Improved Drug Delivery to Brain Metastases by Peptide-Mediated Permeabilization of the Blood–Brain Barrier

Brain metastasis is a frequently reported complication for patients with cutaneous melanoma, where the average survival time, if untreated, is 3 to 5 months. Current treatment strategies involve surgery, systemic therapy, radiotherapy, and/or radiosurgery (1). This can to some extent increase the survival time, yet with a divergent treatment efficacy, emphasizing the need for new treatment options.

It is well known that melanomas are molecularly heterogeneous (2) and immunogenic tumors (3), properties that have been exploited for drug development (4). For instance, it has been shown that the serine/threonine kinase protein BRAF is a key molecular driver of metastatic melanoma, which has led to the development of several BRAF inhibitors (BRAFi). Furthermore, immune checkpoint inhibitors such as PD-1/PD-L1 have shown a strong clinical efficacy in clinical trials for melanoma (5–7). However, a recurring issue is that many of these drugs are too large to cross the intact interface between circulating blood and the brain parenchyma, that is the blood–brain barrier (BBB; refs. 8, 9). The BBB consists of vascular endothelial cells linked by tight junctions, encircled by astrocytic end-feet and pericytes leading to a selective barrier that determines the entry of molecules into the brain (10). This barrier represents in many instances a major obstacle for systemic brain metastasis treatment. Compounds that consist of more than 8 to 10 hydrogen bonds and are larger than 400 to 500 Da are prohibited from entering the brain. All large molecular drugs, such as antibodies, and 98% of small molecular drugs are excluded from the brain by the BBB (11). The brain is thus considered as a sanctuary site for metastatic growth (12) and the exposure to drugs is lower in brain metastases than systemic metastases. This is not only ascribed to the presence of the BBB but also the blood–tumor barrier (BTB; ref. 13). The BTB differs from the BBB in that the vascular system is no longer surrounded by the other, normal BBB components, but tumor cells. Because of this structural difference, the BTB is proposed to be more permeable than the BBB (14). Micro-metastases, that is lesions smaller than 1 mm³, usually have a lower permeability than larger metastases, in which the BTB might be compromised as a result of tumor growth (15). Systemic therapy may therefore show efficacy on larger metastatic lesions, whereas micro-metastases receive subtherapeutic drug concentrations, which can contribute to treatment resistance (16). However, it has also been shown that there is not necessarily a straightforward association between brain metastasis size and drug uptake. Within the same lesion, the distribution can vary up to 10-fold (17). Also, melanoma patients with advanced disease can present with multiple brain metastases of different sizes with varying BTB integrities, further challenging systemic treatment (18). Several strategies have therefore been developed to temporarily disrupt the BBB for improved drug delivery such as focused ultrasound combined with circulating microbubble contrast agents (19–21), hyperosmotic opening (22–24) and radiotherapy (25). Other strategies involve the circumvention of the BBB by convection-enhanced delivery (26), viral-mediated or liposomal delivery (27), carrier molecules (28), and polymer wafers (29). These strategies have shown both strengths and weaknesses, but with limited success and many with apparent side effects (30, 31).

It has previously been reported that the synthetic peptide K16ApoE can carry relatively large compounds into the mouse brain through the low-density lipoprotein receptor (LDLR) pathway (32). The use of K16ApoE in a therapeutic setting in vivo, however, has not been investigated. Here we determined, using advanced MRI techniques, the length of the therapeutic time window of K16ApoE BBB permeabilization in NOD/SCID mice and also its in vivo toxicity profile. Moreover, the morphologic and functional effects of the peptide on cells and tissues were elucidated. In addition, we assessed the ability of K16ApoE to enhance drug delivery of a clinically active BRAFi (dabrafenib) on preclinical brain metastases, which was our main objective with this study. Finally, using PET/CT as an in vivo biodistribution tool for studying brain penetration, we assessed the potential of the peptide to deliver compounds to the brain with a size range corresponding to clinically relevant immune checkpoint inhibitors.

