Systemic HER3 ligand-mimicking nanobioparticles enter the brain and reduce intracranial tumour growth

Recombinant protein production

See Supplementary Methods.

Structural modelling and MD simulation

The HPK monomeric and pentameric structures were generated as previously described⁵⁰. A truncated version of the HPK monomer and pentamer was built (tHPK) to model the effect of pH on the titratable residues in the penton base core of the HPK pentamer. The protein sequence of tHPK monomer was: TGGRNSIRYSELAPLFDTTRVYLVDNKS TDVASLNYQNDHSNFLTTVIQNNDYSPGEASTQTINLDDRSHWGGDLKTILHTNMPNVNEFMFTNKFKA RVMVSRLPTKDNQVELKYEWVEFTLPEGNYSETMTIDLMNNAIVEHYLKVGRQNGVLESDIGVKFDTRNFRLGFDPVTGLVMPGVYTNEAFHPDIILLPGCGVDFTHSRLSNLLGIRKRQPFQEGFRITYDDLEGGNIPALLDVDAYQASLKDDTEQGGGGAGGSNSSGSGAEENSNAAAAAMQPVEDMNDHAIRGDTFATRAEEKRAEAEAAAEAAAPAAQPEVEKPQKKPVIKPLTEDSKKRSYNLISNDSTFTQYRSWYLAYNYGDPQTGIRSWTLLCTPDVTCGSEQVYWSLPDMMQDPVTFRSTRQISNFPVVGAELLPVHSKSFYNDQAVYSQLIRQFTSLTHVFNRFPENQILARPPAPTITTVSENVPALTDHGTLPLRNSIGGVQRVTITDARRRTCPYVYKALGIVSPRVLSSRT.

The pentameric core structure of the tHPK particle was investigated under four different pH conditions (pH 7, 5, 3, 1). The protonation states of the titratable residues were obtained from the propKa server^51,52, yielding a corresponding net charge on the tHPK bioparticle at different pH values: −60 at pH 7, −5 at pH 5, +255 at pH 3 and +300 at pH 1. The tHPK molecular structures at four pH levels (tHPK7, tHPK5, tHPK3, tHPK1) were relaxed using implicit solvent generalized Born MD simulations with the AMBER ff14SB force field⁵³ that are part of the AMBER v.18 simulation package⁵⁴. The protonation states of the titratable residues were set for the specific pH values and not allowed to change during the biophysical simulations. All bioparticle structures were relaxed using 10 ns of simulation time. The pH 5 and 7 (tHPK5/tHPK7) bioparticles were simulated for an additional phase of 50 ns because the pH 1 and 3 (tHPK1/tHPK3) pentameric bioparticles broke apart into monomers within the first phase of 10 ns.

Particle assembly

See Supplementary Methods.

Electron microscopy

The Electronic Imaging Center for Nanosystems at the University of California Los Angeles provided fixation and TEM through a core services voucher.

Electrophoretic mobility shift assay

See Supplementary Methods.

Serum digest (protection) assay

See Supplementary Methods.

Dynamic light scattering

See Supplementary Methods.

Cells

See Supplementary Methods.

Patient-derived tissue

Deidentified surgical specimens of two independent breast cancer tissues and one normal breast tissue were obtained by informed consent under protocol number 29973 which received ethical approval by the Cedars-Sinai Medical Center Institutional Review Board. Resected breast tissues were immediately placed in cold, sterile DMEM after excision and cut into 2–4 mm pieces before undergoing enzymatic and mechanical dissociation using the gentleMACS Octo Dissociator multitissue kit and protocol (Miltenyi Biotec). Resuspended cells were then promptly plated into flasks, multiwell plates or chamber slides for the indicated treatments. Human brain specimens were received from three fresh male cadaver brains, aged 68, 71 and 76 years (Tissue for Research). Samples were preserved in 10% buffered formalin.

Cell surface detection of HER3

See Supplementary Methods.

Receptor binding

See Supplementary Methods.

