Novel lipid nanoparticle provides potent SARS-CoV-2 mRNA vaccine at low dose with low local reactogenicity, high thermostability and limited systemic biodistribution


 Concerns with current mRNA Lipid Nanoparticle (LNP) systems include dose-limiting reactogenicity, adverse events that may be partly due to systemic off target expression of the immunogen, and a very limited understanding of the mechanisms responsible for the frozen storage requirement. We applied a new rational design process to identify a novel multiprotic ionizable lipid, called C24, as the key component of the mRNA LNP delivery system. We show that the resulting C24 LNP has a multistage protonation behavior resulting in greater endosomal protonation and greater translation of an mRNA-encoded luciferase reporter after intramuscular (IM) administration compared to the standard reference MC3 LNP. Off-target expression in liver after IM administration was reduced 6 fold for the C24 LNP compared to MC3. Neutralizing titers in immunogenicity studies delivering a nucleoside-modified mRNA encoding for the diproline stabilized spike protein immunogen were 10 fold higher for the C24 LNP versus MC3, and protection against viral challenge in a SARS-CoV-2 mouse model occurred at a very low 0.25 µg prime/boost dose of the same immunogen in the C24 LNP. Injection site inflammation was notably reduced for C24 compared to MC3. In addition, we found the C24 LNP to be entirely stable in bioactivity and mRNA integrity when stored at 4 ºC for at least 19 days. Storage at higher temperatures reduced both bioactivity and mRNA integrity, but less so for C24 than MC3, and in a manner consistent with the phosphodiester transesterification reaction mechanism of mRNA cleavage. The higher potency, lower injection site inflammation, and higher stability of the C24 LNP present important advancements in the evolution mRNA vaccine delivery.


Introduction
There are 2 completed and 6 ongoing clinical trials for COVID-19 mRNA vaccines 1

. The
BioNTech/Pfizer and Moderna products have received Emergency Use Authorization 2,3 and announced results from phase 3 clinical trials that reported efficacy greater than 94% for reduction of SARS-CoV-2 symptomatic infection after 2 doses of a nucleoside-modified mRNA encoding the spike protein delivered in a lipid nanoparticle (LNP) 4,5 . Recent interim data from a trial by CureVac contained disappointing results with only 47% protection 6 , possibly due to the use of non-nucleoside-modified mRNA, which has higher innate immunogenicity than nucleosidemodified mRNA 7 , and thereby limiting the dose to 12 µg in the CureVac trial versus 30 µg for BioNTech and 100 µg for Moderna. These latter two doses were maximum tolerated doses determined in phase 1 trials where higher doses (250 µg Moderna 8 , 100 µg for BioNTech 9 ) were discontinued due to frequent and severe injection site pain. Data from the clinical trials 10,11 as well as post-approval follow-up data 12,13 indicate the Moderna vaccine has a higher frequency of adverse events (ADEs) and reactogenicity than the BioNTech vaccine which could also be related to its higher dose (100 µg Moderna versus 30 µg BioNTech/Pfizer). Vaccine reactogenicity is thereby dose-related, generating a narrow successful dose range for mRNA vaccines that motivates the identification of more efficient mRNA delivery systems to achieve protection at lower doses.
In addition to injection site pain that is seen in nearly all subjects, systemic adverse reactions are seen in nearly half, and allergy-type reactions can be found in as many as 2% 12 with anaphylactoid reactions in around 1 in 100,000 14 . Cases of myocarditis 15 and thrombocytopenia 16 have also been identified as potentially mRNA vaccine-related. These ADEs and reactogenicity may be due to innate inflammatory responses to the lipids or mRNA, molecular mimicry between viral spike protein and endogenous proteins 15 , and off-target biodistribution and mRNA expression at unintended sites following intramuscular (IM) injection. For example, IM injection of LNPs containing an mRNA-encoded luciferase in mice showed high levels of liver expression at 6 hours ( Fig S4 in 17 ). The current emergency-approved vaccines also generated off-target distribution and expression in liver and other organs in rodent preclinical models 18,19 . Our recent work showed a means to minimize off-target expression by using a less negatively charged mRNA-LNP that is more locally retained in the injected muscle and draining lymph nodes versus trafficking systemically as seen by expression in liver 20 . A final challenge for current mRNA vaccines is the 5 need for frozen storage with relatively short stability times for non-frozen conditions, severely limiting global vaccination. Unfortunately, current literature does not provide any data to illuminate the mechanisms involved in loss of mRNA-LNP bioactivity 21,22 during storage although regulatory documentation suggests a loss of mRNA integrity is involved 18,19 . In order to widely implement mRNA-LNP vaccines for infectious diseases in the future, there is an urgent need to increase their potency and reduce dose, as well as to control biodistribution and increase understanding of the mechanisms determining stability during storage.
