Skip to main content
Premium Trial:

Request an Annual Quote

GSK-led Team Develops Mass Spec Thermal Profiling Method for Studying Protein-Drug Interactions


NEW YORK (GenomeWeb) – A team led by researchers at Cellzome, a part of GlaxoSmithKline's R&D division, has developed a thermal profiling-based method for identifying protein drug targets and measuring protein-drug binding.

Detailed in a paper published this week in Science, the method uses mass spec to track proteome-wide changes in protein stability resulting from drug treatment. Unlike other approaches – including Cellzome's own chemoproteomics techniques – the thermal profiling method can be applied to live cells, allowing researchers to look not only at the effect of treatment on drug targets but at changes in downstream signaling, as well, said Gerard Drewes, Cellzome's head of science and senior author on the study.

In addition, Drewes told ProteoMonitor, the technique allows researchers to better assess a drug's level of target engagement, a key parameter in pharma research that he noted, "is very difficult to measure by any other means."

While the technique is still under development, Drewes said it currently is in use at GSK across a variety of drug development programs.

Underlying the technique is the fact that when bound to a ligand like a drug, proteins tend to have higher thermal stability. Using this information, Drewes and his colleagues devised a process wherein they established thermal profiles for more than 7,000 human proteins using mass spectrometry. They were then able to treat cells with drugs of interest and, by looking for proteins whose thermal profiles shifted in response to treatment, identify the proteins binding to the drug.

Key to development of the method was the multiplexing capabilities of the TMT isobaric tagging reagents (sold by Thermo Fisher Scientific under a license from Proteome Sciences). The recent release of a 10-plex version of these reagents meant that the Cellzome researchers could profile proteomes of interest at 10 different temperatures, allowing them to get enough data points to put together a high-quality denaturation curve.

"In order to get a proper melting curve for each protein you really need these 10 temperatures, so that is why TMT was ideal," Drewes said. "Some years ago when we only had four-plex we couldn't have done it."

Upon denaturation, proteins clump together, which means they are not introduced into the mass spec and therefore are not detected. Using a Thermo Fisher Q Exactive instrument, the researchers built their thermal profiles by tracking the decrease in the amount of a given protein detected – which corresponds to the increase in its denaturation – across the series of 10 temperature points.

With these profiles in hand, they then investigated in K562 cell extract targets of the kinase inhibitors staurosporine and GSK3182571, identifying 51 kinases that showed significant shifts in their thermal profiles.

Among these proteins was the heme biosynthesis enzyme ferrochelatase (FECH), which, the authors noted, was not a known target. The observed interaction suggested that the photosensitivity induced by several kinase inhibitors might be "mediated by FECH," they said, citing this finding as an example of the method's potential for detecting unexpected off-targets.

Perhaps most interesting, though, is the approach's usefulness in determining the percentage of targets occupied by a drug at a given dosage, Drewes said. To do this the researchers flipped the experiment around so that rather than measuring protein denaturation at 10 different temperatures, they kept the temperature constant and measured denaturation at different drug dosages.

"Say your target is melting at around 55 degrees and your compound shifts it to 57 degrees because it gets more stable," Drewes said. "Then we would do an experiment at 56 degrees and run a series of [drug] concentrations, and basically the concentration that shows the half-maximal shift we infer is also the concentration that induces half-maximal occupancy of the target. So by doing the dose response experiment we can actually measure target engagement."

Termed an isothermal dose response experiment, this capability is "a particular strength of [the thermal profiling] method," Drewes said, noting that he and his colleagues are currently applying it to measuring target engagement for various projects within GSK.

The company is also applying the thermal profiling method to the study of compounds identified as interesting through phenotypic screens. Here, he said, the approach is useful in that it allows researchers to quickly identify likely targeted proteins and pathways without having to develop and validate the reagents required for chemoproteomic methods – which typically use labeled versions of compounds of interest to pull down targets for identification.

"It's better than [chemoproteomics] in that you don't need to make a labeled version of your compound," Drewes said. "It is sometimes challenging to get this labeled compound analog and to get the same activity and do all the controls to show that the tagged compound is really behaving like the actual drug."

Another advantage of the technique, Drewes said, is that, unlike other methods, it works not only in cell lysate but in living cells, as well. This, he noted, allows it to identify the downstream effects of compounds, as post-translational modifications caused by signaling changes due to drug treatment also affect proteins' thermal profiles.

This makes the technique easily adaptable to animal studies, he added. "You can think about dosing an animal, taking cells from an organ, heating it up, and doing the profiling."

The method does have several limitations at present, Drewes noted, citing mass spec sensitivity and sample complexity as one main challenge.

"In order to really identify targets in a sensitive way, you have to temperature profile thousands of proteins, and the current limit of our analytical technologies is somewhere between 5,000 and 10,000 proteins, so there is part of the proteome that will be hard to see," he said.

Additionally, the method is not yet optimized for membrane proteins, which are an important class of drug target. Drewes said the researchers are working on this issue, though, and that he expects they will ultimately manage to apply it to these proteins.

However, some proteins – especially very large proteins – do not show thermal shifts upon ligand-binding, and these, he said, "will probably remain sort of the blind spot of this method."