Intracellular trafficking

The intracellular trafficking of HPK was evaluated on HER3 + MDA-MB-435 cells (whose relatively broad cytoplasmic areas are conducive to such studies) following our previously established procedures⁵⁰, with the following modifications: 12-well plates containing 10,000 cells per well plated on coverslips were briefly prechilled and exposed to 7 µg HPK per well in Buffer A for 1 h to promote receptor binding but not internalization. Equivalent samples received 100 nM bafilomycin-A1 in Buffer A for 30 min before adding HPK. Plates were then transferred to 37 °C to promote synchronized uptake and intracellular trafficking. Cells were fixed at indicated time points after warming, processed for the immunoidentification of HPK using an antibody that recognizes the polyhistidine tag (RGS-His antibody; Qiagen 1:100), and counterstained with 4,6-diamidino-2-phenylindole (DAPI). Images were acquired using a high-throughput digital microscope (Molecular Devices ImageXpress Pico Automated Cell Imaging System) using a 40× magnification lens. Exposure times for each fluorescence wavelength remained fixed to compare between treatments and time points. Where indicated, vesicular-like sequestration of HPK was quantified by subtracting the measured integrated density (Int D) of extravesicular (e) from the vesicular (v) regions normalized by v, or (Int D_v − Int D_e)/Int D_v.

Endosome maturation staining antibodies against RAB7 and early endosome antigen 1 (EEA1) were purchased from Abcam (ab50533 1:50 and ab206860 1:100, respectively). Samples were imaged using a Leica SPE laser-scanning confocal microscope with Leica Application Suite X (LAS X) 3.3.0.16799. Acquired images were imported and separated into individual channels. Individual cells in selected channels were delineated, and pixel overlap was evaluated using ImageJ.

Subcellular fractionation

Subconfluent (70% confluency) HER3 + MDA-MB-435 tumour cells grown in complete media were rinsed with 1 × PBS, serum-starved in Buffer A (DMEM containing 20 mM HEPES, pH 7.4, 2 mM MgCl₂ and 3% BSA) for 1 h at 37 °C, rinsed with 1 × PBS, detached with 2 mM EDTA/PBS and neutralized with double the volume of 1 × PBS++. An aliquot containing 6 × 10⁶ cells was washed with PBS and resuspended in 0.7 ml Buffer A containing 5 nM indicated proteins (quantified by Bradford Assay). Cells were incubated with rocking for 1 h at 4 °C to promote receptor binding but not uptake, followed by transfer to 37 °C to promote synchronized cell uptake. At the indicated time points, cells were pelleted (10 min, 5,000 rpm, 4 °C) and washed in a mildly acidic buffer (1 ml of 1 × PBS, pH 6) for 5 min to remove the remaining cell surface protein. Cell pellets were then rinsed with 1 × PBS and processed for subcellular fractionation (Qproteome Cell Compartment Kit, Qiagen) following the manufacturer’s protocol. Indicated fractions were isolated, and protein precipitation was performed by incubation in 4 vol of ice-cold acetone for 15 min, followed by pelleting (10 min, 14,000 rpm, 4 °C), removal of the supernatant and resuspension in storage buffer (10% glycerol and 5% SDS in dH₂O). Samples were subject to reducing sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and immunoblotted using antibodies recognizing recombinant protein (Qiagen RGS·His Antibody 34610, 1:100 in 3% BSA) and corresponding fraction controls: cytosolic (GAPDH; R&D Systems, MAB5718, 1:10,000 in 3% BSA); cytoskeletal (β-actin; R&D Systems, MAB8929, 1:5,000 in 3% BSA); and membrane (TIM23; BD Biosciences, 611222, 1:5,000 in 3% BSA). Primary incubation was performed overnight at 4 °C followed by incubation with anti-rabbit or anti-mouse horseradish peroxidase-containing secondary at room temperature for 2 h (Abcam, goat anti-rabbit AB6721 1:3,000 and goat anti-mouse AB6789 1:2,000, respectively). Immunoblots were imaged using the high-sensitivity setting on a Bio-Rad chemidoc imager.

