Nose-to-brain delivery of IRX4204 by cell-adhesion peptide-functionalized gemini surfactant-phospholipid nanoparticles

Abstract

Efficient delivery of hydrophobic therapeutics to the central nervous system (CNS) remains challenging due to poor solubility, limited blood–brain barrier permeability and systemic toxicity. Here, we report the development of gemini surfactant-phospholipid nanoparticles (GPNPs) for intranasal delivery of the highly potent and preferential retinoid X receptor agonist IRX4204. Physicochemical characterization demonstrated stable nanoparticles (60 nm, +40 mV) with high encapsulation efficiency and long-term stability. Dynamic light scattering, transmission electron microscopy and small-angle X-ray scattering revealed that optimal drug loading in GPNPs is ≤0.5 mg mL⁻¹. The formulations showed the coexistence of two structurally distinct particle populations, flattened micelles and larger vesicular structures. In vivo studies in mice demonstrated that intranasal administration of IRX4204-GPNPs (0.12 mg kg⁻¹) achieved higher CNS concentrations than a 100-fold higher oral dose (12 mg kg⁻¹), with sustained distribution across various CNS regions and enhanced striatal accumulation. Compared to oral IRX4204 suspension administration, intranasal delivery of IRX4204-GPNPs achieved about 1.4-fold higher CNS drug concentrations, 136-fold greater delivery efficiency (0.19% vs. 0.0014% of dose) and 73.5% direct transport percentage (DTP), accompanied by markedly reduced plasma and liver exposure. Pharmacodynamic evaluation confirmed sustained activation of RXR-responsive genes, including SCD1 and LXRα, following intranasal nanoparticle delivery. These findings demonstrate that optimized GPNPs enable efficient CNS targeting and support their use as a promising intranasal platform for delivery of RXR agonists and other hydrophobic therapeutics.

Intranasal retinoid X receptor agonist IRX4204 in gemini surfactant-phospholipid nanoparticles (GPNPs) achieved sustained drug distribution throughout the CNS and reduced plasma and liver concentration compared to oral administration.

Introduction

Delivery of drugs to the central nervous system (CNS) after oral or intravascular administration remains a major challenge due to both the structurally and biochemically restrictive nature of the blood–brain barrier (BBB). The BBB permits only highly controlled transcellular passage through mechanisms such as passive diffusion, carrier-mediated transport, efflux systems, and receptor- or adsorptive-mediated transcytosis.¹ Although BBB permeability can increase in certain pathological conditions, these changes are often transient and regionally variable, resulting in inconsistent or insufficient drug exposure in the brain.²

Nose-to-brain delivery has emerged as a promising non-invasive drug delivery method capable of bypassing the BBB through the olfactory rather than the respiratory epithelium. Intranasally administered nanoparticle (NP) formulations gain access to the CNS directly through the olfactory and trigeminal neural pathways.^3,4 While the precise biological mechanisms underlying intranasal transport are still being elucidated,⁵ multiple nanotechnology-based systems, including liposomes, lipid nanoparticles (LNPs), hydrogels, dendrimers, and nanoemulsions, have shown potential for facilitating drug deposition within the brain.^6–8 These platforms provide versatile design options for encapsulating diverse therapeutic agents and enable enhanced functionalities such as mucoadhesive polymers to prolong nasal residence time, permeation enhancers, and thermosensitive gel-forming excipients to optimize brain delivery.⁹

The retinoid X receptor (RXR) is a central regulator of transcriptional programs relevant to neural repair. The RXR forms homodimers or heterodimers with a broad range of nuclear receptors such as the farnesoid X receptor (FXR), thyroid receptor (TR), vitamin D receptor (VDR), retinoic acid receptor (RAR), pregnane X receptor (PXR), nuclear receptor related 1 and 77 proteins (Nurr1, 77), peroxisome proliferator activated receptor (PPAR), and liver X receptor (LXR).¹⁰ Activation of RXR heterodimers, including RXR-Nurr1, allows RXR agonists such as IRX4204 ((2E,4E)-3-methyl-5-[(1S,2S)-2-methyl-2-(5,5,8,8-tetramethyl-6,7-dihydronaphthalen-2-yl)cyclopropyl]penta-2,4-dienoic acid), to modulate complex transcriptional programs associated with mitochondrial biogenesis, progenitor cell proliferation and immune and glial polarization toward pro-repair phenotypes. These coordinated effects may contribute to reduced inflammation, enhanced debris clearance, and the creation of a microenvironment that supports the differentiation and integration of vascular and neural cell types into newly formed neurovascular units (NVUs), processes that are thought to underlie functional recovery in neurodegenerative diseases (NDDs).^11–13