The design of the ionizable lipid in LNPs is considered the key aspect that determines potency or mRNA delivery efficiency, as well as degradability, toxicity, reactogenicity and the adjuvant properties of the LNP 23 . Early studies using siRNA containing LNPs showed that only 1-2% percent of the siRNA in endosomes is released from the LNP and from endosomes and loads into the RNA-induced silencing complex 24 . The current LNPs in COVID-19 vaccines may increase endosomal release several fold 25,26 but still release only a small fraction of mRNA for cytosolic translation. For this reason, the main design principle guiding improvements to ionizable lipids is to increase endosomal release that occurs due to protonation of the ionizable lipid in the endosome followed by ion pairing with a negatively charged endogenous endosomal phospholipid where this ion pair can open the endosomal membrane to release mRNA into the cytoplasm 27 . The endosomal protonation requirement is well established and is summarized as requiring the pKa of the LNP to be in the 6-7 range 28 . We recently showed that this LNP pKa is different from the ionizable lipid pKa that is in the 8-10 range 20 and this 2-3 point difference as mainly due to proton partitioning between the lipid phase and the aqueous media external to the LNP. This insight has now permitted us in the current study to systematically scan and screen theoretical structures of novel multiprotic head groups of ionizable lipids, since we are able to predict the pKa of the resulting LNP based on the structure of the ionizable lipid.
In addition to endosomal protonation, another principle of ionizable lipid design is the molecular shape hypothesis 29,30 , which states that the lipid tails should have a wider cross-section than the head group, thereby creating a cone-shaped ion pair with an endogenous phospholipid that is not compatible with a lipid bilayer and therefore destabilizes it for endosomal release. The evolution of ionizable lipids displays an increased level of lipid tail branching that augments a cone-shaped morphology. MC3, the standard reference ionizable lipid in the approved Onpattro silencing RNA 6 product from Alnylam has a dilinoleic tail, the Moderna SM-102 ionizable lipid has 3 saturated alkyl branches and the Acuitas ALC 0315 in the BioNTech/Pfizer product has 4 saturated alkyl tails 23 . A third critical component of the ionizable lipid is the linker between the head group and tails that should be degradable in order to permit elimination in the body and limit accumulation 31 .
A favored linker in this respect is a primary ester since it has been shown to degrade quickly in vivo to minimize accumulation and allow repeat administration 31 . This rapid degradability was associated with reduced inflammation at the IM injection site and increased tolerability 26 .
We designed a new ionizable lipid for mRNA-LNP vaccines based on the above criteria. We theoretically screened a broad molecular design space of head groups that were not limited to monoprotic units used in current mRNA LNP vaccines. The incorporation of more than one ionizable nitrogen in the head group can provide an additional dimension to both enhance endosomal protonation and to control charge of the LNP that influences biodistribution. The chosen trivalent head group, 4-methyl-1-piperazinebutanamine, was linked via two degradable primary esters to octyldodecyl tails creating 4 saturated alkyl tails with non-symmetric 8 and 10 carbon lengths. The resulting compound, abbreviated C24, was used along with the standard 3 additional lipids to produce mRNA-LNPs containing a nucleoside-modified mRNA that encoded for either a luciferase reporter or a diproline-stabilized membrane-bound spike protein (S2P) immunogen from SARS-CoV-2 that is equivalent to the immunogen in the current emergency authorized mRNA vaccines. We comprehensively studied the physicochemical properties of the C24 ionizable lipid, the resulting C24 LNP, as well as immunogenicity, protection against lethal SARS-CoV-2 challenge, injection site inflammation and stability of C24 mRNA-LNPs and mRNA integrity during liquid storage. By direct comparison to the standard MC3 LNP, which was used in the first two phase 1 clinical trials of nucleoside-modified mRNA for influenza 32 , the C24 mRNA LNP appears significantly more potent with 10 fold higher neutralizing antibody titers than MC3, protecting against SARS-CoV-2 infection at a low 0.25 µg dose administered twice in mice and is less inflammatory at the injection site while entirely maintaining bioactivity beyond 2 weeks when stored in a liquid format at 4°C.