BBB chip

The organ-on-a-chip is composed of a flexible polydimethylsiloxane elastomer that contains two closely apposed and parallel microchannels (1 mm × 1 mm top channel; 1 mm × 0.2 mm bottom channel)⁵⁵ separated by a porous, flexible polydimethylsiloxane membrane (50 µm thick, with 7-μm-diameter pores, spaced 40 µm apart, resulting in 2% porosity over a surface area of 0.171 cm² separating the two channels) coated with Matrigel in the top channel, and collagen–fibronectin extracellular matrix in the bottom channel. Cell aggregates (‘EZ-spheres’)⁵⁶ were derived from human cells obtained through Institutional Review Board protocol number 21505. To generate BBB chips, EZ-spheres containing iPSC-derived neural progenitor cells were dissociated into single cells using accutase and were seeded into the top channel (‘brain channel’) at a density of 1.25 × 10⁶ cells ml⁻¹ in terminal differentiation media (TDM, containing Rock inhibitor 1:2,000 (Stemgent)). The neural progenitor cells were allowed to settle for 2 h and were then flushed with TDM without Rock inhibitor. Media was replaced with 100 µl TDM every other day. Five days later, human induced pluripotent stem cell-derived brain microvascular endothelial-like cells were seeded into the bottom channel (‘vascular channel”) at 15 × 10⁶ cells ml⁻¹ in S3 BMEC medium containing Rock inhibitor (1:2,000) and inverted for 2 h. A second seeding was performed after 2 h using the same protocol without inversion for 2 h. Following the second incubation period, the BBB chips were flushed with S3 BMEC medium without Rock inhibitor. The following day, chips were flushed with fresh TDM and S4 BMEC medium. The following day, the chips were added to Emulate pods and placed on an active flow of 30 µl h⁻¹. Chips were validated via paracellular permeability assays using dextran–fluorescein isothiocyanate overnight to confirm the barrier function of the chips under flow before release to investigators for experimental testing. Validated BBB chips were treated with NBPs at a concentration equating to 1 µg ml⁻¹ of HPK that was passed through the endothelial channel with or without 10:1 blocking peptide purchased commercially from Sinobiological (10201-H08H). After 4 h of constant flow, the chips were fixed using 4% paraformaldehyde and subjected to immunocytofluorescent staining. Chips were imaged on a Nikon A1R confocal microscope with NIS Elements v.5.42.01 and IMARIS v.10.2.0 software for image acquisition and analysis. Notably, gene silencing via CRISPR/Cas9 has been a technical and viability challenge in these cells⁵⁷ and hence the contribution of HER3 toward extravasation was examined by ligand inhibition.

Sandwich ELISA

See Supplementary Methods.

Animal subjects

Immunodeficient (NU/NU) and immunocompetent (BALB/c) mice were obtained from Charles River Laboratories. All procedures involving mice were performed following protocol numbers 6037 and 5790, which had received ethical approval by the Cedars-Sinai Institutional Animal Care and Use Committee, in accordance with the institutional and national Guide for the Care and Use of Laboratory Animals. The criteria for euthanasia included tumour ulceration, interference with ambulation and access to food and water, or BCS < 2 (emaciation, prominent skeletal structure, little/no flesh cover, visible and distinctly segmented vertebrae)^58,59,60. Blood collected when the animals were killed was processed using serum separator tubes (BD Microtainer tube with serum separator additive/gel, Becton Dickinson) following the manufacturer’s protocol and isolated sera were transferred to an external reference lab (IDEXX BioAnalytics) that provided the measurements of serum analytes. Samples were provided in a blinded/anonymous fashion (sample labelling lacked identifying information). Normal ranges for blood analytes of healthy, tumour-free female NU/NU and BALB/c mice were obtained from Charles River Laboratories^61,62.

