In search of a target: PD-1 and PD-L1 profiling across glioma types

Immune checkpoint inhibitors have recently entered the field of oncology, with several solid and hematological malignancies showing clinically meaningful, sometimes lasting, responses in some patient populations.¹ Monoclonal antibodies targeting programmed death 1 (PD-1) and programmed death ligand 1 (PD-L1) and thereby counteracting tumor-associated immunosuppression have been or will soon be approved in melanoma, non-small-cell lung cancer, renal cell cancer, and other cancers. Several lines of reasoning support the expansion of the clinical development program on PD-1/PD-L1 inhibitors to gliomas: the clear unmet clinical need and the paucity of effective treatment options in this patient population, the long-standing recognition that gliomas have strong immunosuppressive properties, several reports of target molecule expression in various brain tumors, documented activity of immune checkpoint inhibitors in preclinical glioma models, and reports of therapeutic successes in brain metastases and some gliomas.^2,3 Indeed, several clinical trials, including large international phase III studies in newly diagnosed and recurrent glioblastoma are ongoing and will soon report efficacy data. However, while expression of PD-L1 (on tumor cells) and PD-1 (on tumor-infiltrating lymphocytes, TIL) in glioblastoma has been documented by several groups, the expression of these molecules in other glioma types and their relation to relevant genetic and epigenetic alterations was unclear.^4,5 In this issue of Neuro-Oncology, Garber et al addressed this open issue and correlated PD-L1 and PD-1 expression with tumor type and molecular data in a large cohort of 410 glioma specimens.⁶ Although the cohort was composed mainly of glioblastomas and comprised relatively few cases of other glioma types, several interesting observations were made and warrant closer examination and discussion.

Garber et al found differences in the frequency of PD-L1 expression and density of PD-1+ TILs in dependence of tumor types and grades. Glioblastoma specimens presented with a higher frequency of PD-1+ TILs compared with lower-grade gliomas, while tumoral PD-L1 expression did not differ significantly between tumor grades. Among anaplastic tumors (grade III according to World Health Organization 2007 classification), PD-L1 expression was more frequently observed in tumors of astrocytic lineage compared with tumors with an oligodendroglial component. Of note, gliosarcoma seemed to be enriched with cases showing PD-L1–positive tumor cells and infiltration by PD-1–positive TILs. Thus, there seemed to be an association of PD-1/PD-L1 expression with histological phenotype, although it must be acknowledged that the case numbers in most categories were rather small and limited the statistical power of the analyses. Still, differences in PD-1/PD-L1 expression patterns between diffuse glioma types might be of relevance for treatments targeting these molecules. Therefore, further studies investigating a larger number of non-glioblastoma diffuse gliomas would be of interest to substantiate the findings by Garber et al. Importantly, such studies should apply the recently updated World Health Organization classification of CNS tumors, which introduced significant changes in the classification of diffuse gliomas by defining integrated histological and molecular diagnoses.⁷ In the present study, however, neither PD-1 nor PD-L1 expression correlated with the presence of aberrations in isocitrate dehydrogenase 1, phosphatase and tensin homolog, tumor protein 52, BRAF, or epidermal growth factor receptor. There was a negative correlation of O⁶-DNA methylguanine-methyltransferase promoter hypermethylation with the presence of PD-1–positive TILs. Mutational load and neoantigen signatures were not analyzed by Garber et al, but have been proposed as promising biomarkers of response to immunotherapies of several cancer types and may also be of relevance in gliomas.^8–10

Therapies commonly prescribed in glioma patients such as chemotherapy, radiotherapy, anti-angiogenic agents, and corticosteroids may have profound effects on the composition of the tumor inflammatory microenvironment. Unfortunately, data on therapies applied to the patients in the series by Garber et al prior to tissue sampling are not reported, and consequently no conclusions on a potential influence of therapeutic interventions on the observed expression patterns of PD-1 and PD-L1 can be made. A previous study did not find substantial differences in PD-L1 expression between newly diagnosed and matched recurrent glioblastoma specimens in a small series of 18 cases.⁴ However, the longitudinal evolution of immune-related factors and the influence of certain therapies will need to be addressed in further investigations.

The optimal method for assessment of PD-L1 expression is an ongoing controversy across tumor types. There is a large variability of immunostaining protocols and cutoff definitions among studies investigating PD-L1 expression in various tumor types, and no consensus guidelines are available. Garber et al used the commercially available SP142 antibody and reported membranous PD-L1 expression in >5% of tumor cells in 7.8% of glioblastoma specimens. This rate is well below the previously reported frequency of membranous PD-L1 positivity at a 5% cutoff in glioblastoma of 38% by Nduom et al and 37.6% by Berghoff et al.^4,5 These discordant results might be explained, at least to some extent, by the use of different staining protocols, as different PD-L1 antibody clones were utilized in the 3 studies (Nduom et al: Abcam clone EPR1161(2); Berghoff et al: clone 5H1; Garber et al: clone SP142).^4,5,11 Garber et al correctly state that much effort should go into optimization and harmonization of PD-L1 immunostaining protocols to achieve better comparability of expression data between studies. Importantly, the biological and potential clinical significance of various reported expression patterns of PD-L1, including cell membrane-bound, cytoplasmic, and diffuse/fibrillary expression on tumor cells and expression on various immune cell subtypes need to be better understood.^11–13

Overall, the study by Garber et al provides interesting data pointing toward potential differences in immune checkpoint molecule expression patterns across diffuse glioma subtypes and will certainly motivate follow-up investigations on the biological and clinical implications of these findings. The main open question at this point remains whether expression of PD-1 or PD-L1 has a predictive role for response to specific monoclonal antibodies in gliomas. This issue will hopefully be addressed in translational research projects accompanying the ongoing clinical trials exposing glioblastoma patients to PD-1/PD-L1 inhibitors.² Should these studies show that indeed cases with PD-1 or PD-L1 expression derive an increased benefit from these therapeutics, clinical trials enrolling also other PD-1 or PD-L1–positive glioma types should be designed.