All chemicals used are listed in the relevant section below, and were purchased from Sigma Aldrich, unless stated overwise.
R21 adjuvanted vaccine
The virus-like particle R21 vaccine was produced as described previously25 and used at 1 µg per mouse dose41. The liposome-based adjuvant was manufactured by the Vaccine Formulation Institute (VFI) as described previously42, and contains QS-21 saponin and the synthetic TLR4 ligand 3D-6-acyl-PHAD incorporated into neutral liposomes composed of 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) and cholesterol. The adjuvant mouse dose was defined at 5 µg of QS-21 and 2 µg of TLR4 agonist. Due to the supplied concentration and volume restriction during encapsulation process, it was possible to encapsulate only 1/7th of a full adjuvant mouse dose corresponding to 0.7 µg of QS21 and 0.3 µg of 3D6AP into LCSS particles used for immunisation.
Fluorescent labelling of the R21 vaccine
R21 was labelled with Alexa Fluor 647 (AF-R21) to enable the measurement of encapsulation efficiency and particle core size post encapsulation by microfluidics. Buffer exchange of a 0.5 mL R21 stock solution (0.38 mg/mL in Tris Buffer) was performed using an ultra-centrifugal filter unit (30 kDa cut-off, Amicon, Sigma-Aldrich) and centrifuged 3 times at 10,000 g for 10 minutes at 4 °C, using phosphate buffer saline (PBS) as the replacing buffer. R21 in PBS was then mixed in a 0.5 mL Eppendorf tube with 50 μL of 4 mg/mL Alexa Fluor 647 NHS ester (ThermoFisher) suspended in DMSO, and vortexed for 30 seconds at 3,000 rpm. The labelling reaction was incubated on a rotating shaker for 1.5 hours at 30 rpm mix mode, 28 °C, protected from light. Removal of unincorporated dye and washing of the labelled product was done by 6 cycles of diafiltration (30 kDa cut-off, Amicon, Sigma-Aldrich) at 10,000 g for 10 minutes at 4 °C, using PBS. Labelled R21 was stored at -80 °C.
LCSS particle manufacture by microfluidics
Microfluidic chips were produced by common soft lithography and polydimethylsiloxane (PDMS) replication method43. Following computer-aided design of the microfluidic channels (AutoCAD, AutoDesk), the corresponding SU-8 soft lithography positive moulds were manufactured externally by Micrux Technologies. The moulds were made hydrophobic by vapour deposition of trimethylchlorosilane (TMCS, Sigma-Aldrich) to prevent sticking during PDMS replication. PDMS was mixed at a 10:1 ratio of pre-polymer to curing agent (SYLGARD 184, Dow) and poured onto the positive SU-8 mould. After de-gassing in a vacuum chamber until no air bubbles were visible, the PDMS was cured in an oven overnight at 65 °C. The PMDS layer was peeled off the mould and each chip design cut out with a razor blade. Inlets and outlets were punched with a blunt 18G needle, the chips cleaned with isopropanol, dried with nitrogen, and bonded to a flat PDMS slab by plasma exposition in a plasma chamber (HPT-200, Henniker plasma) for 3 min at 100% power. After plasma exposition, the PDMS chip design and PDMS slab were pressed together for 30 seconds and their bonding strengthened on a hot plate at 115 °C for 30 seconds.
Immediately following plasma bonding, spatially constrained hydrophilic coating was performed to ensure the hydrophilicity of the second intersection of the flow focusing design, adapted from previously reported method44.In brief, a coating solution of 1% (w/v) polyvinyl alcohol (PVA, 30-70 kDa Mw, 87-90% hydrolysed) in deionized water (DI water) was injected from the outer inlet at 100 mbar while continuously flushing air was through the inner and middle inlet at 50 mbar to protect the first intersection from the coating solution. The treatment was performed for 5 minutes before flushing all channels with air and drying the chip for 5 minutes at 115 °C on a hot plate. The process was repeated 3 times to ensure durable and reliable hydrophilic coating in the relevant segment of the chip. Following coating, the microfluidic chips were left at 60 °C for a minimum of 48 hours to allow any uncoated parts to recover their hydrophobicity. Chips were then stored at room temperature and used for microfluidic emulsification within 3 months of manufacture.
Fluidics formulation
The following fluid compositions were used for the emulsification processes.
Inner fluid: For in vitro work, fluorescent dextran payloads were dissolved in 1X PBS.
For in vivo work, 100 μL of AF-R21 solution was mixed with 400 μL of R21 solution and 500 μL of adjuvant. The inner core fluid was kept on ice in the dark for the time of production.