Results and Discussion
Trivalent head group of C24 ionizable lipid displays molecular and LNP ionization properties that augment protonation in the endosomal pH range 7 The initial screening of our ionizable lipid design space included more than 100 head groups, 10 linkers and 50 alkyl tails that combine for over 50,000 potential ionizable lipid candidates. We calculated aqueous phase pKas for ~500 candidates that sampled this design space and selected several head groups for synthesis with acrylate bearing alkyl tails using a synthetic procedure that involved only 2 reactions and one catalyst versus more than 5 reactions and 7 catalysts for the ionizable lipids in the current COVID-19 mRNA vaccines 26, 33 . The simplicity of the reaction scheme resulted in much fewer purification steps, much shorter synthetic time, and over 80% yield producing an estimated 10 fold reduction in cost that could faciliate global vaccination campaigns.
More than 30 candidate ionizable lipids were initially synthesized and characterized physicochemically and for certain biological performance indicators resulting in the selection of C24 (Fig 1a) for further investigation in the current study. C24 bears a trivalent 4-methyl-1piperazinebutanamine head group with theoretical aqueous phase pKas ranging from 4 to 8, two of which were close in predicted values (7.7 and 7.8). We synthesized a water soluble analogue (C24-WSA) and measured the pKa of each nitrogen using an established 1 H NMR method 20 confirming theoretical pKas to within 0.4 units (Fig 1b). A slightly stretched deviation from an ideal Henderson-Hasselbalch behavior was observed for the terminal nitrogen (pKa 8.1, red in Fig   1b) consistent with an interaction with the nitrogen atom with a slightly lower pKa (7.5 in Fig 1b) that protonates almost simultaneously. This observation also indicates the potential for proton coordination between these two sites and the formation of a dative bond during initial protonation stages. NMR analyses of the water soluble MC3 analogue revealed a pKa of 9.5 (Fig 1c) that is very similar to the predicted value of 9.4 (Fig 1a). Protonation of LNPs made with C24 and MC3 was assessed using the TNS dye-binding assay that measures surface charge and with zeta potential by electrophoretic mobility that measures net charge of the LNP, as we described recently 20 . The TNS dye-binding assay revealed a higher pKa for C24 (6.77) than MC3 (6.55) and a larger increase in surface protonation when pH drops from 7.4 to 6 (4517 versus 2559 RFU in Fig 1f) indicating greater surface protonation in the endosomal pH range for C24 versus MC3. The lower pH limit in calculating endosomal protonation was taken as 6 for the TNS dye-binding assay since there is no change in surface charge below this pH. Electrophoretic mobility measurements, in contrast to the TNS dye-binding assay, measure net charge of the LNP and show broader changes in LNP net charge reflected by zeta potential increasing down to pH 3 (Fig 1e). MC3 followed very closely the behavior of the extended Henderson-Hasselbalch model originally proposed for 8 polyelectrolytes 34 since it accounts for ionization-state-dependent pKa of closely interacting ionizable sites such as the thousands of dimethylamine MC3 head groups that are in close proximity within the LNP. The extended Henderson-Hasselbalch analyses revealed a zeta potential pKa of 5.33 and a pI of 5.57 for the MC3 LNP (Fig 1f). The ionization behavior of the C24 LNP was more complex where a rapidly rising zeta potential occurred from pH 7.4 to pH 6 corresponding to simultaneous protonation of 2 of the 3 nitrogens in the head group with aqueous phase pKas of 8.1 and 7.5 (Fig 1b) below which the zeta potential continued to rise with the contribution of the central nitrogen with a pKa of 3.7 (Fig 1b). This multiprotic ionization behavior for C24 was not well represented by the extended Henderson-Hasselbalch model (dashed versus solid red line in Fig 1e) that resulted in a zeta potential pKa of 5.21 and pI of 6.12 (Fig 1f). The Size characterization showed C24 LNPs to be slightly larger than MC3 (80 nm versus 64 nm diameter in Fig 2a) and to have a slightly lower fraction of mRNA that is inaccessible to ribogreen dye-binding (68 % versus 79 % in Fig 2b). The fraction of mRNA that is inaccessible to ribogreen is often called encapsulation efficiency, however it is now known that the accessible portion is not a free fraction of mRNA since it may not migrate on a gel-based assay 35 . We used our recently published molecular volume model 20 of the LNP to estimate the number of copies of the Firefly Luciferase (FLuc) encoding mRNA in each LNP finding 4.5 copies in the MC3 LNP and 6 for the C24 LNP (Fig 2c), on average, due to its larger size. CyroTEM analyses revealed a larger size for the C24 LNP consistent with DLS measurements and both LNPs had a similar structure showing a peripheral bilayer 36 and an internal electron dense amorphous core (Fig 2d,e).
