
==== Front
Mol Ther Methods Clin Dev
Mol Ther Methods Clin Dev
Molecular Therapy. Methods & Clinical Development
2329-0501
American Society of Gene & Cell Therapy

S2329-0501(24)00115-3
10.1016/j.omtm.2024.101299
101299
Original Article
Toward a large-batch manufacturing process for silicon-stabilized lipid nanoparticles: A highly customizable RNA delivery platform
Saffie-Siebert Suzanne saffie@sisaf.com
1∗
Torabi-Pour Nissim 1
Gibson Andrew 1
Sutera Flavia Maria 1
Dehsorkhi Ashkan 1
Baran-Rachwalska Paulina 1
Quinn Skye 1
1 SiSaf Ltd, Surrey Research Park, Guildford GU2 7RE, UK
∗ Corresponding author: Suzanne Saffie-Siebert, SiSaf Ltd, Surrey Research Park, Guildford GU2 7RE, UK. saffie@sisaf.com
17 7 2024
12 9 2024
17 7 2024
32 3 10129915 1 2024
12 7 2024
Crown Copyright © 2024 Published by Elsevier Inc. on behalf of The American Society of Gene and Cell Therapy.
2024

https://creativecommons.org/licenses/by-nc-nd/4.0/ This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
While lipid nanoparticles (LNPs) are a key enabling technology for RNA-based therapeutics, some outstanding challenges hinder their wider clinical translation and use, particularly in terms of RNA stability and limited shelf life. In response to these limitations, we developed silicon-stabilized hybrid lipid nanoparticles (sshLNPs) as a next-generation nanocarrier with improved physical and temperature stability, as well as the highly advantageous capacity for “post-hoc loading” of RNA. Nevertheless, previously reported sshLNP formulations were produced using lipid thin film hydration, making scale-up impractical. To realize the potential of this emerging delivery platform, a manufacturing process enabling multikilogram batch sizes was required for successful clinical translation and deployment at scale. This was achieved by developing a revised protocol based on solvent injection mixing and incorporating other process adjustments to enable in-flow extrusion of multiliter volumes, while ensuring sshLNPs with the desired characteristics. Optimized procedures for nanoparticle formation, extrusion, and tangential flow filtration (to remove residual organic solvent) currently enable production of 2 kg finished batches. Importantly, sshLNPs produced via the modified large-scale workflow show equivalent physical and functional properties to those derived from the earlier small-scale methods, paving the way for GMP manufacturing protocols to enable vital translational clinical studies.

Graphical abstract

Saffie-Siebert and colleagues describe a manufacturing method using solvent injection mixing, which enables the kilogram-scale production of silicon-stabilized lipid nanoparticles (sshLNPs), a promising next-generation delivery platform for therapeutic RNA, with equivalent physical and functional properties to sshLNPs produced at small batch sizes via lipid thin film hydration.

Keywords

lipid nanoparticles
silicon nanoparticles
gene therapy
RNA delivery
process development
Bio-Courier
sshLNP
==== Body
pmcIntroduction

Lipid nanoparticles (LNPs) have revolutionized the field of nucleic acid therapeutics by overcoming significant challenges in the cellular delivery of DNA and RNA. They particularly rose to prominence as the mRNA delivery technology in the Moderna and Pfizer/BioNTech COVID-19 vaccines,1 and currently represent the predominant delivery mechanism in over 200 ongoing clinical trials of other RNA-based drugs.2 Despite these impressive successes, current LNP formulations (and, more broadly, liposomal vesicles in general) face some recognized shortcomings that must be overcome to realize improved clinical translation of RNA-based therapeutics.3,4

One crucial limitation concerns the inherent chemical and metabolic lability of RNA, particularly its hydrolytic instability,5 which presents significant challenges for the larger mRNA constructs representing the majority of products in clinical development.2 As a result, most marketed nucleic acid therapeutics contain extensive chemical modifications to enhance stability and efficacy.6,7 When LNPs are used as the delivery vehicle, another issue that emerges is the need to introduce the RNA payload early in the production process. Effectively, LNPs must be formed around the RNA because there is no easy way of encapsulating it later.8,9 This restricts batch sizes in commercial manufacturing due to RNA instability,10 and it can also compromise product quality. Tozinameran, the mRNA-based Pfizer/BioNTech COVID-19 vaccine, is estimated to retain only 70% of initial mRNA integrity at the end of the manufacturing process.11 Moreover, the currently marketed mRNA products (tozinameran, elasomeran, and derivatives) require cold chain distribution and storage to make them clinically and commercially viable. This is not ideal for sustainability or global accessibility in the long run.12,13

We recently developed silicon-stabilized hybrid lipid nanoparticles (sshLNPs), marketed under the name Bio-Courier, as an alternative nanocarrier for therapeutic RNA (Figure 1) that can address the aforementioned challenges. Incorporation of hydrolyzable mesoporous silicon modifies the properties of the resultant nanoparticles in unique and advantageous ways, particularly conferring superior long-term stability14 and readily enabling tuning of the formulation for targeted nucleic acid delivery. Critically, sshLNPs can be manufactured in “empty” form, shipped at ambient temperature (either in liquid suspension or as a lyophilized powder), and loaded with the desired therapeutic RNA at the point of use. This has been established for siRNAs15,16 and is under investigation for mRNA payloads, as reported here. Specifically, previous in vivo studies have demonstrated that the loading step is straightforward and takes only 1 h at room temperature.15,16 We further showed that, after loading, sshLNPs can functionally deliver siRNA to cornea15 and bone16 in mice to achieve robust target gene knockdown. In the latter case, a therapeutic effect was observed through selective knockdown of the mutant Clcn7G213R allele in a mouse model of autosomal dominant osteopetrosis (ADO2).Figure 1 Schematic overview of conventional LNPs (left) and sshLNPs (right), highlighting key features of the latter that are responsible for their unique properties

Successful clinical translation of the Bio-Courier technology requires a reliable manufacturing process that can be deployed at scale. The robust structural integrity of sshLNPs under shear stress14 already makes them compatible with standard fill/finish workflows, unlike conventional LNPs that require process adaptations to avoid aggregation or detrimental physical damage.9,17 However, other aspects of the workflow require improvement. To permit incorporation of the silicon component and produce sshLNPs with the desired properties, we originally developed a process16,18,19 based on the lipid thin film hydration technique that was historically used to prepare standard LNPs.20 Although this method has been scaled up for effective delivery of various active pharmaceutical ingredients, it is unsuitable for manufacture of large batches. This paper charts the development of a revised method for production of sshLNPs on a multikilogram scale, paving the way for future clinical studies with application in both high-demand and personalized pharmaceutical products.

