Ultrasound-mediated delivery of doxorubicin to the brain results in immune modulation and improved responses to PD-1 blockade in gliomas

Abstract

Given the marginal penetration of most drugs across the blood-brain barrier, the efficacy of various agents remains limited for glioblastoma (GBM). Here we employ low-intensity pulsed ultrasound (LIPU) and intravenously administered microbubbles (MB) to open the blood-brain barrier and increase the concentration of liposomal doxorubicin and PD-1 blocking antibodies (aPD-1). We report results on a cohort of 4 GBM patients and preclinical models treated with this approach. LIPU/MB increases the concentration of doxorubicin by 2-fold and 3.9-fold in the human and murine brains two days after sonication, respectively. Similarly, LIPU/MB-mediated blood-brain barrier disruption leads to a 6-fold and a 2-fold increase in aPD-1 concentrations in murine brains and peritumoral brain regions from GBM patients treated with pembrolizumab, respectively. Doxorubicin and aPD-1 delivered with LIPU/MB upregulate major histocompatibility complex (MHC) class I and II in tumor cells. Increased brain concentrations of doxorubicin achieved by LIPU/MB elicit IFN-γ and MHC class I expression in microglia and macrophages. Doxorubicin and aPD-1 delivered with LIPU/MB results in the long-term survival of most glioma-bearing mice, which rely on myeloid cells and lymphocytes for their efficacy. Overall, this translational study supports the utility of LIPU/MB to potentiate the antitumoral activities of doxorubicin and aPD-1 for GBM.

Introduction

The prognosis for patients suffering from glioblastoma (GBM) remains dismal despite extensive molecular characterization. The failure of several drug-based therapeutic approaches may be in part explained by the blood-brain barrier (BBB) that prevents sufficient brain penetration for most agents. For instance, modern antibody-based treatments that have improved the outcomes of many solid tumors do not cross the BBB¹. Indeed, GBM cells are known to migrate and infiltrate brain regions well beyond what is revealed on magnetic resonance imaging (MRI) by contrast uptake where the BBB is impermeable to several systemically delivered agents^2,3. Even when the enhancing tumor region is completely resected, infiltrating residual tumor cells lead to recurrence, with patients almost invariably succumbing to their disease^4,5.

Penetration of different drugs and biologicals in the brain can be achieved through the opening of the BBB with low-intensity pulsed ultrasound (LIPU) in combination with intravenous injection of microbubbles (MB), i.e., LIPU/MB^{6,7,8,9,10,11,12,13,14,15,16}. This technology works by using a skull-implantable device or MRI-guided transcranial focused ultrasound (FUS) that sends ultrasound waves that induce the vibration of MB to open the BBB^{10,11,12,13,16,17}. This technique has resulted in increased brain concentrations of therapeutic agents in preclinical glioma models and patients with either GBM or brain metastases^{6,7,10,12,13,18}. Clinical studies have shown that this technique is safe and effective^10,16,17,19, with ongoing studies further exploring its therapeutic applications.

The failures of recent large randomized clinical trials that evaluated anti-PD-1 immunotherapy (aPD-1) to improve the outcome of patients with newly diagnosed or recurrent GBM^20,21,22 highlight the importance of developing treatment combinations that reach the tumor cells and elicit effective anti-tumoral immunity. Doxorubicin (and liposomal doxorubicin, DOX), a cytotoxic anthracycline that intercalates into the DNA and inhibits topoisomerase type II²³, has displayed immunogenic effects in several cancers^24,25. This chemotherapy activates the pathway associated with cyclic GMP-AMP synthase (cGAS) and its downstream signaling effector stimulator of interferon genes (STING) that senses cytosolic DNA^26,27,28, promoting the expression of type I interferon (IFN) signature²⁹. In addition, DOX induces immunogenic cell death in tumor cells²⁵. This anthracycline also led to an increase in tumor-infiltrating IFN-γ-expressing CD8⁺ and CD4⁺ T cells to sustain anticancer activities in preclinical models of sarcoma, lymphoma, breast, and colon cancer^30,31. The immunological qualities described for DOX have led to its exploration as an immune-modulating agent to enhance the efficacy of immune checkpoint inhibitors for cancer^24,32. This effect has been also demonstrated clinically. A multi-cohort clinical trial evaluated treatment responses when different chemotherapy agents preceded aPD-1 therapy in patients with advanced breast cancer. Induction therapy with DOX showed a doubling of objective response rates compared to treatment with aPD-1 alone (35 vs 17%, respectively)³³. Along with these clinical responses, bulk RNA-seq of tumors exposed to short-term treatment with DOX showed increased tumor expression of several inflammatory gene signatures, including tumor necrosis factor α (TNF-α) signaling through NF-κB and those related to IFN-α and IFN-γ response³³.

Although DOX has been characterized by its immunogenic attributes, the inability to penetrate the CNS limits tumor and immune cells residing in the brain from being affected by these anthracycline-specific immune properties. Additionally, the limited inflow of antibodies into the brain precludes their binding to immune cells in the tumor microenvironment (TME) and peritumoral regions^34,35,36. In this context, we investigated the use of LIPU/MB to increase DOX and aPD-1 concentrations in the brain to promote an antitumoral immune response for GBM.

