key: cord-0895525-mxl335w2 authors: Packer, Meredith; Gyawali, Dipendra; Yerabolu, Ravikiran; Schariter, Joseph; White, Phil title: A novel mechanism for the loss of mRNA activity in lipid nanoparticle delivery systems date: 2021-09-21 journal: bioRxiv DOI: 10.1101/2021.09.21.461221 sha: d6fb3522f2e0526ff455f10b7c89eb09ba4d1524 doc_id: 895525 cord_uid: mxl335w2 Lipid nanoparticle (LNP)-formulated mRNA vaccines were rapidly developed and deployed in response to the SARS-CoV-2 pandemic. Due to the labile nature of mRNA, identifying impurities that could affect product stability and efficacy is crucial to the long-term use of nucleic-acid based medicines. Herein reversed phase ion pair high performance liquid chromatography (RP-IP HPLC) was used to identify a class of impurity formed through lipid:mRNA reactions; such reactions are typically undetectable by traditional mRNA purity analytical techniques. The identified modifications render the mRNA untranslatable, leading to loss of protein expression. Specifically, an electrophilic impurity derived from the ionizable cationic lipid component is shown to be responsible. Mechanisms implicated in the formation of reactive species include oxidation and subsequent hydrolysis of the tertiary amine. It thus remains critical to ensure robust analytical methods and stringent manufacturing control to ensure mRNA stability and high activity in LNP delivery systems. Nucleic acid-based medicines have emerged as promising alternatives to more traditional vaccines and therapeutics 1, 2 . Most notably, mRNA-based vaccines recently developed by Pfizer/BioNTech and Moderna changed the course of the SARS-CoV-2 pandemic, receiving swift emergency use authorization for public use due to efficacy and safety demonstrated in phase 3 trials 3, 4 . The rapid deployment of these vaccines was in part due to the advantages offered by mRNA versus conventional vaccines, including flexibility of mRNA sequence design and scalability of the manufacturing process. Furthermore, the rapid biodegradability of mRNA makes it an appealing modality from a safety and pharmacokinetic perspective 5 ; however, this same intrinsic instability is a key shelf-life limiting parameter and hurdle to effective vaccine delivery through various routes of administration 6 . The effective delivery of mRNA-based vaccines and therapies is enabled by the use of lipid nanoparticles (LNPs), which protect nucleic acid degradation by exo-and endonucleases 7, 8 and facilitate cellular uptake and expression 9, 10 . Used in both the Pfizer/BioNTech and Moderna COVID-19 mRNA vaccines, this delivery system is particularly effective as it leverages LNP surface properties [11] [12] [13] [14] , the ability of LNPs to facilitate endosomal escape through ionization of the amino lipid 15, 16 , and deliver mRNA to specific tissues based on particle size 17 . Together, these features improve vaccine immunogenicity 18 . Although LNP technology is an effective route for mRNA delivery to tissues, the interaction of certain chemical functionalities during storage such as oxidation, hydrolysis, or transesterification can lead to mRNA degradation 19, 20 through backbone cleavage of the mRNA into smaller fragments 21 . Herein we report the discovery, characterization, and identification of another class of mRNA reactivity that leads to loss in activity: the formation of lipid-mRNA adducts through the covalent addition of reactive lipid species to the nucleobase. Importantly, as many lipid-based nucleic acid formulations share common chemical functionalities, particularly those that use an ionizable amino-lipid, mechanisms identified in this study are broadly applicable. These data can inform manufacturing protocols to limit the formation of lipid-mRNA adducts and ensure the high quality of nucleic acid-based products. Reversed phase-ion pair high performance liquid chromatography (RP-IP HPLC) is a wellaccepted method alongside agarose or polyacrylamide gel or capillary electrophoresis (CE) to assess mRNA integrity 22, 23 . As in more conventional RP chromatography modes, separation of phases is driven by hydrophobic interactions between the analyte and stationary phase; however, since the phosphodiester mRNA backbone is highly polar, high salt concentrations are necessary to neutralize the negative charge and allow retention based on hydrophobicity