Results

As noted above, small-scale production of sshLNPs was carried out previously using a workflow based on lipid thin film hydration followed by size extrusion (Figure 2A, method 1). Two successive in-flow extrusion steps through membranes with successively smaller pore size (0.4 and 0.1 μm) were used to reduce operating pressure on the membranes and ensure that the final particles had the desired properties. Trehalose (1 mg/mL) and glycine (0.5 mg/mL) were also included as excipients in the aqueous phase since we found that they could further stabilize the sshLNPs, presumably through coordination to the silicon nanoparticles (SiNPs). Both of these agents are capable of forming hydrogen bonding networks with silanols on the silicon surface, and they favor dispersibility of sshLNPs when used at appropriate ratios. The amino acid excipient is also included to modulate the rate of hydrolysis of silicon, promoting formation of orthosilicic acid rather than insoluble polymeric silicon species.21Figure 2 Removing the first solvent evaporation step in production of sshLNPs

(A) Workflow of the original lipid thin film hydration method (method 1) where organic solvent is evaporated from activated SiNPs, and the revised protocol (method 2) where the activated SiNP suspension is added directly to the aqueous phase. (B) More sedimentation was observed with method 2 (lower image) after lipid film hydration. (C) DLS results indicated lower PDI and higher zeta potential for sshLNPs produced using the revised method. Note: “DLS/Zeta” indicates points at which measurements were made of hydrodynamic size, polydispersity index (PDI), and zeta potential. Panel C report mean ± SD for samples analyzed in triplicates (n = 3).

To develop this method for large-scale application, two solvent evaporation steps needed to be eliminated; namely, removal of MeOH from the SiNP suspension after activation, and from the dissolved lipids to create the thin film. The first of these steps was avoided by mixing the activated SiNP suspension directly into the aqueous trehalose/glycine solution prior to lipid thin film hydration (Figure 2A, method 2). This approach resulted in an initial sshLNP suspension with visibly more sediment than before (Figure 2B), but produced particles with lower polydispersity index (PDI) and slightly higher zeta potential after extrusion (Figure 2C). Ideally, finished sshLNPs should display a zeta potential of +50 to +70 mV, indicating good to excellent colloidal stability and minimal tendency for aggregation (due to charge repulsion).

To replace the lipid thin film hydration step, direct slow injection of the methanolic lipid solution into the aqueous SiNP suspension was attempted (Figure 3A, method 3). Less sedimentation was seen with this approach, and the resultant sample of sshLNPs showed comparable average size (102 ± 0.17 nm), PDI (0.135 ± 0.01), and zeta potential (56.0 ± 0.81 mV) to previous batches (Figure 3B). This established feasibility for the more scalable direct injection mixing technique, which was therefore selected as the method of choice for sshLNP formation moving forward. To ensure reproducibility, we introduced two in-process checks (IPCs) at this stage (Figure 3A): a visual check for complete dissolution of lipids before mixing (IPC 1), and specified reference ranges for post-extrusion dynamic light scattering (DLS) parameters: i.e., average hydrodynamic size 75–140 nm, PDI 0.1–0.2, and zeta potential +50 to +70 mV (IPC 2). These ranges were established based on prior optimization of the lipid thin film hydration method to produce functional sshLNPs. If required, the 0.1 μm extrusion step is repeated following IPC 2.Figure 3 Removing the second evaporation step and optimizing a solvent injection mixing method

(A) Workflow of the initial direct injection mixing approach (method 3), including two in-process checks (IPCs). (B) Less sedimentation was seen compared with lipid thin film hydration. (C) Modified workflows incorporating an additional 0.8 μm extrusion step and other adjustments as indicated. (D) The additional extrusion step (method 4) led to smaller sshLNPs with similar PDI and zeta potential. (E) Prefiltration before extrusion (method 6, right) removed insoluble aggregates that would otherwise accumulate on the 0.8 μm extrusion membrane (method 4, left), although the 0.4 μm membrane appeared to remove additional aggregated material. (F) Introducing the prefiltration step did not impact the properties of the final sshLNPs. Graphs in panels B, D, and F report mean ± SD for Size, PDI and zeta potential for samples assessed in triplicates (n = 3).

Other adjustments were investigated before scale-up (Figure 3C, methods 4–6). Most notably, an additional in-flow extrusion step (0.8 μm pore size) was trialed to remove larger SiNP aggregates and successfully led to sshLNPs with reduced average size (Figure 3D, method 4). Adding the trehalose/glycine solution to the activated methanolic suspension of SiNPs, rather than the other way round, did not affect the results (Figure S1, method 5). Including a 0.8 μm prefiltration step after initial sshLNP formation reduced buildup of aggregates on the extrusion membranes (Figure 3E, method 6), significantly lowering operating pressure on the 0.8 μm membrane (Figure S2) and essentially leading to sshLNPs with unaltered properties (Figure 3F).

It should also be noted that MeOH was used as organic solvent, rather than EtOH, because it leads to superior aqueous colloidal stability of finished sshLNPs. For commercial manufacturing, it is essential that MeOH is removed from the finished product such that any residual levels are below regulatory limits (e.g., <3,000 ppm in pharmaceutical products in Europe22). For this purpose, we employed a tangential flow filtration (TFF) setup as depicted in Figure 4A. Initial feasibility studies (up to 100 mL scale) used a TFF cassette (Figure 4B) with 100 kDa MWCO and 0.02 m2 membrane surface area. Since trehalose and glycine are also removed from the solution under these conditions, a second pump was used to infuse an aqueous diafiltration (DF) solution of 0.1 mg/mL trehalose and 0.05 mg/mL glycine at a constant flow rate matching that of filtrate generation. Performing DF against 10 volumes of this solution lowered the MeOH content from ∼20% (v/v) after extrusion to well within acceptable regulatory limits (∼200 ppm based on 1H NMR) while maintaining the intended levels of trehalose/glycine. The final process included an initial ultrafiltration (UF) step to reduce the input volume by 50% before performing DF, to produce a finished sshLNP suspension with the desired silicon concentration (see below) and low residual MeOH levels (∼450 ppm; Figure 4C).Figure 4 Purification of sshLNPs by TFF

(A) Arrangement of the experimental setup illustrating how a second pump (pump 2) was included to enable both ultrafiltration (UF) and diafiltration (DF) operations. (B) Schematic showing how the TFF cassette functions. (C) Representative result that illustrates successful removal of MeOH from sshLNPs produced via method 5, as judged by 1H NMR. (D and E) The TFF process was found to slightly modify the DLS characteristics (D) and lipid content (E) of the particles, but these variations lie within relevant reference ranges and were therefore not considered problematic. Error bars within the graphs in panel D and E show mean values ± SD based on three independent measurements.

