An mRNA lipid nanoparticle can retain an acceptable average diameter, polydispersity index (PDI) and apparent encapsulation efficiency while losing protein-expression activity. Those measurements show whether the bulk particle population and dye-accessible RNA fraction have changed; they do not establish that the encapsulated transcript remains full-length, chemically unmodified and capable of translation. Conversely, a change in one physical attribute does not identify which lipid, RNA or interfacial event caused the functional loss.
A useful stability study therefore follows several layers of the same stressed samples over time. The analytical plan must distinguish parent-lipid loss from degradant formation, intact mRNA from lipid-adducted or fragmented RNA, and colloidal change from failure of intracellular delivery or translation. The functional assay then tests whether the detected changes are consequential. Without this matched design, a panel of individually valid measurements can still produce an incomplete stability conclusion.
Experimental design for formulated mRNA-LNPs should identify the study arms, time points, and orthogonal measurements required to resolve a degradation pathway. The LNP and exosome lipidomics service can support lipid-composition and degradant measurements within a study that also tracks mRNA integrity, particle attributes, and functional response in matched samples.
mRNA-LNP Stability Study Objectives and Experimental Arms
The study objective determines the stress conditions, sampling density and analytical depth. Formulation ranking, mechanistic forced degradation and storage-condition assessment each require a different stress matrix, sampling density and analytical depth.
| Study Objective | Primary Comparison | Required Design Features | Supported Study Output |
|---|---|---|---|
| Formulation selection | Ionizable lipid, helper lipid, PEG-lipid, buffer or protectant variants | Independent preparations, common mRNA lot, matched concentration and container, several time points | Relative stability ranking and the attribute that differentiates formulations |
| Degradation-pathway identification | Temperature, oxidation, hydrolysis, light or mechanical stress | Graded stress, unstressed control, early time points, lipid-only or lipid–mRNA model mixtures when needed | Candidate pathway supported by parent loss, product formation and an orthogonal RNA or functional readout |
| Storage-condition assessment | Intended frozen, refrigerated or room-temperature condition | Representative formulation concentration, headspace and container; longitudinal sampling; batch representation | Change rate under the tested condition and the leading stability-indicating attribute |
| Handling or excursion simulation | Freeze–thaw, agitation, thawed hold, dilution, transfer or syringe manipulation | Defined cycle number, duration and handling intensity; pre- and post-excursion samples | Sensitivity to the specific handling event and recovery after return to storage |
At minimum, the experimental set should include the time-zero material and a contemporaneous unstressed control. Where the objective is formulation selection, independent LNP preparations are more informative than repeated measurements from one vial because preparation-to-preparation variation can otherwise be mistaken for analytical precision. Where the objective is mechanistic, naked mRNA, individual lipid components, empty LNPs or simplified lipid–mRNA mixtures can isolate whether a signal requires particle assembly or direct chemical contact.
Container and sample presentation are experimental variables. Formulation concentration, fill volume, headspace, dissolved oxygen, vial or syringe contact, and the number of freeze–thaw events can alter the apparent outcome. A stability comparison should hold these factors constant unless one of them is the question being tested. Removing aliquots repeatedly from one vial can also combine storage time with additional mixing and headspace exposure; separately stored time-point aliquots are usually easier to interpret.
Figure 1. An mRNA-LNP stability conclusion requires matched evidence from lipid chemistry, mRNA integrity, particle attributes and protein expression. A stable result in one layer does not exclude degradation in another.
Forced Degradation, Accelerated Stability, and Real-Time Storage
These study types answer different questions. Forced degradation is primarily a method-development and mechanism-discovery tool. Accelerated studies can compare formulations or reveal emerging risks. Real-time storage most closely represents the intended condition. Data from one category should not automatically be promoted to a claim belonging to another.
| Study Type | Typical Purpose | Useful Conditions | Appropriate Interpretation | Main Limitation |
|---|---|---|---|---|
| Forced degradation | Expose plausible failure pathways and challenge analytical selectivity | Controlled heat, oxidation, hydrolytic conditions, light, agitation or freeze–thaw | Identifies which attributes respond to a defined stress and whether degradants can be separated from parent material | Severe stress may create pathways that do not dominate during storage |
| Accelerated stability | Rank formulations and obtain measurable change within a practical interval | Moderately elevated temperature or intensified handling with multiple time points | Compares relative change rates when the degradation mechanism remains comparable | Cannot by itself establish long-term stability at another condition |
| Real-time storage | Monitor the formulation under the intended storage condition | Target temperature, container, concentration and handling configuration | Provides the most relevant time-course evidence for that tested condition | Requires longer observation and may show little early change |
| Excursion or in-use study | Test defined deviations from routine storage | Thawed hold, transport vibration, light exposure, repeated withdrawal or dilution | Establishes sensitivity to the specified event | Results apply only to the excursion intensity and sequence tested |
The stress range should produce partial, time-resolved change. Conditions that immediately destroy RNA, precipitate the formulation or aggregate every particle provide little information about the first failure event. A short pilot can locate a range in which parent material decreases and one or more products emerge without collapsing all readouts at the first post-stress time point. Early sampling is important because particle aggregation, RNA cleavage and covalent adduction may have different onset times.