of the aromatic nucleobases. Alkylammonium salts can increase hydrophobic interactions, driving selectivity based on the number of charges conferred by sequence length and enabling high resolution size-based separations (Figure 1a ). This provides a similar separation to CE, which is driven by size and charge; however, based on the ion pair system used, RP-IP retains some selectivity to variations in mRNA hydrophobicity due to sequence or chemical modifications 24 . When RP-IP HPLC integrity analysis was applied to mRNA extracted from an mRNA-LNP, a late eluting-peak (LP) was detected by HPLC ( Figure 1b ) that was not observed by CE ( Figure 1c ). The elution time for the LP was 21 minutes, distinct from the mRNA elution time of 10- To investigate whether tertiary mRNA structures played a role, size exclusion chromatography (SEC) at ambient conditions was applied to the MP and LP. The extensive intra-and inter-molecular structure of mRNA molecules leads to a sequence-, salt-, and temperature-dependent ensemble of structures; these are typically denatured under RP-IP HPLC or CE conditions but can be resolved by native SEC 25 . SEC analysis revealed that the MP and LP fraction profiles were identical to mRNA extracted from the LNP, with a dominantly monomeric profile ( Figure 2c ). Together, these results eliminated aggregation as the origin of the LP and strongly implicate additional hydrophobicity behind this phenomenon. We next applied compositional analysis to differentiate the LP from MP. The UV spectral data did not distinguish LP from MP ( Figure 1b) . Fourier-transform infrared spectroscopy (FT-IR) revealed minor differences potentially consistent with chemical modification (Supplementary Identification of lipid-modified nucleosides. Nucleoside profiling was performed by enzymatic digestion of the LP and MP fractions and analysis using positive mode LC-MS/MS. Although UV and total ion-current chromatograms showed identical composition of four unmodified nucleobases for the MP and LP fractions, differential analysis revealed several mass-to-charge (m/z) values exclusively found in the isolated LP, with <1% abundance relative to total unmodified nucleosides ( Figure 3a ). Under LC conditions used, elution time of these unique masses was 9.5-11 minutes; this elution time was later compared with unmodified nucleosides impurities of the ionizable lipid rather than chemical reactivity of the lipid itself. Subsequently, a highly reactive ionizable lipid from the chemistry screen was studied to understand the chromatographic behavior of intact adducted mRNA. At 1 day, a discrete shift in retention produced LP peaks at 10-12 minutes; increased tailing from 12.5-22.5 minutes was observed as the MP was depleted (Figure 4d ). This observation strongly suggests a single adduct event can shift the RNA molecule to LP, and accumulation of multiple adducts per mRNA molecule further drives increases in retention time. The binary system was then used to study contributions of the mRNA molecule to adduct formation. A series of mRNA sequences ranging in length from 700 to >4000 nucleotides in length was assessed at equivalent masses in individual binary reactions, resulting in increasing LP with mRNA length (Figure 4e ). As relative UV intensity of the LP peak in the RP-IP HPLC assay is correlated with relative mass, this observation is consistent with a constant rate of reaction on the single base level resulting in greater mass shifts to LP with increasing sequence length. This correlation highlights the relevance of these reactions for mRNA as a high molecular weight nucleic acid polymer. Although comparable lipid systems are used in low molecular weight RNA products such as small interfering RNA (siRNA), the same molar reaction rate observed in mRNA systems would result in lower levels of adduct formation of siRNA on an intact mass basis due to the stochastic nature of adduct formation. Notably, the same adduct peaks can be observed across mRNAs of different lengths in the chromatographic profile with a decrease in retention time of each LP region with increasing mRNA length ( Figure 4f ). We can understand this as an average hydrophobicity of the adducted mRNA sequence; as sequence length increases, the same adducts comprise less of the total molecule, lowering the impact on retention. The combined impact of mRNA