The K16ApoE peptide has the following amino acid sequence: KKKK-KKKK-KKKK-KKKK-LRVR-LASH-LRKL-RKRL-LRDA with a molecular weight (M_W) of 4.521.79 Da. The synthesis and characterization of the peptide is elaborated in Supplementary Materials and Methods. Briefly, a series of 16 lysine residues (K16) was covalently linked to the 20 amino acid part of the low-density lipoprotein receptor binding segment of apolipoprotein E (ApoE).

Five cell lines were used as constituents of the in vitro model system of the BBB, namely Mabin-Darby Canine Kidney (MDCK) cells, MDCK II, rat brain endothelial cells 4 (RBE4), human brain endothelial cells (hCMEC/D3), and human brain astrocytes (SC-1800). In addition, 2 brain metastatic melanoma cell lines were used: H1 (or H1_DL2) and H2 (see Supplementary Materials and Methods). We obtained written consent by the Regional Ethical Committee (#013.09) and the Norwegian Directorate of Health (#9634) before human tumor tissue was collected and stored.

Female NOD/SCID mice were purchased from Envigo. The animals were bred and maintained in our animal facility certified by the Association for Assessment and Accreditation of Laboratory Animal Care International. They were fed a standard pellet diet and provided water ad libitum. Anesthesia was induced with 3% sevoflurane (Abbott Laboratories Ltd.) in oxygen and maintained with 1.5% sevoflurane in oxygen during all procedures unless stated otherwise. The mice were monitored daily and sacrificed when significant morbidity symptoms were observed. The National Animal Research Authority approved all animal procedures prior to all experiments (Application #8093, approved February 13, 2016).

The toxicity of K16ApoE was evaluated by intravenous tail vein injections of increasing concentrations of peptide into 38 NOD/SCID mice as described in Supplementary Materials and Methods and Supplementary Fig. S1A.

Dynamic contrast enhanced MRI (DCE-MRI) was carried out using a 7 Tesla small-animal horizontal scanner (Bruker BioSpin GmbH), using a 72 mm quadrature transmit coil and a 4-channel mouse brain array receive coil. The animals were placed in prone position and body temperature was maintained at 37°C.

T₁ and T₂ weighted spin echo scans were acquired to provide anatomical references by using fast spin echo (FSE) protocols as described in Supplementary Materials and Methods. The mice received a dose of 50 (n = 5), 100 (n = 11), or 200 μg (n = 5) of K16ApoE dissolved in 100 μL 9 mg/mL NaCl administered through the tail vein 10, 30, 60, 120, or 240 minutes before the start of the perfusion scans. Mice in the negative control group received 100 μL 9 mg/mL NaCl.

The perfusion scans were performed using DCE-MRI and analyzed using the Extended Tofts model implemented in nordicICE v2.3 (Nordic NeuroLab) as described in Supplementary Materials and Methods.

RBE4 cells were pretreated with 20 μg/mL rhodamine-conjugated K16ApoE for 45 minutes and exposed to inhibitors of dynamin- and clathrin-mediated endocytosis and studied by flow cytometry. RBE4, MDCK, hCMEC/D3, H1, and H2 cells were incubated with endocytosis inhibitors and pretreated with 20 μg/mL K16ApoE and Alexa Fluor-647–conjugated BSA prior to flow cytometry. Both experiments are elaborated in Supplementary Materials and Methods.

The viability of MDCK, MDCK II, RBE4, and hCMEC/D3 endothelial cells and H1 brain metastasis cells after treatment with K16ApoE was evaluated in vitro using a resazurin proliferation assay and for MDCK cells also a Live/Dead assay. The procedures are described in Supplementary Materials and Methods.

RBE4 and MDCK II cells were incubated with 0, 20, 40, or 80 μg/mL K16ApoE for 45 minutes before they were fixed and prepared for scanning electron microscopy to study the morphology of the endothelial monolayers after peptide exposure. The protocol is described in Supplementary Materials and Methods.

The procedure for the cell adhesion assay carried out prior to in vitro BBB modeling experimental set-ups is described in the Supplementary Materials and Methods.

