Circulating PD-L1 levels change during bevacizumab-based treatment in recurrent glioma

Despite optimal surgical and adjuvant treatment, diffuse gliomas have a recurrence rate of up to 90% due to their infiltrative growth pattern [2]. However, the range of available treatment options at recurrence is very limited, underlining the urgent need for new therapeutic approaches. While immune-modulating therapies have revolutionized oncology, a clinically relevant efficacy in primary brain tumors such as glioma has not been observed so far. The CheckMate-143 trial comparing nivolumab with bevacizumab in recurrent glioma was unable to show a superiority of immune checkpoint inhibition [3]. However, a small subset of patients (objective response rate 7.8%) showed durable responses, underscoring the need for a more profound understanding of inflammatory subgroups in glioma potentially profiting from immune-modulating therapies [4].

Soluble programmed death receptor ligand 1 (sPD-L1) has been shown to correlate with prognosis and the response toward immune-modulating agents in a wide array of solid tumors [5‐9]. Most probably, sPD-L1 is generated by proteolytic cleavage of membrane-bound PD-L1 on both tumor and immune cells. Cleavage may be generated by metalloproteinases such as ADAM10 or ADAM17 [10] whose expression has been described to correlate with poor prognosis in glioma [11]. Previously, we showed that sPD-L1 levels and detectability differ in different brain tumor entities and observed a correlation with survival in lower-grade glioma (LGG) and glioblastoma (GBM) [12]. Specifically, patients with GBM had longer overall survival (OS) in the presence of sPD-1 as compared to their counterparts (median OS 20.9 vs. 8.4 months, p = 0.006). In contrast, patients with LGG presented with worse OS when sPD-L1 could be detected (median OS 38.9 vs. 89.6 months, p = 0.028). These results suggest that the immune phenotype of LGG and GBM might differ, probably due to the immunosuppressive role of the metabolite 2-hydroxyglutarate (2-HG) in isocitrate dehydrogenase (IDH)-mutated tumors [13, 14]. Bevacizumab treatment is frequently used as a salvage treatment in patients with symptomatic glioma progression in need for steroid treatment [15]. Importantly, vascular endothelial growth factor signaling was previously shown to impact the efficacy of immune-modulating therapies in extracranial malignancies [16, 17]. Therefore, we aimed in the present study to investigate the longitudinal sPD-L1 concentrations as a systemic inflammatory marker in patients with recurrent LGG and GBM treated with bevacizumab-based therapy.

In the present study, we analyzed the time course of sPD-L1 as a surrogate marker for systemic inflammation in patients with recurrent glioma receiving bevacizumab-based treatment. As shown previously [12], sPD-L1 is a detectable marker in brain tumors and correlates with local and systemic inflammatory parameters. Here, we found a significant decrease in sPD-L1 levels upon initiation of bevacizumab-based salvage treatment in WHO grade II–III glioma. This change was accompanied by alterations of other established markers of systemic inflammation such as leukocyte counts and CRP levels.

Timely changes of sPD-L1 and other inflammatory parameters in bevacizumab-treated patients with glioma suggest that tumor–immune system interactions are observable on a systemic level. Although gliomas rarely show extracranial metastases and present with a growth pattern restricted to the CNS, several studies previously suggested that systemic inflammation is altered by glioma. Erythrocyte sedimentation rate, CRP levels and the neutrophil-to-lymphocyte ratio were shown to correlate with survival prognosis [20, 21]. Further, an association of systemic inflammation markers such as leukocyte, neutrophil and platelet counts as well as CRP levels with markers of the local tumor microenvironment and overall survival was reported [22]. Indeed, the variability in the systemic inflammation could be an explanation for the so far lacking clinical efficacy of immuno-modulatory therapies in glioma. Systemic inflammation as measured by CRP, neutrophil-to-lymphocyte ratio or recently also sPD-L1 was shown to correlate with the likelihood of response to immune checkpoint inhibitors in extracranial malignancies [23]. Profiling of systemic tumor–immune system interactions could therefore allow for a rational selection of patients who potentially benefit from immunotherapy. Notably, although the CheckMate-143 trial failed to meet its primary endpoint, few durable responses were observed in the nivolumab arm, suggesting that a small subgroup of patients with GBM benefits from immunotherapy [3].

There was higher inter-patient variability of sPD-L1 concentrations in WHO grade II–III glioma than in GBM as determined by the CV. Although not statistically significant, we could detect tendentially higher intra-patient variability in IDH-mt glioma than in IDH-wt tumors. WHO grade II–III gliomas more frequently display IDH mutations leading to an abundant production of the oncometabolite 2-HG exerting local immune-modulating effects. Here, we provide further data on distinct glioma–immune system interactions on systemic level, adding further to the evidence that tumor–immune system interactions differ between glioma subtypes.

In the present study, only patients receiving bevacizumab-based treatment were included. It is well known that therapies targeting the vascular endothelial growth factor (VEGF) axis interact with local and systemic inflammation. Specifically, VEGF inhibition regulates the infiltration of immune cells to the tumor microenvironment by normalizing the tumor vasculature and altering the endothelial expression of cell adhesion molecules [24]. Furthermore, VEGF along with other immunosuppressive factors exerts systemic immunosuppressive effects hampering effective antitumor immune responses [25]. Recently, the results of a phase II trial of pembrolizumab with bevacizumab in recurrent glioma were presented [26]. However, combination therapy was not superior in terms of efficacy as compared to historical bevacizumab monotherapy controls. With regard to our data, it remains elusive whether the observed effects on systemic inflammation are bevacizumab-related or rather due to glioma progression, as bevacizumab exerts only minor antitumoral efficacy in glioma and is mainly used for symptomatic control without significant benefit in terms of overall survival [27].

Clearly, our study has several limitations which are mainly due to its exploratory, retrospective design and the small patient number resulting in limited statistical power to detect significant alterations especially in subgroup analyses. The variability in number of measurements per patient with as few as two measurements available in some patients impeded statistical testing for time-dependent effects. Moreover, although we applied the same sandwich ELISA as previously published [12], the method has not been systemically validated with other assays. Lastly, with only 6 IDH-mutant glioma cases, further analyses with respect to molecular subtypes could not be performed.

In conclusion, our data show that tumor–immune system interactions are observable on a systemic level in patients with recurrent glioma. sPD-L1 concentrations change over time in a cohort treated with bevacizumab-based salvage treatment and correlations with other markers of systemic inflammation could be detected, although the mechanistic foundations remain to be elucidated. Our data underscore the need for comprehensive panels of immune-related biomarkers for the conception of prospective immunotherapy trials in CNS tumors. Furthermore, our data add to the available evidence that tumor–immune system interactions differ between glioma subtypes which should be considered in future studies on immune-modulating agents and combinations thereof in glioma.