Accordingly, improved delivery of such agents to the brain is likely to be a critical determinant of therapeutic efficacy in neurodegenerative diseases. Advances in brain-targeted delivery strategies may therefore broaden the clinical potential of RXR agonists and related therapeutics across diverse conditions, including multiple sclerosis, Parkinson's disease, Alzheimer's disease, and stroke, although robust clinical validation remains to be established.¹⁴

This therapeutic approach is exemplified by IRX4204, is a highly potent and preferential retinoid X receptor (RXR) agonist. However, when administered orally or parenterally, this drug exhibits poor brain penetration and dose-limiting adverse effects.⁹ Hence, intranasal administration could potentially overcome these limitations.

We have developed an intranasal gemini surfactant-phospholipid (GP) NP formulation platform with the goal of enhancing the delivery IRX4204 into the brain and potentially targeting NVUs vulnerable to high oxidative stress in the brain.⁹ In this work, we describe the development, physicochemical properties and ability of intranasally delivered GPNPs containing IRX4204 preferentially deliver this drug to the CNS. Using IRX4204 as an example, our findings support the feasibility of GPNPs as a safe and effective strategy for delivering drugs to the CNS that are otherwise inaccessible using conventional administration routes.

Experimental

Encapsulation of IRX4204 in gemini surfactant-phospholipid NPs

The RXR agonist, IRX4204 ((2E,4E)-3-methyl-5-[(1S,2S)-2-methyl-2-(5,5,8,8-tetramethyl-6,7-dihydronaphthalen-2-yl)cyclopropyl]penta-2,4-dienoic acid; obtained from Io Therapeutics, Inc., under a materials transfer agreement) was incorporated into GPNPs formulated using a combination of a pH-sensitive ionizable gemini surfactant (GS) (18-7NH-18 (N,N′-(azanediylbis(propane-3,1-diyl))bis(N,N-dimethyloctadecan-1-aminium))), and its RGD tripeptide conjugated derivative (18-7N_RGD-18, GS_RGD) by a one-pot self-assembly method using ultrasonication. Briefly, all GPNP components, 18-7N_RGD-18 and 18-7NH-18 (both synthesized in-house¹⁵), cholesterol (Chol) (Croda, England), and 1,2-distearoyl-sn-glycero-3-phosphocholine (DSPC) (Lipoid GmbH, Germany) were suspended in biotech grade water at 1 : 1 : 2 : 10 weight ratio for blank GPNPs, or followed by adding the required amount of IRX4204 (0.1, 0.25, 0.5, 1 and 2 mg mL⁻¹) for IRX4204-GPNPs. The mixtures were processed at 55 °C using a Q500 QSonica Sonicator equipped with a cup horn attachment and a temperature-controlled water bath (Newtown, CT, USA) operating at 30–50% amplitude in pulse mode (40 s on/15 s off) for two hours (method USM-1–2 h). Other compositional ratios, such as 0 : 2 : 2 : 10 with 0, 0.1, and 0.5 mg mL⁻¹ of IRX4204, and 0 : 10 : 10 : 50 with 0 and 5 mg mL⁻¹ of IRX4204, were also prepared for SAXS evaluation. Another series of blank and IRX4204 GPNPs formulations (1 : 1 : 2 : 10, 0.1, 0.25, 0.5, 1, 1.5 and 2 mg mL⁻¹) were also prepared using remote loading (IRX4204 was added to preformed blank GPNPs and processed as above) and used for dynamic light scattering (DLS) encapsulation studies.

Characterization of the physicochemical properties of IRX4204 GPNPs

Particle size distribution, polydispersity index (PDI), and zeta potential (ZP) were determined by DLS using Zetasizer Ultra Red (Malvern Instruments, Worcestershire, UK) at room temperature. ZP was measured in 20-fold diluted samples. Values are reported as an average of three readings ± S.D.

The stability of GPNP (1 : 1 : 2 : 10)-IRX4204 0.1 mg mL⁻¹ stored at 4 °C for six months was assessed by particle size, PDI and ZP measurements using Zetasizer at t = 0, 1, 2, 3, 4, 8, 12, 16, 20 and 24 weeks, samples (diluted 20-fold) were transferred to disposable low-volume polystyrene (ZEN0040) cuvettes for size and PDI or to disposable folded (DTS1070) capillary cells to measure the ZP. Values are reported as an average of three readings ± S.D.

