NEW YORK – A team led by researchers at the University of California, San Francisco and the University of California, San Diego has developed protein-protein interaction maps in several cancer types that could help identify new drug targets.
Described in three studies published this week in Science, the work demonstrates how characterizing changes in protein-protein interactions across cancer cell types can provide new insights into the biology of the disease and improve interpretation of genomic-level data.
The researchers generated their protein interaction data using affinity purification mass spectrometry, a method that uses a set of tagged bait proteins expressed in cells of interest. After expression, these proteins are pulled out of the sample, and they, along with their protein interactors, are quantified using mass spec. Using this approach they analyzed the interactors of dozens of proteins known to be involved in breast cancer and head and neck cancer.
In the breast cancer study, the researchers mapped out interaction networks for 40 proteins commonly altered in breast cancer, with and without relevant mutations, identifying 589 protein-protein interactions. In their work on head and neck cancer, they used 31 protein baits from commonly mutated genes present in 99 percent of head and neck tumors to build a map detailing 771 protein interactions involving roughly 650 proteins.
In the third study, the researchers developed a statistical model using protein interaction data to better interpret how the many genetic mutations identified in genomic sequencing work converge at the level of protein networks.
They also developed a statistical method for calculating what they termed differential interaction scores to assess what cell types were enriched for what interactions.
Nevan Krogan, director of the Quantitative Biosciences Institute at the UCSF School of Pharmacy and one of the senior authors on the studies, highlighted the fact that only a small percentage of the mapped interactions had been observed in previous protein interaction experiments and that the different cells he and his colleagues analyzed varied greatly in terms of the protein interactions they exhibited.
"One of the surprises was to see how different the interactions are as you look at cancerous versus noncancerous cells, and it was very interesting that the majority of interactions we found had never been reported before," he said. "That illustrates the importance of generating these maps across multiple cell types."
At the same time, there were commonalities that indicated shared mechanisms across cancer types, he said. For instance, the researchers identified a pair of point mutations in the protein PIK3CA that strengthened the protein's interaction with HER3 in a way that made mouse models of the disease more responsive to treatment with HER3 inhibitors.
"Those two mutations are in 5 percent of all cancers," Krogan said. "So we think that this finding, if you can inhibit HER3 in these cases, could be very important."
"There are differences across the cells, but there are some themes that I think will have powerful implications across many cancers," he said.
Krogan said that protein interaction mapping across cell types and across different protein variant forms could help collapse the complexity of genomic data where researchers have struggled to tease out the implications of the many mutations discovered through next-generation sequencing.
"When people started sequencing tumors … they quickly realized that the same type of tumors in different individuals might have a couple of [mutated] genes overlapping, but that then there is this long tail of heterogeneity," he said, adding that this genomic heterogeneity becomes easier to interpret in the context of protein interactions.
"Say, for example, you have a five-protein complex, and you have five different individuals with mutations in five different components [of that complex]," Krogan said. "If you just looked at the genes, you might think, oh, there's no overlap. But if you knew that the [mutated] proteins were all in the complex [together] you would say, oh, it's the same mutation, because it has the same effect."
"We need the genomic data, but it's an entry point," he said. "We have to go to the proteins, which are the functional units of the cell, and we need to get to these [interaction] maps."