The Brain Tumor Immune Microenvironment: A Multidimensional Barrier
Glioblastoma (GBM) and other primary brain malignancies represent a uniquely hostile immunological landscape, combining cellular, cytokine, metabolic, and structural elements that collectively establish what is characterized as a profoundly "cold" tumor microenvironment (TME). Understanding these barriers in mechanistic detail is prerequisite to rational therapeutic design.
Cellular barriers are dominated by myeloid lineages rather than effector T cells. Tumor-associated macrophages (TAMs), microglia, and two functionally distinct populations of myeloid-derived suppressor cells (MDSCs)—early progenitor MDSCs (E-MDSCs) enriched in hypoxic pseudopalisading regions, and monocytic MDSCs (M-MDSCs)—can collectively account for up to 30% of total tumor mass 411. Single-cell RNA sequencing of 33 gliomas revealed that E-MDSCs co-localize with metabolic stem-like tumor cells in a reciprocal trophic relationship: glioma stem-like cells secrete chemokines to recruit E-MDSCs, which in turn produce growth factors sustaining tumor cells—a crosstalk absent in IDH-mutant gliomas 4. M-MDSCs are driven largely by the macrophage migration inhibitory factor (MIF)–CD74 signaling axis, and high co-expression of MIF and CD74 in human GBM specimens predicts worse survival (hazard ratio 2.45) 3. Regulatory T cells (Tregs) constitute an additional cellular suppressor layer; elevated intratumoral Treg signatures correlate with poorer prognosis, and these cells actively blunt CD8+ T-cell clonal expansion within the tumor parenchyma 8.
Cytokine-mediated suppression is orchestrated through overlapping networks. Transforming growth factor-β (TGF-β), IL-10, and vascular endothelial growth factor (VEGF) are constitutively upregulated in GBM and collectively suppress lymphocyte function while promoting immunosuppressive cell recruitment 2. Interleukin-6 (IL-6) has emerged as a central node: spatial and single-cell analyses of patient-matched GBM samples reveal that rare responders to checkpoint blockade harbor significantly lower pre-treatment IL-6 levels. In preclinical models, IL-6 blockade has been reported to transiently sensitize GBM to anti-PD-1 therapy by increasing MHC-II+ monocytes, CD103+ migratory dendritic cells, and effector CD8+ T cells, and may also enhance sensitivity to radiotherapy 7.
Metabolic barriers are orchestrated in large part through hypoxia-inducible factor-1α (HIF-1α), which drives angiogenesis, extracellular matrix remodeling, and checkpoint molecule upregulation, while simultaneously recruiting Tregs and MDSCs 5. MDSCs exploit arginine depletion via arginase-1, tryptophan catabolism via indoleamine 2,3-dioxygenase (IDO), and adenosine accumulation in hypoxic regions (signaling through adenosine A2A receptors) to blunt T-cell activation and persistence 6.
Structural barriers include the blood-brain barrier (BBB) and blood-tumor barrier (BTB), which restrict immune-cell trafficking and reduce penetration of systemically administered agents. Even in GBM, where angiogenic disruption creates heterogeneous BBB permeability, deeply infiltrative tumor margins often retain intact barriers. Chemokine–receptor mismatches between tumor-secreted chemokines and receptors on circulating T cells further compound trafficking deficits 19.
How the TME Limits CAR-T Cell Function
Across early phase I clinical trials of CAR-T therapy in GBM, encompassing 128 patients across 13 published studies, CAR-T cell kinetics have generally been transient and heterogeneous, with limited persistence reported in many studies 11. Early intratumoral CAR-T presence within the first three days post-infusion, rather than systemic persistence, is the strongest predictor of survival—reframing the problem from peripheral pharmacokinetics to TME penetration and early local activity 10.