Encapsulation efficiency

Particle size distribution profile was used to detect the presence of unencapsulated drug particles in a series of GPNPs made with increasing concentration of IRX4204. Briefly, blank GPNP (1 : 1 : 2 : 10) and IRX4204 suspension were separately prepared by USM-1–2 h. Then a series of six different concentrations of IRX4204 (0.05, 0.1, 0.5, 1, 1.5 and 2 mg mL⁻¹) were encapsulated in the blank GPNP (1 : 1 : 2 : 10) by adding the required volume of IRX4204 suspension to 40 µL of blank GPNP (1 : 1 : 2 : 10) and sonicating at 55 °C for 15 min (remote loading). After cooling down to room temperature, particle size distribution for each concentration was measured using Zetasizer Ultra Red.

Transmission electron microscopy

Samples of IRX4204-GPNP (1 : 1 : 2 : 10) and blank GPNP (1 : 1 : 2 : 10) were placed on formvar/carbon coated 300 mesh Cu grids (SPI, West Chester, PA, United States) and negatively stained with 0.5% uranyl acetate (SPI). The grids were imaged using Philips CM10 electron microscope (Philips, Amsterdam, Netherlands).

SAXS analysis of IRX 4204 GPNPs

Small-angle X-ray scattering (SAXS) on selected GPNP formulations were carried out at the BL4-2 beamline at Stanford Synchrotron Radiation Lightsource (SSRL, Stanford, USA) at an energy of 11 keV and a sample to detector distance of 1.7 m, wavelength of 1.127 Å, at 20–30 s exposure time. Scattering patterns were recorded over the momentum transfer range q = 0.001–0.5 Å⁻¹, where q = (4π/λ)·sin(θ), λ is the X-ray wavelength, and 2θ is the scattering angle. Scattering intensity versus q plots were obtained by radial integration of the 2D patterns using GSAS II. SAXS curves (intensity (I(q)) vs. scattering vector (q)) were plotted using BioXTAS Raw 2.3.0.¹⁶ Datasets were modeled using SasView 6.1.3.

The critical packing parameter (P) was calculated to assess aggregate geometry and curvature using structural parameters derived from SAXS analysis. The packing parameter is defined as P = v/(a₀ × l_c), where v is the hydrophobic chain volume, a₀ is the effective headgroup area at the aggregate interface, and l_c is the maximum hydrophobic chain length.¹⁷ The effective headgroup area was estimated from SAXS-derived cross-sectional dimensions. Because the aggregates exhibit anisotropic geometry, an effective radius (R_eff) was calculated as the geometric mean of the minor and major radii obtained from elliptical cylinder fits: R_eff = √(R_minor × R_major). The headgroup area was then estimated using a₀ = 2v/R_eff this approach assumes that the dominant scattering contribution arises from the micellar cross-section and provides an effective representation of interfacial packing.¹⁸

The hydrophobic chain volume (v) and maximum chain length (l_c) were estimated using Tanford equations: v = (27.4 + 26.9n_c) ų and l_c = (1.5 + 1.265n_c) Å, where n_c is the number of carbon atoms in the hydrophobic chain.¹⁹ The value of n_c = 18, corresponding to the alkyl chains of the gemini surfactants, yielded v = 512 ų and l_c = 24.3 Å. These values were treated as effective parameters and applied consistently across all formulations to enable comparative analysis of packing behavior. Because SAXS primarily probes cross-sectional structure in anisotropic systems, micelle length could not be reliably determined. Packing parameters were therefore calculated from cross-sectional dimensions and represent effective values that reflect trends in curvature and interfacial organization rather than absolute geometric parameters.

The aggregation number (N_agg) was estimated from SAXS-derived dimensions assuming an elliptical cylindrical geometry, using N_agg = V/v, where V = πR_minor × R_majorL. Because micelle length is poorly constrained by SAXS, representative values informed by TEM were used. Therefore, N_agg values are approximate and are shown for comparative analysis across formulations rather than as absolute quantities.

