Intrathecal delivery of nanoparticle PARP inhibitor to the cerebrospinal fluid for treatment of metastatic medulloblastoma

Engineering NPs for optimal CNS retention

To produce PLA-HPG NPs, we used a single emulsion solvent evaporation technique that has been previously described (17); here, we modified the technique to allow us to monitor their location in vivo by fluorescence microscopy. Particle size and surface charge was determined for a range of particle compositions (Fig. 1, A–C). To synthesize covalently tethered fluorescent dye NPs for effective tracking by microscopy, NPs were produced with blends of PLA-HPG copolymer and a 10% Cy5-PLA conjugate. Unlike dyes, which may leak from NPs and complicate interpretation of measurements, Cy5-PLA conjugation ensures that fluorescent signal originates from the NPs themselves (20). We hypothesized that PLA-HPG NPs would similarly exhibit clear advantages in the CSF environment after direct CSF injection through the cisterna magna (CM). We administered Cy5-PLA-HPG NPs (Cy5-NPs) into the CM of healthy mice and prepared frozen brain and spinal cord sections 24 h and 48 h post-injection. At 48 h post injection, we detected substantial accumulation of Cy5-NPs in the leptomeninges and perivascular spaces, without parenchymal uptake in all coronal sections of the brain. We also observed an even deposition of Cy5-NPs surrounding the outer layer of the spinal cord (Fig. 1, D–E).

Prior reports suggest that the diameter of the administered vehicle is an important element affecting CSF half-life (21, 22), so we assessed CNS retention of three distinct sizes of PLA-HPG NPs: 90 nm, 150 nm, and 210 nm (Table S1). Six hours after intra-CSF delivery by injection into the CM, mouse brain and spinal cord sections were imaged. FITC-conjugated dextran (either M_W= 3- or 3,000 kDa) served as size-fractionated controls (diameter estimated to be <4 nm and >60 nm respectively) (23). Although all mice administered Cy5-NPs showed retention in the brain and spinal cord leptomeninges, the 90 nm NPs covered the greatest percentage of area in fresh frozen coronal brain sections as quantified by fluorescent signal (P < 0.005 when compared to 150 nm, P <0.0001 when compared to 210 nm) (Fig. S1). In comparison, although a strong fluorescent signal was detected for both dextrans at 30 min after injection, no signal was detected in any brain or spinal cord sections of mice that were administered the 3 kDa and 3,000 kDa dextran at 6 hours post-injection (Fig. S2).

Next, we compared the effect of surface charge and surface chemistry of the PLA-HPG NPs. Previously, we reported a method for converting the HPG on the NP surface into an aldehyderich corona with enhanced bioadhesive properties (termed aldehyde-NPs) (17, 24, 25). Previous studies showed that the conversion from NP to aldehyde-NP improved plasma half-life and organ accumulation after intravenous (i.v.) or intraperitoneal (i.p.) administration (18, 24), and an impact on local retention after topical or subcutaneous administration (25–27); this conversion had a negligible effect on CSF distribution in the leptomeningeal region (Fig. 1, F–G). There were no significant differences in CNS retention between Cy5 conjugates of the aldehyde-NPs and NPs over 7 days of ex vivo whole-body imaging with the Xenogen in vivo imaging system (IVIS). Both formulations of NPs were well retained in tumor-free brain and spinal cord of healthy mice (BALB/C) at all timepoints examined with no significant difference in fluorescent radiance flux as measured in vivo and ex vivo (Fig. 1, H). In case there could be differences in accumulation along lymphatic clearance pathways, we assessed the percentage of Cy5-NP positive cells and percentage of Cy5-NP-aldehyde positive cells in the mandibular and deep cervical lymph nodes (mLN and dCLN respectively) of tumor-free mice (BALB/C) and tumor-bearing mice (human MB cell xenografts into J:Nu homozygous Foxn1/F mice). Again, there were no significant differences in the accumulation of NPs in immune cells (CD45+). We looked closer at mononuclear phagocytes, specifically macrophages (CD11b+ CD11c−) and dendritic cells (CD11b− CD11c+ and CD11b+ CD11c+) of the lymph nodes. When considering the effect of surface chemistry on uptake by comparing aldehyde-NPs and NPs, there were no observable differences in NP accumulation within these cell populations; however, we did note more accumulation of the aldehyde-NPs with statistical significance in both dendritic cell subpopulations in mLN in WT mice (P <0.05 for CD11b− CD11c+, P <0.01 for CD11b+CD11c+) and in CD11b+ CD11c+ subpopulation in tumor-bearing mice (P <0.05) (Fig. S3).

