Immunotherapies, such as immune checkpoint blockade, have revolutionized cancer treatment.
In particular, the development of strategies targeting the CTLA-4 and the PD-1 pathways
have resulted in significant improvements in OS (OS) in a variety of challenging cancers,
including melanoma, non-small cell lung cancer, head and neck cancers and others.
Immune checkpoints are regulatory inhibitory pathways involved in the maintenance
of immunologic homeostasis by modulating intensity and duration of immune responses.
Among other functions, such pathways prevent auto-immunity, maintain self-tolerance
in physiologic conditions, and protect tissues from damage during infection[1]. Cancer often highjacks such pathways for evading immune system surveillance, a
vulnerability that has been therapeutically exploited through the development of immune
checkpoint inhibitors (ICPi). A plethora of drugs has been successfully tested and
approved by the FDA in multiple indications. The use of drugs such as ipilimumab,
nivolumab, pembrolizumab, atezolizumab, durvalumab, and other anti-PD-1 and PD-L1
inhibitors have since become widespread in oncology. However, certain cancers, particularly
those with low mutational burden, have been found to be refractory to ICPi.
Glioblastoma and other gliomas are highly attractive targets for immune checkpoint
blockade[2]. Glioblastoma is the most common and aggressive form of primary brain tumor, and
associated with dismal outcomes[3]. Treatment remains restricted to radiotherapy and alkylating agents, which seem
to mostly benefit patients with tumors harboring methylated MGMT promoter, present
in about a third of patients. Glioblastomas notoriously promote immunosuppression
and may evade the immune system through multiple mechanisms. Implicated systemic factors
include decreased T cell responsiveness, increase in Tregs, decreased monocyte and
dendritic cell function, lower levels of immunoglobulins, frequent use of corticosteroids,
and lymphopenia from treatments. Those add to unique local immunosuppressive factors
such as down regulation of MHC molecules, secretion of inhibitory cytokines such as
TGF-Beta, VEGF, PG-E2, IL-10, LLT-1, polarization of microglia and tumor associated
macrophages towards the immunosuppressive M2 phenotype[4], decreased T cell function due to hypoxia, T cell apoptosis through Fas, and, importantly,
infiltration with Tregs and increased expression of immune checkpoints[5].
In 2017, we reported the first prospective clinical trial of ICPi in glioblastoma,
which focused on anti-CTLA-4 and anti-PD-1 monoclonal antibodies[6]. That study was part of clinical trial Checkmate 143, which consisted of multiple
cohorts across different lines of treatment for this disease, and sponsored by Bristol
Myers Squib. In this phase 1 portion of the study, anti-PD-1 monoclonal antibody nivolumab
was given with or without the anti-CTLA-4 monoclonal antibody ipilimumab to patients
with recurrent disease. That study showed that the toxicity profile in this population
was consistent with other cancers, and no new safety signals were identified. Importantly,
there was no evidence of clinically significant neurotoxicity, which was a concern
given the anatomical location in the brain. As expected, the combination of ipilimumab
and nivolumab was more toxic than nivolumab alone, resulting in more frequent and
more severe immune-related adverse events. PD-L-1 expression was high in archived
tumor specimens from those patients (68%), and some signals of activity were observed,
including a few radiographic responses and some patients displaying increased immune
cells infiltrates on tissue biopsy[6].
Although phase 2 data was not available, several large trials of nivolumab were promptly
launched in this disease, encouraged at the time by the excellent track record in
other diseases with high frequency of PD-L1 expression, and the desperate need to
rapidly develop novel therapies for glioblastoma. Because of a more favorable toxicity
profile, and emerging evidence of efficacy as a single agent in other tumor types,
nivolumab was initially selected for further development; accrual to these trials
was fast, highlighting the interest from patients and physicians on this approach.
The first randomized study was the phase 3 portion of Checkmate 143, which randomized
recurrent glioblastoma patients to receive nivolumab or anti-VEGF therapy bevacizumab,
a standard therapy for this disease setting[7]. Among the 369 randomized patients, the median OS was 9.8 months for the nivolumab
arm and 10 months for bevacizumab (HR, 1.04; P = .76). Of note, there was no evidence
of improved efficacy in PD-L1 expressing tumors. Interestingly, nivolumab-treated
patients with MGMT-methylated tumors who had not received baseline corticosteroids
seemed to experience prolonged survival in post-hoc analyses, although the study was
underpowered to properly investigate that question.
