Immunotherapies, such as anti-PD-1 and anti-PD-L1 antibodies, offer promising new treatment options for various malignancies including non-small cell lung cancer (NSCLC). With the advent of these immunotherapies, much interest and energy has been focused on developing a companion predictive biomarker.
Immunotherapies, such as anti-PD-1 and anti-PD-L1 antibodies, offer promising new treatment options for various malignancies, including non-small cell lung cancer (NSCLC). With the advent of these immunotherapies, much interest and energy has been focused on developing a companion predictive biomarker.
Programmed death-ligand 1 (PD-L1) has gained traction as a potential predictive biomarker despite limited data for its use in this regard. While the prospect of a companion predictive biomarker for anti-PD-1 and anti-PD-L1 therapies is both useful and exciting, many questions regarding its potential role remain unanswered.
These questions include, but are certainly not limited to the following:
Currently, only one commercially available PD-L1 antibody (clone E1L3N) has been validated using IHC in formalin-fixed paraffin-embedded (FFPE) tissue. While this antibody has been validated, the utilization of this antibody for predicting response to anti-PD-1 or anti-PD-L1 therapies remains unknown. Clone 5H1, another PD-L1 antibody, has been validated and was used in the initial phase I trials of BMS-936558, but is not commercially available. Interestingly, clone SP142, which has been used to measure stromal PD-L1 expression in clinical trials with anti-PD-L1 (MPDL3280a), recently became commercially available for clinical research purposes. The validation data for this particular antibody, however, is also currently not known. All other clinical trials with anti-PD-1 and anti-PD-L1 therapies are using proprietary antibodies for PD-L1 measurement, meaning validation data for these particular antibodies is also unknown. Which of these antibodies should be used for PD-L1 protein measurement as a potential predictive biomarker therefore remains an open question.
Besides which antibody should be utilized for PD-L1 protein expression, the cutoff for positivity/ negativity is confusing at best. Multiple clinical trials define PD-L1 positivity as 5% expression in the tumor epithelium, but this is arbitrary and has demonstrated mixed results. Still, other studies have defined PD-L1 positivity ranging from 1% to 50% expression in the tumor epithelium, while others have stratified PD-L1 expression in the tumor epithelium into low (1% to 5%), moderate (5% to 10%), or high (at least 10%). These definitions have at times resulted in correlative trends of higher PD-L1 protein expression and overall response rate (ORR), though many PD-L1-positive tumors lack response while some PD-L1-negative tumors respond.
The MPDL3280a trial focused on PD-L1 expression in the stroma (as opposed to tumor epithelium), but again with uncertain results. It remains unclear if PD-L1 expression in the tumor epithelium, stroma, both tumor epithelium and stroma, or another yet unidentified factor, better predicts response to treatment.
Instead of using binary cutoffs for determining positivity/negativity, some researchers have investigated quantitative measurements of PD-L1 expression. Quantitative measurement has proven difficult due to the apparent heterogeneity of PD-L1 expression, but whether a more quantifiable assay is a better method of predicting response to anti-PD-1/anti-PD-L1 therapies remains to be seen. Additionally, other modalities for measuring PD-L1 expression, such as RNA and CyTOF, have been used to determine PD-L1 expression. PD-L1 RNA expression, while not correlating with protein expression, has been validated and correlates with overall survival in NSCLC. Whether PD-L1 RNA expression in tumor samples predicts response to anti-PD-1/anti-PD-L1 therapies is unknown. CyTOF (and other similar strategies) are in development, but distort tissue architecture thereby eliminating compartmentalization of PD-L1 expression within the sample. Quantitative measurement, RNA, CyTOF, and various other methods for PD-L1 measurement not discussed here, should be investigated further in developing a potential companion diagnostic marker.
Finally, other factors in addition to PD-L1 expression almost certainly play a role in predicting response to the aforementioned immunotherapies. TILs have been associated with PD-L1 expression and improved overall survival, but less is known about their role in predicting response to anti-PD-1/anti-PD-L1 therapy. Hypothetically, PD-L1/TIL-positive tumors would be more likely to respond, while PD-L1/TIL-negative tumors would be less likely to respond--though this relationship to therapy response has not been fully elucidated. The utility of measuring other inhibitory components of the PD-1/PD-L1 axis such as PD-1 and PD-L2 is poorly understood. Complicating the picture, immune stimulatory molecules including OX40 are another part of the paradigm for which significant knowledge is lacking. It is clear that much more information must be gathered in the PD-1/PD-L1 axis, but also in TILs and other inhibitory/stimulatory pathways, to fully understand responses, primary resistance, and acquired resistance to immune therapies.
In conclusion, a multitude of questions remain unanswered pertaining to companion predictive biomarkers to anti-PD-1/anti-PD-L1 therapies. This article merely scratches the surface of these questions, but in doing so, hopes to draw attention to the much needed research required in this evolving subset of medical oncology. It may be that multiple companion predictive biomarkers (an immune panel of sorts) measuring components in the PD-1/PD-L1 axis, TILs, and various stimulatory molecules, will be required to predict response to immune therapies. While the idea of a companion predictive biomarker for immune therapies remains compelling, one must recognize the inherent difficulties in attempting to identify such a marker. Inevitably, more information must be gathered for this purpose, which will require ongoing collaborative efforts between researchers, clinicians, academic centers, and industry.