Mono- and co-culture BBB models were constructed using the human astrocyte cell line SC-1800 and the endothelial cell line hCMEC/D3. The resistance values, indicating increased BBB permeability, were recorded using the electric cell substrate impedance sensing (ECIS) system and CellZScope for 2D and 3D modelling respectively, as reported previously (33). The mono- and co-cultures were treated with 0, 20, 40, or 80 μg/mL of K16ApoE and resistance was recorded until recovery of the barrier was observed. Resistance values were obtained in Ω from the ECIS system and Ω.cm² from the automated sensing system, CellZScope. See Supplementary Materials and Methods for a detailed description of the protocol.

To study the biodistribution of the peptide, we injected ¹²⁵I-K16ApoE intravenously into NOD/SCID mice and collected blood samples and a selection of organs and measured these for radioactivity. See Supplementary Materials for further details.

In an initial control experiment, 9 NOD/SCID mice (8 weeks old) were injected with 5 × 10⁵ H1_DL2 cells intracardially in 0.1 mL PBS as described in Supplementary Materials and Methods and divided into 2 groups by simple randomization: 4 animals were injected with 200 μg K16ApoE and 5 with 9 mg/mL NaCl, to exclude any treatment effects of the peptide.

Thirty-six female NOD/SCID mice (8 weeks old) were then injected with 5 × 10⁵ H1_DL2 cells intracardially as described in Supplementary Materials and Methods. The next day, mice were by simple randomization divided into 3 treatment groups: The first group received 200 μg K16ApoE followed by 10 mg/kg dabrafenib (free base, CT-DABRF; ChemieTek) 5 minutes later. The next group received 10 mg/kg dabrafenib and the third group received 9 mg/mL NaCl (vehicle).

All solutions were administered intravenously. The mice were treated twice a week for 6 weeks. See Supplementary Figure S1Β for a detailed description of the animals used.

Contrast enhanced T₁ and T₂ weighted MRI was conducted 4 and 6 weeks after as described in Supplementary Materials and Methods.

Mouse organs such as lungs, heart, liver, kidneys, colon, stomach, spleen, skin, muscle, and brain were harvested after treatment with K16ApoE and fixed using 4% formaldehyde. Paraffin-embedded organs were sectioned and mounted on slides. The sections were deparaffinized and stained with hematoxylin and eosin (H&E) for histologic assessments.

A mass spectrometry (MS) experiment was performed to confirm the presence of dabrafenib in K16ApoE combination treated NOD/SCID mice from the in vivo treatment experiment. The procedures are described in Supplementary Materials and Methods.

The capacity of BBB permeability from K16ApoE was further evaluated by PET/CT using ¹⁸F-albumin (∼67 kDa) and ¹⁸F-IgG (∼150 kDa) to study if also these compounds could enter the brain after administration of K16ApoE. The albumin and IgG labeling procedure prior to PET/CT as well as the dynamic scanning procedures are described in Supplementary Materials and Methods.

The statistical analyses were carried out in Prism 7 for Mac, Version 7.0b. Unpaired t tests were used to evaluate 2 normally distributed groups, whereas Mann–Whitney tests were used to compare nonparametric data. A Mantel–Cox log-rank test was used to analyze survival data from the in vivo treatment experiment. The results are displayed as individual points with mean ± SEM. A 2-tailed P-value ≤0.05 was considered significant.

To determine the maximum tolerated dose of K16ApoE, mice were injected intravenously with increasing concentrations of peptide (50–1,000 μg). For peptide doses up to 400 μg, the mice showed no signs of pain or distress following systemic peptide exposure, and they all recovered from anesthesia within approximately 3 minutes (Fig. 1A). Higher peptide doses led to a respiratory and/or cardiac arrest within 30 minutes (Supplementary Videos S1 and S2). Higher doses were also associated with an abnormal erythrocyte morphology (Supplementary Fig. S2).

DCE-MRI on healthy mice demonstrated a dose-dependent effect of the peptide. K16ApoE concentrations of 50 or 100 μg was insufficient for BBB permeabilization, that is allowing Omniscan to enter into the extravascular, extracellular space (EES) from the blood plasma, as seen by the K^trans values (Fig. 1B). A major leakage of Omniscan contrast agent from blood into the brain tissue was observed when administered 10 minutes after injection of 200 μg K16ApoE. This implicates that the peptide was able to successfully open the intact BBB (Fig. 1C and D). Interestingly, our results also showed that the BBB was partially open for up to approximately 1 hour after K16ApoE injection, reflecting a putative time frame for effective drug administration (Fig. 1C and D). Other DCE-MRI parameters besides K^trans are listed in Supplementary Table S1. Based on the preceding, the toxicity studies above and previous literature (34), we chose to use 200 μg per mouse for further in vivo experiments.