In the newly diagnosed setting, the first published study focused on the exploratory
cohorts of CheckMate 143, which enrolled 136 patients[8]. In that study, the various cohorts investigated nivolumab added to radiotherapy
and temozolomide. Given the lack of efficacy of temozolomide in this phenotype, nivolumab
and radiotherapy were given without temozolomide in various cohorts of MGMT unmethylated
patients. Overall, results demonstrated the feasibility of adding nivolumab to radiotherapy
with or without temozolomide, with more toxicities observed when both temozolomide
and nivolumab were given. The cohorts where temozolomide was omitted had lower incidence
of lymphopenia, and survival results clearly varied according to MGMT methylation
status. The median OS with nivolumab+RT+TMZ was 33 months in patients with methylated
MGMT promoter, although that cohort was small (N=15). Across the different cohorts
for MGMT unmethylated patients, median OS varied from 16.5 and 14.8 months for nivolumab+RT+TMZ
and 14.4 and 14 months for nivolumab+RT.
While results of Checkmate 143 were maturing, two randomized clinical trials were
launched. Checkmate 498 focused on unmethylated MGMT newly diagnosed glioblastoma,
and tested nivolumab as a potential replacement for temozolomide in this chemoresistant
population (Omuro et al.[8]). Unfortunately, results showed that the addition of nivolumab to radiotherapy was
actually inferior to chemotherapy with temozolomide. Among the 560 randomized patients,
the median OS (mOS) was 13.4 months with NIVO+RT vs 14.9 months with TMZ+RT (HR, 1.31
[95% CI, 1.09-1.58]; P=.0037). Results seemed to reflect that temozolomide may have some efficacy in MGMT
unmethylated patients as defined by the standard definitions and cutoffs for MGMT
methylation. However, a nivolumab-related phenomenon of “hyperprogression” leading
to inferior results cannot be entirely excluded, as has been suggested in other tumor
types. Finally, Checkmate 548 was a randomized phase 3 study testing the addition
of nivolumab to standard chemoradiotherapy in MGMT methylated glioblastoma, and similarly
showed no improvements in OS (Weller et al, manuscript submitted). Of note, across
all nivolumab studies, there was no correlation between PD-L1 expression and efficacy.
Trials of the combination of nivolumab and ipilimumab are still ongoing.
Pembrolizumab is another anti-PD-1 antibody that has been tested in glioblastoma.
In a phase 1 study, patients with glioblastoma expressing PD-L1 (>=1%) were eligible[9]. A total of 62% of screened patients displayed PD-L1 expressing tumors and a total
of 26 patients were eventually enrolled. Again, only modest efficacy was observed,
with ORR of 8%, with two partial responses observed. The 6-month progression-free
survival was 37.7% and median OS was 13.1 months. In a randomized phase 2 study, pembrolizumab
was given with and without bevacizumab, and achieved a median OS of 8.8 months (combination
arm) and 10.3 months (pembrolizumab alone); authors concluded that both treatments
were ineffective[10]. Of note, the unfavorable results of the combination arm damped the initial enthusiasm
on the combination of ICPi with bevacizumab, which could potentially afford less corticosteroids
usage and synergistic effects, but also decrease lymphocyte trafficking and cytokine
release. Another phase 1 study[11] investigated pembrolizumab with hypofractionated stereotactic re-irradiation and
bevacizumab for recurrent glioblastoma and anaplastic astrocytoma, and found the combination
to be feasible, with a median PFS of 8 months and OS of 13.5 months in bevacizumab
naïve patients, which is difficult to interpret given the heterogeneity of the population
enrolled.
