Apolipoprotein E Binding Drives Structural and Compositional Rearrangement of mRNA-Containing Lipid Nanoparticles

Emerging therapeutic treatments based on the production of proteins by delivering mRNA have become increasingly important in recent times. While lipid nanoparticles (LNPs) are approved vehicles for small interfering RNA delivery, there are still challenges to use this formulation for mRNA delivery. LNPs are typically a mixture of a cationic lipid, distearoylphosphatidylcholine (DSPC), cholesterol, and a PEG-lipid. The structural characterization of mRNA-containing LNPs (mRNA-LNPs) is crucial for a full understanding of the way in which they function, but this information alone is not enough to predict their fate upon entering the bloodstream. The biodistribution and cellular uptake of LNPs are affected by their surface composition as well as by the extracellular proteins present at the site of LNP administration, e.g., apolipoproteinE (ApoE). ApoE, being responsible for fat transport in the body, plays a key role in the LNP’s plasma circulation time. In this work, we use small-angle neutron scattering, together with selective lipid, cholesterol, and solvent deuteration, to elucidate the structure of the LNP and the distribution of the lipid components in the absence and the presence of ApoE. While DSPC and cholesterol are found to be enriched at the surface of the LNPs in buffer, binding of ApoE induces a redistribution of the lipids at the shell and the core, which also impacts the LNP internal structure, causing release of mRNA. The rearrangement of LNP components upon ApoE incubation is discussed in terms of potential relevance to LNP endosomal escape.

In the following, the procedure to analyze the SANS data collected on the different LNP samples is explained in detail.
The sample MMO was described by a sphere model (Fig. SI2), since the deuteration scheme led to an indistinguishable core-shell structure. For all the LNP samples, except MMO, a preliminary analysis of the data was carried out: the pair distance distribution function (p(r)) and the density profile (d(r)) were obtained from the SANS curves in order to support the choice of model for the fitting. GIFT 1 and DECON 2 were used respectively to determine the pair distance distribution function and the density profile as a function of particle radius. Figure SI3 shows the p(r) and d(r) for sample MMC in different solvent contrasts, and comparing the different curves it is clear the indication toward a core shell structure. For MCH and MCHPC, the fitting procedure was a two-step process, however the first step was common to MMC as well. The first step was a simultaneous fit using the core shell sphere model applied to the 4 (or 5) curves. The core radius, shell thickness and shell scattering length density (SLD) were constrained to be the same amongst the different solvent contrasts, while the core SLD was allowed to vary accounting for solvent in the core (SLDcore=vfsol×SLDsol+(1-vfsol)×SLDdry core). Since the SANS curves for MCH and MCHPC showed a clear peak at q ~0.1 Å-1, the final model was a sum of core shell sphere and broad peak models to better describe the data in the q range above 0.05 Å-1. This broad peak arises from the internal structure in the core of the LNPs. The combined model was applied to each curve separately, keeping constant all the parameters previously optimized for the core shell sphere model and the structural parameters of the broad peak model while the intensity (i.e. contrasts) related parameters where allowed to vary.
The scale factor was fixed to the volume fraction (determined by the sample concentration), the background value was optimized for each curve and the solvent SLD was set as calculated from the mixing ratios of H 2 O/D 2 O.

A B
From the core SLD and shell SLD (Table SI2), the volume fraction of each component can be determined. For each lipid component in the LNP, the molecules partition between shell and core with the distribution being constrained by the molecular volume and the volume of shell and core. The use of molecules with different SLDs allows to refine the partitioning, which is optimized to obtain estimated SLDs that matches the SLD values from the fitting (Table SI2).
The broad peak model accounts for the internal core structure, and becomes relevant at low q as well when overall intensity of the scattering curve is low (i.e. MCH and MCHPC in 39% D 2 O based buffer) since MC3 has a SLD of and it has increasing contrast when 0.08 × 10 -6 Å -2 d-PBS % > 50%. Table SI2. SLD values resulting from the fit of SANS curves collected with LNPs with and without ApoE incubation for 3 hours.