To acquire a mechanistic insight on how K16ApoE facilitates BBB permeability, we first conducted baseline studies for further in vitro experiments. We determined K16ApoE IC₅₀ values for 5 normal endothelial cell lines, and these were all within a relatively narrow range of 30.89 to 86.18 μg/mL (Supplementary Fig. S3A–S3D). For H1_DL2 cells used in the intracardiac metastasis model, the IC₅₀ was 25.75 μg/mL (Supplementary Fig. S3E).

Live/dead assays and scanning electron microscopy images of MDCK and RBE4 cells (Supplementary Fig. S3F–S3H) showed a dose-/time-dependent increase in the number of dead cells over 45 minutes (see also Supplementary Videos S3–S6).

We then applied flow cytometry to assess endocytic activity in RBE4 cells treated with 20 μg/mL rhodamine-labeled K16ApoE. As shown in Fig. 2A, high endocytic uptake of peptide was observed in the RBE4 cells (pink curve in Fig. 2A). By adding chlorpromazine (dark blue curve in Fig. 2A) or dynasore (brown curve in Fig. 2A), which are inhibitors of clathrin- and dynamin-mediated endocytosis, respectively, the peptide uptake was reduced with the strongest effect seen for chlorpromazine.

We then pretreated RBE4 cells with AF647-labeled BSA and incubated with (purple curve in Fig. 2B) and without K16ApoE (green curve in Fig. 2B). When chlorpromazine was added as well, there was a reduction in BSA uptake (black curve in Fig. 2B). The lowest BSA uptake was seen in RBE4 cells incubated with chlorpromazine and no peptide (yellow curve in Fig. 2B). Corresponding experiments were carried out on MDCK, hCMEC/D3, H1, and H2 cells studying BSA uptake after pretreatment with endocytosis inhibitors (Supplementary Fig. S4). The same pattern was seen across all cell lines: the highest BSA uptake was observed for cells pretreated with K16ApoE (purple curves), whereas endocytosis inhibitors reduced this increase, chlorpromazine (yellow curves) to a larger extent than dynasore (blue curves). Dynamin-mediated endocytosis can be serum dependent. We therefore carried out the dynasore experiments with (Supplementary Fig. S4) and without BSA (Fig. 2). To summarize, both clathrin- and dynamin-mediated endocytosis are likely involved in K16ApoE uptake (below IC₅₀ doses).

To show that also other uptake mechanisms of K16ApoE likely are involved, we investigated the uptake of AF647-conjugated BSA in RBE4 cells with (purple curves in Fig. 2C and D) and without (green curves in Fig. 2C and D) pretreatment with K16ApoE. When compared with 37°C, the uptake of BSA was reduced in cells that were kept at 4°C (Fig. 2D), that is at a temperature when endocytosis usually is abolished (Fig. 2C). In conclusion, based on the above data, endocytic mechanisms are involved in peptide uptake, likely in combination with other mechanisms as described in the following.

We then studied changes in endothelial cell monolayer surfaces in vitro following K16ApoE exposure. Scanning electron microscopy images showed that MDCK II and RBE4 cells not exposed to the peptide formed uniform monolayers (Fig. 3A; Supplementary Fig. S3H, respectively) with occasional protruding cells with smooth surfaces. At increasing K16ApoE concentrations, the cell surface lost the uniform morphology, and punctures in the membranes could be observed, indicating dying cells (see inserts in Fig. 3A). There was an association between increasing concentrations of the peptide and the number of punctured, protruding cells. This dose-dependent cell death was verified by the live/dead experiment (Fig. 3B; Supplementary Fig. S3F and S3G).