After TFF, the DLS properties of the extruded sshLNPs were largely maintained (Figure 4D). However, relative to reference values, all three measured characteristics were slightly improved for the combined UF plus DF procedure versus DF alone. We also developed an HPLC method to measure the extent of lipid recovery (expressed as percent of input for each lipid component; Figure S3). The results showed a notable drop in lipid content when applying DF only, of around 10% recovery for each lipid (Figure 4E). This was largely mitigated by applying the modified TFF procedure combining UF and DF, but is not currently a concern as some batch-to-batch variation is seen but lies within acceptable limits.

We also measured the silicon content of final sshLNP samples using inductively coupled plasma-optical emission spectroscopy (ICP-OES). For the batch subjected to TFF with DF only (Figures 4C and 4D), the silicon content was 1.19 mg/L, compared with a target value of 2.1 mg/L based on sshLNPs prepared using the original lipid thin film hydration method (Figure 2A). This explains why the 2× concentration (i.e., UF) step was introduced into the TFF workflow, leading to sshLNPs with a silicon content of 2.18 mg/L. Together, the aforementioned analytical methods were used to guide further optimization efforts on moving toward scale-up. They are now being further developed and validated for phase appropriate quality control during commercial manufacture. As part of this development process, the original 1H NMR method for determining residual MeOH content has been superseded by a headspace GC assay that was recently disclosed elsewhere.16

The first large-scale test run (Figure 5A) used a protocol incorporating the three in-flow extrusion steps (i.e., a scale-up of method 4 using the modification introduced in method 5) but omitting TFF at this stage. Lipid stock solutions were now prepared at 10 mg/mL rather than 5 mg/mL as before, to limit the volume of MeOH in the system. On a 1 L scale, the main issue encountered was greater accumulation of insoluble aggregates on the extrusion membranes (Figure 5B), even though larger membranes were used (47 mm diameter versus 25 mm previously). This resulted in higher operating pressures during extrusion (Figure 5C), although the DLS properties of the sample were not adversely affected (Figure 5D). We proceeded to demonstrate production of a 2 L batch incorporating the two-stage TFF protocol developed earlier (Figure 6A), but in this case using a cassette with larger membrane surface area (0.5 m2). As before, the main issue encountered was during extrusion, where the membranes became saturated with insoluble aggregates (Figure 6B) and had to be changed midway through the procedure due to pressure fluctuations. The post-extrusion sample exhibited suboptimal visual and DLS characteristics (Figure 6C), but closer process monitoring revealed that that average particle size had gradually increased with extrusion volume (Figure 6D) and only exceeded the reference value of 108 nm after around 1.5 L of the total volume of 4 L had been processed. Following TFF, residual MeOH was below detection limits. When a 100 mL aliquot of the sample was subjected to an additional extrusion step through 3 × 0.1 μm membranes, the average particle size dropped from 129 ± 0.44 to 105 ± 0.38 nm, and the PDI from 0.17 ± 0.02 to 0.11 ± 0.01 (both well within reference ranges), confirming overloading of the extrusion membranes as the problem during processing of the large batch.Figure 5 Initial large-scale (1 L) run to produce sshLNPs

(A) Schematic of the workflow used; a scale-up of method 5. (B) Representative images of extrusion membranes and sample aliquots (lower-right quadrant), showing buildup of aggregates despite the use of larger membranes as well as some remaining insoluble material after the first extrusion step. (C) Accumulation of aggregates on the 0.8 and 0.4 μm membranes resulted in substantially higher operating pressure than on a 50 mL scale (method 4). (D) DLS characteristics of the sshLNPs were not adversely affected and were within reference ranges. Panel D error bars are represented as mean ± SD for Size, PDI and zeta potential for samples assessed in triplicates ( n = 3).

Figure 6 Production of a 2 L batch of sshLNPs

(A) Schematic of the workflow used. Note that the TFF step contained two stages: UF to concentrate the sample from 4 to 2 L, followed by DF to remove MeOH and free lipids. (B) The extrusion steps were compromised by significant buildup of aggregates on the membranes, compromising their performance. (C) For the finished batch, average size and PDI were outside the reference ranges. (D) This was due to deterioration of the extrusion membranes with excessive accumulation of aggregates, indicating the need for a prefiltration step after activation of SiNPs when manufacturing sshLNPs at scale. Graphs in panel C and D were measured in triplicates (n = 3) with error bars represented as mean ± SD.

In view of these results, we modified the procedure to reduce sedimentation during sshLNP formation. In a small-scale test run omitting TFF, when the activated suspension of SiNPs was filtered prior to mixing with the trehalose/glycine solution (Figure 7A, method 7), aggregates were successfully removed and no significant deposition of insoluble material on the extrusion membranes was observed (Figure 7B). Operating pressure was lower during extrusion (Figure 7C) and DLS properties of the resultant sshLNPs were well within reference ranges (Figure 7D). To implement this method at scale, two sets of the larger 47 mm extrusion membranes were used, arranged in parallel (Figures 7A and S4), successfully enabling production of a 1 L batch of sshLNPs via the optimized protocol (Figure 7E). In the current implementation, a final 0.2 μm filtration step is additionally used for bioburden reduction. The full workflow has been used for final batches up to 2 L (i.e., 4 L working volume prior to TFF) and further refinements are ongoing to achieve 5 L final batches. With the current workflow, MeOH is undetectable in finished batches by 1H NMR and is present at levels below the limit of detection by headspace GC (approximately 10 ppm).15Figure 7 Prefiltering the activated SiNP suspension improves the process

(A) Schematic of the modified workflow, designated method 7. (B) In this case, there was no significant accumulation of aggregates on the extrusion membranes. (C) This was reflected in considerably lower operating pressures during extrusion, compared with the protocol without prefiltration. (D) The modified procedure also led to sshLNPs with DLS properties that were well within reference ranges, both on a small scale and for a 1 L batch size. Panel D graphs report mean ± SD for samples analyzed in triplicates (n = 3).