Luciferase expression of C24 LNP is 4 fold higher than the MC3 LNP upon intramuscular injection and displays 6 fold less off-target expression in liver than MC3 9 In vivo expression of the mRNA FLuc encoding reporter after intramuscular (IM) administration of the LNPs in mice at a relatively high dose of 5µg mRNA showed 2 fold higher expression at the injection site for C24 at 4hrs and 24hrs (Figs 3a and 3b). Systemic biodistribution and offtarget expression in liver was high for MC3 (Figs 3). In contrast, C24 at this high dose virtually eliminated systemic biodistribution with a 6 fold reduction of off-target expression in liver compared to MC3 (Fig 3C), an important finding since systemic reactogenicity and adverse events associated with mRNA-LNP vaccines may be linked to systemic biodistribution and off-target expression in sites other than the injection site and draining lymph nodes. The high off-target expression seen here for MC3 is consistent with previous findings for MC3 17 and regulatory documentation 18,19 suggest it also occurs in rodent models for the current emergency authorized COVID-19 mRNA vaccines. The mechanism that significantly limits systemic biodistribution for the C24 LNP could be related to its rapid increase in surface charge and net charge near neutral pH (Fig 1d and 1e), since we found previously that a less negatively charged LNP was more locally contained upon IM adminsitration 20 . IM administration of Fluc-encoding mRNA-LNPs at a lower 0.5µg dose that is more representative of vaccination doses in mice showed C24 to express 4 fold higher than MC3 at the injection site (Fig 3d and 3e). Off target expression could not be detected by IVIS at this 10 fold lower dose due to low signal to noise at the liver site. Daily imaging of mice showed an initial burst of expression lasting 48 hours with a gradual decline to baseline over 5 days (Fig 3f).
Immunogenicity towards mRNA-encoded SARS-CoV-2 spike protein shows higher binding titers and 10 fold higher pseudoneutralization titers for C24 LNP versus

MC3 LNP against the original Wuhan strain and against two prominent variants
LNPs were assembled with nucleoside-modified mRNA encoding for the diproline-stabilized membrane-bound spike protein immunogen (S2P) that is in the current emergency-authorized COVID-19 mRNA vaccines. We performed immunogenicity studies in Balb/c mice by IM administration of two immunizations with dose ranging from 0.1 µg to 1 µg with 3 weeks between prime and boost and serum analyses for binding antibodies to the receptor-binding domain of the spike protein, as well as neutralization assays to a SARS-CoV-2 pseudovirus. Optical density of the ELISA binding assay at transitional dilutions showed binding antibodies were significantly higher for the C24 LNP versus MC3 at all doses (Fig 4a and 4b). Neutralization assays to a SARS-CoV-2 pseudovirus showed C24 LNPs with significant ~10 fold increases of neutralization titers versus MC3 at all doses after the boost (Fig 4C). A very similar assay done in Balb/c mice with an identical mRNA-encoded immunogen (other mRNA structures differed) in the SM-102 LNP of Moderna revealed titers similar to those of MC3 (1,000 at 1 µg) 37 suggesting that the C24 LNP may also be more potent than the SM-102 LNP. The 1 µg dose delivered in the C24 LNP was also capable of inducing neutralization after a single dose where MC3 did not produce neutralization (Prime in Fig 4c). Finally, we found that serum from animals vaccinated with the highest 1 µg dose were capable of neutralizing two variants of SARS-CoV-2 and that titers of C24 were higher than those of MC3 for the tested variants (Fig 4d). Two of the five animals in each group were sacrificed on day 5 to assess viral titers in the lung and the remaining 3 followed until euthanasia