Middle fluid: Poly(D,L-lactic-co-glycolic acid) polymers (PLGA, Resomer, Evonik) were dissolved in dimethyl carbonate (DMC, Sigma-Aldrich) solvent, in a glass vial at room temperature, at the maximum observed compatible concentration with the microfluidic setup before clogging (Extended Data Table 1). The glass vials were agitated on a rolling shaker for 2 hours to ensure complete dissolution of the polymer.
Outer fluid: 3%(w/v) PVA (9-10k Mw, 80% hydrolysed, Sigma-Aldrich) was dissolved in DPBS (1X, Gibco).
Microfluidic double emulsification
To perform microfluidic Water-in-Oil-in-Water (W/O/W) emulsification, a pressure driven microfluidic setup was assembled. Compressed air was generated from a compressor (6-4 quiet running compressor, Jun-Air) and fed to a microfluidics dedicated pressure controller (OB1, Elveflow). Separated reservoirs containing the inner, middle and outer fluids, supplied by Darwin Microfluidics or custom manufactured in Delrin, were precisely pressurised by the pressure controller and their liquid content pushed into the microfluidic chip through PTFE tubing (1/32” outer diameter, 300 μm inner diameter, Darwin microfluidics) and ¼ 28” fittings. The emulsification process was monitored using an inverted microscope (DMi8, Leica) operated in brightfield, and a high-speed camera (Nova S16 from Phantom). The injected pressures were set respectively to 200, 300, and 400 mbar initially, to prefill all channels and reduce risk of backflow. After 2 minutes the flow pressures were changed to the required pressure and monitored using the high-speed camera. Videos of the production were taken, and the core and outer droplet diameter and frequency measured using a dedicated software (DMV, developed at Wayne State University45) and ImageJ. The pressures were adjusted manually until the droplets reached the desired diameter and maximum generation frequency. Double emulsions were collected in a glass petri dish and transferred to a glass vial (22 mL) filled with outer fluid. Extraction of the solvent was performed over 48 hours at room temperature on a roller shaker, replacing the PBS extraction media every 12 hours. Resulting LCSS particles were then filtered using mini cell strainers (100 μm, pluriSelect) and stored in 2 mL of PBS in Eppendorf tubes at 4 °C until use. Quality control was performed by fluorescent microscopy imaging, as described in the corresponding section.
Batch emulsification
To produce particles by batch emulsification (BE), 100 μL of inner fluid was pipetted into 500 μL of middle fluid. The mix was vortexed at 3,000 rpm for 1 minute in a 2 mL LoBind Eppendorf. The resulting water-in-oil emulsion was transferred by pipetting into 10 mL of the outer fluid in a 15 mL glass vial, and vortexed at 2,500 rpm for 1 minute. The resulting water-in-oil-in-water emulsion was left to rest for 15 minutes. Extraction of the solvent was performed over 48 hours at room temperature on a roller shaker, replacing the PBS extraction media every 12 hours. Particles were then filtered using mini cell strainers (100 μm, pluriSelect) and stored in 2 mL of PBS in Eppendorf tubes at 4 °C until use.
In vitro release assay
Produced particles encapsulating dextran-TRITC were divided equally into 3-4 aliquots, suspended in 0.01% (w/v) Tween 80 (Sigma-Aldrich) in PBS in 2 mL LoBind Eppendorf tubes, and incubated in a rotating incubator (Roto-Therm, Benchmark Scientific) kept at 37 °C at 20 rpm. The combination of the full rotation and the presence of Tween 80 surfactant in the release media prevented particle aggregation and sedimentation. At relevant timepoints, depending on the PLGA formulation, particles were temporarily sedimented by centrifugation for 30 seconds at 50 g, and 1 mL of supernatant was replaced with fresh releasing media to provide sink conditions. 200 μL of the removed media were used for quantification of the released payload. Image release monitoring and SEM imaging were also performed at relevant timepoints. The effect of different LCSS formulation parameters on the in vitro release was tested separately. For each parameter, the release was compared to a common reference formulation. Details are listed in the Extended Data Table 2.
Release quantification
The amount of dextran-TRITC released from the particles was measured by fluorescent spectroscopy (FLUOstar Omega, BMG Labtech), using a standard curve. The release was normalised to the initial total encapsulated content for each replicate by measuring the remaining unreleased dextran-TRITC at the experimental end point, after mechanically breaking up the particles with a cell homogeniser for 30 seconds at 30,000 rpm (IKA T 10 Basic Ultra Turrax Homogenizer, Cole-Parmer) and centrifugation to remove PLGA debris, at 15,000 g for 10 minutes (Extended Data Fig. 13). The amount released at each timepoint was corrected for the dilution resulting from the partial replacement of the releasing media.