Temperature is not a universal acceleration factor. If elevated temperature changes particle phase behavior, exposes RNA to bulk water, activates a new hydrolysis pathway or shifts the dominant lipid degradant, its kinetics may not represent refrigerated or frozen storage. Freeze–thaw studies require their own design because the events at ice formation, solute concentration and thawing are not captured by a simple temperature-time integral. Light and agitation should likewise be tested as defined stresses rather than folded into a generic “accelerated” arm.
Lipid Degradation and Lipid–mRNA Adduct Analysis
Parent-Lipid Loss and Degradant Formation
The lipid portion of the study should follow both sides of the reaction. Interpret a decrease in an ionizable-lipid peak together with plausible oxidation, hydrolysis or cleavage products, extraction recovery and detector response. For nonchromophoric formulation lipids, charged aerosol detection can support parent–degradant reconciliation when response behavior has been calibrated or otherwise qualified for the compounds being compared; uncorrected CAD peak area is not a universal molar mass balance. LC–high-resolution MS provides structural specificity for low-level degradants, and targeted LC-MS/MS becomes appropriate after the relevant products and transitions have been established.
The monitoring panel should be derived from the actual structures in the formulation. Relevant pathways can include tertiary-amine oxidation, peroxide formation at unsaturated chains, ester hydrolysis, loss of hydrophobic tails, phospholipid conversion to lysophospholipids and changes in PEG-lipid recovery. The parent lipid, major identified degradants and class-relevant internal standards should be measured in the same extraction series. A targeted lipidomics strategy is most useful here as a structure-specific follow-up: the panel should be built around the actual ionizable lipid and plausible products rather than assuming that every tertiary-amine lipid produces the same N-oxide, aldehyde profile or degradation rate.
Time zero has a special role. A reactive species present in the starting lipid lot is a raw-material or process-related contribution, whereas a product that increases during formulated storage is a stability-related contribution. Comparing lipid raw material, freshly formulated LNP and stored LNP can separate those origins. When two lipid lots produce different mRNA-adduct kinetics despite similar nominal purity, a simplified binary incubation of lipid and mRNA can test whether the difference is chemically transferable before the full particle system is investigated.
Detecting Lipid–mRNA Adducts
Reactive lipid impurities or degradants can covalently modify nucleobases. Packer and colleagues resolved a late-eluting region by ion-pair reversed-phase HPLC after extracting mRNA from LNPs and connected the modification to reduced translation. Conventional capillary electrophoresis did not resolve the same class of adduct with equivalent sensitivity. The methods provide complementary selectivity: electrophoresis resolves RNA size and fragmentation, whereas IP-RP-HPLC can reveal later-eluting, hydrophobic RNA species compatible with lipid-derived modification.
The distinction is supported by recent work on (4S)-KEL12 LNPs. Chen and colleagues found that elevated temperature promoted both mRNA adduct accumulation and loss of apparent mRNA integrity, but the two processes followed divergent kinetics at 25 °C. The aldehyde degradant Z4 generated approximately four times more adduct than the N-oxide Z1, while Z4-adducted RNA retained its apparent integrity by capillary electrophoresis yet showed poor protein expression in 293T cells. Nucleoside-level mass spectrometry localized the detected modification to cytidine-containing products. For stability design, these results support tracking fragmentation, adduct formation and translation as separate response variables rather than using electrophoretic integrity as a proxy for chemical or functional RNA integrity.
A staged confirmation strategy is appropriate:
- Analyze extracted mRNA by a method capable of separating the intact main peak, shorter products and later-eluting hydrophobic species.
- Demonstrate that the late-eluting signal increases with storage time or the relevant stress.
- Compare the signal with ionizable-lipid degradant formation in the same samples.
- Where structural confirmation is required, digest the RNA and characterize modified nucleosides by LC-MS/MS, or reproduce the reaction with a defined candidate degradant.
- Test whether the isolated or enriched adduct-containing fraction shows impaired protein expression.
