Liquid crystalline inverted lipid phases encapsulating siRNA enhance lipid nanoparticle mediated transfection

RNA therapy relies on the delivery of exogenous RNA molecules (mRNA, siRNA) to control disease-relevant gene expression. Due to their charge, immunogenicity and membrane impermeability, delivery systems are required. Ionizable lipid nanoparticles (LNPs) are state-of-the-art vehicles that package, protect and release RNA inside cells, enabling clinical translation (e.g., Onpattro) and effective mRNA vaccines against SARS-CoV-2. LNPs are multicomponent systems (ionizable lipid, phospholipid, cholesterol, PEG-lipid, nucleic acid payload) formed by rapid microfluidic mixing. At low pH, ionizable lipids electrostatically complex anionic nucleic acids and self-assemble into core–shell nanostructures (~30–150 nm). Cytosolic delivery relies on endosomal acidification, protonation of ionizable lipids, interaction with endosomal membranes, and disruption of LNP and endosomal structures; yet ≥98% of RNA often remains trapped in endosomes/lysosomes. Empirical diversification of LNP components (ionizable lipids, helper/PEG-lipids) has improved transfection, and biophysical characterizations have begun linking lipid organization to biological activity. However, defined lipid structures in LNPs and mechanistic links to endosomal interaction and delivery remain elusive. Here, using a bottom-up rational design, we generate LNPs with defined lipid superstructures (lamellar, liquid crystalline inverted hexagonal, and mixed) encapsulating siRNA to study structure–activity relationships. With cryoTEM, cryoET and SAXS, we identify and characterize these structures, their thermal stability, and their interactions with anionic membranes, and demonstrate that pre-programmed inverse hexagonal phases enhance transfection and bypass lamellar-to-hexagonal transitions during delivery.

The study builds on extensive work optimizing LNP components to enhance RNA delivery, including diversification of ionizable lipid chemistry and variation of helper and PEG-lipids to improve transfection efficiency. Prior biophysical studies have correlated LNP lipid organization with biological activity, but clear identification of defined internal lipid phases and mechanistic understanding of how they affect endosomal interactions and cytosolic delivery have remained limited. Earlier work on permanently cationic lipoplexes showed inverse hexagonal phases related to nucleic acid release, and molecular simulations suggested mechanisms for nucleic acid escape from hexagonal assemblies. Nonetheless, analogous pre-programmed non-lamellar phases within clinically relevant ionizable LNPs and their impact on delivery had not been systematically resolved.

• LNP design and formulation: Varied DOPE content to program lipid phase (10, 30, 49 mol% DOPE), substituting cholesterol, with constant ionizable lipid DODAP and PEG-lipid DMPE-PEG2000. Formulations termed 10PE-LNP, 30PE-LNP, 49PE-LNP. siRNA cargo included the Patisiran sequence and anti-GFP siRNA for functional tests. LNPs assembled by microfluidic mixing (3:1 siRNA:lipid flow, siRNA in 50 mM citrate pH 4; lipids in ethanol) followed by dialysis (either 10 kDa MWCO to retain free RNA for EE% measurement or 1 MDa MWCO to remove unencapsulated RNA) into PBS. Nitrogen-to-phosphate (N/P) ratios: no RNA, N/P=6, N/P=1.
• Physicochemical characterization: DLS for hydrodynamic radius and PDI; zeta potential. SAXS (Cu Kα, q=0.0129–0.687 Å−1; 6 h exposures) to detect Bragg reflections and phase spacings; temperature-dependent SAXS at 25 °C and 37 °C including time courses. CryoTEM for nanoscale structural visualization; FFTs to extract lattice spacings; analysis of single particles (n>75 per condition). CryoET on Titan Krios for 3D visualization and surface rendering of internal structures; model fitting of lipid–siRNA arrangements.
• Thermal stability control: Designed rigid lamellar control by substituting DODAP with saturated DSDAP (10PE-DS-LNP-NP1) to increase transition temperature and test lamellar stability at 37 °C by SAXS and cryoTEM.
• LNP–membrane interaction assays: Prepared anionic large unilamellar vesicles (LUVs) mimicking endosomal composition (PC:PE:PS:Chol:PI = 50:27:10:10:3 mol%). FRET lipid mixing assay using PE-NBD/PE-LR (1.5 mol% each) in LUVs; mixing LNPs (250 µM) with LUVs (125 µM) at pH 6.0 and 7.4 at 37 °C, monitoring dequenching of NBD over 25 min. Structural analysis of mixed LNP–LUV samples by cryoTEM at 1 h and 7 h and SAXS over 6 h at 37 °C.
• Cell-based transfection: HeLa-CD63-eGFP and U2OS-UBE21-eGFP cell lines. LNPs included 0.1 mol% DiD dye to quantify uptake. Cells incubated with siRNA-loaded LNPs (0–83 nM encapsulated siRNA) for 24 h; medium replaced and cultured 48 h more. FACS at 72 h to quantify uptake (DiD MFI) and GFP knockdown (GFP MFI relative to PBS). Cell viability assessed by alamarBlue/MTT-type assay at 72 h. Dose–response curves used to derive IC50 values.

