Predicting If an Immune Checkpoint Drug Will Work


Drugs that activate the immune system to attack cancer in a process known as immune checkpoint blockade (ICB) are a focus of intense investigation. A number of them are already approved by the U.S. Food and Drug Administration (FDA) for various cancers; namely, the anti-CTLA4 antibody ipilimumab (Yervoy), two anti-PD-1 antibodies: pembrolizumab (Keytruda) and nivolumab (Opdivo), and three anti-PD-L1 drugs: atezolizumab (Tecentriq), avelumab (Bavencio) and durvalumab (Imfinzi). These ICB drugs have the potential to induce durable cancer regressions, but the majority of cancer patients just do not respond to them at all.

Biomarkers, signature molecules in the blood or other tissue, can sometimes be used to predict a patient’s response to a given treatment. But no reliable biomarkers exist for ICB, and this is a serious concern. Patients who may really benefit from ICB could be overlooked, and patients who are not likely to respond may receive useless (and very expensive) ICB treatment.

Most potential response predictors that have already been identified are not yet useful for one or all of the following reasons: they are not extensively validated, their significance is still uncertain and may differ from one cancer (or even one patient) to another, or they are technically challenging for routine use. These markers are addressed below.

Predictive markers of ICB efficacy within tumors

Tumor-infiltrating T cells: This is probably the oldest treatment response predictor, but it seems to have fallen out of fashion even though it is time-tested as a prognostic marker. The presence of T cells (a type of white blood cell) in a tumor indicates that the immune system has been alerted but is unable to attack the tumor because of the many mechanisms used by the tumor to inactivate the immune response. Anti-PD-1/PD-L1 checkpoint inhibitor drugs are capable of increasing the activity of T cells that are already recruited to tumors.

PD-L1: This protein is sometimes found on tumor cells and is the second-oldest predictive marker of response to drugs that target PD-1 or PD-L1. Its presence signals a higher probability of response to ICB and serves as an FDA-approved companion diagnostic test for pembrolizumab in some indications for this drug. However, the predictive power of PD-L1 is limited even in cancers for which it is an FDA-approved companion diagnostic, and is not obvious in some other cancers. Still, pembrolizumab is FDA-approved for only highly PD-L1-positive cancers in some instances (first-line NSCLC, third-line gastric cancer). The power of PD-L1 testing is obviously limited.

Tumor mutational burden (TMB): TMB was recognized as a strong predictive marker of response to ICB three years ago in a seminal study of NSCLC patients. TMB is a measure of how many mutations are present overall in the DNA of a tumor. The presence of many random mutations shows that many mutant proteins are made in the tumor, and these could be potentially recognized by the immune system. For example, smoking causes gradual accumulation of numerous random mutations in lung cells and is associated with high TMB and a better response to ICB. Melanomas with high levels of mutations induced by sun exposure also respond better to ICB than other cancers do. Several companies offer testing for TMB, most notably Foundation Medicine, and development of new tests will hopefully lead to wider use.

Microsatellite instability (MSI): An inherited condition known as Lynch syndrome causes MSI, in which a certain process to repair DNA damage (mismatch repair) is disabled due to a deficiency in one of the enzymes involved. Lynch carriers have a greater chance of developing colorectal (and some other) cancers. About 5% of all colorectal cancers have high MSI and respond fairly well to ICB, whereas the common, non-Lynch-associated colon cancers are essentially non-responsive. Other types of cancer may have sporadic (not hereditary) MSI, and both anti-PD-1 drugs (pembrolizumab and nivolumab) were FDA-approved in 2017 for any type of MSI-high cancer.

Other tumor features associated with response to ICB: Several publications have reported other features of genomic changes in tumors that may be predictive of response to ICB. A common type of kidney cancer is associated with a fairly high response rate to ICB even though PD-L1 is rarely present, TMB is typically low, and MSI is not observed. It turns out that a certain type of DNA mutation that involves small insertion or deletions is common in kidney cancer. The presence of these particular alterations is associated with a response to ICB.

Mutations in DNA damage repair genes are also associated with better response to ICB, presumably by allowing accumulation of mutations cause by inefficient DNA damage repair. On the other hand, large losses of genomic material in melanoma are associated with poor response to ICB.

Most interestingly, recent reports suggest that alterations in single genes may be predictive as well; for example, mutations in the PBRM1 gene are predictive of treatment response in kidney cancer, and in other cancers as well. Mutations in the gene SMARCA4 in a rare type of ovarian cancer (hypercalcimic) dramatically enhance the immune environment and are key in excellent responses to ICB.

Systemic indicators of response to ICB

The characteristics of cancer itself are obviously critical in how the immune system recognizes tumors. However, this is only part of the larger picture. Attempts to reactivate dormant T cells may be futile in patients whose immune systems are ill equipped to handle cytotoxic (cell-killing) responses to tumors.

HLA diversity: One very important finding emerged recently, and it concerns a group of proteins known as human leukocyte antigen (HLA) type I proteins. These normal molecules are absolutely required to present mutant or “foreign” proteins to T cells—an important step in the immune system’s attack of tumors. Humans possess a great variety of HLA subtypes, and some people have low diversity of HLA proteins while others have many different variants. Patients whose HLA subtypes are less varied have a worse response to ICB. A greater variety of HLA allows for presentation of more mutant proteins to T cells, and plays a very important role in the efficacy of ICB. HLA typing is a fairly routine procedure widely used in matching donors and recipients of tissue and bone marrow transplants for a variety of diseases.

Microbiome: Late last year, highly publicized studies reported an association between the composition of bacteria living in the gut (the microbiome) and the efficacy of immune drugs in melanoma and lung cancer. We wrote about this fairly recently. The diversity of a patient’s microbiome and the presence of certain bacterial species strongly influence response to ICB.

Other blood cells: The composition of the immune system in responders versus non-responders before they receive ICB has been studied extensively. A recent study showed that melanoma patients with high levels of monocytes (a common type of myeloid cell in the blood) have a better response to anti-PD-1 drugs. A skewed ratio between monocytes and other cells is assumed to be due to the movement of T cells from the blood to tumors even before treatment. Measuring this ratio via a blood test would be a different way to evaluate the number of T cells found within or very near tumors, which is the oldest predictive marker mentioned at the beginning of this post.

The importance of being able to predict, with a reasonable degree of confidence, a potential benefit of ICB is obvious. This will most likely involve integrating a number of different tests rather than looking just at one marker, be it PD-L1, TMB, or HLA subtype.