Tumour models

Peripheral breast tumour models used for the biodistribution and therapeutic efficacy studies were established in 6-week-old female mice. For xenograft models, immunodeficient (NU/NU) mice received bilateral flank implants of JIMT-1 human HER2+ breast tumour cells (1 × 10⁷ cells per implant). For immune-competent models bearing peripheral TNBC tumours, BALB/c mice received bilateral mammary fat pad injections of 4T1LucGFP cells (1 × 10⁴ cells per injection in 0.1 ml PBS). Bioluminescences of 4T1LucGFP tumours were acquired as described later below. Primary tumour volumes (height × width × depth) were monitored approximately three times per week under single-blinded conditions (treatment groups unknown to the individual acquiring measurements). For evaluating therapeutic efficacy, mice were randomly allocated at tumour establishment (≥100–150 mm³) into separate treatment groups (n = 5 mice per group) and received the indicated therapeutic reagents or controls by tail vein injection twice per week for 4 weeks. HerDox and lipodox dosages equated to 0.2 mg kg⁻¹ based on doxorubicin content. Additional cohorts shown in the Supplementary Materials received 0.02 mg kg⁻¹ HerDox. Empty particles (lacking the doxorubicin) equated to the 0.2 mg kg⁻¹ dose. Saline (mock) treatments were administered at equivalent volumes as the experimental reagents. HerGa and S2Ga dosages equated to 0.2 mg kg⁻¹ based on gallium corrole content. Because the HPK protein mediates the receptor-targeted delivery of NBPs and targeting to tumours is largely dependent on and limited by HER3 cell surface levels, in vivo dosage concentration, dosing number and frequency were determined based on the following parameters: circulating blood concentration of drug was based on the minimally effective concentration for reducing HER3+ tumour cells (4T1 mouse TNBC) but not HER3− non-tumour cells (NIH3T3 mouse fibroblasts). From the minimally effective concentration, we could then determine the therapeutically effective ratio of drug molecules to cell number (drug to cell ratio) and extrapolate this to the estimated cell number in the tumours in vivo, using primary tumour size as an initial gauge⁶³. For tumours measured based on bioluminescence, cell numbers were estimated from a calibration curve plotted from the bioluminescence measurements of known tumour cell titrations implanted in control mice. The total desired accumulation of drug in the tumour (drug to cell ratio) informed the total number of doses to administer at the determined circulating blood concentration of drug. Frequency of dosing is based on the time course of biodistribution and tissue clearance in tumour models. These same parameters were used to design the treatment regimen in mice with IC tumours as described below.

For the implantation of IC tumours, anaesthetized 4-week-old female mice were positioned in a stereotactic frame, and a burr hole was created in the skull using a steel bit 2 mm right of the sagittal and 2 mm anterior to the lambdoid suture. A stereotactic frame was used to guide a Hamilton syringe, and 10,000 cells in 2 µl were implanted at 4 mm depth. After implantation, bone wax was used to seal the hole, and the incision was sealed with surgical staples, which were removed 5 days later. Immunodeficient NU/NU mice were used for evaluating HerDox and Lipodox (n = 12 for HerDox and Lipodox; n = 14 for mock). Treatments were delivered via the tail vein at a dose of 0.004 mg kg⁻¹ based on doxorubicin content at a regimen meeting the parameters described earlier. Immunocompetent BALB/c mice were used to evaluate HerGa and S2Ga (n = 5 per treatment group), each delivered at 0.2 mg kg⁻¹ based on gallium corrole content at a regimen meeting the parameters described earlier. Tumour growth was monitored via luciferase beginning on day 4 after implantation and randomized before treatments.

To ensure rigour, all measurements were collected in blinded fashion with cohort identities unknown to the researcher.

Bioluminescence and NIR fluorescence acquisition

Luciferase monitoring was performed every 4 days, beginning on day 4 postimplantation. Mice received 200 µl intraperitoneal of 30 mg ml⁻¹ d-Luciferin (Caliper) dissolved in Dulbecco’s PBS 10 min before imaging using an in vivo imaging system (IVIS). d-Luciferin was allowed to circulate in the animals for 15 min, followed by imaging using a PerkinElmer IVIS Lumina Spectrum. Total flux (photons s⁻¹) signals were quantified using equally sized regions of interest (ROIs) centred around the cranial region in LivingImage software v.4.8.0.

NIR image acquisition was performed on freshly excised organs at indicated time points. Average radiant efficiencies ([p/s/cm²/sr]/[µW/cm²]) were quantified using ROIs outlining individual organs in LivingImage software.