• Programmable internal phases: Increasing DOPE content shifted structures from lamellar (10PE) to mixed (30PE) to non-lamellar inverse hexagonal (49PE). SAXS peaks shifted to lower q with higher DOPE, indicating larger lattice spacings and non-lamellar phases.
• Lattice spacings (cryoTEM/FFT): • 10PE-LNP-NP1: lamellae only; spacings at 6.38, 3.19, 2.08 nm; distribution 4.20–6.90 nm; median 6.18 nm (n=84, 5.88±0.65 nm; median 6.18 nm). • 30PE-LNP-NP1: coexisting lamellae and inverse hexagonal views (hexagons, straight lines). Spacings: straight lines 5.78±0.14 nm (n=56); hexagons 5.77±0.14 nm (n=34); lamellae 6.26±0.18 nm (n=78). Coexistence within some particles. • 49PE-LNP-NP1: inverse hexagonal only (no lamellae). Spacings: hexagons 6.03±0.12 nm (n=46); straight lines 6.10±0.13 nm (n=35). Lattice expansion from 30PE to 49PE attributed to higher DOPE and reduced cholesterol.
• Empty vs filled hexagonal tubes (49PE): • 49PE-LNP-noRNA: single population spacing 4.78±0.11 nm (n=63) indicating empty tubes; also inverse micellar spheres (~8–10 nm) coexisting with tubes by cryoET. • 49PE-LNP-NP6: bimodal spacings 4.84±0.16 nm (n=50, empty) and 6.20±0.17 nm (n=27, filled). 49PE-LNP-NP1: 6.06±0.13 nm (n=81, filled). Spacing increase (~1.3–1.4 nm) consistent with accommodating ~2 nm siRNA diameter in tubes. • Particle sizes (cryoTEM): filled are smaller and more monodisperse (D≈117–118 nm) vs empty (D≈224–257 nm).
• CryoET 3D architecture: 49PE-noRNA shows coexisting inverse micelles (spherical aqueous cores at periphery) and empty hexagonal tubes. 49PE-NP1 shows liquid crystalline inverse hexagonal phases throughout core with peripheral amorphous/polymorphic zones; model fitting places siRNA density in tube cores.
• Thermal stability: 10PE-LNP-NP1 lamellar Bragg peak present at 25 °C is abolished at 37 °C and does not recover upon cooling (irreversible disorder). 30PE and 49PE profiles are stable between 25 °C and 37 °C. Rigid lamellar control (10PE-DS-LNP-NP1) maintains lamellar SAXS signal at 37 °C over 12 h; cryoTEM shows lamellae in spherical/disk-like LNPs persisting after 1 h at 37 °C.
• LNP–membrane interaction (FRET lipid mixing): At pH 6.0 and 37 °C, lipid mixing kinetics and extent increase with DOPE content: 49PE > 30PE > 10PE; rigid 10PE-DS lowest. No mixing at pH 7.4. Trends align with cellular silencing outcomes.
• Structural evolution upon mixing with anionic LUVs: 10PE-LNP-NP1 shows decreased lamellar SAXS signal and formation of inverse hexagonal structures at LNP–LUV interface (lamellar-to-HII transition). 30PE becomes dominated by HII-like signal. 49PE retains well-defined inverse hexagonal peaks and cryoTEM shows only crystalline HII structures at 1 h and 7 h (pre-programmed HII preserved).
• Functional delivery (FACS, IC50 for GFP silencing): • U2OS-UBE21-eGFP: 49PE-LNP-NP1 IC50=9.9 nM; 30PE-LNP-NP1 IC50=22.8 nM; 10PE-LNP-NP1 IC50=33.0 nM; rigid 10PE-DS-LNP-NP1 IC50=55.7 nM. Uptake similar up to 50 nM; viability unaffected. • HeLa-CD63-eGFP: 49PE-LNP-NP1 IC50=9.0 nM; 30PE-LNP-NP1 IC50=15.3 nM; 10PE-LNP-NP1 IC50=57.2 nM; 10PE-DS-LNP-NP1 IC50=39.7 nM. Uptake and silencing trends similar; some cell-dependent differences between lamellar variants. • DiD dye inclusion (0.1 mol%) did not alter hexagonal structure or transfection; negative control siRNA showed no background knockdown.
• Mechanistic model: Acidic pH protonates ionizable lipids enabling LNP–membrane interaction. Lamellar LNPs undergo in situ lamellar-to-inverse hexagonal transition at the membrane interface to disrupt membranes and release RNA (two-step pathway), whereas pre-programmed HII LNPs bypass the transition and directly engage membranes in a one-step mechanism, improving lipid mixing and transfection, although complete fusion is not achieved and some siRNA remains in stable complexes.