Taken together with data presented in Supplementary Figs. S2 and S3, this indicates that cell lysis is likely involved especially at higher peptide concentrations, likely due to a cationic effect leading to electrostatic interactions with negatively charged cell membranes (35), in addition to the endocytosis mechanisms indicated by flow cytometry (Fig. 2).

Cellular adhesion was measured prior to in vitro BBB modeling using crystal violet in 96-well plates after treatment with 0, 20, 40, or 80 μg/mL K16ApoE for 45 minutes. The strongest cellular adhesion was observed for untreated hCMEC/D3 cells. The adhesion potential was significantly reduced during peptide treatment in a dose-dependent manner (Supplementary Fig. S5A and S5B). However, although not statistically significant, there was a tendency that the adhesion potential started to recover with increasing recovery times after 45 minutes of peptide exposure. Endothelial cells exposed to 40 and 80 μg/mL showed a statistically significant increase in adhesion potential after 60 minutes (Supplementary Fig. S5A).

In both human in vitro BBB models assessed, the cells were exposed to 0, 20, 40, or 80 μg/mL of the peptide for 45 minutes, before they were allowed to recover for as long as deemed necessary in a mono- and coculture model. In the mono-culture model, the endothelial cell monolayers restored their integrity within 3 hours (Fig. 3C), whereas in the co-culture model with endothelial cells and astrocytes, the integrity of both cell layers was restored after 15 hours (Fig. 3D).

The activity of ¹²⁵I-labeled K16ApoE in blood plasma was rapidly reduced over the total measured time of 30 minutes. The most prominent decline was observed within the first minute after the peptide was injected into the tail vein, whereas subsequent values quickly reached a baseline. The curve corresponds to a half-life of K16ApoE in blood of approximately 1 minute (Supplementary Fig. S6B).

The biodistribution of ¹²⁵I-K16ApoE was analyzed in numerous selected organs. The highest values of ¹²⁵I-K16ApoE accumulation were seen in the liver (164,570 cpm), kidney (160,107 cpm), and spleen (136,889 cpm), whereas the lowest counts were observed in colon (17,290 cpm), femur (14,720 cpm), and muscle tissue (13,608 cpm). Intermediate values were observed for ventricle (92,533 cpm), lungs (76,554 cpm), skin (38,835 cpm), and heart (28,482 cpm). The high activity seen in the kidneys, liver, and spleen suggested that elimination occurred through all of these organs (Supplementary Fig. S6C).

Because ¹²⁵I was labeled to the only histidine residue present in the peptide, the potency of the final ¹²⁵I-K16ApoE construct only allowed us to inject a concentration of 25 μg K16ApoE without exceeding the maximum volume possible to inject intravenously in a NOD/SCID mouse. Further, because the measurements were carried out more than 30 minutes after the peptide was injected, this is a time window that does not allow major remnants of the peptide to be detected in the brain. For these reasons combined, we did not focus on the amount of peptide in the brain in this experiment, as both the concentration and time window likely is too small to see any significant uptake, as observed by DCE-MRI (Fig. 1).

Histologic analysis was carried out in 2 separate experiments (Supplementary Fig. S1). In the first experiment, healthy NOD/SCID mice were subjected to a one-time exposure of 0, 200, 400, 600, 800, or 1,000 μg K16ApoE (Supplementary Fig. S7A). In the second experiment, tumor-bearing animals were subjected to 200 μg K16ApoE twice a week over a period of 6 weeks (Supplementary Fig. S7B). Representative H&E sections from brain, lungs, kidneys, liver, spleen, skin, muscle tissue, colon, stomach, and heart did not reveal any pathologic changes in animals from either treatment group, across both experiments (Supplementary Figs. S7A and S7B).

We then studied whether a combined use of K16ApoE with dabrafenib (537.6 Da) could increase the therapeutic effects in a human brain metastasis animal model, compared with dabrafenib treatment alone.

A small control experiment was initially carried out with injections of only K16ApoE or vehicle (saline) to evaluate whether the peptide itself had any therapeutic effects on tumor burden. No such effects were observed (Supplementary Fig. S8A–S8C). A larger study was then done with 3 treatment groups: Dabrafenib, K16ApoE, and dabrafenib and vehicle. To minimize the number of animals used, the K16ApoE alone group was not repeated.