Evaluation of IRX4204 intranasal brain delivery in mice

Adult female C57BL/6 mice (12–14 weeks; Charles River, Canada) were used in all studies. All procedures were approved by the Dalhousie University Committee on Laboratory Animals and followed Canadian Council on Animal Care guidelines. Mice were housed under a 12 h light/dark cycle with food and water ad libitum and acclimated for one week prior to experiments. At study end, mice were euthanized by intraperitoneal sodium pentobarbital (200 µL, 34 mg mL⁻¹) followed by intracardiac PBS perfusion (10 mL, pH 7.4). For intranasal administration, mice were anesthetized with 2% isoflurane and placed supine. GPNP formulation, IRX4204 suspension, or PBS was administered as 6 µL aliquots alternated between nostrils at 1 min intervals to a total volume of 24 µL. Mice were maintained supine for 1 min after the final dose before recovery.

At the designated endpoint, mice were euthanized and following loss of pedal reflex, about 400 µL blood was collected via cardiac puncture into 10 µL of 3.2% sodium citrate (pH 7.9), then centrifuged (1500g, 10 min, 4 °C) to obtain plasma. Spinal cords were isolated by hydraulic extrusion, brains were dissected and regionally sectioned (forebrain/hindbrain or olfactory bulb, striatum, cortex, hindbrain), and livers were collected. Samples were extracted with acetonitrile (1 : 4, tissue mass per volume of acetonitrile). Tissues were homogenized using 1.5 mm zirconium beads (BeadBug™) and centrifuged (12 000g, 15 min, 4 °C); supernatants were clarified by a second centrifugation and diluted 60 : 40 by volume with water prior to analysis.

LC/MS analysis

Quantification of IRX4204 was performed using a Vanquish Flex UHPLC coupled to a Q Exactive™ Orbitrap mass spectrometer (Thermo Fisher Scientific, MA, USA). Separation was achieved on an ACE C18-PFP column (50 × 2.1 mm, 5 µm) at 35 °C with a 20 µL injection. The mobile phase consisted of 0.4% formic acid in water (A) and 0.4% formic acid in acetonitrile (B) at 0.75 mL min⁻¹ using a gradient from 60% to 95% B (2.0 min), held to 2.4 min, then returned to initial conditions (total run time 3 min).

Electrospray ionization (positive mode) was operated at 3.5 kV with a capillary temperature of 300 °C (sheath gas 45, auxiliary gas 10, auxiliary heater 250 °C). Parallel reaction monitoring (PRM) was used with precursor m/z 353.24751 (IRX4204) and 349.21621 (TTNPB, internal standard). Collision energies were 25 eV and 30 eV, respectively. Transitions used for IRX4204 quantification were m/z 307.2420, 293.2264, 283.1693, and 265.1587; for TTNPB, m/z 333.1849, 293.1356, 279.1379, and 265.1223. Resolution was 35 000, AGC target 2 × 10⁵, and maximum injection time 200 ms. The method was validated by identifying spectral peaks for these transition ions in samples spiked with TTNPB and/or IRX4204 in each matrix. To calculate drug levels in tissue, increasing concentrations of IRX4204 were used in samples spiked with 10 nM TTNPB. IRX4204 standards included 1 nM, 3 nM, 5 nM, 7 nM, 10 nM, 30 nM, 50 nM, 70 nM, 100 nM, 300 nM, 500 nM, 700 nM, and 1 µM. The following standard curve properties were used for quality control: R² > 0.995, lower limit of quantification ≥ 10 nM, and limit of detection ≥ 3 nM. All values below 3 nM were considered 0 nM. All samples were processed on the day of tissue extraction.

Quantitative RT-PCR (qRT-PCR)

Spinal cords were isolated by hydraulic extrusion, and brains were dissected and sectioned into forebrain and hindbrain. Livers were also collected. Tissues were flash-frozen, homogenized in pureZOL™, and total RNA was extracted using Aurum RNA kits per manufacturer protocols. RNA quality and integrity were confirmed spectrophotometrically and by electrophoresis, and only samples with RIN ≥ 7.5 were used. cDNA was synthesized using the iScript™ kit and stored at −20 °C. RT-qPCR was performed using SoFast EvaGreen Supermix on a Bio-Rad CFX96 system under optimized conditions, with melt curve analysis. Gene expression was calculated using the 2^−ΔΔCq (Livak) method. HPRT1 and GAPDH were used as reference genes. Data reflect averages of triplicate technical replicates. Forward and Reverse Primers sequences were as follows:

GAPDH(F) AGGTCGGTGTGAACGGATTTG; GAPDH(R) GGGGTCGTTGATGGCAACA; HPRT1(F) TCAGTCAACGGGGGACATAAA; HPRT1(R) GGGGCTGTACTGCTTAACCAG; SCD1(F) TTCTTGCGATACACTCTGGTGC; SCD1(R) CGGGATTGAATGTTCTTGTCGT; LXRα(F) CTCAATGCCTGATGTTTCTCCT; LXRα(R) TCCAACCCTATCCCTAAAGCAA.