Once within the tumor, CAR-T cells face adaptive resistance: their own cytotoxic activity releases interferon-γ (IFN-γ), which paradoxically upregulates PD-L1 on tumor cells and exhaustion markers (PD-1, LAG-3, TIM-3) on infiltrating T cells 13. In relapsed tumors, intratumoral T cells demonstrate an exhausted or anergic phenotype with high PD-1 expression and failure to recognize fresh tumor cells ex vivo 10. Concurrent CAR-T activity also triggers recruitment of IL-10-secreting, FoxP3+ Tregs and expands TAM/MDSC populations that directly inhibit T-cell activation and cytotoxicity through contact-dependent and soluble mechanisms 11.
Antigen heterogeneity compounds these immune evasion mechanisms. GBM exhibits profound spatial and temporal variation in target antigen expression; following CAR-T therapy targeting antigens such as EGFRvIII or IL13Rα2, recurrent tissue specimens in some studies have shown decreased target antigen expression, supporting immunoediting and antigen-loss escape as important resistance mechanisms 11.
Biological Rationale for CAR-T Plus Checkpoint Blockade
The layered immunosuppression of the brain TME provides compelling mechanistic rationale for combining CAR-T cells with immune checkpoint inhibitors (ICIs). First, blocking PD-1/PD-L1 signaling directly addresses the IFN-γ-induced adaptive resistance triggered by CAR-T activity, potentially sustaining effector function during and after the initial cytotoxic wave 13. Second, CTLA-4 blockade operates earlier in the T-cell activation cascade and may enhance endogenous T-cell priming and broader antitumor immunity, although its direct impact on CAR-T cell costimulation within the tumor remains less clearly established 213. Third, neoadjuvant rather than adjuvant PD-1 blockade in recurrent GBM has shown a meaningful survival signal (13.7 vs. 7.5 months) with evidence of enhanced IFN-γ signaling and increased TCR clonal diversity, validating the principle that timing of checkpoint intervention critically modulates immunological outcomes 2.
Beyond PD-1 and CTLA-4, second-generation checkpoint targets—TIM-3, LAG-3, and TIGIT—are gaining preclinical traction. In diffuse midline glioma models, dual PD-1 and TIM-3 blockade combined with stereotactic radiosurgery produced a 100-day survival benefit versus controls, far exceeding the 33-day benefit from anti-PD-1 monotherapy 20. A phase I trial (NCT03961971) is now evaluating MBG453 (anti-TIM-3) plus spartalizumab (anti-PD-1) combined with radiosurgery in recurrent GBM 20.
A particularly elegant engineering solution involves checkpoint reversal receptors (CPRs) co-expressed within CAR constructs. Bicistronic CARζ/CPR-4-1BB constructs are designed to convert inhibitory PD-1 signaling into 4-1BB-mediated costimulation intrinsically and have shown improved immune synapse formation, metabolic profiles, and antitumor activity in GBM xenograft models; whether this approach reduces toxicity risk in patients remains to be established 1.
Preclinical and Clinical Evidence
The most compelling preclinical proof-of-concept for combination therapy comes from orthotopic GBM models using CAR-NK cells with systemic anti-PD-1. Advanced tumors refractory to CAR-NK monotherapy were durably controlled by the combination, which drove a transcriptional shift toward pro-inflammatory programs—including IFN-γ, CD3ε, and CD4 upregulation—and increased CD4+ T-cell and NKT-cell infiltration. Paired tissue analyses from two patients in the CAR2BRAIN clinical program showed concordant increases in NKT cells post-treatment 13.
On the clinical side, a notable recent phase I advance is the CART-EGFR-IL13Rα2 program in recurrent GBM: intracerebroventricular delivery of bivalent CAR-T cells targeting EGFR and IL13Rα2 was reported in 18 patients, and 8 of 13 patients (62%) with measurable disease at infusion experienced tumor regression, with one confirmed partial response by modified RANO criteria. One patient maintained stable disease for more than 16 months. Among 7 patients with at least 12 months of follow-up, 43% remained alive at one year—a clinically meaningful signal given that median survival in recurrent GBM is typically 6–10 months 1617. CAR-T cells remained detectable in spinal fluid up to one year post-infusion in some patients, and repeat surgical specimens showed T-cell infiltration and macrophage-mediated tumor clearance, indicating sustained immune surveillance 17.
