NEW YORK – Researchers have identified a gene expression signature of the tumor microenvironment in patients with acute myeloid leukemia (AML) that may predict both resistance to chemotherapy and response to immunotherapy.
"This will definitely help us tailor treatment options in AML," said Sergio Rutella of Nottingham Trent University in the UK, who led the research identifying the predictive signature and is an author of the publication detailing its development in Science Translational Medicine. "The ambition here would be to have an algorithm to stratify patients [because] you may want to target these different subgroups of patients with different options."
The signature has specifically been shown to predict patients' response to flotetuzumab, an investigational agent developed by Rockville, Maryland-based biotechnology firm MacroGenics that recognizes both CD3 and CD123 in order to redirect patients' T lymphocytes to attack CD123-expressing cancer cells. Flotetuzumab recently completed a Phase I/II dose-escalation study in patients with chemotherapy refractory or early-relapse AML, and is expected to advance to a registration study in the coming year for the treatment of patients with refractory disease who have not responded to multiple prior lines of chemotherapy and other agents.
Although the signature has the potential to guide treatment decisions for patients with AML, MacroGenics CEO Scott Koenig said that it needs to be developed further. As such, MacroGenics does not have plans at this point to use it as a companion diagnostic to advance flotetuzumab through the FDA approval process.
"In his analysis, Rutella came up with a number of genes that fit this [predictive] profile," Koenig said. "But what would be even better would be to try to reduce this to a smaller number of genes to get greater precision … and it's going to take time to do that."
Instead of hitting pause on the flotetuzumab development program until the signature is ready to be used as a companion diagnostic, Koenig said that MacroGenics plans to continue to work with Rutella and colleagues to improve the signature while simultaneously moving the therapy through clinical trials.
"I'm a little concerned that if we step aside and try to get [a diagnostic] approved first, it slows down the process of getting a drug out there that will benefit these patients," he added.
Identifying the signature
To identify the predictive signature, Rutella and colleagues began with the hypothesis that patients with higher levels of T cell infiltration would experience worse clinical outcomes, as the tumor infiltration would lead to increased production of interferon gamma, and, in turn, the upregulation of negative immune checkpoints.
To test this hypothesis, the researchers analyzed a total of 442 archival bone marrow samples from three independent cohorts of patients with newly diagnosed AML. Initially, they performed targeted gene expression profiling on the samples using NanoString Technologies' nCounter PanCancer Immune Profiling Panel, which measures messenger RNA of 770 genes representing 14 immune cell types. They then performed gene set enrichment analysis to identify the genes associated with interferon-gamma response and inflammatory response.
The researchers were able to stratify patients into immune-infiltrated and immune-depleted AML subtypes. Interferon-related genes were differentially expressed within the subtypes; patients with immune-infiltrated AML tended to have higher levels of interferon-related gene expression, while the opposite was true of patients with immune-depleted AML.
The researchers then correlated gene expression scores with clinical outcomes among patients in the three cohorts, including overall survival, relapse-free survival, and chemotherapy resistance, and found a significant link when it came to chemo resistance. They defined chemotherapy resistance as patients' failure to achieve complete response or as relapsing within three months of achieving complete response. Among patients resistant to chemotherapy, the majority were those with the interferon-high gene expression scores.
"We found a nice correlation between increased interferon gamma signaling and resistance to chemotherapy," Rutella said. He highlighted that the current "gold standard" score for predicting chemotherapy response in AML is the ELN score, which has around a 0.65 predictive capability as measured statistically by the AUROC curve (for which a score of 1.0 would denote perfect predictability and 0.5 would denote no predictive ability). In comparison, the gene expression signature that Rutella and colleagues identified had a predictive capability of 0.85, a significant improvement on the ELN score.
The researchers further validated their gene expression signature's ability to predict chemotherapy resistance in silico using publicly available transcriptomic data in AML cohorts, including from The Cancer Genome Atlas, the Beat AML Master Trial, and the Dutch HOVON Data Center. During this process, they also pinpointed driver gene mutations — including TP53 and RUNX1 mutations — that were particularly enriched in patients classified as immune infiltrated.
Predicting flotetuzumab response
After identifying and validating the gene expression signature and its correlation with chemotherapy resistance, Rutella and his colleagues moved into the translational portion of their study, during which they sought to test another hypothesis: whether this signature that predicts chemotherapy resistance, could, inversely, predict for response to immunotherapy agents.
To evaluate this predictive capability, the researchers used patient data from the Phase I/II clinical trial of MacroGenics' flotetuzumab. They first used NanoString's PanCancer IO360 gene expression assay to profile bone marrow samples that had been collected from 30 patients prior to treatment with flotetuzumab. They then correlated the gene expression profiles of these samples with patients' response to flotetuzumab.
Of the samples taken from patients who went on to demonstrate evidence of flotetuzumab antileukemic activity on the clinical trial, 92 percent had immune-infiltrated tumor microenvironments, suggesting that the higher tumor infiltration score could be used to determine patients who are more likely to respond to flotetuzumab.
As MacroGenics goes on to evaluate flotetuzumab for efficacy in a pivotal registration trial, Rutella and colleagues will continue their translational research in a larger cohort of 200 patients with relapsed, refractory AML to refine the signature and further validate its predictive capability.
"This is going to be a unique opportunity to extend our findings to a larger cohort of patients receiving flotetuzumab," Rutella said. "And the long-term goal of all of this is … to stratify patients upfront [to] spare unnecessary chemotherapy to patients who are less likely to benefit."
Additionally, Rutella and Koenig both noted the need to investigate whether this signature has the same predictive capabilities in pediatric AML patients as it does in adult patients.
"It's interesting to note that these interferon-gamma gene signatures are higher as people age," Rutella said, explaining that pediatric patients with AML have been historically less likely to respond to immunotherapy than adults, possibly because of the lower levels of interferon gamma signaling seen in children. Through further research, Rutella said he hopes to address the question of whether there would be a correlation between this tumor inflammation signature and responses in children.
"We may need different predictors in children [or] age-adjusted RNA profiles to predict responses," Rutella said.