NEW YORK – A team led by researchers at the Technical University of Munich (TUM) has developed a proteomic workflow for measuring the dose- and time-dependent response of protein post-translational modifications (PTMs) to different drugs.
Called decryptM, the approach could be useful for better determining drugs' modes of action as well as the impacted biological pathways, and could also help researchers better understand the role of PTMs whose functions are currently unknown, said Bernhard Kuster, professor of proteomics and bioanalytics at TUM and the leader of the effort.
He added that the approach might also provide information that could help doctors tailor cancer treatments to individual patients based on their proteomic profiles.
In a paper published this month in Science, Kuster and his coauthors used the approach to measure the effects of 31 cancer drugs spanning six drug classes in 13 cell lines, producing 1.8 million dose-response curves. They identified 47,502 regulated phosphopeptides, 7,316 regulated ubiquitinylated peptides, and 546 regulated acetylated peptides.
The team has made this data publicly available in the ProteomicsDB database that Kuster and TUM established roughly a decade ago for the sharing of proteomic datasets.
The researchers generated their data on a Thermo Fisher Scientific Fusion Lumos Tribrid instrument using TMT labeling, which allowed them to multiplex the different doses and time points for each drug-cell line combination in a single mass spectrometry experiment. This was followed with either immunoprecipitation or immobilized metal affinity chromatography to enrich PTMs of interest.
While a number of experiments have looked at the effects of different drugs on cell line proteomes and PTMs, few have analyzed them in a dose- and time-dependent way, Kuster said.
"The mainstay has been to just take an arbitrary high concentrate of the compound and then basically record the effect [on the proteome of interest]," he said. "You do it in a couple of replicates and then record [expression] fold-changes and do a statistical evaluation of those fold-changes for whatever proteins or PTM sites you can measure."
However, treating the sample with a single dose of a drug doesn't account for the fact that "drugs exert their effects in a dose-dependent fashion," Kuster said, noting that without having this dose-dependent data it can be difficult to distinguish between meaningful changes and less biologically important effects.
"It helps us a great deal with interpreting the data because you can sort the effects by drug potency, and you can focus your attention on those effects that are very potentially modulated by the drug, and you wouldn't spend much time, at least initially, on those effects that happened but only at high doses," he said.
Kuster said that collecting time- and dose-resolved data also helps the researchers better identify what proteins and PTMs are potentially being modulated together, allowing them to more fully flesh out the pathways impacted by a drug and assign modified proteins of unknown function to particular pathways.
"It informs the mechanism of action of these drugs because it becomes possible to delineate the potent effects from the not-so-potent effects," he said. "You essentially measure target engagement and pathway engagement with these profiles."
Kuster suggested that improvements in mass spec instrument performance and multiplexing had made collecting such data more feasible, though he said he found it somewhat surprising how rare such datasets were.
"It seems like a blatantly obvious thing to do, and yet there's not a ton of literature about it," he said.
Gordon Mills, director of precision oncology at the Oregon Health & Science University Knight Cancer Institute, echoed Kuster's comments, noting that collection of "the detailed titration curves" presented in the Science study "really hasn't been done by any other group."
Mills, who was not involved in the research, called it "a major addition to the information that is available to the community."
"This really adds to RNA data and other proteomic data — which is much more sparse than this — to really understand how drugs work, what their targets are, and how cells adapt to the stress of drugs," he said. "I think it’s a major new resource that will be used extensively."
Mills noted that cancer drug profiling work is increasingly moving beyond cell lines and into models like tumor organoids and patient-derived xenografts, where researchers can look at not just tumor cells themselves but also the tumor microenvironment and the spatial relationship of its components. The data being produced by Kuster and his colleagues will be "a wonderful resource to help interpret" data generated in these systems, he added.
Kuster and his lab have continued to build out the dataset. He said it is now roughly 10-fold larger than what was presented in the Science paper, with much of that work done by a single graduate student in his lab. He said his lab aims to have this additional data uploaded to the ProteomicsDB resource by the end of the year.
"It's entirely possible to do in any reasonably well-funded laboratory," he said. "You don't need 10 mass specs to do it; you don't need 20 people."
Kuster said his lab is also collaborating with the German Cancer Research Center to explore whether the drug activity profiles could be useful in tailoring cancer treatments. He and his collaborators are profiling the phosphoproteome of cancer patients with the goal of then identifying drugs that hit the targets and pathways altered in those patients.
While targeted cancer drugs are commonly targeted to patients based on genetic mutations, "there is a large number of cancer patients for whom [best] treatments may not be so clear based on the current state of mutational profiling," Kuster said. "Therefore, we're trying to look at the PTM [profiles] of patients to try to find vulnerabilities, but that is only meaningful if we can pair it with an agent that we know modulates the levels of those PTMs."
He said he and his collaborators have profiled around 500 patients thus far and have a steady flow of patients moving through their profiling pipeline, though Kuster added it will still take some time to determine whether this information will prove useful for guiding patient treatment.
A number of researchers and companies have explored whether patient PTM profiles — primarily phosphoproteomics profiles — could help guide cancer therapies, though such approaches remain little used compared to genomic profiling. As Kuster and Mills noted, however, detailed dose-dependent data on the impact of particular drugs on the phosphoproteome has been lacking.
"Looking at the phosphoproteomes of patients and knowing which of those phosphorylation sites might be modulated by a drug actually opens up the possibility of matching the two things together," Kuster said.