By Julia Karow
In a proof-of-principle study, researchers at St. Jude Children's Research Hospital in Memphis, Tenn., have shown that even limited DNA sequencing can uncover important mutations that drive relapse in acute lymphoblastic leukemia patients and provide clues for more effective treatments. More comprehensive genome sequencing, they say, might deliver additional clinically relevant results.
For their study, published in Nature last month, the researchers sequenced 300 genes in matched diagnosis and relapse samples from 23 ALL patients by Sanger sequencing and found one gene that was mutated in almost a fifth of ALL patients that relapsed.
"There may be other mutations that are driving relapse in ALL, so the hunt is on now, using broad-based next-gen [sequencing] approaches, to comprehensively identify all sequence variants in ALL," said Charles Mullighan, assistant member in the Department of Pathology at St. Jude and the first author of the study. "Likely, we will find other potentially clinically useful hits that could be translated into therapy."
The results of the study are "very exciting," according to Wendy Stock, a professor of medicine at the University of Chicago and a leukemia expert. "The findings suggest that these mutations result in aberrant trancriptional and epigenetic regulation of glucocorticoid response genes," she told CSN by e-mail. "Since glucocorticoids are the most important drugs in the treatment of ALL, this paper sheds new insights into potential mechanisms of drug resistance that may be common to many of the different molecular cytogenetic subsets examined."
Leukemia is still a common cause of death in children, though it has been studied for a long time. Generally speaking, the samples are purer and genetically less complex than those for many adult tumors, "so in many ways, they are a little more tractable for these genomic approaches," Mullighan said.
Previous studies using genotyping microarrays identified genes that are associated with treatment failure and relapse and showed that many genetic changes occur between diagnosis and relapse. "We were interested in doing a more extensive project profiling DNA sequence alterations to try and identify new changes" occurring between diagnosis and relapse, Mullighan said.
At the time they were planning the study a couple of years ago, next-generation sequencing was still relatively expensive, so the researchers decided to start with a few hundred genes as a proof of concept, which they had sequenced by Beckman Coulter Genomics using Sanger capillary electrophoresis technology.
They selected genes known to be mutated or deleted in ALL, genes in the same pathway, genes known to be mutated in other cancers, and other genes of interest, such as tyrosine kinases.
After sequencing those genes in matched diagnosis and relapse samples from 23 patients, they found 52 somatic non-synonymous mutations in 32 genes, many of them novel.
The most striking result was mutations in CREBBP, or CBP, a gene that encodes a transcriptional co-activator and histone acetyltransferase. When they analyzed a total of 71 diagnosis-relapse cases and 270 ALL samples that did not relapse, these mutations were present in almost 20 percent of the relapse cases but were rarely seen in patients that did not relapse.
The mutations were often found only at relapse but not at diagnosis, but in many cases, when the researchers reanalyzed samples obtained at the time of diagnosis, they were able to pick them up in minor cell populations. "It suggests that these mutations confer some property to the cells that render them less sensitive to therapy," Mullighan said.
[ pagebreak ]
"It will be very interesting to see whether these mutations are also noted in relapse of adult ALL — if so, one wonders whether the mutation rate might even be higher, given the higher number of relapses in adult ALL," said Stock.
Functional studies by Mullighan's team showed that the mutations of the CBP gene impair the function of its protein, which is crucial for a common ALL treatment, corticoid steroids, to work. Mullighan explained that these drugs kill cancer cells by activating a gene expression program, but mutations in CBP interfere with this response, so the cells become resistant to the drug. "If we could somehow restore the loss of function of CBP in some way, that might enhance the responsiveness of these cells to treatment," he said.
One way to counteract CBP's deficiency would be to use a histone deacetylase inhibitor, and experiments in cell lines with CBP mutations successfully rendered them sensitive to corticoid steroids.
Other lines of evidence have also suggested that histone acetylase inhibitors might be useful in treating ALL, and clinical trials for testing their use are already being considered, Mullighan said.
Besides CBP, the St. Jude researchers found other genes recurrently mutated in relapsed ALL, but none were as common as CBP mutations. Many of those genes are involved in epigenetic modification, suggesting that this pathway is important in driving treatment responsiveness, he said.
The study results could also help predict which patients are likely to relapse. Previous studies had already found genetic changes at the time of diagnosis that are associated with relapse, according to Mullighan, but this study helped to identify other mutations that could be used to diagnose risk of relapse.
One problem, however, is that these mutations may only be present in a subclone at the time of diagnosis. "We have to get a lot more clever and sophisticated in the way that we are able to pick up these mutations," he said. "Potentially, it's feasible, but one would need to have a next-gen approach where one has very deep coverage of a gene, or regions of a gene, in order to pick up the variants."
Mullighan and his colleagues are now focusing their efforts on broader-based sequencing studies to pick up more relapse mutations, as well as more functional studies. They are now using exome sequencing, while other groups working on ALL have moved on to whole-genome sequencing. "We are already finding some very interesting additional hits that we had not identified before from our pre-next-gen approaches, sequencing and microarray-based profiling," he said.
Eventually, whole-genome analyses will be needed, he said, because targeted sequencing and exome sequencing only address a fraction of the genetic variation present in tumors. "If we ask to identify all genetic changes that are driving relapse, we do need whole-genome approaches that can pick up structural variations and rearrangements," he said. But whole-genome sequencing has its own challenges, he added, including its cost, getting enough coverage to pick up all variants accurately, not missing repetitive or GC-rich regions, and analyzing the data.
Mullighan is also a collaborator in the National Cancer Institute's Therapeutically Applicable Research to Generate Effective Treatments, or TARGET, initiative for ALL. In addition, he studies other subtypes of ALL that are associated with high risk of treatment failure, "where we know, essentially, nothing about the genetics."
But there are a number of challenges to these large-scale sequencing studies, he said. One is to obtain pure enough tumor samples from both the initial diagnosis and relapse.
Another is the sequencing itself, which is "not trivial, although it's becoming much more standard now," Mullighan said. St. Jude is equipped with several Illumina GAIIx and HiSeq instruments but also collaborates with the Washington University Genome Center on other projects, for example a pediatric cancer genome project that will sequence 600 samples (IS 1/26/2011).
[ pagebreak ]
For that project in particular, the partners have already sequenced 50 tumors and are nearing completion of the analysis. "We have already found some very interesting recurring hits in novel genes that have not been previously identified in some subtypes of leukemia and other disorders," he said. "And this has given us major insights into mechanism, potential new pathways causing high-risk disease, and some of these are also druggable," he said.
The data analysis in particular can be tricky. "We have found for many of our studies at St. Jude that accurately calling variants is not necessarily straightforward and requires quite a sophisticated bioinformatics pipeline" that is able to identify both single-nucleotide variants and indels, Mullighan said.
Overall, large-scale cancer sequencing projects will likely provide important insights into disease mechanisms, he said, cautioning that many potential targets "are going to be very tough to drug — it's going to be very difficult to rapidly take them into the clinic."
On the other hand, "we are already finding genetic changes that are druggable," either by existing agents or by novel agents that can intervene in the pathway the mutation is part of. "I think there is a lot of potential from these approaches."
Will every cancer patient have their tumor sequenced anytime soon? "There are barriers for whole-genome sequencing on the individual patient at the moment, including the time of sequencing, the cost, the bioinformatics," Mullighan said, though he added that "it's likely that all those will be solved.
"I don't think it is ready for widespread use at the moment," he said. "But likely, that will come."
Have topics you'd like to see covered in Clinical Sequencing News? Contact the editor at jkarow [at] genomeweb [.] com.