criteria were met. We found C24 mRNA LNPs completely protected mice at 0.25 µg with one of 3 animals at 0.1 µg dose also surviving versus MC3 where protection occurred at 0.5µg dose (Fig 5a). Neutralization titers in a plaque assay showed C24 LNPs achieved titers at 0.25 µg that were equivalent to those of MC3 at 1 µg dose after both the prime and the boost (Fig 5b and c). Lung viral titers examined 5 days after infection found C24 mRNA LNPs entirely blocked lung infection at 0.5µg dose and that MC3 did not block infection entirely even at the highest 1 µg dose. Taken together, these results indicate that C24 LNPs achieve protection against infection at a dose~4X lower than MC3. The 0.25 µg protective dose applied twice in our study is 60X lower than the single effective dose (15 µg) found in an LNP 38 containing a self-replicating mRNA for the spike protein and 10X lower than another single dose found with a different LNP 39 also containing a self-replicating mRNA. Although these latter studies with self-replicating mRNA are single dose, the required dose is much greater and at a 11 level that would be expected to generate unacceptable reactogenicity in humans. The study that is most similar to ours is prime/boost nucleoside modified approach using the S2P immunogen in the SM-102 LNP of Moderna 37 where doses as high as 1 µg of the S2P immunogen in the SM-102 LNP were not capable of blocking lung infection, although they did reduce lung infection compared to PBS controls. The potency of the SM-102 LNP for blocking lung infection in mice therefore appears similar to that of MC3 in our study supporting with the similar potency we found for SM-102 and MC3 neutralization titers against a SARS-CoV-2 pseudovirus (both at ~1,000 Fig   4c). Taken together these results suggest the C24 LNP system exceeds the potency of MC3 and of SM-102 LNP in mice models. Often, translation to non-human primates and humans is not predictable from mouse models 40 so that results in larger animal models and clinical studies are required to further assess C24 mRNA LNP vaccines.

Injection site inflammation for the C24 LNP is milder than for the MC3 LNP
Injection site inflammation for MC3 mRNA LNPs was visually evident 24 hrs post IM administration by macroscopic swelling and a high level of stiffness of the injected leg. Injection site inflammation of C24 mRNA LNPs injected at a 5 µg dose was macroscopically lower than the swelling of MC3 LNP injected sites. We therefore fixed and processed the injected legs for standard histological analyses (Paraffin and H&E staining). We found that the injected 50 µL depot of mRNA LNPs was not inside the muscle tissue but was rather deposited mainly in the facial plane between the medial and lateral gastrocnemius (IS in Fig 6a, d,g) and was difficult to distinguish from adipose tissue at these sites. This non-intramuscular site for the LNP depot was due to the standard 3mm injection depth passing through the medial gastrocnemius that is ~2 mm thick. The volume of the medial gastrocnemius is only ~150 µL 41 so that the standard 50 µL LNP volume would not likely be contained inside the muscle even if the injection depth were reduced.
In comparing C24 histology to MC3 we observed greater levels of inflammation for MC3, for example in the synovium which was multicellular and thickened for MC3 versus a normal appearance for C24 (Fig 6 e,f versus b,c). In addition to infiltration of leukocytes and vasculature at the injected site, we observed the presence of mixed cell-type lymphoid structures (Fig 6 a,g,i) at 24 hrs post administration for both C24 and MC3 LNPs that were absent in control uninjected legs. These rapidly forming (in 24 hrs) lymphoid structures may play an important role in the immune response in mice and are being further characterized.