Treat an association between lipid oxidation and an mRNA late peak as pathway evidence. Structural assignment requires orthogonal confirmation because oxidation can generate multiple secondary products and more than one product may modify RNA. Reporting should distinguish an “adduct-compatible late-eluting species,” a confirmed lipid–mRNA adduct and a specific nucleoside modification.
mRNA Integrity, Encapsulation, and Particle Stability
Encapsulation efficiency, RNA integrity and RNA activity are separate measurements. A dye-exclusion assay can indicate the fraction of RNA protected from an externally added fluorescent reagent, yet intact and adducted transcripts inside the particle may both be counted as encapsulated. Reinhart and colleagues observed stable RiboGreen-derived mRNA content and limited change in common particle measurements for several nonfrozen formulations even when protein-expression behavior diverged. Stability assessment therefore needs RNA-resolving and functional measurements in addition to loading.
Where the study objective requires fuller attribution of RNA quality loss, 5′-cap integrity and poly(A)-tail length or distribution can be added as project-specific endpoints. Their inclusion should be driven by the transcript design, suspected pathway and expression result rather than applied as a universal panel to every LNP time point.
| Attribute | Practical Method | Analytical Question | Key Limitation |
|---|---|---|---|
| Total and accessible RNA | Fluorescent RNA assay with and without particle disruption | Has RNA content or apparent encapsulation changed? | Does not establish transcript length, chemical integrity or translation competence |
| mRNA size and integrity | Capillary or microchip electrophoresis | Has the intact RNA peak decreased or fragmented material increased? | May not resolve hydrophobic lipid–RNA adducts |
| Hydrophobic mRNA modification | IP-RP-HPLC after validated extraction | Have later-eluting, adduct-compatible RNA species appeared? | Peak identity requires orthogonal confirmation |
| Particle diameter and PDI | DLS | Has the intensity-weighted particle population shifted or broadened? | Averages can underrepresent low-abundance subpopulations and do not show chemical degradation |
| Particle concentration and subpopulations | NTA, AF4-MALS or another orthogonal sizing method | Are particle number or size-resolved populations changing? | Results depend on method-specific detection and separation limits |
| Morphology or internal organization | Cryo-TEM, SAXS or related structural methods | Has particle architecture reorganized? | Mechanistic method rather than a routine substitute for time-course assays |
| Lipid composition and molar balance | LC-CAD and/or LC-MS with appropriate calibrators | Has one component been preferentially lost or degraded? | Nominal feed ratio is not a measured stored-sample composition |
The four principal lipid components should be quantified using a response model suited to their different ionization and detector behaviors. Phospholipid analysis, combined with appropriate assays for the ionizable and PEG lipids, can determine whether apparent particle stability masks helper-phospholipid hydrolysis, preferential PEG-lipid loss or decline of intact ionizable lipid. These data should be interpreted as measured composition at each time point, not as a restatement of the formulation recipe.
Orthogonal particle methods are best deployed in response to a defined analytical gap. For example, an increase in DLS Z-average without a clear PDI shift may be investigated using a separation-based size method or imaging to determine whether a small aggregate population is influencing the intensity-weighted result. Applying every available structural platform to every time point can consume sample without improving the stability decision.
Figure 2. Stability study arms should share baseline material and matched time points while retaining study-specific stresses and controls. The same analytical core enables comparison without treating forced degradation as real-time shelf-life evidence.
Linking Stability-Indicating Attributes to Protein Expression
Protein expression provides the functional bridge between analytical change and LNP performance. It should be measured on aliquots corresponding to the same time points used for chemical and particle analysis. Running a functional assay only at the final time point can confirm loss of activity but cannot identify whether expression began to decline before or after an analytical attribute changed.
The cell system must provide a reproducible dose–response window for the formulation and encoded protein. Cell density, exposure time, serum or apolipoprotein conditions, mRNA dose, viability and readout timing should be controlled because each can shift apparent expression independently of sample stability. A time-zero reference and a multi-point dose curve help distinguish a genuine loss of potency from an assay performed above a saturated response range. Reporter constructs are useful for formulation development, whereas a product-relevant encoded protein may be required when sequence, RNA length, folding or secretion changes the response.
Measure cellular uptake and protein expression as separate endpoints. If stored LNPs enter cells but produce less protein, the leading hypotheses include impaired endosomal escape, damaged mRNA or reduced translational competence. If both uptake and expression fall, particle surface or structural changes become more plausible. A labeled-particle or labeled-RNA uptake experiment can therefore be used as a focused follow-up, provided that the label itself has been shown not to alter the formulation.