NPs with the most optimal characteristics for CNS retention exhibited a spherical morphology under electron microscope analysis (Fig. S4), with an average hydrodynamic diameter of 90–100 nm (Fig. 1, B). This is consistent with earlier findings on NP delivery, which indicates that smaller particles are less readily cleared by NP-macrophage interactions (28–30). As such, we opted to use unmodified NPs for drug encapsulation; their zeta potential averaged around −10 mV (Fig. 1, A–C).

NPs show higher CNS retention compared to small molecules

As microscopy studies suggest that PLA-HPG NPs have potential as carriers for the delivery of drugs by IT administration, we sought to clarify the whole-body biodistribution kinetics of the NPs in the CSF space and in vital organs. To image our NPs with PET, we developed a radiolabeled NP PET probe with analogous properties. We chose positron-emitting zirconium-89 (⁸⁹Zr) to label PLA-HPG NPs, due to its stability and long half-life (t_1/2 = 78.4 h), and availability of a well-characterized chelator, deferoxamine (DFO) for ⁸⁹Zr (31). We first functionalized the NP surface with DFO-mesylate through a Schiff-base reaction, and detected no change in hydrodynamic diameter, but did observe a small change in zeta-potential from −10 mV to −5 mV (Fig. 1, C). The DFO-conjugated NPs were subsequently used to chelate ⁸⁹Zr (Fig. S5).

We measured the quantitative biodistribution of our [⁸⁹Zr]Zr-DFO-NPs into all major organs over time by PET imaging in mice. We first performed a 2 h dynamic scan, to determine the distribution and timing of NP after IT administration, in BALB/C mice without tumors, using free [⁸⁹Zr]Zr-DFO as a control. Within 5 min of injection, indicated as the 0 h time point, we observed NPs distributing from the site of injection (CM) into the brain and cervical regions of the spinal cord (Fig. 2, A–B). The amount of [⁸⁹Zr]Zr-DFO-NP recorded in the CNS dropped slightly and then remained constant for the next 2 h, with limited distribution of signal to the systemic circulation or other organs. In contrast, [⁸⁹Zr]Zr-DFO distributed immediately throughout the subarachnoid space, similar to the NPs, but then distributed from the CNS into systemic circulation, with less than 30% ID/g detectable in the CNS by 2 h post injection of the tracer. We determined that the [⁸⁹Zr]Zr-DFO had a CNS half-life of ~60 min, which is comparable to known half-lives of small molecules following IT delivery (32, 33). In comparison, the amount of [⁸⁹Zr]Zr-DFO-NPs were relatively stable in the brain and spinal cord during the first 2 h of injection. Additionally, [⁸⁹Zr]Zr-DFO enters the interstitial parenchyma to a greater extent than [⁸⁹Zr]Zr-DFO-NPs (Fig. 2, C). We think this is related to the size difference between the two materials (100 nm vs < 1 nm). We note that we observed an accumulation of [⁸⁹Zr]Zr-DFO-NP in the bladder as early at 5 min post injection (Fig. 2, D), which is addressed in Supplementary Material (Fig. S6).

Next, we administered [⁸⁹Zr]Zr-DFO-NPs (7.4 × 10¹⁰ Bq/g) IT to animals without tumors (BALB/C mice) and performed static PET/CT imaging at predetermined time points up to 12 days post injection (n=4) (Fig. 2, D–E). At 3 h, we found that at least 60% of the signal was in the brain, and at least 20% of the signal was in the spinal cord (>80% in the CNS). Activity in the brain gradually decreased over the first 4 d, eventually reaching 30%. This degree of activity was observed in the brain for at least 12 d, our last day of imaging. In the spinal region, the decline in activity was slower over the first 4 d, with similar activity observed at these time points, about 25%. On day 7 and 12, there was a decrease in signal to ~20% and ~10%. Ex vivo gamma counting after 12 d confirmed the in vivo imaging findings, with the brain exhibiting the highest Bq/g signal (44% of total injected dose per gram of tissue, or ID/g), then the liver (16%), spleen (34%), and cervical lymph nodes (3%) (Fig. S7).