length and lipid lot is shown in Figure 4g , demonstrating that the simplified binary system can be used as a predictive model for the final LNP. (Figure 6b ). When mRNA integrity was assessed using RP-IP HPLC, a correlation between relative mRNA integrity (including the adduct as an impurity) and protein expression was found (Figure 6b ). In contrast, mRNA integrity measured by CE only weakly correlated with protein expression; all samples had a narrow range of relative purity despite the decrease in protein expression (Figure 6b ). Extrapolating these correlations show a total loss in protein expression despite mRNA purity >60% by CE. These data demonstrate the potential of such adduct reactions to reduce activity of mRNA-LNP products and the inadequacy of CE to determine mRNA quality in LNP formulations. The introduction of impurities into the vaccine manufacturing process could potentially impact vaccine stability even under refrigerated conditions. The loss of mRNA purity to adduct formation under refrigeration (5°C) was examined for 2 vaccine formulations utilizing the same mRNA sequence and lipid system. One vaccine had poor controls in place to limit adduct formation (Vaccine A) and the other had rigid controls (Vaccine B; Figure 6c ). Data for Vaccine A show an initial delta of 15%, indicating rapid adduct formation during product processing or prior to testing. This is followed by substantial loss in mRNA integrity by almost 50% to adduct Herein we report identification of a lipid-modified class of mRNA impurities generated by electrophilic degradants and impurities originating in the ionizable lipid; these impurities disrupt mRNA translation and negatively impact the activity of LNP-formulated mRNA products. Additionally, such impurities are difficult to identify using traditional techniques used to assess mRNA integrity in mRNA-LNP such as CE. RP-IP HPLC provides specificity and sensitivity to detect adducted mRNA molecules, which are otherwise difficult to identify due to the low rate of modification. Indeed, data suggest that RP-IP HPLC can detect even single adduct events on intact mRNA, whereas in contrast, even purified LP is not distinguished by the conventional CE methodology. Furthermore, it is critical that formation of these mRNA-lipid adducts are examined during formulation design, clinical evaluation, and commercialization by using appropriate RP-IP HPLC methodologies to ensure there is no loss in mRNA integrity. A potential source of impurities is through hydrolysis of N-Oxide to aldehydes; this reaction is broadly relevant to tertiary amines commonly used in LNP formulation of siRNA and mRNA 28, 30 . It remains highly probable that formation of this class of adducts has been missed by historically applied analytical technologies such as CE, thus overlooking an important critical quality attribute of mRNA-LNP. This represents a gap in quality control of mRNA-LNPs during manufacturing, particularly as it pertains to consistency and activity of the resultant drug product. In this study, we report the ability of RP-IP HPLC to resolve aldehyde species differing only by a short alkyl chain, highlighting the selectivity of the RP-IP HPLC method to detect minute differences on the intact mRNA molecule. The pathway reported herein identifies an important class of impurities that may be present in all lipid-based mRNA systems. These reactions can be mitigated through raw material control, manufacturing process parameters, formulation design, and LNP storage conditions. It thus remains critical to monitor and control lipid adduct formation during the research, development, and manufacture of LNP-formulated nucleic acid products. Importantly, this study highlights the need for advanced methods such as RP-IP HPLC to ensure the quality, consistency, and efficacy of any pharmaceutical product. All mRNA and LNP formulations used in this work were representative of GMP-grade material. Several mRNA molecules were used throughout this work; all were 5-prime capped, 3-prime poly-adenylated sequences with modified uridine chemistry (N1-methyl-pseudouridine) encoding different vaccine targets and other proteins. mRNA was extracted from the mRNA-LNP formulation or lipid binary mixture by isopropanol precipitation. 