Storage of LNPs at 4°C reveal C24 LNPs are stable for at least 19 days while storage at higher temperatures induces loss of bioactivity and mRNA cleavage that is
consistent with the phosphodiester transesterification reaction mechanism of mRNA cleavage mRNA LNPs generally require frozen storage and can be stored at refrigerated temperatures for up to 30 days according to instructions for use of the emergency authorized LNPs. Recent reviews 21,22 have highlighted a nearly total lack of information on mechanisms contributing to instability during storage of mRNA LNPs although European regulatory documentation 18,19 states that instability is temperature-dependent and involves a loss of mRNA integrity as well as changes in LNP size and generation of impurities. We therefore stored C24 and MC3 FLuc mRNA LNPs for 2 weeks at one of 2 temperatures, 4°C or room temperature (RT≈22°C) in PBS and tested bioactivity by Luciferase expression in vitro. We found that bioactivity was stable (to within the high variability of this bioactivity assay performed on cells seeded on different days) for both C24 and MC3 LNPs for 2 weeks when stored at 4°C but declined by 20-40% when stored at room temperature (Fig 7a). We did not find any change in LNP size or accessibility to Ribogreen for either storage temperature over 2 weeks (Fig 7b,c). We calculated the loss of mRNA integrity of the 2,061 nucleotide long FLuc over time using a model derived from data characterizing the temperature-and pH-dependence of the base-catalyzed phosphodiester transesterification and cleavage of mRNA backbone (Eq e from 42 ). The model predicted half-lives of 2,300 days at 4°C and 125 days at 22°C (RT) and 11 days at 37°C, all at pH 7.4 (Fig 7d). mRNA backbone cleavage is thereby expected to be exquisitely sensitive to temperature and these trends are qualitatively consistent with previous findings for free mRNA (Fig E2 from 43 and Fig 2 from 44 ). We then stored additional mRNA LNPs in PBS at 4°C and at 22°C (RT) as well as at 37°C to accelerate degradation and extracted mRNA in chloroform/methanol to analyze mRNA integrity on a microfluidic electrophoresis device. We found mRNA integrity to be maintained for at least 19 days when stored at 4°C while the higher temperatures could produce mRNA cleavage (Fig 7e). mRNA integrity was estimated by the area under the curve (AUC) corresponding to the FLuc mRNA peak normalized to the day 0 value of C24 and MC3 LNPs taken together. The C24 LNP appeared to maintain higher levels of mRNA integrity than the MC3 LNP at the higher temperatures although sample numbers were too low to permit statistical analyses (Fig 7f).
Increased mRNA cleavage at higher temperatures is consistent with the model of based-mediated 13 phosphodiester transesterification reaction mechanism of mRNA cleavage 42 (Fig 7d) but may exhibit a dependence on the LNP environment and on the specific structure of the ionizable lipid in the LNP. The LNP environment may for example accelerate mRNA degradation by the proton partitioning phenomena we found to be responsible for the difference in pKa of the ionizable lipid in aqueous media versus in the LNP. Namely, the higher proton solvation energy in the lipid environment will exclude protons and raise the pH in the LNP compared to the external aqueous phase and could thereby accelerate base-mediated mRNA cleavage as predicted by the pHdependence of the model of base-catalyzed phosphodiester transesterification and cleavage of the mRNA backbone 42 .

Conclusions
Through a rational design process that includes a theoretical ionization assessment of ionizable lipid candidates, followed by structural considerations that include high levels of branching alkyl tails and quickly degradable primary esters, we identified and synthesized a new ionizable lipid, C24, with a multiprotic trivalent head group. The ionization behavior of the C24 LNP was consistent with molecular ionization characteristics reflecting the multiple protonation stages of the multivalent head group, resulting in an overall doubling of protonation in the endosomal pH range compared to the MC3 reference mRNA LNP. Intramuscular administration in mouse models showed that mRNA translation was 4 fold higher for C24 versus MC3 and this translated to 10 fold higher neutralizing titers in immunogenicity studies and a significantly greater protective capacity in a SARS-CoV-2 challenge model with protection at a low 0.25 µg prime/boost dose in mice. Additional important observations for C24 were a 6 fold lower level of systemic biodistribution compared to MC3 and lower levels of inflammation at the injected site indicating the potential for lower reactogenicity and a wider therapeutic window if these results translate to larger animal models and to human studies. Finally, we found the C24 LNP to maintain bioactivity and mRNA integrity for at least 19 days. When stored at higher temperatures our data suggests base-mediated mRNA cleavage is likely responsible for loss of mRNA integrity and bioactivity in LNPs. The important improvements in potency, reactogenicity and storage properties of the C24 LNP motivate further preclinical studies for eventual use in human mRNA vaccine studies.