NPs show higher CNS retention in xenograft tumor models

We next analyzed the CNS retention of [⁸⁹Zr]Zr-DFO-NPs in tumor-bearing mice (J:Nu homozygous Foxn1/F). We developed a mouse model with leptomeningeal metastatic MB by IT injection of DAOY cells stably expressing luciferase into the CM. By day 14, IVIS showed tumor growth throughout the CNS, including widely disseminated tumors in both the cerebellum and spinal cord (Fig. S8, A). To establish that the presence of tumors did not result in BBB breakdown, we injected mice with Gd-DTPA MR contrast agent intravenously and imaged with MRI to confirm no BBB penetration occurred as compared to control (Fig. S8, B). Two mice were injected intravenously with Gd-DTPA and imaged by MRI after 20 minutes. There was no detectable Gd-DTPA signal in the brain, confirming an intact BBB. However, we observed acute hydrocephalus in vivo in both mice, most likely due to tumor obstruction of CSF flow. H&E images of the brain ex vivo revealed a high tumor burden in the brain cerebellum, ventricles, and in the spinal leptomeninges (Fig. S8, C–D).

We evaluated the biodistribution and uptake of [⁸⁹Zr]Zr-DFO-NPs in tumors and normal tissue throughout a 21-day period using PET/CT imaging in two cohorts of mice (n=6 total) (Fig. 3, A–G). The first cohort (n=2) was scanned continuously for 120 min after administration. Within the first 5 min, 30% of the total injected dose was detected in the tumor, which gradually increased to 40% over the next 2 h. The next cohort (n=4) were scanned at predefined intervals (6 h, 24 h, 4 d, 7 d, 12 d, 21 d). We saw greater uptake in the tumor than in any other normal tissue, with 50% of total injected dose identified in the tumor site within the cerebellum (roughly 3.5 times higher % ID/g compared to the brain at day 21). Over the next 21 days, uptake in the brain and spinal cord remained stable at roughly 15% and 10% of injected dose, respectively. We observed greater accumulation in peripheral organs with the tumor bearing mice compared to healthy mice, specifically seeing accumulation in the cervical lymph nodes, peripheral lymph nodes, liver, and spleen. At no point did any of these peripheral organ accumulation exceed 20% of overall activity. In general, tumor bearing mice showed longer retention of NPs in the CNS, most likely due to uptake of NPs by tumor cells. At day 21, 40% of total ID/g remained detectable in the tumor site, which could be visualized using maximum intensity projections of the PET/CT image, which is a summed representation of radioactivity in the whole body (Fig. 3, G).

We also identified NP accumulation at the cellular level by IT administration of Cy5-NPs in tumor-bearing mice and observing brain sections by microscopy (Fig. 3, H). We detected dense Cy5-NP accumulation and uptake in tumor foci at 24 h after injection; NP retention in this area of the cerebellum was not observed in healthy mice (Fig. 3, I). We also saw a lower density of NP accumulation in the leptomeninges of the brain and spinal cord of tumor-bearing mice than in healthy animals, implying that the NPs circulate through the perivascular space in a similar manner, but accumulate at significantly higher density in sites with tumor foci than in healthy brain or spinal tissue (P <0.001).