The directly relevant combinatorial trial—NCT04003649—is evaluating IL13Rα2-targeted CAR-T cells with or without nivolumab (anti-PD-1) and ipilimumab (anti-CTLA-4), representing the first prospective clinical comparison of CAR-T monotherapy versus dual checkpoint combination in brain tumors 18. In analogous CNS disease, a large retrospective study of 472 patients with melanoma brain metastases demonstrated that first-line nivolumab plus ipilimumab produced a median OS of 36.0 months versus 18.8 months with anti-PD-1 monotherapy (HR 0.47, 95% CI 0.34–0.67), with greatest benefit in asymptomatic patients 14. Additionally, a phase I trial of re-irradiation combined with nivolumab, ipilimumab, and bevacizumab in 26 recurrent high-grade glioma patients demonstrated feasibility (no unexpected CNS toxicities, no delayed radiation necrosis), an objective response rate of 64%, and median OS of 15.6 months 15.
However, early clinical experiences combining EGFRvIII-targeted CAR-T cells with pembrolizumab did not clearly improve outcomes over CAR-T alone, suggesting that single-agent PD-1 blockade may be insufficient when myeloid and metabolic suppressive circuits dominate 11.
Translational Challenges and Limitations
| Challenge | Mechanism | Potential Mitigation Strategy |
|---|---|---|
| Antigen escape | Post-CAR-T antigen downregulation (EGFRvIII, IL13Rα2) | Dual/multi-target CAR constructs; sequential antigen targeting |
| Poor CAR-T trafficking | BBB/BTB barriers; chemokine-receptor mismatch | Intratumoral/intrathecal delivery; radiation-guided chemokine augmentation |
| T-cell exhaustion | PD-L1 upregulation by IFN-γ; checkpoint pathway dominance | Intrinsic CPR engineering; multi-checkpoint blockade (TIM-3, LAG-3) |
| Myeloid dominance | MDSC/TAM-mediated suppression; IL-6, MIF-CD74 signaling | IL-6 blockade, ibudilast, arginase-1 or IDO inhibition |
| Neuroinflammation/ICANS | IFN-γ-driven CNS inflammation; CRS risk | CAR-NK platforms (lower IL-6 risk); localized IL-12Fc delivery |
| On-target/off-tumor toxicity | Target antigen expression on normal CNS tissue | Logic-gated CARs; dose optimization; local delivery |
Safety is a paramount concern. Grade 3 neurotoxicity occurred in 56% of patients in the dual-target CAR-T trial 16, and 141 severe adverse events were recorded across 13 early-phase GBM CAR-T trials 11. Localized immune stimulation strategies—such as FcRn-silenced IL-12Fc, engineered to prolong brain retention and limit systemic exposure—demonstrate that compartmentalized immune activation can enhance efficacy while minimizing systemic inflammatory toxicity 9.
Future Directions
Durable therapeutic success in brain tumors will likely require simultaneous address of multiple suppressive nodes: (1) engineering CAR-T constructs with intrinsic checkpoint resistance (CPR-4-1BB); (2) co-targeting myeloid suppression via IL-6 blockade, MIF-CD74 inhibition, or adenosine antagonism; (3) deploying dual or multi-antigen targeting to preempt antigen escape; (4) optimizing intratumoral delivery timing relative to peak intratumoral CAR-T expansion; and (5) establishing biomarker-driven patient selection using baseline IL-6 levels, MDSC burden, and hypoxia signatures to identify patients most likely to benefit 21. Prospective, preferably randomized and biomarker-informed trials with rigorous neuroinflammation monitoring are needed before this strategy can be considered for broader clinical adoption.