic these leukaemias will be diagnosed as different diseases, whether it’s T/myeloid MPAL or ETP-ALL, but we found that actually about 30-40% of this biologically similar group of leukaemias were T/myeloid MPAL and 35% were ETP-ALL. So this suggests that our ability to identify biological distinctions among leukaemias just based on immunophenotype or morphology is insufficient.

What did the chromatin topology analysis demonstrate?

This is digging into the molecular underpinnings of this group. So, as I said, we identified structural alterations targeting the BCL11B locus and this gene expression group. These alterations result in BCL11B, the gene… so the breakpoints spare the entire coding gene, I think that’s important to indicate. So this results in BCL11B being placed in other regions in the genome. Because we saw monoallelic expression we immediately hypothesised that these alterations are targeting BCL11B expression. So they’re not perturbing the gene itself, like we do see mutations in this gene in canonical T-ALL, but here we hypothesised that it was what’s known as enhancer hijacking which is essentially where a gene that’s normally repressed is aberrantly or atopically activated because it’s now placed in proximity to a strong transcriptional activator. These are enhancers and there are hundreds of thousands of them in the genome and they are active at very specific times during development.

So we saw that all of the rearrangements that we observed in this group placed BCL11B near a very strong enhancer that’s active in haematopoietic stem and progenitor cells. One of the implications of that observation is again supporting that the cell of origin for these leukaemias is a similar progenitor cell because BCL11B is normally only expressed in the T lineage. So this suggests that this alteration occurred before the onset of T-cell differentiation.

So the chromatin topology analysis was aimed to first prove that these rearrangements place this BCL11B gene near active enhancers. So one benefit of the approach we took, which was called HiChIP, is that it simultaneously identifies enhancers by virtue of being able to tell you where H3K27 acetylation is located in a genome. This is a post-translational histone modification that’s associated with transcriptionally active chromatin. It also gives you information on the other regions of the genome that regions with this marker are interacting with. So if you want to identify which genes a given enhancer is interacting with you could use this approach to get at where that enhancer is and what are the genes that it’s interacting with.

So the take-home message is that this approach demonstrated that these rearrangements place BCL11B near active enhancers that correspond to active haematopoietic stem and progenitor cell enhancers. Importantly, these enhancers are forming chromatin interactions, so physically looping to the BCL11B gene. That provides the strongest line of evidence that we can provide in primary patient samples that the mechanism of T-cell [?] expression is through enhancer hijacking.

How can these results impact future research?

That’s a great question. There are a couple of things that we would like to say about that. The first is that we believe that our results support that taking transcriptional and genomic approaches to better classify leukaemias is a very valid approach and that we need to continue doing this to continue identifying these subtypes that are present within homogenous diagnostic entities. So that can lead to our better ability to classify leukaemias and then start to correlate these different occurrences with treatment response and outcome. So that’s a first step that needs to be taken.

We acknowledge that identification of this alteration is based on next generation sequencing and so its immediate utility to identify this group in the clinical setting may not be immediately applicable. But hopefully within a few years that will definitely be a reality.

We also believe that it’s important now to start to engineer and utilise preclinical models, faithful preclinical models, using both engineering mouse models and patient-derived xenografts to model these alterations. I didn’t mention this but we observed that 80% of these samples also had FLT3 alterations, most commonly internal tandem duplications. So one of the immediate clinical implications is whether FLT3 inhibitors would be specifically effective in this group. So that’s something we would like to test in preclinical models.