Cell-based expression remains an experimental model. It should not be described as a direct substitute for in vivo exposure, biodistribution or biological efficacy. Its purpose in a stability study is narrower: to test whether storage-induced analytical changes correspond to loss of translation under a controlled cellular condition. Patel and colleagues showed that cell line, assay window, PEG-lipid composition, mRNA mass and particle size can all influence the measured response, reinforcing the need to qualify the functional assay for the formulation rather than adopting a generic transfection endpoint.
Degradation Kinetics and Attribute-to-Function Analysis
Modeling Individual Stability Attributes
Each stability-indicating attribute should first be analyzed as its own time course. Useful outputs include parent-lipid remaining, degradant concentration, intact mRNA fraction, adduct-compatible RNA fraction, particle diameter, PDI, free or accessible RNA and relative protein expression. Plotting all results only as percent of time zero can aid comparison, but the report should retain the original units and method-specific uncertainty.
Zero-order or first-order models should be selected from the observed trajectory, residuals and plausible mechanism rather than imposed uniformly. Barros and colleagues found that the apparent reaction order and degradation rate of mRNA varied with temperature and analytical method. This is a practical warning against treating a half-life calculated from one assay as an assay-independent material constant. Rate constants should be reported with confidence intervals and the fitted range, particularly when few time points are available.
Parent loss and product gain provide a stronger kinetic argument than either alone. If the parent ionizable lipid declines but the expected degradants do not increase, the investigation should consider extraction recovery, adsorption, unmonitored products or a detector-response change. Likewise, an mRNA late peak that increases without corresponding lipid chemistry may require examination of other formulation components or sample-processing artifacts.
Temperature Extrapolation and Mechanism Consistency
Arrhenius analysis is appropriate only when the same dominant process is being accelerated across the fitted temperatures. A new chromatographic degradant, a change from monodisperse particles to aggregation, a formulation phase transition or a switch in reaction order indicates that the datasets should not be combined into one extrapolation. Frozen conditions also require separate consideration because reaction in a freeze-concentrated phase is not described adequately by the nominal storage temperature alone.
Accelerated results can still be valuable without shelf-life extrapolation. They can rank formulations, identify an early chemical marker and determine which real-time assay deserves the greatest sensitivity. The report should state whether the accelerated and target-condition samples share the same product pattern and attribute sequence before comparing their rate constants.
Connecting Analytical Change with Protein Expression
Attribute-to-function analysis should retain time, stress and formulation identity. Pooling different stresses into a single correlation can create a strong statistical relationship even when each stress has a different causal pathway. Correlations should therefore be examined within a defined stress and confirmed across an independent batch or formulation where possible.
Several interpretable result patterns are common:
| Matched Observation | Leading Interpretation | Focused Follow-Up |
|---|---|---|
| Particle size and encapsulation remain stable; lipid degradants and mRNA adducts rise; expression falls | Chemical modification precedes gross particle failure | Confirm adduct structure or test the implicated lipid degradant in a simplified system |
| Size and PDI rise; accessible RNA increases; expression falls | Aggregation, leakage or particle disassembly | Examine particle subpopulations and morphology; review agitation, freeze–thaw or protectant conditions |
| mRNA integrity declines while lipid composition and size remain stable | RNA cleavage or sequence-specific instability | Review buffer, pH, water activity, RNase control and transcript-specific structural risk |
| Lipid degradants increase but expression remains stable over the observed interval | Potential early chemical marker or noncritical product at the measured level | Extend the time course and establish whether the degradant continues to accumulate |
| Routine analytical attributes remain stable while expression declines | Unmeasured mRNA modification, altered delivery, endosomal escape or assay-specific effect | Add IP-RP analysis, uptake measurement or a mechanistic cell assay rather than repeating DLS alone |
Tong and colleagues observed nonlinearity in the relationship between heat-induced mRNA degradation and cellular potency. A project-specific change point can be useful, but it should not be converted into a universal acceptance threshold for unrelated mRNA sequences, lipid chemistries or cell assays. The decision should remain tied to the tested formulation and study objective.
Figure 3. Attribute-to-function interpretation uses matched time points to determine which analytical signal precedes functional decline. Correlation is evaluated within a defined stress before a candidate failure mechanism is assigned.
mRNA-LNP Stability Data Interpretation for Formulation Selection and Root-Cause Analysis
A stability report should allow the research team to decide what to change in the formulation or study, not simply identify statistically significant time points. The core output is a time-aligned dataset in which sample identifiers, storage history, stress exposure, freeze–thaw count and analytical results can be traced to the same aliquot or matched vial.