Fluorescent microscopy imaging for particle characterisation and monitoring of in vitro release
For each image, 20 μL of particle suspension were mixed with 200 μL of 0.01% (w/v) Tween 80 in PBS in Cellview slides (Greiner) and multi-tile imaging was performed with an inverted microscope (TiE2, Nikon, 10X objective) in the brightfield and TRITC fluorescent channel. The image tiles were processed with MATLAB Image Processing Toolbox to measure the core and shell fluorescence and the diameter of each particle. Briefly, each image was smoothed by applying a bilateral filter or a gaussian filter for the brightfield and fluorescent channels respectively. Background fluorescence was removed from the fluorescence channel, and contrast was enhanced in the brightfield channel. Each individual droplet or particle was detected using a Hough transform algorithm and separated from each other by watershed segmentation in case of touching. Masking techniques were used to associate each particle in brightfield with its core in fluorescence, and the diameter and fluorescence of the different regions of each particle were measured for different timepoints, including immediate post-production quality control.
For each particle, the fluorescence profile was computed by averaging the fluorescence intensities of pixels at equal distance (normalized radius) from the core center.
Coefficient of variation (CV) for the core and outer particle diameter distributions was computed as the ratio (%) of the Standard Deviation and the Mean.
As this is a single core-shell system, the encapsulation efficiency was computed as the ratio (%) between the number of cores and the particles detected.
Fluorescence microscopy images of LCSS particles containing dextran-TRITC as the model payload were produced by enhancing the contrast by contrast stretching, allowing 0.35% of pixel saturation, and overlaying the TRITC and brightfield channels.
SEM imaging of the in vitro release
The particle morphology immediately after production or at different timepoints of in vitro degradation was imaged using a field emission gun scanning electron microscope (FEG-SEM, Zeiss Sigma 300) with an acceleration voltage of 2 kV. Before imaging, 10 μL of particles in suspension in PBS or release media was pipetted and spread onto a conductive adhesive carbon tape attached to a SEM pin stub and a thin film of Pd/Au coating was sputtered onto the sample (Q150R ES Dual Carbon/Sputter Coater, Quorum Technologies).
R21 loaded LCSS particle formulation and quality control
Quality control of the produced LCSS particles was performed by microscopic image quality control, as described in the corresponding section, using the AF-R21 as the fluorescent marker within the core. Quantification of the encapsulated R21 was performed by mechanically breaking the particle shells: 300 μL of the produced particle batch was added to 700μL of PBS and homogenised for 30 seconds with a cell homogenizer (Cole-Parmer) to break up the particles. Released R21 was isolated from the PLGA particles debris by centrifugation at 15,000 g for 10 minutes, and the concentration of R21 in the supernatant determined using the micro BCA Protein assay (ThermoFisher). From this, the number of LCSS particles required to achieve a vaccine mouse dose was calculated (in the order of 100,000 LCSS particles per mouse).
Depending on the regimen, the correct number of particles, antigen in solution and adjuvant were resuspended in 0.25% (w/v) medium viscosity carboxymethylcellulose (CMC, Sigma-Aldrich). CMC was added to prevent early sedimentation of the particles.
Mouse immunisations
Mice were maintained at the Wellcome Centre for Human Genetics or the John Ratcliffe Hospital, University of Oxford, housed under Specific Pathogen Free (SPF) conditions and in accordance with the recommendations of the UK Animals (Scientific Procedures) Act 1986 and ARRIVE guidelines. Protocols were approved by the University of Oxford Animal Care and Ethical Review Committee for use under Project Licenses P9804B4F1 or PP0984913B granted by the UK Home Office.
Six to 10-week-old female inbred BALB/c mice (Envigo, UK) were immunised intramuscularly with a total volume of 50 μL in the tibialis muscle under 3% isoflurane anaesthesia. All injections were performed with 100 μL Gas Tight glass syringes (Hamilton), with 1.40 mm barrel inner diameter to improve the particle injectability46.