The mean number of tumors as well as the mean total tumor volume in animals treated with a combination of K16ApoE and dabrafenib decreased at weeks 4 and 6, compared with control animals (vehicle) or animals treated with dabrafenib only (Fig. 4A–C). No statistically significant differences could be found in tumor numbers or tumor volumes between the dabrafenib group and the control group at week 4 and 6, indicating that dabrafenib treatment alone was not effective.

Kaplan–Meier curves revealed no difference in survival between vehicle- and dabrafenib-treated animals whereas the combinatorial treatment group had a significant survival benefit (Fig. 4D).

A pilot mass spectrometry imaging experiment revealed that dabrafenib was taken up in the brain in mice treated with a co-injection of peptide and dabrafenib, as shown in Supplementary Fig. S9A. A negative control animal treated with vehicle is presented in Supplementary Fig. S9B, where no uptake is detected. Dabrafenib was fragmented in 2 segments, namely at m/z 480.1 and 344.1 as seen in Supplementary Fig. S9C.

To test the potential of K16ApoE to carry even larger molecules than dabrafenib across the BBB, we injected 200 μg K16ApoE followed by ¹⁸F-albumin (∼67 kDa) or ¹⁸F-IgG (∼150 kDa) and performed subsequent dynamic PET/CT brain imaging. Over 30 to 60 minutes, we observed a significant increase in average standardized uptake values (SUV_mean) in peptide-injected mice as compared with vehicle-treated mice for both ¹⁸F-albumin (Fig. 5A) and ¹⁸F-IgG (Fig. 5B). This implies leakage of the radiolabeled molecules from the blood plasma and into the EEC. Thus, K16ApoE facilitates the delivery of compounds with a molecular weight of up to at least 150 kDa.

The delivery of therapeutic drug concentrations across the BBB and into metastatic lesions remains a critical issue in the treatment of brain metastases. In the smallest lesions not detected by clinical MRI, the BBB is presumably intact. As the brain metastases progress, the barriers develop a heterogeneous permeability to different-sized molecules, still with many metastases showing minor or no permeability at all (15). Thus, efficient drug delivery to the brain lesions is often compromised, necessitating the need for strategies to increase the leakiness of the BBB (30). Here, we describe a treatment strategy using a BBB permeabilizing peptide and thereby improving the delivery of dabrafenib (537.6 Da), which previously has demonstrated profound effects towards extracranial melanomas with BRAF mutations. Although the anatomical distribution of dabrafenib is superior to several other BRAFis (9), the drug does not readily penetrate into the brain parenchyma if the BBB is intact. Here we show that our treatment strategy inhibits the progression of BRAF^V600E mutated melanoma brain metastases, ascribed to an improved drug delivery across the BBB.

In a previous study with K16ApoE, Evans blue (M_W ∼0.96 kDa) was injected 10 minutes after administering the peptide, and the results indicated that a therapeutic window of approximately 60 minutes was facilitated (34). We used DCE-MRI which is a quantitative and highly sensitive MRI technique to validate these results. The observed increase in K^trans values demonstrates a genuine pharmacodynamic increase in BBB permeability and thus perfusion (36), and this was directly attributable to peptide action.

However, although it has been shown by us and others that K16ApoE is able to permeabilize the BBB, the specific mechanisms responsible for this effect have not been firmly established (32, 34, 37). Meng and colleagues have previously seen that K16ApoE likely promotes endocytosis into endothelial cells (37). Our findings indicate that there are likely several mechanisms involved, including loss in cell membrane integrity and lytic properties in addition to endocytosis, especially at higher peptide concentrations. We show that the cellular uptake of K16ApoE at lower doses (i.e. 20 μg/mL) was reduced by dynasore and chlorpromazine, inhibitors of dynamin-dependent and clathrin-mediated endocytosis, respectively. Our data may suggest that clathrin-mediated endocytosis is the most active one of these 2 mechanisms, which is also in line with the literature (38). We found that peptide-mediated uptake of BSA was still present at low temperatures (4°C) when endocytosis is significantly reduced, although at reduced levels. This adds to the hypothesis that also other cellular uptake-mechanisms may be involved. For instance, our scanning electron microscopy data indicated that the peptide exerted a dose-dependent, toxic effect on the cells, which has also been reported previously (37). At increasing concentrations, more cells with punctured cell membranes were observed (Fig. 3A), suggesting also a lytic effect at higher concentrations. These findings were supported by our fluorescence time lapse study, which also indicated that the cells were relatively unaffected by the peptide during the first minute of exposure. In addition, the half-life of K16ApoE in the blood was approximately one minute, which taken together implies that relatively large doses of peptide (above the IC₅₀ doses) may be injected safely into the bloodstream. This is also supported by the in vivo toxicity study summarized in Fig. 1A.

