NEW YORK – New research published last week by scientists from Germany's Hummingbird Diagnostics and collaborators suggests that blood-borne microRNAs (miRNAs) may enable better personalization of cancer immunotherapy.
In the study, appearing in the Nature partner journal Precision Oncology, the investigators used whole-blood miRNA profiling to develop a five-miRNA risk score that is predictive of overall survival in stage IV non-small cell lung cancer patients following immunotherapy — outperforming PD-L1 expression, which is currently the most established biomarker for guiding the use of anti-PD-1 agents.
Hummingbird now hopes to commercialize a test based on this biomarker signature to guide immunotherapy in lung and potentially other cancers.
Hummingbird was founded in 2016, with a focus on miRNA biomarker development for the early detection of lung cancer. More recently, the firm started a new program using the same discovery platform but aiming for a signature that could predict response to immunotherapy.
The company is currently small, with about 20 people located in Heidelberg, though it plans to set up a US lab to commercialize the miRNA immunotherapy predictor as a laboratory-developed test.
According to Bruno Steinkraus, Hummingbird's chief scientific officer, the firm's focus on microRNAs reflects the fact that these molecules are "very good monitors of what's going on in the body."
"They can be thought of as sort of molecular rheostats" akin to electrical variable resistors, controlling gene expression linked to everything from physiology to homeostasis to disease, Steinkraus said. "It's why they're often called master regulators," he added.
Being nucleic acid-based, they also lend themselves to high-throughput profiling, he added. They can also be studied without the degradation that other RNA molecules are known for. "Normally RNA gets very quickly degraded outside of cells. But these microRNAs are actually remarkably stable," Steinkraus said.
Steinkraus said that evidence has also emerged that miRNAs can reflect immune activity. Like many in the precision oncology sphere, the Hummingbird team became cognizant of the significant unmet need for biomarkers to inform more precise and personalized use of cancer immunotherapies.
Patient stratification for these often dramatically effective therapies is currently performed using measurements of tumor PD-L1 expression. However, the study authors estimated that PD-L1 is an accurate predictor of response in only about 30 percent of cases.
Other tools, like the assessment of microsatellite instability or tumor mutational burden have also emerged to identify patients likely to respond. Meanwhile, other gene-expression based techniques have emerged; investigators are exploring germline factors; and sequencing of the immune cell repertoire is uncovering yet more components related to drug response. Still, available methods remain imperfectly predictive, and leaders in the field increasingly argue that multi-modal strategies are likely to be necessary to best guide the use of these drugs and their combination with other treatments.
In their study, Steinkraus and colleagues analyzed three well-characterized cohorts consisting of a total of 334 stage IV NSCLC patients divided into a monotherapy test group treated with either Merck's Keytruda (pembrolizumab) or Bristol Myers Squibb's Opdivo (nivolumab), a complementary training group, and a third control group receiving chemotherapy in addition to immunotherapy.
Investigators performed whole-blood miRNA profiling, coupled with probe-based methods to correct for ubiquitous background molecules. "There is a lot of signal coming from red blood cells, and we have a way to work around that by blocking these highly abundant microRNAs," Steinkraus said.
Comparing outcomes in an initial training set, the team narrowed down to its five-miRNA signature predictive of better versus worse survival, which remained predictive in the test set, but was not predictive in the control chemotherapy-treated group. The authors added that they saw "significantly superior performance" of their blood-based "miRisk" score compared to tissue-based PD-L1 testing.
"What was quite surprising to us is that the five miRNAs in the signature actually come largely from myeloid cells," Steinkraus added. "When we look at the oncology space, everyone's all about T cells, so it was interesting that we saw our signal coming from myeloid cells with some lymphoid contribution."
The peripheral immune system has been relatively underexamined, he added. "It's been very tumor-centric with PD-L1 expression on tumor [cell] membrane and tumor mutational burden, but there hasn't been so much done in the periphery."
The myeloid component observed by the Hummingbird team "dovetails with where we think that some of the academic research is going," he said. "We're beginning to understand that myeloid cells are actually very important in the tumor microenvironment and can sort of be first responders that then configure this space locally and thereby influence how well the T-cell response will actually function."
Steinkraus said that the team is now working to collect functional data that further validates this potential mechanism.
Caveats include the study's small size and the fact that the current data haven't cemented that the miRNA signature is truly predictive of drug response and not just prognostic of overall outcome.
"It's not entirely 100 percent crystal clear from the paper that [our miRNA signature] is predictive rather than prognostic in general," Steinkraus said. "However, when we do an analysis just in the PD-L1 cases that have expression greater than 50 percent, we do see it becoming predictive … so that is the first use case we're putting forward at the moment — in PD-L1-high patients."
Under this plan, when implemented commercially, the firm's miRisk test would come into play after PD-L1 immunohistochemistry, when a clinician is trying to decide whether a patient should receive immune-monotherapy or immunotherapy combined with chemotherapy. "This is where we want to take this because it is currently a clinical dilemma that cannot really be addressed on PD-L1 alone," Steinkraus said.
This complementarity could certainly also function with other tools like tumor mutation burden, immune cell analyses, and other gene expression-based signatures.
"We don't think we'll replace the biopsy," Steinkraus said. "We consider this a sort of complementary puzzle piece to the question of who's responding."
In terms of next steps, the company is now preparing to start a follow-up trial, which will include additional endpoints apart from overall survival. Eventually, Steinkraus said, the company wants to pursue a dual route to the clinic, launching the test initially as an LDT through a planned US lab, while also building evidence necessary to seek US Food and Drug Administration approval.
Other practical next steps include studies of other cancer types, starting with bladder cancer, for which the company is beginning a new trial.
"In theory, since the microRNAs are not coming from the tumor and we think it's really a host signature … it should function in other tumor entities, as well, where immunotherapies play a role," Steinkraus said.