After developing the kilogram-scale manufacturing protocol, it was also essential to establish equivalent properties between sshLNPs produced by the original lipid thin film hydration technique (method 2), at smaller scale, and the optimized large-batch process (method 7). We found that scale-up led to slightly smaller particles on average (Figure 8A), but the PDI and zeta potential were essentially identical (Figures 8A and 8B). Both batches showed excellent mRNA encapsulation efficiency, as determined by the RiboGreen assay (Figure 8C). In cellular transfection experiments, sshLNPs produced by either route efficiently introduced an mRNA encoding firefly luciferase (fLuc) into HEK293 cells, and significantly higher transgene expression was observed after 24 h compared with Lipofectamine 3000 as reference (Figure 8D). The higher luminescence observed with large-batch versus small-batch sshLNPs may reflect their slightly smaller average size, which could conceivably improve transfection efficiency (although this requires further investigation to elucidate the relevant factors). When used for topical ocular delivery of siRNA, small-scale and large-scale sshLNPs demonstrated comparable in vivo corneal uptake in a mouse model (Figure S5). Interestingly, pilot experiments employing systemic administration in ADO2 model mice have suggested that siRNA-loaded large-batch sshLNPs (produced using method 7) may achieve superior gene knockdown efficiency in femur compared with sshLNPs derived from method 2, although this has yet to be confirmed. Overall, sshLNPs manufactured by the large-scale process developed here can be regarded as physically and functionally equivalent to those produced via the original method.Figure 8 Bio-Courier sshLNPs produced at small scale (using lipid thin film hydration, method 2) and large scale (using method 7) exhibited similar physical and functional properties

(A) The large-scale process produced significantly smaller (p < 0.0001) final particles with comparable PDI. (B) Zeta potential was unaffected by scale-up. (C) Both methods gave sshLNPs with equivalent RNA encapsulation efficiency. (D) Large-scale manufacture also maintained transfection efficiency of HEK293 cells with an mRNA encoding firefly luciferase (fLuc). Luminescence was measured to determine fLuc expression levels at 24 h post-transfection. Lipofectamine 2000 was used as positive control and untreated cells as negative control. (A–C) Mean ± SD for samples analyzed in triplicate, while (D) shows mean ± SD for three independent biological replicates.

Discussion

Compelling proof of principle for the therapeutic potential of sshLNPs as an RNA delivery mechanism was previously established through functional studies in mice. Topical application of siRNA-loaded sshLNPs elicited robust gene silencing in cornea,15 while knockdown of mutant Clcn7 in bone was achieved with intraperitoneal administration in ADO2 model mice16 and fully rescued the bone phenotype. Nevertheless, clinical translation to realize the full potential of sshLNPs requires a reliable manufacturing process capable of producing particles with the relevant properties on a multikilogram scale. This study set out to develop such a process and is particularly timely given the flexibility of the Bio-Courier platform.

The convergent manufacturing approach produces platform intermediate (i.e., “empty”) sshLNPs, amenable to subsequent nucleic acid loading prior to fill/finish operations that can be separated by considerable time and distance from original manufacture. This capability is thought to be conferred by the SiNPs, some of which remain accessible at the surface and stabilize the lipid membrane. Their presence mitigates the well-known tendency for LNP aggregation or fusion over time (i.e., Ostwald ripening), which is a limiting factor for shelf life of current formulations8,23 and was seen in previous experiments comparing the stability of particles formulated without silicon14 with that of sshLNPs. Incorporation of SiNPs also allows for omission of the cholesterol component of conventional LNPs that accounts for ∼40 mol % of the lipid content in marketed RNA-LNP products.24 In sshLNPs, the interaction of the phospholipids with silicon provides enhanced structural integrity, reducing the risk of rupture during extrusion. Intriguingly, it also appears to produce final particles with an incompletely sealed lipid bilayer where the interior remains accessible for nucleic acid loading.

Another advantage of the silicon component is that it permits more flexibility in Bio-Courier formulations, in the sense that the lipid composition can readily be modified without having to make significant adjustments to the manufacturing process. For example, sshLNPs can be formulated without the PEGylated lipid which is an essential component of conventional LNPs, but is sometimes associated with safety and efficacy concerns. It is well known that PEGylation can induce anti-PEG antibodies, potentially provoking premature release of the RNA payload through antibody binding.25,26 Thus, non-PEGylated sshLNPs may be a clinically useful option for individuals with a history of PEG hypersensitivity reactions.27,28

Moreover, the compatibility of sshLNPs with point-of-care RNA loading minimizes degradation of nucleic acid drugs while providing the opportunity for late-stage customization and personalization of therapeutic formulations. In fact, this “post-hoc loading” concept has recently been highlighted as a potential strategy toward overcoming remaining challenges in the RNA medicines field.29 Low delivery efficiency, short shelf life, and high market barriers to entry (i.e., cost of development and manufacture of optimized LNP formulations) are particularly recognized as current limitations that hinder the growth of the field,4 and sshLNPs can potentially answer all of these challenges.

Since 2020, priority seems to have been given to addressing the cold chain requirements and limited shelf life encountered with COVID-19 vaccines. Notable recent innovations have resulted in improved lyophilized mRNA-LNP formulations that maintain stability and transfection efficiency at 4°C (or even room temperature) for several weeks.30,31,32 However, commercial scalability of these workflows has yet to be established, and they do not resolve the issue of having to incorporate the RNA component early in the production process. Extensive optimization of the LNP formulation may therefore still be necessary on a case-by-case basis. As an example, identifying the optimal formulation for patisiran—the first FDA-approved RNA therapeutic delivered by LNPs—required screening of over 300 ionizable lipids alone.33 In contrast, our study with ADO2 mice required an initial screen of just seven sshLNP formulations to identify a promising lead candidate for siRNA delivery to bone.34 Thus, the Bio-Courier platform shows promise for accelerated clinical translation of RNA therapeutics.

Building on this point, the stabilizing effect of sshLNPs is also strong enough to render chemical modification of the RNA unnecessary, as illustrated by the aforementioned in vivo studies15,16 that used unmodified siRNA. In contrast, all of the current FDA-approved oligonucleotide therapeutics contain chemical modifications,35 including the five siRNA-based products.36 Use of unmodified (or minimally modified) RNAs with sshLNPs is another factor that could drive increased translation, by reducing the considerable time and costs to develop current constructs that often require complex series of chemical modifications.

For clinical application, establishing safety is essential. In prior rodent studies,15,16 no treatment-emergent adverse events were encountered for either topical or systemic administration of siRNA delivered by sshLNPs, with repeat treatment over several weeks. A detailed safety evaluation in siRNA-treated ADO2 model mice found no gross abnormalities in major organs or common serum toxicological markers (alanine transaminase, alkaline phosphatase, and urea) for the full duration of the experiment.16 Also, effective silencing of target genes has been demonstrated in cornea with topical delivery15 or in femur with subcutaneous or intraperitoneal administration of siRNA-loaded sshLNPs.16,37 For mRNA delivery, preliminary animal experiments show similar safety and efficacy profiles and the results will be reported in due course. We believe that exploring the in vivo transfection efficiency of sshLNPs compared with conventional LNPs would be an intriguing avenue for further investigation. Moreover, the first clinical study with sshLNPs is ongoing and concerns a phase 2b trial in alopecia areata, where our platform is being used to reformulate a small-molecule drug for enhanced topical delivery.38

In sum, sshLNPs show significant promise as a platform technology for improved RNA delivery, but previously reported methods for their production are not easily amenable to scale-up. Here, we have documented the development of a kilogram-scale manufacturing process to facilitate clinical studies with our in-house pipeline of RNA therapeutics38 and beyond. Further modifications for GMP compliance are ongoing to support wider clinical translation. In particular, we recognize considerable potential for sshLNPs in “kit-based” applications with therapeutic RNA, allowing growth of the field not only in vaccines but also in personalized medicine through point-of-care formulation and administration.