The endothelial cell integrity after exposure to the peptide was studied using in vitro BBB models and by investigating peptide action on proteins important in the assembly and maintenance of tight junctions. The peptide reduced the endothelial barrier integrity as measured using both ECIS (Fig. 3C) and cellZScope (Fig. 3D). Within 3 hours, the endothelial monolayer was restored, whereas the co-culture system remodeled the barrier integrity after 15 hours. This time discrepancy may be explained based on the fact that the methods are not directly comparable to each other. In the ECIS system, the output values are normalized before the resistance values can be regarded as absolute, whereas the measured resistance in cellZscope is directly attributable to the cell layers. Also, in the ECIS measurements there was only one layer of cells while a co-culture was constructed for the cellZScope. Thus, both cell lines had to recover for the barrier to be intact. Nevertheless, the results from both systems indicate that the BBB is restored after being transiently exposed to K16ApoE. In addition to its added mechanistic information, these experiments also validate the peptide action in a 3D human system.

In the in vivo treatment study summarized in Fig. 4, 200 μg K16ApoE was administered intravenously into each mouse, and with a blood volume of approximately 1.5 mL, this corresponds to a peptide concentration of around 133 μg/mL blood. No side effects were observed after these injections, presumably due to the quick clearance from the blood, which occurred mainly through the liver, spleen, and kidneys. Meng and colleagues argued that there was a positive correlation between toxic effects of the peptide at higher concentrations and BBB permeabilization. They also indicated that interactions of the peptide with adjacent erythrocytes resulted in the formation of microthrombi, which could be the underlying mechanism of toxicity (37). Taken together with our findings of hemolysis after treatment with K16ApoE, the injection of 200 μg K16ApoE per mouse thus represents a compromise between a favorable BBB permeable effect of the peptide and potential unwanted toxic side effects.

Our drug delivery strategy involved intravenous administration of a drug that normally is given orally to patients, thus several potential effects on the drug from the gastrointestinal tract such as degrading enzymes, low pH, or heterogenous blood perfusion were not taken into consideration. We also administered 10 mg/kg to the mice twice a week, while in the clinic, patients are commonly given 2 × 150 mg dabrafenib daily, which corresponds to 4 mg/kg for a 75 kg patient. Although the concentrations are not directly comparable, our findings clearly suggest that dabrafenib represents an effective treatment for melanoma brain metastases, provided successful entry through the BBB. The amount of injected drug that penetrated the BBB and accumulated within the mouse brains was not quantified. However, a pilot MS experiment of brain tissue harvested from control mice and K16ApoE + dabrafenib treated mice was performed, showing the presence of dabrafenib within the brains after peptide administration. Dabrafenib was not found in brains from control mice. Further experiments should be carried out aiming to quantify the detected amount of drug when coinjected with K16ApoE.

The histopathologic examination performed by an experienced neuropathologist showed that no changes in organ histology could be found after administering a single high-dose (1,000 μg) of peptide to the mice, or at the end of the dabrafenib study, when the mice had been given 12 injections of 200 μg K16ApoE over 6 weeks.

As a final experiment, we used dynamic PET/CT to study whether the BBB was permeable, following K16ApoE exposure, to even larger molecules. We detected ¹⁸F-labeled albumin (M_W ∼67 kDa) and IgG (M_W ∼150 kDa) in brain tissue. Although drug uptake across the BBB or BTB is not only limited to size, our results indicate that K16ApoE can facilitate the delivery of substances in the size range of immune checkpoint inhibitors to patients with brain metastases. Examples include ipilimumab (M_W ∼148 kDa), which targets cytotoxic T-lymphocyte antigen 4 (CTLA4) or inhibitors of PD-1/PD-L1 such as for instance nivolumab (M_W ∼143 kDa) and atezolizumab (M_W ∼145 kDa), respectively. Thus, for future investigations, we aspire to carry out in vivo treatment experiments with immune checkpoint inhibitors on melanoma brain metastases.