Prior ApoE incubation
Upon ApoE incubation SLD dry core /10 -6 Å -2 SLD shell/10 -6 Å -2 SLD dry core /10 -6 Å -2 SLD shell /10 -6 Å -2 MCH  (2.49)** ± 0.01 NA * The sphere model was used for modelling MMO and hence SLD dry core corresponds to the SLD of the sphere, since there is a lack of contrast between shell and core in the experimental conditions used. ** The increase in SANS intensity (I(q)) upon ApoE incubation for MMO can be described by both an increase or decrease of SLD dry core, since the I(q) is proportional to the square of the SLD dry core -SLD solvent and we have only measured MMO+ApoE in one contrast.    Average 20 ± 2 55 ± 2 23 ± 4 3.0 ± 0.1 21 ± 3 79 ± 3 3.3 ± 0.6 10 -2 * particle radius and shell thickness are smaller than prior to incubation. ** particle radius and shell thickness have same sizes as before incubation.

Protocol for LNP immobilization on QCM-D gold sensors and Binding isotherm for
ApoE3 to LNP Au-sensors (Q-sense) were cleaned in base piranha for 5 minutes at 75°C ( µl/min, then rinsed with 1 ml PBS pH 7.4, 1 ml of AntiPEG-biotin 5 µg/ml was injected, then rinsed with 1 ml PBS, 1 ml of 50 µg/ml BSA was injected as blocking agent and a final 1ml of buffer was flown. The coverage of AntiPEG-AB is found to be (2.07 ± 0.14) × 10 12 mol • as will be reported in an upcoming publication (manuscript in preparation). cm -2 A peristaltic pump (IPC 4, Ismatec) was used for injection of all solutions.
LNP stock (mRNA concentration 0.1 mg/ml, all hydrogenous components) was diluted to 20 µg/ml (mRNA concentration) and 1ml was injected to each of the four flow modules. After injection, the solution was left for 1 hour prior to PBS rinsing ( Figure SI4). Quick binding was observed accompanied by the formation of a soft layer (high increase in dissipation).
ApoE3 stock solution (0.5 mg/ml) was diluted to 0.5, 2.5, 5 and 12.5 µg/ml. After rinsing the LNP immobilized, 1ml of each ApoE dilution was injected to each sensor and left for about 10 minutes, then rinsed with PBS. Further adsorption occurred that did not induce a further increase in dissipation. Thus, no further softening of the adsorbed layer occurred upon ApoE binding which justifies the use of the Sauerbrey equation for determination of ApoE adsorbed amount. 3

Estimation of ApoE to LNP
From the wet mass adsorbed upon ApoE addition an upper limit for the number of molecules was determined by dividing the mass for the molecular weight (38 kDa).
Considering an LNP of 60 nm in diameter to have a packing at the sensor surface to be random (60% coverage) or to have a hexagonal arrangement (90% coverage), we can determine the total LNP available surface on the sensor for ApoE binding and calculate the number of ApoE per LNP. From the number of ApoE per LNP we can determine the weight ratio estimating the LNP mass to be about 68 MDa for 60 nm diameter particle. For example 1:10 ApoE:lipid %w corresponds to about 180 ApoE per LNP. Figure SI4. Frequency shift (left y-axis, black) and dissipation (right y-axis, grey) recorded in the binding experiment of ApoE 12.5 µg/ml to immobilized LNP.
Additional SANS data: ApoE4 and HSA binding to LNP In order to determine if the effect of ApoE3 binding to LNP is specific or general, we measured SANS data with ApoE4 and Human Serum Albumin (HSA).     The 1 H and 2 H NMR spectra of cholesterol-d45 (Fig. SI12B) shows deuteration across all nonlabile protons. By comparing the 1 H and 2 H NMR spectra and considering the relative ratios of the peaks, it is concluded that the alkene single proton at 5.3 ppm (C6) was deuterated more than the single proton at the C3 carbon next to the alcohol group at 3.5 ppm. The deuterium substitution (% D) of these positions and some others, shown in the structure below, were determined by integrating the 13 C signals in the 13 C { 1 H, 2 H} NMR spectra, which are isotopically shifted due to deuterium incorporation (Fig. SI13) using the method of Darwish et al.. 4 The rest of the positions were determined to be at 88% D based on the average deuterium level calculated for the whole molecule from the ESI-MS (87%D).