NEW YORK (GenomeWeb) – Researchers from the National Cancer Institute's Clinical Proteomic Tumor Analysis Consortium (CPTAC) have completed a proteogenomic analysis of breast cancer patient-derived xenografts.
Published last week in Nature Communications, the study identified and validated a number of gene-based cancer subtypes and genomic drug targets while also discovering several regulatory processes and possible drug targets apparent only at the proteomic and phosphoproteomic level.
The findings indicate the potential utility of proteomic data for clarifying and prioritizing the many and highly complex genomic features present in breast cancer and other cancers," said Matthew Ellis, professor of oncology at Baylor College of Medicine and an author on the study.
"Genomic data hasn't been very successful in, for example, separating patients out for treatment with drugs like MTOR inhibitors or PI3 kinase inhibitors," he said. "Those pathways are affected by mutations at multiple levels: loss of tumor suppressors; gain of function in catalytic subunits; loss of function in regulatory subunits; downstream gain of function mutations. The number of potential mutations in those pathways are myriad."
This has led researchers, in CPTAC and elsewhere, to investigate whether proteomic and, specifically, phosphoproteomic data could better help physicians predict a patient's response to a targeted therapy.
In the Nature Communications study, Ellis and his colleagues looked at 24 patient-derived mouse xenografts, in which patient breast tumors were transplanted and then grown in mice. Using mass spec analysis on a Thermo Fisher Scientific Q Exactive instrument, they identified and generated relative abundances for 10,068 proteins and 36,609 phosphorylation sites across the samples.
This protein-level data recapitulated several established genomic breast cancer subtypes and identified known drug targets. It also showed correlation with genomic events like copy number variations and mRNA up-regulation. For instance, HER2 protein and phosphoprotein expression were upregulated in the HER2-E xenograft models WHIM8 and WHIM35.
However, the mass spec data also highlighted differences at the genomic and proteomic levels that could inform treatment decisions. For instance, a mouse xenograft WHIM14, which had low HER2 expression unexpectedly responded to treatment with the HER2 inhibitor lapatinib. Phosphoproteomic analysis found that this tumor had high levels of EGFR phosphorylation, suggesting that the response to lapatinib was likely due to inhibition of EGFR, rather than HER2. Looking at proteomic data from other CPTAC breast cancer samples, the researchers found three others with outlier EGFR expression, suggesting, the authors noted, that the potential of EGFR as a therapeutic target in breast cancer requires more exploration.
Looking at PI3k signaling, Ellis and his colleagues attempted to add proteomic and phosphoproteomic data to improve targeting of combined PI3K and MTOR inhibition. As Ellis indicated above, PIK3CA mutation status by itself did not predict the efficacy of this treatment, but adding the protein-level information they found that drug response appeared to be linked to the overexpression and phosphorylation of downstream signaling targets in the PI3K pathway. Specifically, the tumors that responded most strongly showed overexpression of the proteins AKT1 and AKT2 as well as phosphorylation of the serine 122 and 475 sites on AKT1.
Looking across the full set of PDX samples in the study, the researchers found that more than 20 percent exhibited both copy number amplification of PI3KCA at the genomic level and overexpression of AKTs at the protein level, indicating, they said, that this "may be a common signature of breast tumors."
"When you think about it, next-generation sequencing has not been hugely successful yet, with a lot of controversy in terms of drug choice based on genomic principles," Ellis said. "There are a few nice examples, ALK translocations in lung cancer and EGFR mutation in lung cancer and maybe HER2 mutation in breast cancer, but there are all very small populations. For the bulk of these cancers, genomic analysis alone has not provided us with elegant predictive markers."
"So, when we look at that [Nature Communications study], there is a lot of emphasis on methodology, but also you see the beginnings of this new deeper profiling where we are [using] proteomics to guide us towards more accurate therapeutic hypotheses than the genome alone."
Even in the case of well-established disease subtypes like HER2-positive breast cancer, there is reason to believe proteomic and phosphoproteomic data can improve treatment, Ellis said. "HER2-positive breast cancer is a hugely heterogeneous disease, and we all know there are patients there who are getting Herceptin and aren't benefiting and patients who should have gotten Herceptin and didn't get it. And I think these are all things that can be addressed using proteomics."
Ellis said that he and his colleagues in the CPTAC program hope to tackle more specific, narrowly defined clinical questions in the third stage of the initiative. The NCI announced in November that it was soliciting applications for this stage, which will include funding for three Proteogenomics Translational Research Centers where researchers will use proteogenomics to investigate drug response in specific clinical contexts.
This, Ellis said, will allow the initiative to tackle questions of the actual clinical utility of the proteogenomic methods developed and used in the second stage of the initiative.
"Are we or are we not going to have mass spec-based proteomics as a clinical tool?" he said. "I think that is what CPTAC 3 is now going to address."
For his part, Ellis said he believed mass spec-based proteomics would find a place in the clinic for guiding cancer therapy, likely within the next five years.
What exactly those assays will look like remains to be seen, he said, though he suggested the most likely route would be targeted measurements on high-resolution instruments, such as the parallel-reaction monitoring approach a number of researchers have been exploring as an alternative to traditional multiple-reaction monitoring assays.
" I think there are all sorts of very intelligent [mass spec] technologies that can produce some very useful clinical profiles," Ellis said.