In conclusion, the use of K16ApoE seems to be a promising strategy to improve drug delivery across the BBB. Potential toxicity issues preclude direct translation into the clinic as of today, warranting further studies with the peptide, which is also the case with several other methods of permeabilizing the BBB. Our strategy using K16ApoE serves as an easy, noninvasive, and reliable tool to establish treatment effects in vivo with agents that otherwise do not penetrate the BBB.

H. Baghirov reports receiving a commercial research grant from and is a Roche Postdoctoral Research Fellow at Hoffmann-La Roche Ltd. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S.N. Aasen, G.J. Pilkington, T. Sundstrøm, R. Bjerkvig, F. Thorsen

Development of methodology: S.N. Aasen, H. Espedal, O. Tenstad, Z. Maherally, A.V. Eikeland, H. Baghirov, D.E. Olberg, R.B. Jenkins, R. Bjerkvig, F. Thorsen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.N. Aasen, H. Espedal, C.F. Holte, O. Keunen, T.V. Karlsen, O. Tenstad, Z. Maherally, A.V. Eikeland, H. Baghirov, D.E. Olberg, R.B. Jenkins, F. Thorsen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.N. Aasen, H. Espedal, C.F. Holte, O. Keunen, O. Tenstad, Z. Maherally, H. Miletic, A.V. Eikeland, H. Baghirov, G.J. Pilkington, T. Sundstrøm, R. Bjerkvig, F. Thorsen

Writing, review, and/or revision of the manuscript: S.N. Aasen, H. Espedal, O. Keunen, Z. Maherally, G.J. Pilkington, G. Sarkar, R.B. Jenkins, T. Sundstrøm, R. Bjerkvig

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z. Maherally, T. Hoang, D.E. Olberg, G.J. Pilkington, G. Sarkar, R.B. Jenkins, R. Bjerkvig

Study supervision: T. Sundstrøm, R. Bjerkvig

Other (provided data): T. Hoang

Other (peptide synthesis): D.E. Olberg

Other (developed and supplied the transporter peptide used in the study): G. Sarkar

The authors thank Hege Avsnes Dale (Molecular Imaging Center, University of Bergen) for valuable assistance with confocal and light microscopy, Anne Karin Nyhaug (Molecular Imaging Center, University of Bergen) for help with tissue sample preparation before electron microscopy, Linda Sandven (Molecular Imaging Center, University of Bergen) for H&E staining, Miro Eigenmann (Institute of Biomedicine, University of Bergen) for input on the development of the biodistribution methodology, Mari-Ann Jørstad Davidsen (Department of Clinical Medicine, University of Bergen) for help with animal procedures, Brith Bergum (Department of Clinical Science, University of Bergen) for assistance with flow cytometry, Jubayer Hossain (The Department of Biomedicine, University of Bergen) for assistance with animal work, Marjo Yliperttula (University of Helsinki) for providing us MDCK II cells, Tilo Wolf Eichler (Department of Clinical Medicine, University of Bergen) for providing us MDCK cells, Michael Aschner (Vanderbilt University) for the RBE4 cells, Tom Christian Holm Adamsen (Department of Chemistry, University of Bergen) for valuable input on PET/CT methodology, and Tina Pavlin (Molecular Imaging Center, University of Bergen) for help with the MRI. The electron microscopy, confocal, and small animal imaging was performed at the Molecular Imaging Center, Department of Biomedicine, University of Bergen. The study was funded by the Western Norway Regional Health Authority (to F. Thorsen), the Kristian Gerhard Jebsen Foundation (to R. Bjerkvig), the Norwegian Cancer Society (to F. Thorsen), Animal Free Research UK (to G. Pilkington), and Brain Tumor Research (to G. Pilkington).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.