On the shuttling across the blood-brain barrier via tubule formation: Mechanism and cargo avidity bias

Tubule formation acts as cargo transport across the blood-brain barrier.

Once the FA2 compound was obtained, PtA2 was then synthesised. Briefly, FA2 (0.53 g, 1.3 mmol) and potassium tetrachloroplatinate(II) (0.50 g, 1.2 mmol) were mixed in acetic acid (150 mL). The liquid suspension was heated to 400 K for 72 hours under nitrogen atmosphere. A yellow-green residue was obtained after filtration. The intermediate, without further purification, was dissolved in dimethyl sulfoxide (5 mL) and refluxed for 30 minutes. Then, the reaction solution was treated with deionised water (50 mL). A yellow powder was obtained after filtration and dried under vaccum conditions. The purified product was obtained by column chromatography (neutral alumina, hexane:ethyl acetate)c = 5:1) as a yellow powder. Yield was found to be 53%. Preparation of PtA2-loaded POs PtA2 (1 mg) and copolymer (10 mg) were individually dissolved into a mixture of methanol and dichloromethane (5 mL) in separate vials. Both solutions were then mixed uniformly into a 10 mL flat-bottom glass bottle, and rotaevaporated at 40 ∘ C to obtain a thin film. The resulting film was hydrated with PBS (5 mL) and left to stir at room temperature for at least 2 weeks. Morphology and PtA2 encapsulation were evaluated by TEM and fluorescence spectrophotometry, respectively.
Transcytosis model Nanoparticles were randomly seeded within the aqueous phase in the starting states. Nanoparticles in the aqueous phase moved according to Brownian motion with a time step of 0.00005 seconds. Dynamic viscosity was given as 0.00078 Pa −1 , which corresponds to the viscosity of the cell medium DMEM at a standard temperature of 37 ∘ (310 K) [51]. The boundaries of the transwell, the top of the aqueous phase and the edge of the cell layer were treated as reflective boundaries. The characteristic length for LRP1-angiopep-2 was estimated to be 6.5 nm based on the molecular weight of LRP1 (the extracellular chains that forms part of the LRP1 heterodimer) using the methods of Erickson [52,53]. Nanoparticles in close enough proximity to the top of the cell layer are able to form bonds with the cell. Temporary ligand agents were created randomly outside of these nanoparticles to the required density as this method would be computationally less expensive than adding rotational diffusion and updating the positions of the ligands each iteration. Moreover, for the time step used for receptor binding (0.01 seconds) and the radius of the nanoparticles, the contact surface is effectively randomised each iteration by rotational diffusion so that recreation of the ligands would not adversely effect the robustness of the model compared to incorporating full rotational diffusion. Therefore, this method was adapted to give a simulation of binding and unbinding across discrete steps. Decuzzi and Ferrari investigated the binding of nanoparticles in a static linear flow, they gave the probability of a nanoparticle adhering as: is the receptor density on the cell surface, 0 is the association constant at zero load per receptor-ligand pair and is the contact area between nanoparticle and cell surface.
The interfacial contact surface area, the receptor density and the ligand density dictate the number of bonds between the particle and a cell. The target cell determines the receptor density whilst ligand density is an adaptable property of the nanoparticles. The interfacial surface area is the surface area of the nanoparticle that is within a set binding distance from the cell. The characteristic length of the receptor-ligand bond. We then build an agent-based model of nanoparticle-cell binding unit based on the transwell in vitro model of the BBB to investigate how adapting the ligand density and nanoparticle size can alter transcytosis efficiency (Fig.S5).