More interestingly, two small studies evaluated pembrolizumab in recurrent glioblastoma
given as a “neoadjuvant therapy” prior to surgery, which allowed for analysis of post-treatment
tumor specimens[12],[13]. In both studies, an increase in T cell infiltration and antigen-reactive clonal
expansion was observed in the post-treatment tumor microenvironment. One of those
studies[12] addressed the potential clinical benefit of the “neoadjuvant” usage, which has been
suggested to provide an optimal setting for ICPi based on a priming effect that is
magnified following tumor removal, as observed in other tumor types. In that study,
35 patients with recurrent, surgically resectable glioblastoma were randomized to
receive neoadjuvant pembrolizumab, with continued adjuvant therapy following surgery
(neoadjuvant arm), versus pembrolizumab given only after surgery without a neoadjuvant
dose (adjuvant arm). The median OS was 13.7 m in the neoadjuvant arm, vs 7.5 m in
the adjuvant arm (p=0.04). Moreover, neoadjuvant PD-1 blockade was associated with
upregulation of T cell- and interferon-γ-related gene expression, downregulation of
cell-cycle-related genes, focal induction of PD-L1 in the tumor microenvironment,
enhanced clonal expansion of T cells, decreased PD-1 expression on peripheral blood
T cells and decreasing monocytic population- all signs indicating improved immunological
response. Another single arm trial of neoadjuvant pembrolizumab[14] found an interesting OS of 20 months, although analysis of tumor tissue found a
paucity of immune activation markers, and an abundance of CD68+ macrophages. Overall,
the small sample sizes of studies to date preclude definitive conclusions on the clinical
benefits of neoadjuvant ICPi, and further studies are needed to investigate this intriguing
concept.
Additional trials have investigated anti-PD-L1 antibody atezolizumab. In a small phase
1 trial, that drug was found to be safe, but efficacy was difficult to discern given
the heavily pre-treated population enrolled[15]. Other trials are awaiting publication of results.
Finally, some attempts have been made to identify optimal candidates for anti-PD-1
therapy. An exploratory retrospective study[16] found enrichment of PTEN mutations associated with immunosuppressive expression
signatures in tumors non-responsive to pembrolizumab or nivolumab, and an enrichment
of MAPK pathway alterations (PTPN11, BRAF) in responders. Responsive tumors were also
associated with branched patterns of evolution from the elimination of neoepitopes
as well as with differences in T cell clonal diversity and tumor microenvironment
profiles. However, these intriguing results are limited by the retrospective nature,
the uncertain definition of responders vs non responders, and lack of adequate controls.
In summary, experience with ICPi to date has overall yielded disappointing results.
However, there is evidence of activity in select patients, in the form of radiographic
responses, increased signs of immune response ([Figure 1]), and occasional patients achieving longer survival, perhaps with an advantage for
the use in the neoadjuvant setting. Because immunotherapy is based on more stable
targets than anti-tumor targeted therapies, this approach remains of high interest
in Neuro-Oncology. Future directions include ongoing research to identify best candidates,
while improving our mechanistic understanding of response and resistance to ICPi.
At this time, it is unclear how much of the lack of activity derives from unique mechanisms
pertaining to tumor growing in an immune privileged site such as the CNS, versus intrinsic
mechanisms of resistance potentially shared with other solid tumors that are unresponsive
to ICPi. Additional challenges posed by the location in the brain include the blood-brain
barrier, which may limit leukocyte trafficking, as well as the unique anatomical and
molecular specialization that allows only T cells that have been activated in the
periphery to enter the CNS. T cells that populate the CNS have characteristics of
tissue-resident memory T cells and are enriched for viral specificities; only limited
antigen presentation in therefore possible in CNS tumors[5]. Indeed, the defective lymphatic drainage within the CNS has been found to be a
crucial aspect of the poor T cell activation observed in glioblastoma models[17].
Figure 1 H/E staining of an MGMT unmethylated glioblastoma resected from a patient after exposure
to anti-PD-1 therapy, showing intense inflammatory reaction and lymphocyte infiltration.
Courtesy of Dr Marc Rosenblum.
As for future directions, ongoing or planned trials are now seeking to investigate
whether patients with increased tumor mutational burden and enhanced epitope landscapes[18], such as gliomas with acquired or primary mismatch repair deficiency, could constitute
better candidates for these therapies. Other trials are focusing on targeting new,
alternative ICPi such as anti-TIGIT, LAG-3, CD137, TIM3 and other antibodies, both
as single agents and in combinations. Other combinations being explored include the
addition of alternative immune modulators and microenvironment modifiers, various
types of vaccines, oncolytic viruses, epigenetic modifiers, chimeric antigen receptor
T cells (CAR-T cells), as well as alternative treatment modalities such as hypofractionated
radiotherapy and laser interstitial thermal therapy. However, success of such approaches
remains contingent upon a better understanding and targeting of the mechanisms of
T cell dysfunction and glioblastoma-associated immunosuppression, and of the unique
barriers to successful immune responses posed by the CNS environment.
