Skip to main content
Premium Trial:

Request an Annual Quote

Researchers Use PDX Models to Profile Proteome of Tumor Microenvironment


NEW YORK (Genomeweb) – A team led by researchers at Washington University in St. Louis and New York University has generated proteomic profiles of the tumor microenvironment of a series of patient-derived xenograft (PDX) breast cancer models.

Described in a study published this week in Science Signaling, the profiles provide one of the most comprehensive looks to date at how tumor cells alter the proteomes of surrounding stromal tissue and could ultimately aid in development of new cancer therapies, said Jason Held, an assistant professor at WUSTL and senior author on the paper.

While the tumor microenvironment has long been an area of research interest, there have been relatively few large-scale proteomic efforts to study this tissue. Held said this was due to a variety of reasons, foremost among them the difficulty of distinguishing between stromal and tumor tissue and the challenge of gathering enough samples to enable a useful analysis.

"Say you have a patient tumor — it's very difficult to discern what is tumor and what is stroma, especially [in the case of] the invading stroma," he said.

And, given that each combination of patient tumor and microenvironment is unique, "it is difficult to say if [an observation] is something that occurs consistently or repeatedly or if it is just kind of a fluke in that one particular sample," he added.

Held and his colleagues tackled these issues by using PDX models in which patient breast tumors were implanted in mice. Because the tumor tissue was human and the stroma was mouse, they were able to distinguish between the two by identifying measured peptides as being either mouse or human.

Additionally, because they were able to implant different tumors into multiple mice, they were able to replicate their studies to determine whether the phenomena they observed were reproduced in multiple subjects, Held said.

The approach was not without its challenges, though, noted David Fenyo, professor of biochemistry and molecular pharmacology at NYU and co-author on the study. PDX models "are often very expensive," he said, which has tended to limit the number of replicates researchers use. Additionally, the need to distinguish between mouse and human peptides in a reproducible way while maintaining depth of coverage makes for a more computationally complicated experiment than a straightforward single-species proteomic analysis.

Because many mouse and human peptides have shared sequences, it is in many cases not possible to determine which organism a peptide came from. According to Held, the researchers tossed out around 50 percent of the peptides they identified because they were shared between the two species, which translated into a loss of around 30 percent of their protein identifications.

This, he said, meant that they had "to dig pretty deep to get enough identifications to start to do pathway analysis."

It also raised the question of whether these shared peptides might be biased in any way because, for instance, they are more evolutionarily conserved. Held said that he and his colleagues are currently looking at that question but that, thus far, they hadn't detected any such issue.

"We don't see any classes of proteins yet that appear particularly problematic," he said. "For instance, we were particularly interested in what happened to kinases, and we don't see any bias towards kinases being thrown out because they are maybe more conserved or more interesting."

The researchers looked at seven breast cancer PDXs representing basal, HER2-enriched, luminal, and claudin-low subtypes. They implanted each patient tumor into three mice, giving them three biological replicates for each. After letting the tumors develop in the mice, they dissected them and used mass spec with 10-plexTMT isobaric labeling to look at the stromal and tumor proteomes.

In total, they identified and quantified 4,784 human (tumor) proteins and 1,721 mouse (stroma) proteins.

One of the most striking findings was the breadth of the stroma protein changes they observed, Held said. "We found that 35 percent of the mouse proteins we detected were consistently altered by one of the tumors, which is far greater than we anticipated."

Also notable was how tightly linked the stroma proteome changes were to the type of the implanted tumor. Each of the seven tumors brought about a distinct set of changes in the proteomes of the surrounding stroma. At the same time, these tumor-specific changes were highly reproducible across the three biological replicates the researchers looked at.

"It was highly individualized [by tumor type], but each tumor was very consistent when you implanted it in a different mouse," Held said. "And it wasn't just that we saw proteins consistently changing, they were really organized along different pathways."

Among the proteins the researchers found to be enriched in certain stromal subtypes were molecules linked to the structuring of the extracellular matrix; to cancer-associated fibroblasts; to the complement system, which the authors noted "is known to facilitate tumor survival;" and to myeloid-derived suppressor cells, which act as immunosuppressors.

They also found that stromal proteome changes were associated with tumor stage, noting, for instance, that members of two stromal clusters exhibited higher protein quantities in stage I tumors.

The researchers also used data from the National Cancer Institute's Cancer Genome Atlas project to determine if the stromal changes they observed in their PDX models were consistent with the stromal changes in the tumor microenvironment of the primary tumor. While they were not able to distinguish between tumor and stromal tissue in the TCGA data, they were able to approach this question by looking at particular proteins known to be primarily expressed in stroma and not typically found in the tumor itself. They found high correlation for these proteins between their data and mRNA and proteomic data generated by TCGA, suggesting that the PDX models and primary tumors influenced their microenvironments in similar ways.

Beyond providing basic insights into the interaction between tumors and surrounding stroma, the research points towards potential therapeutic strategies, Held said.

"The implication is that these [stroma protein] changes are probably very important for tumor growth," he said. "These tumors are kind of finding their way around, and they have to highjack the stroma in order to make it suitable for growth. People aren't thinking too much right now about how to drug the stroma, but here we are providing molecular information that could be potentially useful for [such an effort.]"

He said that he and his colleagues have begun exploring this idea.

"We are thinking about ways to start to drug the stroma and using this information to assess which stromal proteins may be important for tumor growth and devising strategies to maybe co-target tumors and the stroma," he said.