Molecular dynamics simulations
A meshless solvent-free coarse-grained (CG) membrane model was coupled with molecular dynamics (MD) simulations to capture membrane topological changes and nanoparticle aggregation dynamics at the required length and time scales. In the model, the membrane was discretised into beads that each represent a CG membrane surface patch (Fig.S5). Different beads were used to represent a membrane surface patch with no receptors ('inert' bead) and a membrane surface patch with receptors ('receptor' bead). The beads self-assembled into a membrane and replicated biologically relevant properties using a soft-core pairwise inter-particle potential [53]. An equilibrated spherical membrane of 20162 membrane beads of diameter 1 (where is the MD unit of length) was used in the simulations. Typically, nanoparticles were represented as beads of diameter 4 in the model. 105 nanoparticles were randomly distributed in a spherical shell surrounding the membrane with a radius + 6 , where is the undeformed average radius of the membrane. A minimum distance of 5 between the centers of the nanoparticles was enforced in this initial distribution. The repulsive branch of a 12-6 Lennard-Jones potential was used for the volume exclusion of the nanoparticles. The nanoparticle-receptor affinity was modelled using the attractive branch of a 12-6 Lennard-Jones potential. This attractive potential between the nanoparticles and 'receptor' membrane beads was cutoff at = 3.75 . Therefore for each < , the 12-6 Lennard-Jones potential between nanoparticle and receptors is: ( where is the depth of the potential well, 0 the distance at which the potential is zero, and the distance between particles. For the membrane mechanics, we adopted a meshless model. Meshless models can capture membrane topological changes and time dynamics naturally when coupled with MD simulations. However, a careful choice of potential between membrane beads is essential to ensuring a faithful replication of biologically relevant membrane properties. We employed the soft-core pairwise inter-particle potential developed by Yuan, et. al. [54]. A 4-2 Lennard-Jones (LJ) type potential was used for the repulsive branch of the potential that ensured particle volume exclusion and a cosine function potential was used for the attractive branch of the potential for driving membrane self-assembly. The potential was used for interactions between and among membrane and receptor beads. The molecular dynamics simulations were carried out in the NVE ensemble, where is the total number of system particles, is the volume of the simulation box and is the total energy of the system. A Langevin thermostat was applied to the system components to model interactions with an implicit background solvent. The simulations were typically carried out for 2,000,000 time steps, with a time step of 0.01ŁŁ (where ŁŁ is the MD unit of time). The simulations were implemented with the LAMMPS package [55].
Immuno-fluorescence Polarised bEnd3 were washed twice with PBS, fixed in 4% (w/v) paraformaldehyde (PFA) for 15 minutes, permeabilised with 0.1% (w/v) Triton X-100 in PBS for 10 minutes and incubated with 5% (w/v) BSA in PBS for 1 hour at room temperature. Afterwards, cell monolayers were incubated with primary antibodies diluted in 1% (w/v) BSA and 0.01% (w/v) Triton X-100 in PBS overnight at 4 ºC, followed by washing with PBS and incubation with the corresponding secondary antibodies for 2 hours at room temperature. Nuclei was counterstained by incubation with DAPI for 10 minutes. Transwell membranes were excised using a scalpel and mounted on coverslips with Vectashield Mounting Media. Antibodies used in our studies are listed on Table S1.
Competition assay For the compeptition assay, FITC-angiopep-2 was dissolved in ultrapure water at 1.75 pM, which is the concentration equivalent to the ligand functionalisation inA 22 -P. Angiopep-2 or Cy5-labelled A 22 -P were added separately or together to the apical side of the transwell and cells were incubated at 37 ∘ C. After 10 and 60 minutes, cells were washed twice with PBS and fixed in 4% PFA. Transwell membranes were excised with a scapel and mounted on coverslips using VectaShield Mounting Medium with DAPI for confocal imaging. Quantification of intracellular fluorescence of Cy5-labelled A 22 -P or FITC was performed via ImageJ normalising fluorescence to the number of cell nuclei (DAPI) (Fig.S6).

Small molecule pharmacological inhibitors Polarised bEnd3 cells were incubated with CellMask
Deep Red for 10 minutes at 37 ∘ C and then rinsed with PBS. Dynasore (40 M) was added to the cell media and pre-incubated for 10 minutes at 37 ∘ C, followed by the addition of A 22 -P (100 g mL −1 ) for 60 minutes. After incubation, cells were washed and media was replaced with serum free FluoroBrite DMEM. For the dynasore recovery experiments, cells were washed with PBS at pH5 before the addition of the imaging media. In the inhibition experiments with N-ethylmaleimide (NEM), cells were incubated with NEM (0.5 mM) for 5 minutes before the addition of the A 22 -P (100 g mL −1 ) for 60 minutes. Live-cell imaging was performed using Leica TCS SP8 confocal microscope (Fig.S6).