The presence of Cy5-NPs in areas of parenchymal tumor, and the accumulation of NPs in the deep and superficial cervical LNs as detected by PET raised the possibility of several transport pathways of the NPs into the tumor microenvironment and to the cervical LNs. We first examined whether NP accumulation at tumor sites was aided by the presence of immune cells. We injected Cy5-NPs into mice with tumors, and after 48 h, collected and sectioned the head before staining for F4/80 (a marker for macrophages) and Iba1 (a marker for microglia). At 48 h, we could detect the presence of activated microglia (red) and tumor associated macrophages (green) within the tumor bulk (Fig. 4, A–B) and detected colocalization of NPs with both cell types. In WT mice without tumors, we saw no significant NP accumulation in the brain parenchyma, and only in CSF-bathed regions such as the choroid plexus, without any microglia or macrophage association (Fig. 4, C). These results suggest that NP penetration and transport into the tumor bulk is aided by tumor-associated immune cells. We also observed a portion of NPs in the tumor microenvironment that were not associated with either macrophages or microglia.

We sought to examine the drainage pathway by which NPs are cleared from the CNS. To accomplish this, the meninges of the tumor-bearing mice were isolated at 48 h after i.c.m. injection of Cy5-NPs, and stained with the endothelial cell marker CD31, lymphatic vessel marker LYVE-1, and immune cell marker CD45. NPs accumulated throughout the entire meninges and particularly along the venous sinuses, especially the transverse sinus (Fig. 4, D). Along the transverse venous sinuses, NPs co-localized with lymphatic vessels (LYVE-1⁺CD31⁺) although other clusters of NPs were found not associated with meningeal lymphatic vessels or blood vessels (Fig. 4, E). We observed that NP accumulation frequently co-localized with macrophages (LYVE-1⁺ CD45⁺)(Fig. 4, F–G).

Polymeric encapsulation of BMN-673 alters the toxicity profile of drug in animal studies

As a proof-of-concept, we tested whether we could improve the pharmacokinetic (PK) properties and activity of PARPi. As our model drug, we chose BMN-673 for its potent PARP trapping properties and toxicity at very low doses. BMN-673-loading in NPs to obtain BMN-673encapsulated NPs ((BMN)NPs) varied from 1% to 5% (w/w) depending on the solvent ratios and drug to polymer ratios used during NP preparation (Fig. S9, A–B). We selected the NPs without the aldehyde-modified surface due to their similar persistence in the brain (Fig. 1, G–H). The BMN-673 release rate from these NPs was similar in both artificial CSF (aCSF) and PBS at 37 °C, averaging 60% release over 3 days (Fig. S9, C–D) in a bi-phasic manner, similar to the kinetics identified in prior studies (34–36). The relative cytotoxic activity of free BMN-673 and (BMN)NPs were determined on 3 MB cell lines: DAOY, D341, and D283. Although both agents were cytotoxic at a range of 10 nM to 1 μM, the NPs were more potent with a lower IC50 value than that of free BMN-673 (Fig. S10).

To characterize the safety of (BMN)NPs in vivo, we conducted toxicology studies in healthy female nude mice. We first determined the maximum tolerated dose (MTD) for single IT dosing of (BMN)NPs or free BMN-673 in nude mice (Fig. 5, A–B). Increasing doses of either (BMN)NP or free BMN-673 were administered to animals; body weight loss and overall animal health were closely monitored. MTD values for (BMN)NPs were tenfold higher than that of free BMN-673. The median lethal dose for IT administration of free BMN-673 was 0.125 mg/kg. Animals treated with greater than 0.06 mg/kg developed a variety of toxicity-related symptoms, including lethargy, labored breathing and, in some cases, death. The MTD (with a single dose) was determined to be 0.05 mg/kg. A slightly lower dose, 0.03 mg/kg twice a week, was tolerated with less than 10% body weight loss. By comparison, (BMN)NPs were well tolerated at all doses tested at or under 0.5 mg/kg, which was the maximum dose allowed in one infusion due to IT volume dosing limits. To achieve IT doses of more than 0.5 mg/kg, the mice were dosed multiple times in the same day, within 3 h, and the lethal dose of 1.25 mg/kg was determined. At 1.25 mg/kg, we noticed a delay in the onset of acute toxicity symptoms, which was presumably due to delayed drug release from the NPs.

