NEW YORK (GenomeWeb) – Researchers at the University of Washington have completed an in-depth proteomic analysis of triple-negative breast cancer (TNBC).
Looking at 20 human derived breast cancer cell lines and four patient tumor samples, the study, which was detailed in a paper published today in Cell Reports, identified a variety of protein-defined breast cancer subtypes as well as potential therapeutic targets.
The work also highlighted the high heterogeneity of TNBC and the resulting challenges of treating the disease.
Breast cancers are commonly classified according to a patient's receptor status – the presence or absence of estrogen receptor, progesterone receptor, or epidermal growth factor receptor. One or more of these receptors are expressed in more than 80 percent of breast cancers and can be used as therapeutic targets.
In triple negative breast cancer, however, these receptors are absent, leaving clinicians without targeted treatments for the disease. Additionally, TNBC is typically more aggressive and has worse prognosis than receptor positive forms of the disease. Therefore, there is great need for better, targeted approaches to treatment.
However, many of the breast cancer targets identified via genomic work are typically less enriched in triple-negative cases, Robert Lawrence, a University of Washington researcher and first author on the paper, told GenomeWeb.
"So we need new targets, and I think the proteome is a good place to look," he said.
While some gene expression work has been done on the triple-negative cell lines investigated in the study, less has been done on the proteomics side,Lawrencesaid. Using a Thermo Fisher Scientific Q Exactive instrument, he and his colleagues profiled the proteomes of 16 triple-negative lines, three receptor-positive lines, and one non-tumorigenic line. They also profiled four tumor samples taken from patients with metastatic TNBC.
In total, they identified 12,775 proteins from 11,466 different genes, with at least 9,000 proteins identified in each cell line. Each cell line was analyzed in duplicate with an average of 80 percent of the proteins identified in both replicate.
Taking this proteomic data, the researchers generated subtypes of the different cell lines, finding that the proteomic subtypes largely tracked with those identified from gene expression data as well as subtypes based on information like cell morphology and cell invasion assays, Lawrence said.
The UW researchers also examined the proteomic data for potential drug targets, looking, for instance, for proteins over-expressed in the triple-negative lines. They also looked for proteomic signatures associated with genetic mutations known to drive cancer development. This, Lawrence noted, allowed them to identify not only the proteins directly affected by the mutations but also related or interacting proteins that could, in some case, be more easily druggable.
"So, for instance, with loss of function mutations, it's hard to think about how you actually target that molecular event even though it is probably causal," he said. "But there might be proteins associated with that mutation that are targetable. So that is something we were going after."
Lawrence noted that another benefit of measuring targets at the protein level is the fact that this allows clinicians to better account for how different expression levels of a protein target impact patient sensitivity to a targeted drug.
"If you're targeting a protein for some mechanistic reason and you never really consider the level that it is expressed in the cells you are looking at, that can have a very big impact on the [tumor's] sensitivity," he said.
"That is obvious, but the importance of this fact is sometimes underestimated," he added, noting that, in fact, currently most companion diagnostics for guiding targeted treatments are based on detecting the presence of the genetic mutation being targeted, which gives little information as to how much of the mutated protein is actually being expressed in the targeted cells.
"For example, PI3 kinase mutations are very common in breast cancer, so there is a whole group of drugs that have been developed to target this kinase," Lawrence said. "But [clinicians] are using just the gene, whether it is mutated or not, as the marker for choosing a PI3 kinase inhibitor as the drug."
To investigate questions of drug sensitivity, the UW researchers screened the 16 TNBC cell lines against a library of 160 drug compounds at eight different concentrations. They found that 123 out of the 160 drugs had a measurable effect on at least one cell line and that each cell line was sensitive to at least five agents.
Using hierarchical clustering to look at patterns of drug sensitivity and protein expression, they found that proteins that were part of the same pathways or complexes also clustered together in terms of drug sensitivity, suggesting, the authors noted, that such an analysis could be valuable for identifying important new protein targets within targeted pathways.
The study data is a resource that "provides multiple [protein] candidates for further validation that people can go after," Judit Villén, a UW researcher and senior author on the paper told GenomeWeb.
The search for new drug targets is "relevant for any cancer, but in particular for triple-negative, because there are no targeted therapies for that one," she said. She added that she and her colleagues envision a future where, rather than assessing the status of only a few receptors or genes, "many different mutations will be measured and the levels of many different proteins will be measured, and from that we will be able to infer the best drug."
Such an approach has been explored to a limited extent in breast cancer, most notably by researchers involved in the Side-Out Foundation, which in 2013 presented data from a pilot study in which researchers including Nicholas Robert, an oncologist at Virginia Cancer Specialists, used genomic and proteomic analyses of 25 metastatic breast cancer patients to guide late-stage treatment.
In that study, the molecularly guided therapies extended progression-free survival by more than 30 percent compared to the patient's last treatment regimen, a result that Robert told GenomeWeb at the time well surpassed the researchers' criteria for success.
The Side-Out study used for its proteomic analyses reverse phase protein arrays run by George Mason University researchers Emanuel Petricoin and Lance Liotta. RPPA has the ability to look at small sample sizes with high sensitivity, but it is typically limited to relatively small panels of targets – in the range of 200 proteins.
By using mass spec, the UW researchers aimed to vastly expand the number of candidate proteins their were able to investigate. Now, Lawrence noted, they need to further investigate a number of proteins that according to their study can serve to guide selection of a drug therapy.
"I think the next step is to actually demonstrate the causality of some of these correlations," he said. "To manipulate the expression of the protein [in the cell line], and demonstrate that it affects the drug sensitivity."
That, he said, would pave the way for development of multiplex assays to panels of proteins linked to sensitivity of specific drugs.
For this multiplex panel development, the researchers will likely turn to the parallel-reaction monitoring high-resolution targeted quantitation approach, Villén said.
PRM uses the upfront quadrupole of a Q TOF or Q Exactive machine to isolate a target precursor ion, but then monitors not just a few, but all of the resulting product ions using its time-of-flight or Orbitrap analyzer.
This offers several potential advantages over conventional triple quad-based mulitple-reaction monitoring assays. For example, because PRM monitors all product ions instead of just a pre-selected few, researchers don't have to determine upfront what the best transitions to monitor will be, significantly reducing assay development time. Additionally, the larger number of product ions monitored via PRM should improve the specificity of the analysis, since more transitions will be available to confirm a peptide ID. It might also reduce the effects of co-isolating background peptides.
Villén said the researchers are also exploring data-independent acquisition mass spec approaches on some of their projects, though the Cell Reports TNBC study did not involve use of DIA. DIA has gained users in recent years particularly for quantitative experiments comparing large numbers of samples due to its potential advantages in terms of reproducibility.
Villén said that she and her colleagues had, in fact, found that DIA was more reproducible than standard data dependent acquisition (DDA) mass spec approaches, but that "the software tools are not there to accurately extract all the information."
"We are still more successful with DDA, but DIA is out there and coming, and it has huge potential for targeted and large-scale proteomics, I think," she said.