MRI

A 9.4 T MRI system (BioSpec 94/20USR, Bruker) was used for imaging tumour locations and volumes. The tumour perimeters were visually highlighted by intravenous injections of 7.5 µmol gadovist contrast agent. Mice were imaged under inhaled 1.7% isoflurane anaesthetic. Images were collected with an in-plane resolution of 70 µm using an acquisition matrix of 256 × 196 and zero filling in the phase encoding direction to 256, using a field of view of 1.80 cm × 1.80 cm. Twenty consecutive 0.7 mm slices covered the tumour-implanted region. Two averages were collected, with a repetition time of 750 ms and an echo time of 8.77 ms, for a total scan time of 4.9 min using a mouse four-channel brain array coil (T11071V3, Bruker) for reception and a whole-body transmission coil (T10325V3, Bruker) for excitation. Volume calculations for positive contrast brain regions were determined by integration over the entire tumour by a blinded analyser. For each slice containing tumour-enhanced regions, an ROI was drawn to encompass the area. The area of each slice was multiplied by the slice thickness, and the individual volumes were summed to determine the area of the tumour. Bruker Paravision 5.1 software was used for the analysis. MRI core staff performed quantifications in a blinded fashion.

Biodistribution

For the quantification of particle delivery by ICP-MS, each mouse was administered a single tail vein injection of HPK bioparticles loaded with gallium(III)-metallated corrole (HPK-S2Ga or HerGa)⁶⁴ or S2Ga alone, at 1.5 nmol S2Ga per injection. At 6 h after injection (n = 2 per treatment), mice were killed, and the major organs (brain, heart, kidneys, liver, lungs, spleen and tumours) were excised, weighed and transferred to the University of California Los Angeles ICP-MS core facility. Samples were digested overnight and processed by ICP-MS to measure the tissue content of gallium(III) metal.

Each mouse receiving NBPs (n = 5) was administered a single tail vein injection of HPK bioparticles loaded with NIR Alexa Fluor 680-labelled oligonucleotides at a dose equal to 1.50 nmol oligonucleotide per injection. Mice were monitored by epifluorescence imaging at the indicated time points after injection using an IVIS Spectrum (PerkinElmer), followed by tissue harvesting and the acquisition of average radiant efficiency ([p/s/cm²/sr]/[µW/cm²]) per tissue. Where indicated, the relative tissue content of NIR-OND was determined based on the extrapolation of measurements acquired from extracted tissue against a standard curve of known NIR-OND titrations in tissue lysates. Tumours were implanted 8 days before mice received systemic NBPs, allowing sufficient time for damaged vessels to repair, and NIR-OND alone was used to detect the possibility of vascular leakage.

Mice (n = 5 per treatment) receiving directly labelled protein or particles (NIR-labelled HPK capsomeres, Tz or BSA) were injected with the indicated treatment through a single tail vein injection at 12 nmol of labelled protein per injection. Proteins were labelled with Alexa Fluor 680 at primary amines and isolated from the unconjugated dye by size-exclusion chromatography using a commercial protein-labelling kit, following the manufacturer’s protocol (LifeTechnologies).

Immunogenicity assay

Female BALB/c mice (∼6 weeks old; Charles River) received tail vein injections of empty (no drug) NBPs at 0.5 mg kg⁻¹ HPK per injection (equating to ~0.2 mg kg⁻¹ HerDox) twice per week for four weeks. Serum isolated from blood collected at indicated time points underwent serial dilutions to 1 × 10⁻⁴ dilution and was processed by ELISA using either empty NBPs (5 μg ml⁻¹, 0.5 μg per well) or Ad5 (5 × 10⁶ PFU per well) as capture antigens. Mouse Ig was detected using horseradish peroxidase-conjugated anti-mouse antibody, and ELISAs were performed according to standard procedures. Serial dilutions of mouse Ig were used as reference antibody titres.

Immunocytofluorescence and immunohistofluorescence

See Supplementary Methods.

TUNEL assay

See Supplementary Methods.

Statistics and reproducibility

Except where indicated, in vitro data are presented as the mean of triplicate samples ± s.d. from at least three independent experiments. For normally distributed in vitro data, significant differences were determined by one-way analysis of variance (ANOVA) followed by Tukey’s post hoc analysis, unless otherwise indicated. In vivo data are presented as mean ± s.d. Non-parametric analyses were used to determine significant differences within in vivo experiments, where appropriate, using Kruskal–Wallis tests followed by Mann–Whitney post hoc analysis. All measurements were taken from distinct samples.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.