To further test systemic toxicity differences between (BMN)NPs and free BMN-673, we monitored complete blood count parameters, which measure red blood cell (RBC), white blood cell (WBC), and platelet (PLT), in mice at day 3 and day 7 following two bi-weekly administrations at MTD (Fig. 5, C). Mice treated with free BMN-673 at 0.05 mg/kg/week had progressive leukopenia and thrombocytopenia at day 3 which did not improve appreciably at day 7. NPs dosed tenfold higher, at 0.5 mg/kg/week, maintained considerably more normal WBC, RBC, and PLT counts over the course of treatment, and all cell counts except for eosinophils returned to normal ranges by day 7.

BMN-673-encapsulated NPs show superior activity compared to free BMN-673 in xenograft tumor model

We sought to translate the improved therapeutic index of (BMN)NPs over free BMN-673 to improve effectiveness in J:Nu tumor xenograft models. For in vivo studies, we performed intra-cisternal transplantation of DAOY cells that stably express luciferase. A surgical catheter was implanted into their CM and used for both cellular implantation and IT dosage during treatment. Mice were treated when their tumoral luminescence burden was detectable at 10⁵ BLI (p/sec/cm²/sr), 7 days post-implantation. Mice were treated once, at the same dose of 0.1 mg/kg with either (BMN)NPs or free BMN-673 (Fig. 6A). (BMN)NP-treated tumors grew at a slower rate than free BMN-673-treated tumors. Tumor BLI signal was most reduced at a week after the treatment, which delayed tumor growth in subsequent weeks in contrast to the free BMN-673 group (Fig. 6B and C). Even though the dose was more than the MTD for free BMN-673, a tumor reduction benefit was observed in only 20% of mice. Consistent with the IVIS findings, mice treated with (BMN)NPs lived significantly longer than those treated with free drug alone (P <0.0001), with enhanced median survival of 56 days (Fig. 6D). Furthermore, the free BMN-673-treated mice lost more weight than the (BMN)NP group (around 20% vs. non-notable) (Fig. 6E).

We repeated this study using the same mouse model, but with twice-weekly dosing of 0.25 mg/kg of (BMN)NPs or 0.03 mg/kg for free BMN-673 (Fig. 7A), which is the amount of drug encapsulated in the BMN(NP) dosage. These equitoxic doses are equivalent to roughly 50% of MTD, and we observed similar extents of mean weight loss in both treatment groups (such as 3% at week 1 and 5% at week 2, Fig. 7B). In the NP-treated group, we observed tumor regression. The NP treatment had a substantial antitumoral impact in all the mice, especially one week after the doses were administered (Fig. 7C). (BMN)NPs led to no detectable tumor burden in several treated animals. In contrast, tumors continued to progress with treatment of free drug. In NP-treated mice where the tumor progressed, we still observed a median survival benefit of 5 weeks compared to the free drug-treated group, and a survival benefit of 6 weeks compared to the untreated control group (Fig. 7A). In this study, median survival in all treatment groups generally reflected the tumor growth rates evaluated by bioluminescence, suggesting that the mice succumbed to cancer rather than drug-induced side effects. In addition,(BMN)NPs, and to a lesser degree, free BMN-673 (P <0.05), significantly reduced the incidence of CNS metastases (P <0.0001) (Fig. 7D). We observed this improvement in rates of distant metastatic dissemination in the leptomeninges when measured by IVIS as well (Fig. 7E). In comparison, over 80% of untreated animals developed evident spinal cord dissemination and required euthanasia around week 4, due to severe hydrocephalus.

BMN-673-encapsulated NPs synergize with Temozolomide when given in conjunction in a xenograft model

To demonstrate sensitivity of the MB cell lines to TMZ and BMN-673 synergy, we first treated DAOY, D341, and D283 cells with free BMN-673 and (BMN)NPs and performed conventional IC₅₀ analyses. When we calculated combinatorial indices using these two drugs, cells treated with incremental BMN-673 increased in the presence of TMZ demonstrated reduced viability, as quantified with the CellTiter-Glo viability assay, compared to cells treated with BMN-673 alone. When we calculated BMN-673 and TMZ combinatorial index values using a classical Loewe synergy model, we detected high synergy. Drug interaction assays with other commonly used, brain penetrant chemotherapy agents, such as Lomustine, Topotecan, Irinotecan, Cyclophosphamide, did not show consistent synergistic action with BMN-673 across all cell lines (Fig. S11).