In this work, we report the immune response and pharmacokinetics related to the use of LIPU/MB to enhance the penetration and therapeutic effect of both DOX and aPD-1 in mouse GBM models, as well as in a cohort of 4 recurrent GBM patients. These patients had a skull-implantable ultrasound device (SonoCloud-9; SC9, Carthera, Lyon, France) and received aPD-1 and DOX under single-patient expanded access programs (EAP). They had also previously participated in another clinical trial (NCT04528680)¹⁷. Additionally, we present preclinical and clinical evidence that supports the use of the LIPU/MB technology in combination with DOX and aPD-1 to induce the upregulation of MHC class I and II in GBM cells and generate IFN-γ responses mediated by myeloid cells and T cells.

Results

LIPU/MB increases the penetration of DOX into the human and murine brain

Four patients with GBM, who failed two lines of therapy including standard of care (radiation and temozolomide) and LIPU/MB-based opening of the BBB with concomitant albumin-bound paclitaxel¹⁷, were treated with LIPU/MB with intravenous administration of DOX and aPD-1 at recurrence (Fig. 1a). Treatment included an induction cycle of low-dose liposomal DOX (30 mg) alone, followed by a 2^nd dose of DOX and aPD-1 10–14 days later, followed by subsequent cycles of DOX and aPD-1. All treatment cycles that included DOX and aPD-1 were delivered in conjunction with the activation of the SonoCloud-9 device. DOX was administered immediately after sonication, over 30 minutes in all cases. aPD-1 was administered before sonication. All 4 patients underwent surgery after 2–8 cycles of DOX/aPD-1 as clinically indicated tumor debulking (growing mass effect) or biopsy. In this context, our tissue analysis included paired specimens prior to treatment with DOX and aPD-1 (pre-treatment GBM samples) and tumor tissue resected after DOX and aPD-1 delivered with LIPU/MB (on-treatment GBM samples) (Fig. 1a) (Supplementary Table 1). In two of these patients, as per standard neurosurgical technique, we resected and were able to sample peritumoral brain regions that were subjected to LIPU/MB (sonicated) with concomitant administration of DOX and aPD-1 and peritumoral brain regions that were outside the sonication field (non-sonicated). The determination of whether the peritumoral brain was sonicated or non-sonicated was based on the location relative to the ultrasound emitters, as illustrated in Fig. 1b. All surgical samples were acquired 2 days after treatment with liposomal DOX and aPD-1 with concomitant LIPU/MB. The DOX concentrations quantified across all multiple samples from each patient in non-sonicated peritumoral brain regions ranged from 0.0278 μg/g (0.0479 μmol/kg) to 0.112 μg/g (0.194 μmol/kg). On the other hand, DOX concentrations ranged from 0.05 μg/g (0.0872 μmol/kg) to 0.405 μg/g (0.698 μmol/kg) in multiple samples from sonicated peritumoral brain regions (Supplementary Fig. 1). Overall, LIPU/MB led to a 2-fold increase (95% CI of mean:1.406-2.659) in DOX concentration in sonicated peritumoral brain samples compared to non-sonicated peritumoral brain samples two days post-sonication (P = 0.012, chi-squared; Fig. 1c).

To further determine the effect of LIPU/MB-based BBB disruption on DOX penetration within the brain, we quantified the concentration of DOX in the brains of mice obtained two days after treatment with DOX delivered with and without LIPU/MB (Fig. 1d). We found increased DOX concentrations in mouse brains after treatment with DOX delivered with LIPU/MB compared to brains obtained from mice treated with DOX without LIPU/MB (P = 0.003, t-test; Fig. 1e). Consistently, LIPU/MB led to a 3.92-fold increase (95% CI of mean: 2.015-5.826) in DOX concentration at 48 h with LIPU/MB compared to without LIPU/MB (P = 0.003, t-test; Fig. 1f). These results demonstrate the ability of LIPU/MB to increase systemically delivered DOX in the brains of mice and humans.

DOX upregulates antigen-presenting molecules in GBM cells

DOX has previously been shown to enhance the expression of MHC class I (MHC I) in preclinical models of ovarian and colorectal carcinoma^37,38. Thus, we investigated whether human GBM samples treated with DOX and LIPU/MB exhibit upregulation of antigen-presenting molecules by GBM cells. We analyzed pre-treatment and on-treatment tumor samples in the cohort of 4 recurrent GBM patients who had treatment with DOX and aPD-1 delivered with LIPU/MB (Fig. 1a). For this purpose, we employed multiplex immunofluorescence to evaluate the abundance of SOX2⁺ cells (tumor cell marker³⁹) expressing HLA-ABC and HLA-DR (Fig. 2a). This analysis was performed in tumor regions delineated by a neuropathologist (Supplementary Fig. 2a). We found increased cell density of SOX2⁺ HLA-ABC⁺ cells (P = 0.0079, chi-squared; Fig. 2b) and SOX2⁺ HLA-DR⁺ cells (P = 0.024, chi-squared; Fig. 2c) in GBM samples treated with DOX compared to pre-treatment GBM samples. To evaluate whether the upregulation of HLA-ABC and HLA-DR was a result of tumoral progression irrespective of the therapy with LIPU/MB with DOX and aPD-1, we analyzed paired tumor samples obtained at the first and second recurrence of a cohort of 8 GBM patients that did not receive this immunotherapy (Supplementary Fig. 2b). In this control cohort, we did not find differences in the cell density of SOX2⁺ HLA-ABC⁺ cells (P = 0.2582, chi-squared; Supplementary Fig. 2c) and SOX2⁺ HLA-DR⁺ cells (P = 0.7491, chi-squared; Supplementary Fig. 2d) between the first and second recurrent tumors….