Materials and methods

SiNPs were sourced commercially as electrochemically etched powder (≥98% purity, from American Elements, Los Angeles, CA) as undoped silicon. All experiments documented in this paper employed undoped SiNPs, although it should be noted that an identical manufacturing process can be used with doped silicon. Trehalose and glycine (USP/PhEur specification) were purchased from Merck. The following lipids were supplied by Lipoid, Ludwigshafen, Germany: 1,2-dioleoyl-3-trimethylammonium propane chloride (DOTAP chloride) (CAS no. 132172-61-3); 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) (CAS no. 4004-05-1); and N-(carbonyl-methoxypolyethylenglycol-2000)-1,2-distearoyl-sn-glycero-3-phosphoethanolamine, sodium salt (mPEG2000-DSPE) (CAS no. 147867-65-0). All procedures used nuclease free water that was produced in-house. Measurements of mean hydrodynamic particle size, PDI, and zeta potential were made using a Zetasizer Pro instrument (Malvern Panalytical, UK) as reported in previous studies.15,16 In brief, size measurements were made at 25°C after diluting 10 μL of test sample with 990 μL water (i.e., 1:100 dilution) in compatible disposable cuvettes (DTS0012, Malvern Panalytical). Backscatter detection was used for data acquisition at an angle of 173°. To improve signal-to-noise ratio, five scans were performed and the best three selected for further analysis (n = 3). Particle size (Z average, nm) and PDI were determined by fitting the correlation function using the cumulant method. Zeta potential was measured by diluting samples 1:10 with water to a final volume of 1 mL, then transferring into a folded capillary zeta cell (DTS1070, Malvern Panalytical) via a 1 mL syringe. As before, five scans were performed at 25°C and the best three selected for analysis (n = 3). Zeta potential values were computed using the Zetasizer software, which applies the Henry equation (assuming the Smoluchowski model) to measurements of electrophoretic mobility obtained via laser Doppler velocimetry.

Formation of sshLNPs through lipid thin film hydration

In method 1 (Figure 2, SiNPs were activated by suspending them in MeOH (20 mg/mL) and stirring for 0.5 h followed by slow evaporation, as described previously.15 They were then added at 1 mg/mL into an aqueous solution of trehalose (1 mg/mL) and glycine (0.5 mg/mL), and the mixture was sonicated for 1 h at 50°C. Separately, lipid solutions of DOTAP chloride, DOPE, and mPEG2000-DSPE were prepared at 5 mg/mL in MeOH and sonicated for 0.5 h at 40°C. Aliquots of these solutions (1.44, 1.46, and 0.29 mL, respectively) were transferred to a 10 mL round-bottomed flask and mixed. The solvent was removed by rotary evaporation for 0.5 h at 40°C to generate the lipid thin film that was subsequently hydrated by adding 1 mL of the aqueous solution described above and incubating for 5 min at 60°C (i.e., lipid thin film hydration or dehydration-rehydration method). Water was added up to a final volume of 10 mL, then the mixture was extruded through 5 × 0.4 and 5 × 0.1 μm polycarbonate hydrophilic membranes (Whatman Nucleopore) arranged in series, using an Avanti Polar Lipid Extruder in combination with manual injection via a gastight Hamilton syringe. The membranes were prewashed with MeOH and water (20 mL each) prior to use. In method 2, the MeOH was not evaporated from activated SiNPs before further use. Instead, an aliquot of the activated suspension was added directly into an aqueous trehalose/glycine solution to achieve the same final concentrations as above: 1 mg/mL SiNPs, 1 mg/mL trehalose, and 0.5 mg/mL glycine. Formation and extrusion of sshLNPs were then carried out exactly as above.

Formation of sshLNPs through direct slow injection of lipids

In method 3 (Figure 3, SiNPs (20 mg) were activated in MeOH (1 mL) as above then combined with a solution of trehalose (20 mg) and glycine (10 mg) in water (19 mL). The mixture was sonicated for 1 h at 50°C. Separately, lipid stock solutions were prepared and combined as in method 1. The combined lipid solution was added to 1 mL of the aforementioned aqueous solution at 3.2 mL/min using a PSNE100 syringe pump (ProSense, Munich, Germany). The volume was made up to 10 mL with water and the solution was stirred for 0.5 h prior to extrusion, which was carried out as above.

Modified method with added 0.8 μm extrusion step

In method 4 (Figure 3D), an identical procedure was followed as for method 3, except that it was performed at 5× the scale (i.e., final volume 50 mL) and also incorporated 3 × 0.8 μm polycarbonate hydrophilic membranes (Whatman Nucleopore) in series before the 3 × 0.4 and 3 × 0.1 μm membranes during the extrusion step.

Other refinements to the small-scale protocol

In method 5 (Figures 3E and S1), the addition order used to prepare the aqueous solution was varied. The preprepared aqueous solution of trehalose/glycine was added directly to the activated methanolic suspension of SiNPs, rather than the other way round as before (in method 4). The final concentrations of SiNPs (1 mg/mL), trehalose (1 mg/mL), and glycine (0.5 mg/mL) in the intermediate aqueous solution were unchanged, and the rest of the procedure was performed as before. Method 6 was carried out at double the scale of method 4 (i.e., final volume 100 mL) and introduced two new variations. First, the lipid stock solutions were prepared at 10 mg/mL rather than 5 mg/mL, then aliquots of the DOTAP chloride (7.20 mL), DOPE (7.30 mL), and mPEG2000-DSPE (1.45 mL) were combined and added to the preprepared aqueous solution as above. After adjusting the volume to 100 mL with water, the mixture was prefiltered through a 0.8 μm hydrophilic polyethersulfone (PES) syringe filter and extruded as in method 4.

Diafiltration using a TFF setup

A SIUS PD 0.02 m2 (LP) HyS 100 kDa cassette (XP100LP2L; Repligen, Waltham, MA) was used for volumes up to 100 mL, and a SIUS 0.5 m2 100 kDa TFF cassette (NC1095082; Fisher Scientific, Waltham, MA) was employed for larger volumes. After an initial UF step where indicated (typically to 0.5× input volume), samples were diafiltered against 10 volumes of an aqueous solution containing 0.1 mg/mL trehalose and 0.05 mg/mL glycine, to preserve the concentration of these two excipients while removing MeOH and unbound lipids. This was achieved by using a second pump to introduce the trehalose/glycine solution into the flow path at the same flow rate as the main system pump (Figure 4A). Figures 4D and 4E show results for repeat runs of method 5 followed by TFF with DF only, or TFF with UF (to half of the initial volume) followed by DF.