We verified the synergy observed in vitro using a D341 xenograft model, where the cells were inoculated into J:nu mice in a manner identical to the DAOY cell line. In the D341 model, most mice developed spinal metastasis that was detectable by IVIS imaging within 7 days of inoculation. We thus tested whether (BMN)NPs and free BMN-673 were efficacious against existing spinal metastasis by starting treatment at day 7 in the presence of leptomeningeal metastasis (Fig 8A). 4 repeat doses of (BMN)NPs (0.25 mg/kg) were efficacious in reducing cerebral and spinal tumor burden in the D341 model, whereas the free BMN-673 treatment (at an equitoxic dose to the NPs) showed no effect. We sought to improve this therapy with a temozolomide (TMZ) combination approach. The synergy between TMZ and PARPi is well documented in vitro, as PARP1 trapping sensitizes cells to the DNA methylation mechanism of TMZ (37, 38). However, in vivo, a reduction in the PARPi dose is necessary to prevent acute toxicity (39). The (BMN)NP and TMZ combination was well-tolerated in mice without a necessary dose-reduction and led to tumor clearance in 4 out of 6 mice tested (Fig 8B) with similar extents of mean weight-loss across all treatment groups (Fig 8C). We were unable to test the synergy between free BMN-673 and TMZ, as the combination was not tolerated at even the lowest doses of BMN-673 (0.03 mg/kg). Overall, these results imply that intra-CSF delivery of nanoparticle encapsulated drugs is a viable treatment regimen for MB with a meaningful advantage over intrathecal administration of free drug alone in terms of pronounced and sustained in vivo activity.

Study design

The aim of this study was to develop nanoparticles that encapsulate PARPi and to evaluate their efficacy in preclinical models of medulloblastoma. First, we characterized NPs and the effect of size, charge, surface chemistry on uptake and retention in the CNS, before evaluating biodistribution through PET imaging of healthy and tumor-bearing mice. Having determined optimal NP settings, we then determined dosage of free drug and NP to be used for preclinical studies. We then evaluated the impacts of single dosing, repeated dosing, and synergistic dosing of (BMN)NPs in our preclinical model (J:nu mice inoculated with luciferase-expressing DAOY cells) in vivo.

Replicates varied depending on experiment. Synthesized NP characterization was done in technical triplicates, and the effect of NP size, charge, and surface chemistry on uptake in the leptomeningeal and perivascular space after i.c.m. administration of dye-conjugated PLA-HPG NPs had at least biological triplicates (n=6 for 3 h, 24 h, 5d, 7 d). Evaluation through imaging of brain and spinal cord sections under fluorescence microscopy after 24 h and 48 h involved taking 10–15 representative slices per mouse (n=5). Evaluation of NPs in the CNS compartment in healthy mice by PET activity was done over 2 hours (n=2), while whole body biodistribution was observed over 12 days (n=4). These experiments were repeated in tumor-bearing mice (n=2 for CNS compartment evaluations, n=4 for whole body distribution). Cy5-NP quantification in the brain was evaluated 24 hours after i.c.m. administration between healthy and tumor-bearing mice (n=6 per group). Preliminary NP and free-drug dosage (n=1) was done before evaluating MTD for free drug and NPs at day 3 and 7 (n=6 per group). Survival with single dose was evaluated over 17 weeks (n=5 for control, n=5 for free drug, and n=6 for NPs). Changes in body weight were tracked, along with relative bioluminescence intensity of whole brain to evaluate tumor. Repeat dosing was similarly evaluated (n=9 for control, n=7 for free drug, n=8 for NP). Lastly, synergistic treatment of (BMN)NP and TMZ was evaluated (n=8 for control, n=7 for BMN free drug, n=9 for (BMN)NP, n=8 for (BMN)NP+TMZ). All animal procedures and protocols were approved by the Yale University Institutional Animal Care and Use Committee. Mice were grouped such that the average weight would be similar across the different conditions. Conditions were assigned randomly to groups. Study was not blinded.