For formulation selection, the comparison should identify the earliest discriminating attribute and whether it is chemical, physical or functional. A candidate that retains particle size but accumulates an mRNA-reactive degradant has a different development risk from a candidate that aggregates during the first freeze–thaw cycle. Ranking by a composite score can obscure this distinction; attribute-specific trajectories and failure modes are more informative.
For root-cause analysis, the report should separate observation, inference and confirmation. “Ionizable-lipid N-oxide increased during storage” is an observation. “Oxidation may generate an mRNA-reactive species” is a mechanistic inference. Detection of the corresponding RNA modification, reproduction with a defined degradant and associated loss of expression provide progressively stronger confirmation. This sequence of confirmation prevents a high-resolution mass feature from being presented as a complete causal explanation.
The final interpretation should state:
- which attribute changed first and at what tested condition;
- whether parent loss and degradant formation were reconciled;
- whether RNA fragmentation and lipid–mRNA adduction were analytically distinguished;
- whether particle change preceded, accompanied or followed chemical degradation;
- whether the functional assay responded within its qualified range;
- which additional experiment would confirm the proposed failure mechanism; and
- which formulation, storage or handling variable can be modified in the next study.
An integrated design does not require every method at every time point. A compact core panel can be applied across the full study, with HRMS structural work, morphology, uptake, or nucleoside-level confirmation reserved for selected samples. Chemical, physical, and functional conclusions should be linked through matched samples and predefined levels of confirmation.
References
- Packer, M., Gyawali, D., Yerabolu, R., et al. A novel mechanism for the loss of mRNA activity in lipid nanoparticle delivery systems. Nature Communications. 2021;12:6777. https://doi.org/10.1038/s41467-021-26926-0
- Kamiya, M., Matsumoto, M., Yamashita, K., et al. Stability study of mRNA-lipid nanoparticles exposed to various conditions based on the evaluation between physicochemical properties and their relation with protein expression ability. Pharmaceutics. 2022;14(11):2357. https://doi.org/10.3390/pharmaceutics14112357
- Raffaele, J., Loughney, J. W., & Rustandi, R. R. Development of a microchip capillary electrophoresis method for determination of the purity and integrity of mRNA in lipid nanoparticle vaccines. Electrophoresis. 2022;43(9–10):1101–1106. https://doi.org/10.1002/elps.202100272
- Reinhart, A.-G., Osterwald, A., Ringler, P., et al. Investigations into mRNA lipid nanoparticles shelf-life stability under nonfrozen conditions. Molecular Pharmaceutics. 2023;20(12):6492–6503. https://doi.org/10.1021/acs.molpharmaceut.3c00956
- Patel, N., Davis, Z., Hofmann, C., et al. Development and characterization of an in vitro cell-based assay to predict potency of mRNA-LNP-based vaccines. Vaccines. 2023;11(7):1224. https://doi.org/10.3390/vaccines11071224
- Tong, X., Raffaele, J., Feller, K., et al. Correlating stability-indicating biochemical and biophysical characteristics with in vitro cell potency in mRNA LNP vaccine. Vaccines. 2024;12(2):169. https://doi.org/10.3390/vaccines12020169
- Parot, J., Mehn, D., Jankevics, H., et al. Quality assessment of LNP-RNA therapeutics with orthogonal analytical techniques. Journal of Controlled Release. 2024;367:385–401. https://doi.org/10.1016/j.jconrel.2024.01.037
- Barros, C. H. N., Alfaro, M., Csiki-Fejer, A., et al. Comparative analysis of mRNA degradation kinetics using chromatographic and electrophoretic methods. Molecular Pharmaceutics. 2025;22(6):3061–3072. https://doi.org/10.1021/acs.molpharmaceut.4c01543
- Kang, R. S., Wang, Z., Sapa, H., et al. Spotlighting the criticality of lipid quality control through a mechanistic investigation of mRNA activity loss in lipid nanoparticles. Communications Chemistry. 2025;8:313. https://doi.org/10.1038/s42004-025-01699-5
- Kim, J., Sim, J., Park, S., et al. Comprehensive analytical characterization of mRNA influenza vaccine under stress conditions: implications for stability and quality control. Pharmaceutical Research. 2026. https://doi.org/10.1007/s11095-026-04175-4
- Chen, H., Gong, J., Wu, W., et al. Characterization of storage-induced mRNA modifications in (4S)-KEL12 LNP: adduct formation kinetics, mRNA decay, and translational competence. Molecular Pharmaceutics. 2026;23(3):1496–1504. https://doi.org/10.1021/acs.molpharmaceut.5c00950