Methanol content analysis

A calibration curve was constructed by measuring 1H NMR peak area (δ = 3.34 ppm39) in standard solutions of known MeOH concentration in water containing 10% D2O: 50, 100, 300, and 500 ppm (v/v; Figure 4B, inset). Pre-TFF and post-TFF samples were analyzed in H2O/D2O (90:10) and the MeOH content estimated by reference to the calibration curve. Due to their higher MeOH content, the pre-TFF samples were diluted by a factor of 1:1,000 before acquiring NMR spectra.

Lipid recovery analysis

Quantification of lipids relied on HPLC analysis using a Waters XBridge BEH Phenyl Column (130 Å, 5 μm, 4.6 × 150 mm) in conjunction with a charged aerosol detector. Buffer A was 40 mM NH4OAc and Buffer B was 100% MeOH, and the following elution gradient was used: 25% A for 1 min, ramp to 5% A over 6 min, hold at 5% A for 18 min, ramp to 25% A over 0.1 min, then hold at 25% A for 4.9 min (flow rate 1 mL/min). To enable quantification, calibration curves were constructed for DOTAP chloride, DOPE, and mPEG2000-DSPE (which exhibited retention times of 12.8, 15.1, and 20.8 min, respectively; see Figure S3). For DOTAP chloride and DOPE, standard solutions were prepared at 0.1, 0.2, 0.5, 0.7, and 0.9 mg/mL; and for mPEG2000-DSPE, the reference solutions had concentrations of 0.01, 0.05, 0.15, 0.2, and 0.3 mg/mL. Peak area was plotted against concentration to enable determination of lipid concentrations in test samples (see Figure S3 for an example). Lipid recovery was calculated as the ratio of the observed concentration to the theoretical concentration if all lipid had been incorporated into the sshLNPs, expressed as a percentage.

First large-scale test run to produce sshLNPs (1 L scale)

As illustrtaed in Figure 5, SiNPs (100 mg) were suspended in MeOH (5 mL) in a sterile 50 mL Falcon tube and activated by incubating for 0.5 h at room temperature. Then, a solution of trehalose (100 mg) and glycine (50 mg) in water (95 mL) was added directly to the SiNP suspension and the mixture was sonicated for 1 h at 50°C. Lipid stock solutions were prepared by dissolving DOTAP chloride (750 mg), DOPE (750 mg), and mPEG2000-DSPE (200 mg) in MeOH at 10 mg/mL and sonicating the resultant mixtures for 0.5 h at 40°C. Then, aliquots of the DOTAP chloride (72.5 mL), DOPE (73.0 mL), and mPEG2000-DSPE (14.5 mL) solutions were combined (i.e., to achieve final concentrations of 0.725, 0.730, and 0.145 mg/mL, respectively). The preprepared aqueous solution (100 mL) was transferred to a sterile 1 L bottle and the methanolic mixed lipid solution (160 mL) was injected at 6 mL/min using a Knauer K501 HPLC pump. After addition was complete, the volume was made up to 1 L with water and in-flow extrusion was performed using 3 × 0.8 μm, 3 × 0.4 μm, and 3 × 0.1 μm polycarbonate hydrophilic membranes (47 mm diameter, Whatman Nucleopore) arranged in series as above, at 60°C and using a flow rate of 85 mL/min.

Second large-scale test run to product sshLNPs (2 L scale)

As illustarted in Figure 6, SiNPs (400 mg) were activated in MeOH (20 mL) as described above. A solution of trehalose (400 mg) and glycine (200 mg) in water (380 mL) was prepared in a sterile bottle, then the suspension of activated SiNPs was added directly and the resultant dispersion sonicated for 1 h at 50°C. Stock solutions of DOTAP chloride (3.00 g), DOPE (3.00 g), and mPEG2000-DSPE (600 mg) were prepared at 10 mg/mL in MeOH as described above. The aqueous solution (400 mL) was transferred to a 5 L sterile bottle, then the combined lipid mixture (290, 292, and 58 mL, respectively, of the DOTAP chloride, DOPE, and mPEG2000-DSPE stock solutions; total volume 640 mL) was injected into the bottle at 6 mL/min using an HPLC pump. Water was added up to a final volume of 4 L and stirring was continued until the mixture was homogeneous. The crude mixture was stored at 4°C overnight, then in-flow extrusion was performed as described for the previous example. The TFF setup (as shown in Figure 4A, using a SIUS 0.5 m2 100 kDa cassette) was first depyrogenated with 0.2 M NaOH for 1 h, then the 4 L sample was concentrated to 2 L via UF and diafiltered against 10 volumes (i.e., 20 L) of 0.1 mg/mL trehalose and 0.05 mg/mL glycine in water. Final and intermediate sshLNP samples were analyzed using procedures as described above, except that for lipid content analysis, pre-TFF and post-TFF samples were diluted 1:2 and 1:4, respectively, with MeOH to match the range of the calibration curves.

Modified procedure for sshLNPs including prefiltration of aggregates

Method 7 was the same as method 5 (50 mL scale), except for an additional prefiltration step of the activated SiNPs (see Figure 7). Specifically, SiNPs (20 mg) were suspended in MeOH (1 mL) and activated by stirring for 0.5 h at room temperature, then the mixture was manually filtered through a 0.8 μm hydrophilic PES syringe filter. A preprepared solution of trehalose (20 mg) and glycine (10 mg) in water (19 mL) was added directly to the SiNP suspension and the dispersion was sonicated for 1 h at 50°C. Separately, stock solutions of DOTAP chloride, DOPE, and mPEG2000-DSPE were prepared at 10 mg/mL in MeOH with sonication for 0.5 h at 40°C. Aliquots of these solutions (3.63, 3.65, and 0.73 mL, respectively) were combined, and the mixed lipid solution was injected into a 5 mL aliquot of the aqueous solution at 3.2 mL/min using a syringe pump. The volume was made up to 50 mL with water and stirred for 0.5 h at room temperature prior to extrusion, which was carried out as for method 4.

