CHICAGO – An immune classifier comprising 53 genes may help physicians guide therapy for patients with triple-negative and hormone receptor (HR)-positive HER2-negative breast cancer by balancing the benefits of immunotherapy with the risks.
In a presentation at the American Society of Clinical Oncology (ASCO) annual meeting on Saturday, Denise Wolf, a researcher at the University of California, San Francisco, reported on the analysis of immune markers in five immune-oncology therapy arms in the I-SPY2 trial. In the study, a classifier developed by Wolf and colleagues, dubbed ImPrint, predicted response and showed potential to help doctors balance the benefits and risks of serious immune-related adverse events and decide whether to prioritize immunotherapy over other treatments for patients.
Neoadjuvant immunotherapy with chemotherapy is the standard treatment for early-stage triple-negative breast cancer, but not all patients respond to immuno-oncology drugs, and these agents carry risks for toxicities, some causing permanent damage. Those immune-related toxicities include thyroid toxicity, adrenal insufficiency, pneumonitis, colitis, hepatitis, and inflammation of the pituitary gland. Meanwhile, for HR-positive breast cancer, there is currently no immunotherapy approved.
In previous trials, biomarkers such as PD-L1 expression, tumor-infiltrating lymphocytes, tumor mutational burden, and various gene expression signatures have been associated with pathologic complete response to standard treatments. However, no studies have differentiated between the biomarkers that predict response to just chemotherapy versus those associated with response to chemotherapy plus immunotherapy. In this new I-SPY2 analysis, researchers identified biomarkers of response when immunotherapy is added to chemotherapy that, when confirmed with correlative studies, may fill that need.
In 2022, Wolf and colleagues published a study in Cancer Cell describing the development of additional breast cancer subtypes using biological features beyond HR and HER2 status that might better predict drug response. They found that the combinations of biomarkers that performed best included immune, DNA repair, and HER2 luminal phenotypes. This work formed the basis of researchers' current efforts to evaluate ImPrint within I-SPY2.
In I-SPY2, an adaptive platform trial launched in 2010, researchers aim to rapidly evaluate and select potential therapeutics added to chemotherapy for Phase III studies in the neoadjuvant setting for patients with high-risk stage II or III breast cancer. The primary endpoint of interest is pathologic complete response, defined as no residual invasive cancer in the breast or lymph nodes.
Patients enrolled in I-SPY2 are stratified by HR and HER2 status and tested on Agendia's 70-gene expression MammaPrint test to determine if they're at high- or ultra-high risk for recurrence. Therapy combinations that have more than an 85 percent predictive probability of success in a subsequent Phase III trial in the most responsive patient subset "graduate" the trial. Overall, I-SPY2 has tested 22 therapies in more than 2,500 patients.
In the five arms of I-SPY2 included in the current analysis, investigators tested anti-PD-1 and anti-PD-L1 therapies combined with chemotherapy, as well as an anti-PD-L1 agent with a PARP inhibitor; an anti-PD-1 agent with a TLR9 inhibitor; and an anti-PD-1 agent with a LAG3 inhibitor in triple-negative and HR-positive, HER2-negative breast cancer patients. Wolf's group evaluated 32 continuous qualifying gene expression biomarkers, including seven immune checkpoint target genes, 14 immune-cell subpopulation signatures, three T-cell/B-cell prognostic signatures, five tumor immune signaling signatures, TGF-β, an estrogen/progesterone receptor signature, and a proliferation signature.
The biomarkers were derived from pretreatment biopsies collected from the 342 patients who participated in these treatment arms, and researchers looked for associations with pathologic complete response. According to Wolf, the analysis showed clear differences between trial arms and subtypes, with more predictive signals in the HR-positive group than the triple-negative group.
"Interestingly, all of the predictive signals in the triple-negative group are positive, meaning that high [biomarker] levels associate with [pathologic complete response], whereas in the ER-positive group, we see some negative signals such as mast cells and ESR1, where high levels associate with resistance," Wolf said. The tumor immune signatures prominently featuring chemokines and cytokines were most consistently associated with pathologic complete response across immuno-oncology arms and across receptor status.
Using a multiplexed immunofluorescence assay, the researchers used spatial metrics to locate cell types in the tumor and characterize their spatial relationships. Wolf said the assay showed that tumor immune signatures most predictive of response across subtypes and arms were highly correlated with more complex immunofluorescence spatial proximity measures, revealing high spatial colocalization of PD-1-positive immune cells and PD-L1-positive tumor cells, particularly in the triple-negative group.
Wolf and colleagues used the data to develop a research-grade immune predictor signature which they integrated with other biomarkers to identify response-predictive subtypes for breast cancer that, if used to guide treatment strategies, would result in higher levels of response and better outcomes. They then partnered with Agendia to develop that signature into ImPrint, a clinical grade immuno-oncology response predictor, which garnered an investigational device exemption (IDE) by the US Food and Drug Administration, allowing for its use in human trials.
In 200 patients with HR-positive disease across the five treatment arms in I-SPY2, researchers found that 29 percent were ImPrint positive. The rate of pathologic complete response in those patients was 76 percent, compared to 16 percent among ImPrint-negative patients. In 142 patients with triple-negative disease, 51 percent were ImPrint positive and their pathologic complete response rate was 75 percent, versus 37 percent in the ImPrint-negative group.
The highest pathologic complete response rate, 93 percent, was among ImPrint-positive HR-positive patients treated with an anti-PD-L1 agent and a PARP inhibitor. In the triple-negative subgroup, the highest pathologic complete response rate for ImPrint-positive patients, 86 percent, was observed on just an anti-PD-1 agent. Overall, combining HR-positive and triple-negative patients on all therapy combinations, 38 percent were ImPrint positive, and their pathologic complete response rate was 75 percent.
Wolf concluded that tumor immune signatures dominated by chemokines and cytokines can predict response to immuno-oncology agents in both triple-negative and HR-positive HER2-negative patients, reflecting co-localization of PD-L1-positive tumor cells and PD-1-positive immune cells. This study shows, she added, that the ImPrint classifier appears to predict response to a variety of immuno-oncology regimens and identifies a previously uncharacterized subset of HR-positive, HER2-negative patients with a very high likelihood of responding to immunotherapy. "ImPrint may inform prioritization of [immune-oncology therapy] versus other treatments to best balance likely benefit versus the risk of serious immune-related adverse events," said Wolf.
Commenting on the analysis, Elizabeth Mittendorf, a Dana-Farber Cancer Institute oncologist, said, "Clearly we are in need of a biomarker to identify patients that are going to benefit from the addition of [immuno-oncology therapy] to chemotherapy, and I will tell you, it's not going to be PD-L1 expression."
Speculating on whether ImPrint is "ready for prime time," Mittendorf said, "I would suggest if prime time means it's ready for the identification of patients for trials or stratification in a trial, then yes." And now that ImPrint has received an IDE from the FDA, it is being used to randomize patients in the next iteration of I-SPY2. "But if prime time [means it] is for clinical care, then the answer is, 'Not yet.'"
Mittendorf opined that Wolf and her collaborators would need to refine ImPrint to better differentiate response to immunotherapy added to chemotherapy versus response to chemotherapy alone. "I would also ask the authors, if they have an opportunity, to develop a signature that will predict toxicity," she said.