100 µL of mRNA-LNP or binary was diluted 10-fold in 900 µL ammonium acetate (60 mM) in isopropanol, vortexed briefly, and centrifuged at 14,000g for 15 minutes at 4 °C. The supernatant was discarded and the pellet was washed with 1 mL isopropanol, vortexed, and centrifuged at 4 °C; the pellet was dried in vacuo and resuspended in 100 µL RNase-free water at room temperature. Sample separation was performed on a DNAPac RP column with 4-µm particles and dimensions of 2.1 × 100 mm (Thermo Fisher Scientific) at a flow rate of 0.35 mL/minute and column temperature of 65 °C. Mobile phase A consisted of dibutylammonium acetate (50 mM; TCI America) and triethylammonium acetate (100 mM; Sigma-Aldrich) and mobile phase B consisted of 50% acetonitrile (Sigma-Aldrich), dibutylammonium acetate (50 mM), and triethylammonium (100 mM). Separation was accomplished by step-gradient with an initial 1.5-minute hold at 25% B, a 3.0-minute gradient from 25-50% B, a 14.5-minute gradient from 50-56% B, and a 0.5-minute gradient and hold at 100% B. Modified gradient conditions for LP resolution were performed with an initial 1.5-minute hold at 25% B, a 1-minute gradient from 25-45% B, a 12.5-minute gradient from 45-100% B, and a 0.5-minute gradient and hold at 100% B. Approximately 2 µg of mRNA were injected. mRNA was detected by UV at 260 nm. LP is quantified as the relative percent of the total chromatographic peak area. Sample separation was performed on a Fragment Analyzer (Agilent Technologies), an automated multiplexed CE system equipped with an LED light source and charged-coupled device detector. RNA was quantitatively and qualitatively analyzed using the RNA Analysis Kit (Agilent Technologies DNF-489-0500). The RNA separation gel was mixed with an intercalating dye (AATI) at a v/v ratio of 10,000:1 for use as the separation matrix. RNA was denatured at 70 °C for 2 minutes and cooled on ice prior to analysis. Denatured RNA samples were electrokinetically injected at 5 kV for 6 seconds, and electrophoresis was performed for 40 minutes at 8 kV. An RNA ladder (AATI) was similarly analyzed as a calibrator for nucleotide size. Data were analyzed using PROSize 2.0 software (Agilent Technologies). Sample separation was performed on a Zenix SEC-300 150 × 4.6 mm protein SEC column (Sepax) on a Waters H-Class UPLC (Waters). The mobile phase condition was 100 mM Tris acetate/2.5 mM EDTA pH 8 with an isocratic flow of 0.25 mL/minute and UV detection at 260 nm. mRNA-lipid binaries were formed by mixture of mRNA and ionizable lipid. Unless otherwise noted, a standard 2000-nucleotide mRNA (0.135 mg/mL) was prepared sodium acetate (37.5 mM; pH 5.3; Sigma Aldrich) and mixed at a 3:1 ratio with an ionizable lipid solution (4 mg/mL) in ethanol, followed by incubation at room temperature for 24 hours. RNA was extracted from the binary using isopropanol precipitation as described above prior to further analysis. Total nuclease digestion was performed as previously described 31 Narrow ion isolation width (~1.3 m/z) was used for isolation. An N-oxide standard of ionizable lipid was generated. The N-oxide compound was dissolved in ethanol at 4 mg/mL. One milliliter of the N-oxide solution was added dropwise to 3 mL of sodium acetate (37.5 mM; pH 5.3; Sigma Aldrich), followed by incubation at room temperature for 24 hours. Prior to LC-CAD-MS analysis, aminooxy-PEG (1 mM; Thermo Fisher) was used as a labeling agent to detect and identify aldehydes present in the system. Synthetic aldehyde standards corresponding to the 17-carbon linear chain and 25-carbon branched chain of a representative ionizable lipid (heptadecan-9-yl 8-((2-hydroxyethyl)(6-oxo-6-(undecyloxy)hexyl)amino)octanoate) were generated. Aldehyde solutions (4 mg/mL) were individually prepared in ethanol and mixed with an ionizable lipid solution (4 mg/mL) in ethanol to various target compositions, listed in % w/w of the aldehyde to ionizable lipid. Binary preparations, intact mRNA HPLC analyses, and digested nucleoside LC/MS analyses were then performed as described above. Lipids were eluted by step-gradient with an initial 1.5-minute hold at 5% B, a 4.5-minute gradient from 5-48% B and 4-minute hold, a 1-minute gradient from 48-56% B and 12-minute hold, and an 8-minute gradient from 56-96% B and 2-minute hold. Lipids were detected by CAD using an evaporator temperature of 35 °C and analytical gas regulation mode. Assay measurements were taken from distinct samples. This is a descriptive study and as such no Da, corresponds to the monoisotopic residue mass of ribose. 