Statistical evaluations

All statistical analyses were performed using GraphPad Prism 8.0. One-way ANOVA followed by Dunnett's multiple comparisons test was used to analyze the effect of IRX4204 on SCD1 and LXRα. mRNA levels. Timepoint zero was used as the control group for each respective mode of delivery. For biodistribution studies comparing intranasal and oral groups two-tailed Welch's t-tests were used at 2 and 24 h with brain-targeting parameters (K_p, DTE, and DTP) calculated from the 6 h data where plasma concentrations were quantifiable and provided the most stable comparison between routes.

Results and discussion

Physicochemical characterization of IRX4204-GPNPs

The components, schematic representation and physicochemical properties of GPNPs are summarized in Fig. 1A–D. Incorporation of gemini surfactants (mixture of GS_RGD and GS) with the other lipid components enabled the use of a one-pot ultrasonication method (USM-1–2 h) to prepare both IRX4204 loaded GPNPs (IRX4204-GPNPs 1 : 1 : 2 : 10, 0.1 mg mL⁻¹ drug), and blank GPNPs 1 : 1 : 2 : 10. In this approach, all components, including the drug, were combined in aqueous medium and subjected to ultrasonication to achieve encapsulation.

Fig. 1. Formulation, characterization and stability of IRX4204-GPNPs. (A) Schematic representation of the components of GPNPs. (B) Transmission electron micrographs of the blank and IRX4204-GPNP 1 : 1 : 2 :10, 0.1 mg mL⁻¹ formulations prepared by one-pot mixing and ultrasonication, and (C) corresponding DLS profile of IRX4204-GPNP 1 : 1 : 2 : 10, 0.1 mg mL⁻¹ formulation. The DLS profile includes the pink shaded region where the peak corresponding to IRX4204 crystals would appear if present (see also E). (D) Physicochemical parameters of four independent batches of blank and IRX4204-GPNP 1 : 1 : 2 : 10, 0.1 mg mL⁻¹ including main peak average hydrodynamic diameter (HD), polydispersity index (PDI), zeta potential (ZP). Data represent mean ± SD (n = 3). (E) DLS profiles of blank GPNP, IRX4204 suspension (0.1 mg mL⁻¹) (represented by the peak in the pink shaded area) and IRX4204-GPNP 1 : 1 : 2 : 10, 0.05 mg mL⁻¹ as the first proof-of-concept formulation prepared by remote loading process. Also shown is the TEM image for the IRX4204-GPNP 1 : 1 : 2 : 10, 0.1 mg mL⁻¹ formulation showing the similarity to the original method.

TEM imagining of both IRX4204-GPNPs and blank GPNPs showed predominantly spherical to slightly flattened (disc) shaped particles with narrow size distributions (Fig. 1B) consistent with DLS measurements (Fig. 1C). The GPNPs were about 60 nm, with PDI of about 0.3 and ZP of 40–45 mV with acceptable batch-to-batch reproducibility (Fig. 1D).

A second approach, complementary to the USM-1–2 h method, was also evaluated. In this strategy, the drug was added to preformed blank GPNPs followed by further ultrasonication (‘remote-loading’). DLS and TEM results showed that GPNP 1 : 1 : 2 : 10 formulation is suitable for remote loading of IRX4204. Separately prepared blank GPNPs and IRX4204 suspension (obtained by ultrasonication using the same protocol) showed distinct peaks with particle sizes of approximately 60 nm and 400 nm, respectively, corresponding to GPNPs and drug crystals (Fig. 1E). Upon mixing at 1 : 1 v/v ratio (final drug concentration 0.05 mg mL⁻¹) followed by ultrasonication, IRX4204 suspension (0.1 mg mL⁻¹) incorporated into the GPNPs with no detectable evidence of free drug crystals (Fig. 1E). The resulting DLS profile showed one main peak corresponding to IRX4204-GPNPs with approximately 60 nm (Fig. 1E). Consistent with these findings, the TEM imaging of the combined formulation confirmed the absence of crystalline drug (Fig. 1E) and similar morphology to the formulation prepared by direct mixing USM-1–2 h method, confirming that the two methods produced similar results. Next, a series of IRX4204-GPNPs 1 : 1 : 2 : 10 with increasing concentration of drug (0.05–2 mg mL⁻¹) were prepared by remote loading, in which IRX4204 suspension was added to the preformed blank-GPNPs and processed by ultrasonication. The resulting DLS profiles showed particle size distribution patterns indicative of the encapsulation efficiency in this type of GPNPs (Fig. 2A).