Membrane cholesterol depletion
Prior to the cholesterol depletion assay, media of polarised bEnd3 was changed to serum-free DMEM. Methyl--cyclodextrin (CD, 10 mM) was added either to the apical or basal side of the transwell and cells were incubated for 15 minutes at 37 ∘ C. Apical and basal media was collected and used for the quantification of free cholesterol. A 22 -P (100 g mL −1 ) were added to the apical side of the transwell and incubated for 60 minutes at 37 ∘ C. Afterwards, cells were washed twice with PBS and fixed in 4% PFA for 15 minutes, followed by incubation with 0.3% (w/v) Triton X-100 and 10% BSA in PBS for 30 minutes. Cells were incubated with anti-caveolin-1 antibody for 2 hours at room temperature followed by incubation with an appropriate secundary antibody. Cholesterol quantification in the media was performed by using a colorimetric cholesterol quantification kit according to the supplier's instructions (Fig.S7).
Lentiviral shRNA silencing Short hairpin RNA (shRNA) lentiviral particles for syndapin-2 were used according to supplier's instructions. Briefly, bEnd3 cells were seeded on a 6-well plate at a density of 100,000 cells per well and grown overnight. At 50% of confluence, cells were treated with the shRNA lentiviral particles in DMEM supplemented with polybrene (5 g mL −1 ) and then, incubated overnight. On the next day, the media was replaced, and cells were further incubated for 2 days. Stable clones expressing shRNA were selected by puromycin (5 g mL −1 ) and silencing of syndapin-2 was then confirmed by Western blot. Control shRNA lentiviral particles were used as a negative control (Fig.S8).
Transmission electron microscopy of brain tissue Brains from mice treated with PtA2-loaded POs were cut into 1x1x1 mm size cubes. Brain tissue was fixed in fresh 3% (v/v) glutaradehyde in phosphate buffer propylene/araldite resin leaving the samples with the mixture overnight at room temperature. Afterwards, speciments were moved into Araldite resin for 6-8 hours at room temperature with a change of resin after 3-4 hours. Finally, specimens were embedded in fresh Araldite resin for 48-72 hours at 60 ∘ C. Semi-thin sections of approximately 0.5 m were cut on a Leica Ultramicrotome and stained with 1% Toluidine blue in Borax. Ultra-thin sections of 70-90 nm tick were cut on a Leica Ultramicrotome and stained for 25 minutes with saturated aqueous uranyl acetate followed by staining with Reynold's lead citrate for 5 minutes. Sections were examined using FEI Tecnai TEM at an accelerating voltage of 80kVv. Electron micrographs were taken using a Gatan digital camera (Fig.S9).       Quantification of Cy5-labelled A 22 -P fluorescence intensity across confocal z-stack images of endothelial cells before and after pre-treatment with dynasore for 10 minutes. Dynasore recovery condition represents the cells that were washed with PBS after the incubation with dynasore. In the graph, zero represents the beginning of the pores, negative values above the transwell membrane and positive values within the transwell membrane. Data is represented as mean ± SD ( = 3) (g).   Movie S1: 3D rendering as function of time reconstructed from confocal laser scanning micrographs over 40 minutes of brain endothelial cells exposed to LRP1 targeting A 22 -POs (red). The cell nucleus is stained with DAPI (blue) and the cell membrane with CellMask (cyan).
Movie S2: Fast (50fps) 3D rendering as function of time reconstructed from confocal laser scanning micrographs collected over 5 minutes of brain endothelial cells exposed to LRP1 targeting A 22 -POs (red). The cell membrane is stained with CellMask (green).
Movie S3: Top view of a 3D rendering as function of time reconstructed from fast Stimulated Emission Depletion (STED) microscopy micrographs collected over 6 minutes of a section of a single brain endothelial cell exposed to LRP1 targeting A 22 -POs. The structure emerging from the interaction between the POs are coloured according to their depth expressed in micrometers.
Movie S4: Side view of a 3D rendering as function of time reconstructed from fast Stimulated Emission Depletion (STED) microscopy micrographs collected over 6 minutes of a section of a single brain endothelial cell exposed to LRP1 targeting A 22 -POs. The structure emerging from the interaction between the POs are coloured according to their depth expressed in micrometers.
Movie S5: Bottom view of a 3D rendering as function of time reconstructed from fast Stimulated Emission Depletion (STED) microscopy micrographs collected over 6 minutes of a section of a single brain endothelial cell exposed to LRP1 targeting A22-POs polymersomes. The structure emerging from the interaction between the POs are coloured according to their depth expressed in micrometers.