Despite the availability of numerous therapeutic substances that could potentially target CNS disorders, an inability of these agents to cross the restrictive blood–brain barrier (BBB) limits their clinical utility. Novel strategies to overcome the BBB are therefore needed to improve drug delivery. We report, for the first time, how Tumor Treating Fields (TTFields), approved for glioblastoma (GBM), affect the BBB’s integrity and permeability. Here, we treated murine microvascular cerebellar endothelial cells (cerebEND) with 100–300 kHz TTFields for up to 72 h and analyzed the expression of barrier proteins by immunofluorescence staining and Western blot. In vivo, compounds normally unable to cross the BBB were traced in healthy rat brain following TTFields administration at 100 kHz. The effects were analyzed via MRI and immunohistochemical staining of tight-junction proteins. Furthermore, GBM tumor-bearing rats were treated with paclitaxel (PTX), a chemotherapeutic normally restricted by the BBB combined with TTFields at 100 kHz. The tumor volume was reduced with TTFields plus PTX, relative to either treatment alone. In vitro, we demonstrate that TTFields transiently disrupted BBB function at 100 kHz through a Rho kinase-mediated tight junction claudin-5 phosphorylation pathway. Altogether, if translated into clinical use, TTFields could represent a novel CNS drug delivery strategy.

In a recent study, we showed in an in vitro murine cerebellar microvascular endothelial cell (cerebEND) model as well as in vivo in rats that Tumor-Treating Fields (TTFields) reversibly open the blood–brain barrier (BBB). This process is facilitated by delocalizing tight junction proteins such as claudin-5 from the membrane to the cytoplasm. In investigating the possibility that the same effects could be observed in human-derived cells, a 3D co-culture model of the BBB was established consisting of primary microvascular brain endothelial cells (HBMVEC) and immortalized pericytes, both of human origin. The TTFields at a frequency of 100 kHz administered for 72 h increased the permeability of our human-derived BBB model. The integrity of the BBB had already recovered 48 h post-TTFields, which is earlier than that observed in cerebEND. The data presented herein validate the previously observed effects of TTFields in murine models. Moreover, due to the fact that human cell-based in vitro models more closely resemble patient-derived entities, our findings are highly relevant for pre-clinical studies.

Introduction: Although a number of effective drugs are available to treat central nervous system (CNS) disorders, their ability to breach the tight regulation of the blood brain barrier (BBB) still remains a major challenge. Recently, the use of tumor treating fields (TTFields) has become an effective treatment approach for glioblastoma. Furthermore, its combination with chemotherapy significantly improved overall patient survival. Nonetheless, how TTFields could affect the BBB has not yet been studied. Our recent findings exhibit the potential of TTFields administration to open up the BBB in vitro with an optimal frequency of 100 kHz. Consequently, in this study, we therefore aimed to validate our data in vivo. Experimental procedures: Subsequent to 100 kHz TTFields or heat treatment for 72 h, rats were i.v. injected with Evan´s Blue (EB). Next, they were sacrificed to extract and quantify EB from the brain. In the same manner, rats were injected with TRITC-dextran (TD), after which permeation was visualized in sectioned brains. Cryosections of rat brains were also prepared post-TTFields treatment. These were stained for intercellular junction proteins claudin-5, occludin and PECAM-1 as well as immunoglobulin G (IgG) to assess vessel structure. Finally, serial dynamic contrast-enhanced (DCE) MRI with gadolinium (Gd) contrast agent was performed before and after TTFields administration. Results: Permeation of both EB and TD was observed in the brain after TTFields application. Moreover, brain cryosections displayed claudin-5 and occludin delocalization but not PECAM-1. Accumulation of IgG in the brain parenchyma was also noted. Confirming these observations, increased Gd in the brain was shown by DCE-MRI post TTFields treatment. A reversion to normal conditions was, however, detected 96 h after end of treatment demonstrated by no difference in contrast enhancement between control and TTFields-treated rats. Conclusions: Administration of 100 kHz TTFields in rats led to alterations in BBB integrity and permeability, which signal its opening. The subsequent recovery of the BBB at the end of treatment demonstrates transient effects, hence presenting TTFields as a possible novel clinical strategy to open the BBB for enhanced and more effective drug delivery strategy for CNS disorders. Citation Format: Ellaine Salvador, Almuth F. Kessler, Julia Hörmann, Malgorzata Burek, Catherine T. Brami, Tali V. Sela, Moshe Giladi, Ralf-Ingo Ernestus, Mario Löhr, Carola Förster, Carsten Hagemann. Blood brain barrier opening by TTFields: a future CNS drug delivery strategy [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 6251.

Machine learning classifiers with high test accuracy often perform poorly under adversarial attacks. It is commonly believed that adversarial training alleviates this issue. In this paper, we demonstrate that, surprisingly, the opposite may be true -Even though adversarial training helps when enough data is available, it may hurt robust generalization in the small sample size regime. We first prove this phenomenon for a high-dimensional linear classification setting with noiseless observations. Our proof provides explanatory insights that may also transfer to feature learning models. Further, we observe in experiments on standard image datasets that the same behavior occurs for perceptible attacks that effectively reduce class information such as mask attacks and object corruptions.

No abstract