Genetic Variability Affects How Tumors Respond to Immunotherapy New

From eye color to whether cilantro tastes like soap, genetic variability influences many factors in a person’s life. These inherited differences can also determine a person’s risk of developing a particular type of cancer and possibly even the response to certain kinds of cancer therapies. For example, while immune checkpoint inhibitors represent a huge advance in cancer treatment, they only succeed in a subset of patients.¹ However, the precise role of genetics is difficult to study because a person’s immune system and the therapy itself also affect the tumor.

In physics, a dilemma called the three-body problem makes it nearly impossible to calculate the exact trajectories of three interacting objects. “We have almost the equivalent of the three-body problem with immune checkpoint inhibitors in cancer,” said Edison Liua cancer biologist at The Jackson Laboratory.

To study the role of genetic variability in a tumor’s response to immunotherapy, Liu and his team used a mouse model called the Collaborative Cross, which was designed to be genetically diverse while maintaining statistical reproducibility.² Using these animals, the researchers showed gene-specific effects on the outcome of a specific tumor treated with an immune checkpoint inhibitor. The findings, published in Cell Reportscan help determine which patients are most likely to benefit from this type of therapy.³

In humans, genome-wide association studies can help scientists find regions of the genome that might be involved in cancer risk, but the genetic variability in humans makes conducting these studies difficult and can mask small effects. “They’re very expensive (and) take all kinds of people and time,” said Daniel Skellya statistical geneticist at The Jackson Laboratory and study coauthor.

For this reason, scientists use animal models, often mice, to explore possible interactions. However, these come with other caveats. “One of the challenges with mouse models has been that, usually, the mice are all genetically identical to each other, so you can study cancer, but you don’t really get much of a feel for how variation across individuals comes into play with those types of models,” said Hannah Cartera computational biologist who studies how genetic variation influences diseases like cancer at the University of California, San Diego. Carter was not involved with the study.

This led Liu, Skelly, and their colleagues to take advantage of the genetically variable but consistent Collaborative Cross mice. “The power of this genetic resource, the Collaborative Cross, and in derivatives that we formed, is in part that we were able to do this on a much more feasible scale,” Skelly said.

The researchers mated each of their Collaborative Cross strains with a standard mouse strain that had the same genetic background as the tumors that they then implanted into the mice. They assessed whether genetic variation influenced how the immune checkpoint inhibitor, anti-programmed-death 1 receptor (anti-PD-1), successfully reduced the growth of an implanted colon cancer tumor. They saw a range of responses to this therapy across their mouse strains, pointing to differences in genes having an effect.

Comparing the differences in genes between responsive and nonresponsive mice across 32 Collaborative Cross strains, the team identified one locus on chromosome 15 that strongly correlated with a positive anti-PD-1 response. However, they saw weak associations in other chromosomes, so they crossed a responsive with a nonresponsive strain to increase the genetic variability in their population, which would help them identify more involved loci. In these offspring, they saw two new loci on two other chromosomes emerge, while the effect from the locus on chromosome 15 reduced. “Leveraging the diversity to try to narrow down what parts of the genome were involved in anti-PD-1 response, I thought, was a very, very cool approach,” Carter said.

Next, Liu and his team searched for differentially expressed genes within these loci between their top three responsive and nonresponsive animals. In terms of genes involved in immune function, they saw the greatest effect in pathways related to antigen processing and presentation as well as those related to allograft rejection, graft versus host disease, and viral infection and autoimmune diseases. This reflected increased anti-tumor activity in response to anti-PD-1 therapy and cytotoxic T lymphocyte responses.

To study differences in cell composition of the tumor-immune microenvironment between responsive and nonresponsive animals, the team performed single-cell RNA sequencing on these tumors 48 hours after anti-PD-1 treatment. They observed two distinct populations of cytotoxic lymphocytes in their responder mice. One of these exhibited exhaustion markers, but the second displayed genes involved in cytotoxicity and expressed the highest amount of interferon γ. They further saw that a cluster of macrophages with increased expression of the ligand for PD-1, PD-L1, and a protein complex involved in antigen presentation co-localized more with these cytotoxic lymphocytes. As responsive mice contained more of these interactions than nonresponsive animals, it pointed to a significant role of interferon γ-responsive macrophages and cytotoxic lymphocytes in the tumor response.

The researchers then investigated whether their mouse model reflected observations of human cancer therapy responses. Previously, one study identified a population of macrophages, defined by their expression of the genes Cxcl9 and SPP1associated with positive anti-PD-1 responses and survival.⁴ The team observed this same ratio of these genes elevated in the tumors of responsive mice and saw that treatment with anti-PD-1 further increased this ratio.

Finally, to explore how genes and their products affected anti-PD-1 responses, the team blocked two factors whose genes are on chromosome 15 in one responsive mouse strain: the receptor for granulocyte-monocyte colony stimulating factor (GM-CSF) and the receptor for the cytokine interleukin-2 (IL-2Rβ). In addition to their location on chromosome 15, the team also identified these as differentially expressed genes. GM-CSF blockade followed by anti-PD-1 therapy reduced the ratio of Cxcl9/SPP1 to levels comparable with animals treated with a control therapy, pointing to an overall reduced tumor response. Meanwhile, blocking IL-2Rβ further decreased the expression of this population to levels seen in nonresponsive animals.

While Carter pointed out that future work will need to address how well these findings translate into humans, she said that the study provided a good proof of concept for using the Collaborative Cross. “This model can point us to some really interesting things that we can start to look at in humans, where things are often messier and noisy, and if you don’t know where to look, sometimes you can miss important signals.”

Skelly and Liu agreed that this study is only the beginning of potential avenues for this experimental model. “We want to explode this into the universe of combinatorics that’s there,” Liu said.