The study addresses the longstanding challenge of inefficient cytosolic delivery by demonstrating that internal lipid phase programming within ionizable LNPs critically determines delivery performance. By increasing DOPE content to pre-form liquid crystalline inverse hexagonal (HII) phases that encapsulate siRNA, LNPs achieve superior lipid mixing with anionic membranes at endosomal pH and significantly lower IC50 values for GFP silencing across two cell lines compared to lamellar or mixed-phase LNPs. Structural analyses (cryoTEM/cryoET/SAXS) show that lamellar formulations must undergo an in situ lamellar-to-HII transition at the membrane interface to mediate delivery, whereas pre-programmed HII formulations retain their structure upon interaction with membranes, enabling a more efficient, one-step delivery pathway. The thermostability of HII phases at physiological temperature, contrasted with thermolabile lamellae, further supports their functional advantage. These findings mechanistically link internal LNP architecture to endosomal interactions and delivery outcomes, providing a structural basis for designing more potent RNA nanomedicines. Nonetheless, the observation that complete fusion does not occur and that a substantial fraction of siRNA remains complexed highlights persistent barriers to endosomal escape, consistent with low escape efficiencies reported in the field.

By rationally tuning helper lipid content to program internal phases, the authors created LNPs with liquid crystalline inverse hexagonal cores that encapsulate siRNA and enhance transfection compared to lamellar counterparts. Multimodal structural characterization established the presence, stability, and membrane-interaction behavior of these phases, and cell studies quantified functional gains (notably, IC50 ≈ 9–10 nM for 49PE-LNP-NP1). The work provides a mechanistic framework wherein pre-programmed HII structures bypass in situ phase transitions, facilitating more efficient delivery. Future directions include retrofitting this design principle to diverse ionizable lipids and components, expanding to cargos with more complex structures (e.g., mRNA, sgRNA), and leveraging bottom-up biophysical assessments to predict LNP potency prior to biological testing.

• Biological model simplification: LUVs used to mimic endosomal membranes lack the full compositional and dynamic complexity of intracellular environments; results may not capture all factors influencing endosomal escape in vivo.
• Incomplete fusion and escape: Even with HII phases, complete LNP–membrane fusion was not observed, and a substantial fraction of siRNA remains in stable complexes, aligning with generally low endosomal escape efficiencies.
• Structural resolution constraints: SAXS peak overlap limits precise deconvolution of coexisting phases; cryoTEM identification of small (~10 nm) micellar structures in smaller LNPs can be challenging.
• Assembly mechanisms: Exact pathways of LNP assembly and how siRNA content biases empty vs filled tube formation are hypothesized but not fully resolved.
• Generalizability: While the design principle should extend to other ionizable lipids and cargos, optimal compositions may vary and were not exhaustively mapped here.