26 The fragment ion of m/z 394.27 corresponds to the lipid-modified cytosine, which further undergoes fragmentation to m/z 376.26 (loss of water) and 204.08 (at the internal ester), confirming the identity of the lipid chain. This cytidine modification is provided as a representative example, but similar characteristic neutral mass losses of 132.03 were observed for lipid modifications across all four nucleobases. Representative data from 4 repeat experiments are shown. aldehyde (red), and 2% aldehyde (green). g and h, Binary reactions with the pure 17-carbon aldehyde from the N-oxide degradation pathway were analyzed for single nucleoside modifications. Binaries were prepared with no aldehyde (black), 0.5% aldehyde (red), 1% aldehyde (blue), and 2% aldehyde (green) spike. mRNA was extracted, enzyme-digested, and analyzed by liquid chromatography with tandem mass spectrometry (LC-MS/MS). The mass-tocharge ratios (m/z) corresponding to aldehyde-cytidine adducts (m/z 526.3 and 540.3) increased with aldehyde spike level. A representative ionizable lipid system, heptadecan-9-yl 8-((2hydroxyethyl)(6-oxo-6-(undecyloxy)hexyl)amino)octanoate is used here. Representative data from ≥3 repeat experiments are shown. Nanomedicines to Deliver mRNA: State of the Art and Future Perspectives Delivering the Messenger: Advances in Technologies for Therapeutic mRNA Delivery Efficacy and Safety of the mRNA-1273 SARS-CoV-2 Vaccine Safety and Efficacy of the BNT162b2 mRNA Covid-19 Vaccine The many pathways of RNA degradation Addressing the Cold Reality of mRNA Vaccine Stability mRNA vaccine delivery using lipid nanoparticles Role of nanotechnology behind the success of mRNA vaccines for COVID-19 Image-based analysis of lipid nanoparticle-mediated siRNA delivery, intracellular trafficking and endosomal escape Lipid Nanoparticle Systems for Enabling Gene Therapies Advances in Lipid Nanoparticles for mRNA-Based Cancer Immunotherapy Enhanced Delivery of siRNA to Retinal Ganglion Cells by Intravitreal Lipid Nanoparticles of Positive Charge Helper lipid structure influences protein adsorption and delivery of lipid nanoparticles to spleen and liver Membrane-destabilizing ionizable phospholipids for organ-selective mRNA delivery and CRISPR-Cas gene editing Cationic lipid saturation influences intracellular delivery of encapsulated nucleic acids A Novel Amino Lipid Series for mRNA Delivery: Improved Endosomal Escape and Sustained Pharmacology and Safety in Non-human Primates Engineered ionizable lipid nanoparticles for targeted delivery of RNA therapeutics into different types of cells in the liver Targeted mRNA Therapy for Ornithine Transcarbamylase Deficiency Bioinspired Alkenyl Amino Alcohol Ionizable Lipid Materials for Highly Potent In Vivo mRNA Delivery Chemical stability of nucleic acid-derived drugs HPLC methods for purity evaluation of man-made single-stranded RNAs RNA analysis by ion-pair reversed-phase high performance liquid chromatography W. technology) (Date accessed Combining ion pairing agents for enhanced analysis of oligonucleotide therapeutics by reversed phase-ion pairing ultra performance liquid chromatography (UPLC) Factors affecting DNA damage caused by lipid hydroperoxides and aldehydes In vitro DNA/RNA Adductomics to Confirm DNA Damage Caused by Benzo[a]pyrene in the Hep G2 Cell Line Continuous flow synthesis of amine oxides by oxidation of tertiary amines N-demethylation of N-methyl alkaloids with ferrocene Alcohol and endogenous aldehydes damage chromosomes and mutate stem cells Nanoparticle-Mediated Cytoplasmic Delivery of Messenger RNA Vaccines: Challenges and Future Perspectives Impact of mRNA chemistry and manufacturing process on innate immune activation process control led to high LP in Vaccine A, but adduct was well-controlled in Vaccine B. Representative data are shown for 3 repeat experiments for Figure 6a and multiple repeat experiments for Figures 6 b and c The authors declare that the data supporting the findings of this study are available within this Article and its Supplementary Information. The data that support the findings of this study are available from the corresponding authors upon reasonable request. All authors are employees of and shareholders in Moderna Inc. The authors declare no other competing interests.