Demonstration of a large batch with aggregate prefiltration (1 L scale)

Following the large-0scale workflow outlined in Figure 7A, SiNPs (100 mg) were activated by stirring in MeOH (5 mL) for 0.5 h at room temperature. To remove large aggregates, the suspension was manually filtered through three successive 0.8 μm hydrophilic PES syringe filters. The resultant sample was combined with a preprepared solution of trehalose (100 mg) and glycine (50 mg) in water (95 mL), with vigorous magnetic stirring (500 rpm) for 1 h at 50°C. Separately, stock solutions of DOTAP chloride (750 mg), DOPE (750 mg), and mPEG2000-DSPE (160 mg) were prepared at 10 mg/mL in MeOH with sonication for 0.5 h at 40°C. Aliquots of these solutions (72.5, 73.0, and 14.5 mL, respectively) were combined, then the mixed lipid solution was injected into the aqueous solution (in a 1 L polycarbonate container) at 6 mL/min using a Knauer BlueShadow 80P HPLC pump. The volume was made up to 1 L with water (added using the same pump at 100 mL/min, with stirring), and stirring was continued for 0.5 h at room temperature for homogenization. In-flow extrusion was performed using two sets of extrusion membranes (three each of 0.8, 0.4, and 0.1 μm pore size) arranged in parallel, as shown in Figure S4, at 60°C and a flow rate of 85 mL/min. After the TFF setup was depyrogenated with 0.2 M NaOH for 1 h and rinsed with water, MeOH was removed by DF as outlined above.

Assessment of RNA encapsulation efficiency

The Quant-iT RiboGreen Assay Kit (Fisher Scientific, UK) was used in accordance with the manufacturer’s instructions, as reported previously,16 after RNA loading for 1 h at room temperature using a 1:24 (w/w) ratio of mRNA to sshLNPs in water. Encapsulation efficiency was calculated as the ratio of encapsulated (i.e., total minus free) to total RNA, expressed as a percentage, after measuring free and total mRNA concentrations as follows. To determine the free RNA concentration, a standard curve was constructed using known mRNA concentrations (from 20 ng/mL to 1 μg/mL), then test samples of mRNA-loaded sshLNP formulations were diluted in 1× TE buffer to fall within the calibration range. Each sample (100 μL) was mixed with an equal volume of RiboGreen reagent in a 96-well plate, then a Varioskan LUX microplate reader (Thermo Fisher Scientific) was used for analysis (excitation 480 nm, emission 520 nm). In the case of total RNA concentration, standards and test samples were prepared as above in 1× TE buffer containing 2% Triton X-100, which ensured full extraction of mRNA from sshLNPs. Fluorescence intensity measurements were made as before, except that a baseline value corresponding to a blank sample (1× TE containing the same concentration of Triton X-100) was subtracted from all readings before further calculations.

Cellular transfection experiments

Cell culture and transfection were carried out as described elsewhere,16 with the latter using sshLNPs loaded with fLuc mRNA (Fisher Scientific) by prior incubation at room temperature for 1 h.15,16 Luminescence was recorded at 24 h post-transfection.

Data and code availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Supplemental information

Document S1. Figures S1–S5

Document S2. Article plus supplemental information

Acknowledgments

The work described in this paper was fully funded by SiSaf Ltd. Part of the process development work described in this paper was performed at Ardena Holdings NV.

Author contributions

S.S.-S. is the inventor of the Bio-Courier technology and led its development and scale-up. N.T.-P. devised various scale-up concepts including the direct injection method described in this paper. A.G., F.M.S., and A.D. directed and supervised the process development work by contract manufacturer Ardena. A.D. performed Bio-Courier formulation and characterization and S.Q. performed the in vitro transfection studies. S.S.-S., N.T.-P., A.G., F.M.S., P.B.-R., and A.D. contributed to drafting, revision, and final approval of the manuscript.

Declaration of interests

All authors are employees of SiSaf Ltd. S.S.-S. is CEO, board member, and shareholder of SiSaf Ltd. S.S.-S., N.T.-P., and A.D. are named inventors in a patent application filed by SiSaf Ltd. on the manufacturing process for silicon-stabilized lipid nanoparticles.

Supplemental information can be found online at https://doi.org/10.1016/j.omtm.2024.101299.
==== Refs
References

