The search for biomarkers of response to immunotherapy is high on the list of cancer-related research efforts. A new classifier in triple-negative breast cancer is showing promise and is currently being validated in the I-SPY2.2 trial, according to Laura van ’t Veer, PhD, Professor of Laboratory Medicine and Program Leader of the University of California, San Francisco, Helen Diller Family Comprehensive Cancer Center Breast Oncology Program. Dr. van ’t Veer described the development and testing of ImPrintTN at the 2024 European Breast Cancer Conference.1
Laura van ’t Veer, PhD
“The ImPrintTN single-sample classifier for triple-negative breast cancer predicts response as well as nonresponse to a variety of immunotherapy regimens tested in the I-SPY2 program. Within the ImPrintTN-negative subset, rates of pathologic complete response are very low and are similar to regimens not containing immunotherapy,” she said. “Our current data suggest that ImPrintTN may help inform the prioritization of immunotherapy vs other treatments for patients with triple-negative disease, to best balance the likely benefit vs risk of serious immune-related adverse events.”
Need for the Test
The I-SPY2 trial is evaluating a number of inhibitors of PD-1 and PD-L1 in an adaptively randomized neoadjuvant trial of stage II or III locally advanced breast cancer. The program aims to match drugs with subtypes of cancer most likely to respond to them. Agents and combinations of agents “graduate” for efficacy if they are shown to have more than an 85% predictive probability of success in a subsequent phase III trial, based on pathologic complete response.
I-SPY2 has tested 23 agents in 2,118 patients; the immunotherapy arms have generally showed a two- to threefold improved efficacy relative to control therapy in triple-negative tumors. “We know from the first 10 arms of the study that if you have a pathologic complete response, this associates with a 5-year 95% event-free or distant recurrence–free survival, so our aim is to get every patient there,” said Dr. van ’t Veer. But it is also a fact that these drugs can trigger immune-related toxicities, some of which are permanent. “There is, therefore, a high need to understand who really benefits and who is most at risk for the side effects, so we can have better clinical benefit…. We need clinical-grade predictive biomarkers to support immunotherapy allocation decisions,” she added.
To this end, Dr. van ’t Veer and her team developed a large biomarker database that includes transcriptional and protein expression data for more than 900 tumors from the different arms of I-SPY2. From these data, they created predictive signatures for the different treatments, using them to evaluate 990 tumor samples from 10 arms of I-SPY2, stratified by estrogen receptor and HER2 status.2 They came up with five “response predictive subtypes,” defined by a combination of gene-expression signatures, information on intrinsic subtypes, and the MammaPrint gene-expression test.
Dr. van ’t Veer described how pathologic complete response rates can be improved when treatment is selected based on these subtypes. Although the observed rate with standard chemotherapy was 20% to 25%, this increased to 35% to 40% with the addition of a targeted agent for the optimal receptor subtype, and it increased much further—to 60% to 70%—when treatment paired the targeted agent with the appropriate response predictive subtype.
Immunotherapy Response Predictor
Dr. van ’t Veer and her team have developed a clinically applicable immune classifier called ImPrint that can predict response to immunotherapy in triple-negative and hormone receptor–positive tumors. Assessing the performance of ImPrint in the immunotherapy arms, they concluded it performed well in hormone receptor–positive, HER2-negative disease: pathologic complete response rate was 75% in ImPrint-positive patients compared with 17% in ImPrint-negative patients with this subtype. In triple-negative disease, this first version of the classifier was still predictive, “but we noticed it could be further improved by reducing the false-negative rate…. So we went back to the drawing table,” said Dr. van ’t Veer. In triple-negative disease, pathologic complete response rates were 75% in ImPrint-positive patients, but also 34% in ImPrint-negative patients, which was considered too high.
The investigators subsequently refined the ImPrint signature for triple-negative disease using a machine-learning cross-validation approach based on genes from the original ImPrint classifier plus a refined selection of immune-signaling and checkpoint markers. They tested the newer ImPrintTN classifier on 150 I-SPY2 patients from five immunotherapy arms (pembrolizumab, durvalumab, and cemiplimab-rwlc with or without various targeted agents) and 128 patients in the control arms. Using pretreatment mRNA data, they ultimately classified patients as ImPrintTN-positive, ie, likely sensitive to immunotherapy (66% of patients), or ImPrintTN-negative, ie, likely resistant (34% of patients). “Two-thirds of triple-negative patients have a disease biology that we predict will respond to an immunotherapy drug,” she noted.
In an independent test set, pathologic complete response rates were 71% in ImPrintTN-positive patients and 22% in ImPrintTN-negative patients—an improvement of 49% with the use of the refined classifier. Similar performance was observed in the aggregate of all five of the I-SPY2 immunotherapy arms, where the rate was 74% vs 16%, respectively—a 58% improvement. For the 128 patients in the control arms of these studies, pathologic complete response rates were 30% in ImPrint-positive patients and 15% in ImPrint-negative patients. “Standard chemotherapy is more effective in ImPrint-positive biology tumors, which makes sense, because we know chemotherapy triggers the immune system,” she said.
The test is now being prospectively validated in I-SPY2.2.
Expert Point of View
Marleen Kok, MD, PhD, medical oncologist and group leader of the Netherlands Cancer Institute, Amsterdam, praised the I-SPY trial program for endeavoring to identify which patients will benefit from which immunotherapy drugs. “I-SPY is a really important trial, to recruit so many patients and test so many different drugs with a very smart strategy,” Dr. Kok said. The current effort, reported by Dr van ’t Veer, to further refine the prediction of response, is much needed, she added.
Marleen Kok, MD, PhD
Using the ImPrintTN classifier, the investigators determined that patients labeled ImPrintTN-positive are highly likely to benefit from immunotherapy, whereas those who are ImPrintTN-negative have less chance of response. “We can conclude that not only is their chemotherapy benefit going to be less, but the addition of immunotherapy will have less effect,” she said. “I think this is bad news for the patients with the highest clinical need, because for those who respond poorly to chemotherapy, we really need something better to give.”
Dr. Kok noted that low levels of tumor-infiltrating lymphocytes and PD-L1 have similarly been associated with a poor response to immunotherapy. The optimal treatment for these patients is not clear. Research should focus on improving outcomes for patients with any of these factors that portend lack of benefit to immunotherapy and a poor outcome, she said. She estimated that thousands of papers have been published on potential biomarkers for chemotherapy efficacy, “but none have made it to the clinic,” and Dr. Kok expressed hope that I-SPY and other such initiatives will be more successful.
DISCLOSURE: Dr. van ’t Veer is cofounder of, part-time employee for, and holds stock in Agendia NV; and is a consultant and stockholder for Exai Bio. Dr. Kok reported financial relationships with Natera, Alderaan Biotechnology, AstraZeneca, Bristol Myers Squibb, Domain Therapeutics, MSD, Roche, and Gilead Sciences.
REFERENCES
1. Wolf DM, et al: 2024 European Breast Cancer Conference. Abstract 2LBA. Presented March 20, 2024.
2. Wolf DM, et al: Cancer Cell 40:609-623, 2022.