ELISA for measurement of antibody titres
NANP or Cterm-specific IgG ELISAs to detect antibodies against the central repeat (NANP) or the C-terminal (Cterm) regions of CSP were performed as previously described25. Serum was obtained by collecting blood from the lateral tail vein in a microcuvette tube. Blood was allowed to clot for 2 hours at room temperature before centrifugation at 13,000 rpm for 4 minutes and sera removed and stored at −20 °C until use. Nunc-Immuno Maxisorp 96 well plates were coated with 2μg/mL NANP6 peptide (Mimotopes, NANPNANPNANPNANPNANPNANPC) or 1μg/mL Cterm peptide (ProImmune, EPSDKHIKEYLNKIQNSLSTEWSPCSVTCGNGIQVRIKPGSANKPKDELDYANDIEKKICKMEKCS) in carbonate-bicarbonate coating buffer overnight at 4 °C. Plates were washed with PBS-Tween (0.05% v/v) and blocked with 2% BSA in PBS-Tween for 1 hour at RT. Sera were serially diluted 3-fold between 1:100 and 1:24,300 in 1% BSA in PBS-Tween and added to plate in duplicates at the appropriate dilution for the vaccine regimen and time. Plates were incubated for 1.5 hours at room temperature and then washed as before. Fc-specific goat anti-mouse IgG conjugated to alkaline phosphatase (AP) (1:5000, Sigma-Aldrich, A1418), was added for 1 hour at room temperature. Following the final wash, plates were developed by adding p-nitrophenylphosphate at 1 mg/mL in diethanolamine buffer and OD was read at 405 nm. Total IgG concentrations in sera were calculated as ELISA Units (EU) by interpolation against a standard curve of monoclonal 2A10 (MR4, MRA-183A) for NANP-specific or a serum pool of previously R21-immunised mice for Cterm-specific ELISA. Sera from naïve mice were used as negative control.
Malaria challenge
Malaria challenge was carried out using transgenic Plasmodium berghei (P. berghei) parasites carrying an additional copy of the P. falciparum CSP gene at the 230p locus under the control of the P. berghei UIS4 promoter47, produced as described previously25,41. For all experiments 1,000 transgenic P. berghei sporozoites were injected intravenously (i.v.) in a total volume of 100 µL into the lateral tail vein. From day 5 post challenge mice were monitored for infection by thin-film blood smear (fixed in methanol and stained in 10% Giemsa for 30 min) and culled when >1% parasitaemia was observed. If no parasites were detected on day 12 after challenge, mice were considered sterilely protected.
Statistical Analysis
All statistical analysis was performed in R (version 4.2.1). Results were deemed to be statistically significant if p<0.05. For antibody titre, non-inferiority (NI) was declared if the 95% confidence interval for a fold change was entirely above the non-inferiority threshold, which was set at 0.5 in accordance with previous reports48,49.
Analysis of in vitro kinetic profiles. Time to 50% release from particles in vitro was estimated as the mean of time to 50% release (obtained for each replicate by linear interpolation). Confidence intervals are calculated using the corresponding t-distribution.
Analysis of in vivo immune kinetic profiles. Following previous reports50,51 and recommendations52, longitudinal antibody titres were fitted, after log-transformation, with a generalised additive model (GAM), using the mgcv package (version 1.8-40). Penalised cubic regression splines were used to model the time covariate for each regimen, with 10 knots placed evenly throughout its range. Regimen was modelled as a fixed effect, and random effects for subject and subject by time interaction were also included to account for within-subject correlations.
Marginal sampling distributions of the value and time at 50% peak antibody titre, and titre fold change and additional delay compared to the control regimen were generated for each regimen, following53, by repeatedly: 1) sampling model coefficients from the multivariate normal posterior distribution, 2) predicting log titres at 0.01 week steps, 3) finding value and time at 50% peak titre, and 4) calculating titre fold change and additional delay. This process was repeated 10,000 times per regimen. Geometric mean titre fold change and mean additional delay, and their 95% CI, are then estimated from the corresponding sampling distributions.
Analysis of antibody titres at the time of challenge. Geometric mean fold changes were calculated from the measured titres by bootstrapping differences of mean log-transformed titres and back-transforming the mean of the bootstrap distributions. Bias-corrected and accelerated (BCA) confidence intervals are reported. Bootstrapping was performed using the boot package (version 1.3-28) with 10,000 resamples.
Time to infection analysis. Time to infection (defined as 1% parasitaemia) was determined for each mouse by linear interpolation of parasitaemia between the last blood sample below 1% and the first blood sample above. The time to infection data were analysed by Cox regression using regimen and challenge time as covariates. The proportional hazard assumption was confirmed graphically by plotting log-log survival curves and Schoenfeld residuals, as well as by testing the correlation of Schoenfeld residuals with time. Hazard ratios are adjusted for challenge time. Cox regression was performed using the survival package (version 3.2-13), Kaplan Meier plots were generated using the survminer package (version 0.4.9).