1 Wang C. Zhang Y. Dong Y. Lipid nanoparticle–mRNA formulations for therapeutic applications Acc. Chem. Res. 54 2021 4283 4293 34793124
2 Curreri A. Sankholkar D. Mitragotri S. Zhao Z. RNA therapeutics in the clinic Bioeng. Transl. Med. 8 2023 e10374
3 Moss K.H. Popova P. Hadrup S.R. Astakhova K. Taskova M. Lipid nanoparticles for delivery of therapeutic RNA oligonucleotides Mol. Pharm. 16 2019 2265 2277 31063396
4 Verma M. Ozer I. Xie W. Gallagher R. Teixeira A. Choy M. The landscape for lipid-nanoparticle-based genomic medicines Nat. Rev. Drug Discov. 22 2023 349 350 36627441
5 Schoenmaker L. Witzigmann D. Kulkarni J.A. Verbeke R. Kersten G. Jiskoot W. Crommelin D.J.A. mRNA–lipid nanoparticle COVID-19 vaccines: Structure and stability Int. J. Pharm. 601 2021 120586
6 Bege M. Borbás A. The medicinal chemistry of artificial nucleic acids and therapeutic oligonucleotides Pharmaceuticals 15 2022 909 35893733
7 Egli M. Manoharan M. Re-engineering RNA molecules into therapeutic agents Acc. Chem. Res. 52 2019 1036 1047 30912917
8 Nag K. Sarker M.E.H. Kumar S. Khan H. Chakraborty S. Islam M.J. Baray J.C. Khan M.K. Mahmud A. Barman U. DoE-derived continuous and robust process for manufacturing of pharmaceutical-grade wide-range LNPs for RNA-vaccine/drug delivery Sci. Rep. 12 2022 9394 35672337
9 Cameau E. Zhang P. Ip S. Mathiasson L. Stenklo K. Process & analytical insights for GMP manufacturing of mRNA lipid nanoparticles. Cell Gene Ther Insights 8 2022 621 635
10 Catignol P. Lim H. Overcoming bottlenecks in mRNA manufacturing (Pharma’s Almanac) https://www.pharmasalmanac.com/articles/overcoming-bottlenecks-in-mrna-manufacturing 2022
11 Daniel S. Kis Z. Kontoravdi C. Shah N. Quality by Design for enabling RNA platform production processes Trends Biotechnol. 40 2022 1213 1228 35491266
12 Andoh E.A. Yu H. A two-stage decision-support approach for improving sustainable last-mile cold chain logistics operations of COVID-19 vaccines Ann. Oper. Res. 328 2023 75 105
13 Khairi L.N.H.M. Fahrni M.L. Lazzarino A.I. The race for global equitable access to COVID-19 vaccines Vaccines 10 2022 1306 36016194
14 Saffie-Siebert S. Gibson A. Dehsorkhi A. Bio-Courier: A new delivery technology to increase the global accessibility of RNA therapeutics (Manufacturing Chemist) https://www.manufacturingchemist.com/news/article_page/Bio-Courier_a_new_delivery_technology_to_increase_the_global_accessibility_of_RNA_therapeutics/212116 2023
15 Baran-Rachwalska P. Torabi-Pour N. Sutera F.M. Ahmed M. Thomas K. Nesbit M.A. Welsh M. Moore C.B.T. Saffie-Siebert S.R. Topical siRNA delivery to the cornea and anterior eye by hybrid silicon–lipid nanoparticles J. Control. Release 326 2020 192 202 32653503
16 Maurizi A. Patrizii P. Teti A. Sutera F.M. Baran-Rachwalska P. Burns C. Nandi U. Welsh M. Torabi-Pour N. Dehsorkhi A. Saffie-Siebert S. Novel hybrid silicon–lipid nanoparticles deliver a siRNA to cure autosomal dominant osteopetrosis in mice. Implications for gene therapy in humans Mol. Ther. Nucleic Acids 33 2023 925 937 37680985
17 Beyond lipids: The science of LNP manufacturing and fill finish (Berkshire Sterile Manufacturing) https://berkshiresterilemanufacturing.com/resources/blog/beyond-lipids-the-science-of-lnp-manufacturing-and-fill-finish/ 2023
18 Saffie-Siebert R.S. Ahmed M. Sutera F. Compositions Comprising Doped Silicon Particles, and Related Methods (WIPO Patent No. WO 2023/002222 A1) 2023
19 Saffie-Siebert R.S. Sutera F. Protection of Biological Species from Degradation (WIPO Patent No. WO 2023/002224 A1) 2023
20 Hou X. Zaks T. Langer R. Dong Y. Lipid nanoparticles for mRNA delivery Nat. Rev. Mater. 6 2021 1078 1094 34394960
21 Saffie-Siebert R. A Delivery System Comprising a Silicon-Containing Material (WIPO Patent No. WO 2011/012867 A1) 2011 https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2011012867
22 European Medicines Agency. ICH Q3C (R8) residual solvents – scientific guideline https://www.ema.europa.eu/en/ich-q3c-r8-residual-solvents-scientific-guideline 2022
23 Gindy M.E. Feuston B. Glass A. Arrington L. Haas R.M. Schariter J. Stirdivant S.M. Stabilization of Ostwald ripening in low molecular weight amino lipid nanoparticles for systemic delivery of siRNA therapeutics Mol. Pharm. 11 2014 4143 4153 25317715
24 Sun H. Zhang Y. Wang G. Yang W. Xu Y. mRNA-based therapeutics in cancer treatment Pharmaceutics 15 2023 622 36839944
25 Shi D. Beasock D. Fessler A. Szebeni J. Ljubimova J.Y. Afonin K.A. Dobrovolskaia M.A. To PEGylate or not to PEGylate: Immunological properties of nanomedicine’s most popular component, polyethylene glycol and its alternatives Adv. Drug Deliv. Rev. 180 2022 114079
26 Senti M.E. de Jongh C.A. Dijkxhoorn K. Verhoef J.J.F. Szebeni J. Storm G. Hack C.E. Schiffelers R.M. Fens M.H. Boross P. Anti-PEG antibodies compromise the integrity of PEGylated lipid-based nanoparticles via complement J. Contr. Release 341 2022 475 486
27 Chen W.-A. Chang D.-Y. Chen B.-M. Lin Y.-C. Barenholz Y. Roffler S.R. Antibodies against poly(ethylene glycol) activate innate immune cells and induce hypersensitivity reactions to PEGylated nanomedicines ACS Nano 17 2023 5757 5772 36926834
28 Ibrahim M. Ramadan E. Elsadek M.E. Emam S.E. Shimizu T. Ando H. Ishima Y. Elgarhy O.H. Sarhan H.A. Hussein A.K. Ishida T. Polyethylene glycol (PEG): The nature, immunogenicity, and role in the hypersensitivity of PEGylated products J. Control. Release 351 2022 215 230 36165835
29 Li D. Liu C. Li Y. Tenchov R. Sasso J.M. Zhang D. Li D. Zou L. Wang X. Zhou Q. Messenger RNA-based therapeutics and vaccines: What’s beyond COVID-19? ACS Pharmacol. Transl. Sci. 6 2023 943 969 37470024
30 Ai L. Li Y. Zhou L. Yao W. Zhang H. Hu Z. Han J. Wang W. Wu J. Xu P. Lyophilized mRNA-lipid nanoparticle vaccines with long-term stability and high antigenicity against SARS-CoV-2 Cell Discov. 9 2023 9 36683074
31 Meulewaeter S. Nuytten G. Cheng M.H.Y. De Smedt S.C. Cullis P.R. De Beer T. Lentacker I. Verbeke R. Continuous freeze-drying of messenger RNA lipid nanoparticles enables storage at higher temperatures J. Control. Release 357 2023 149 160 36958400
32 Shirane D. Tanaka H. Sakurai Y. Taneichi S. Nakai Y. Tange K. Ishii I. Akita H. Development of an alcohol dilution–lyophilization method for the preparation of mRNA–LNPs with improved storage stability Pharmaceutics 15 2023 1819 37514007
33 Kulkarni J.A. Witzigmann D. Chen S. Cullis P.R. van der Meel R. Lipid nanoparticle technology for clinical translation of siRNA therapeutics Acc. Chem. Res. 52 2019 2435 2444 31397996
34 Maurizi A. Patrizii P. Teti A. Sutera F.M. Hadainamrei R. Dehsorkhi A. Saffie-Siebert S. Development of a potential clinical product for treating CLCN7-dependent autosomal dominant osteopetrosis type 2 (ADO2) BoneKEy Rep. 16 2022 101199
35 Bost J.P. Barriga H. Holme M.N. Gallud A. Maugeri M. Gupta D. Lehto T. Valadi H. Esbjörner E.K. Stevens M.M. El-Andaloussi S. Delivery of oligonucleotide therapeutics: Chemical modifications, lipid nanoparticles, and extracellular vesicles ACS Nano 15 2021 13993 14021 34505766
36 Friedrich M. Aigner A. Therapeutic siRNA: State-of-the-art and future perspectives BioDrugs 36 2022 549 571 35997897
37 Saffie-Siebert S. Alam I. Sutera F.M. Dehsorkhi A. Torabi-Pour N. Baran-Rachwalska P. Iamartino L. Teti A. Maurizi A. Gerard-O’Riley R.L. Effect of allele-specific Clcn7G213R siRNA delivered via a novel nanocarrier on bone phenotypes in ADO2 mice on 129S background Calcif. Tissue Int. 114 2024 advance online publication
38 Pipeline (SiSaf Ltd) https://sisaf.com/pipeline/ 2023
39 Gottlieb H.E. Kotlyar V. Nudelman A. NMR chemical shifts of common laboratory solvents as trace impurities J. Org. Chem. 62 1997 7512 7515 11671879