NMR measurement of pKa of water-soluble ionizable lipid analogues
The pH-dependence of proton NMR chemical shifts was used to measure the pKa's of the ionizable lipid water-soluble analogues (WSA) following published methods 5,6 . Chemical shifts of piperazine, imidazole, 2-chloroacetic acid and acetic acid were used as internal pH indicators. ionization-state-dependent in a way similar to a polyelectrolyte 34,51 . TNS data did not require the extended model since TNS dye binding only detects LNP surface charge. The isoelectric pI was the pH found by interpolating zeta potential to zero.

Cryoelectron miscroscopy
Grids for electron microscopy were plunge-frozen using a Vitrobot in 50µl injected into the medial gastrocnemius muscle The two injections were spaced 3 weeks apart and blood was collected through the retro-orbital route one day prior to the first injection (Pre-bleed), prior to the second injection (Prime) and 2 weeks after the second injection (Boost).
Serum was separated from blood following an incubation period of 30 minutes at room temperature, and samples were centrifuged at 10,000 g for 5 minutes in a non-refrigerated were challenged 2 weeks after the second immunization with SARS-CoV-2 by intranasal administration at dose of 5X10 5 pfu and followed until death or euthanasia criteria were met.
Weight and temperature were recorded daily. Two of the 5 mice in each group were sacrificed on day 5 post-challenge to assess lung viral titers by plaque assay. Lung tissues were homogenized and spun down. The supernatant recovered for assessment of viral load by plaque assay. The remaining three mice were monitored daily for signs of morbidity and mortality. Weight and temperature reading were also recorded daily for the surviving mice until the end of the study.
For PRNT50 assays, mouse sera were diluted 1:10 in DMEM (supplemented with 5% FBE, 1% L-glutamate and 20 U/mL penicillin, and 20 μg/mL streptomycin). Serial two-fold dilutions were then prepared from the 1:10 dilution and mixed with 100 pfu of SARS-CoV-2 virus and incubated at 37˚C for 1h. The sera/virus mixture was then overlayed onto a confluent layer of Vero cells in a 12-well plate format and incubated for 1h at 37˚C incubator with 5% CO2. The inoculated wells were then overlaid with a 1:

Statistics and Reproducibility
The    b) The ribogreen assay showed slightly lower levels of mRNA inaccessible to ribogreen for C24 versus MC3. We performed this assay using 2 standard curves, with and without triton. When only one standard curve with triton is used (such as in most publications to date) the mRNA inaccessible to ribogreen increases by ~8% shown in the yellow hatched portions. This assay is often interpreted as % encapsulation efficiency versus % mRNA inaccessible to ribogreen shown here. The latter is more accurate since it has been shown that dye accessibility of mRNA does not indicate that the mRNA is free and unencapsulated 35 . It may however reflect a different packing and internal structure of the LNP. c) mRNA copy number was calculated using a previously published molecular volume model 20    half-lives at pH 7.4 of 2,300 days at 4°C, 125 days at RT and 11 days at 37°C. e) mRNA was extracted from LNPs using chloroform/methanol just after being produced on day 0 as well as after 19 days of storage in PBS at 4°C, RT and 37°C and analyzed by microfluidic electrophoresis. No detectable degradation was found after 19 days at 4°C (compared to day 0) while storage at higher temperatures (RT and 37°C) produced increasing amounts of mRNA cleavage that was qualitatively consistent with higher degradation at higher temperature predicted by the model in (d). The least degraded example of the 3 LNPs measured for each condition is shown in (e). The peak below 20s is a standard while the peak near 40s above the main mRNA Fluc peak is the product of an untemplated extension that is not amenable to elimination using the cellulose method used to purify the mRNA 45 . mRNA integrity was estimated by the area under the curve (AUC) corresponding to the FLuc mRNA peak and was normalized to the average day 0 value for C24 and MC3 LNPs taken together. f) mRNA integrity was stable for at least 19 days at 4°C and declined at higher temperatures but less so for C24 than for MC3. N=3 or for some N=2 due to technical irregularities in the trace of some samples that precluded quantification.