Fig. 2. Physicochemical evaluation of blank and IRX4204-GPNPs with increasing drug concentrations prepared by remote loading. (A) DLS profiles of GPNPs with increasing IRX4204 concentration. (B) Detailed physicochemical parameters of blank IRX4204-GPNP 1 : 1 : 2 : 10, 0 : 2 : 2:10 and 0 : 10 : 10 : 50 formulations indicating the mean hydrodynamic diameter of individual particle populations (main Peak 1 and Peak 2 by Intensity%), PDI and ZP. (C) Transmission electron micrographs of IRX4204-GPNP 1 : 1 : 2 : 10, 2 mg mL⁻¹. (D) Physical stability of IRX4204-GPNP 1 : 1 : 2 : 10, 0.1 mg mL⁻¹ at 4 °C for 6 months assessed by hydrodynamic diameter (HD), polydispersity index (PDI), zeta potential (ZP). Data represent mean ± SD (n = 3).

Since IRX4204 is highly hydrophobic, assessment of encapsulation efficiency requires indirect methods. The free drug suspension (in aqueous solution) exhibited a well-defined peak in the DLS profile at around 400 nm which could be used to evaluate the presence of unencapsulated drug. In the formulations, the IRX4204-GPNP population (about 70 nm) and IRX4204 drug crystals (about 400 nm when subjected to the sonication procedure similarly to the GPNPs) exhibited clearly distinct size distributions. Thus, the absence of drug crystal peak in the GPNP formulations provided a semi-quantitative indication of encapsulation. Hence, DLS was used as a formulation screening tool to detect and compare the presence of non-incorporated drug within GPNPs. Other frequently used methods for determining encapsulation efficiency, such as centrifugation, dialysis or filtration, are not suitable for IRX4204-GPNPs due to co-sedimentation or retention of both GPNPs and drug crystals or because they disrupt GPNPs resulting in drug release. In the sample series with increasing drug concentration, drug crystals were not detected until reaching 1 mg mL⁻¹. The encapsulation of IRX4204 in GPNPs 1 : 1 : 2 : 10 was complete (can be considered 100%) up to 0.5 mg mL⁻¹ (Fig. 2A). At or beyond this concentration there may be saturation and at 1 mg mL⁻¹ and higher, there was evidence of large aggregates and possibly free drug crystals which can be seen in both the DLS profile (Fig. 2A) and TEM images (Fig. 2C). These results suggest that the encapsulation capacity of the GPNP system is saturated at approximately 0.5 mg mL⁻¹, beyond which excess drug remains as insoluble crystals. Additional compositions were also evaluated to test whether using GS only (not the mixture of GS_RGD and GS) and increasing the overall lipid-surfactant concentration would enhance encapsulation (Fig. 2B). For these types of GPNPs the particle size were similar, however, they were more polydisperse and higher lipid concentration did not result in proportionately higher drug encapsulation.

The selected formulation, IRX4204-GPNPs 1 : 1 : 2 : 10, 0.1 mg mL⁻¹, exhibited good colloidal stability during storage, with zeta potential remaining essentially unchanged over 24 weeks. Using a predefined acceptance criterion of ±20% relative to the initial particle size (47.36–71.04 nm) the formulation remained within this specification for 20 weeks but slightly exceeded the limit at 24 weeks (Fig. 2D). The mean particle size increased from 59.2 to 71.8 nm over 24 weeks, which corresponded to an increase of approximately 21%. The PDI increased from 0.230 ± 0.020 to 0.376 ± 0.066, indicating progressive broadening of the particle size distribution during storage although this is still within the acceptable range of <0.4. The increase in PDI likely reflects progressive broadening of the nanoparticle size distribution due to the formation of a minor population of larger particles during storage. Because zeta potential remained essentially constant throughout the study, the observed increase in PDI is unlikely to result from the loss of electrostatic stabilization and could be due to instead gradual particle association, micellar restructuring, Ostwald ripening or drug-induced heterogeneity within the nanoparticle population.

Structural characterization of GPNPs by SAXS

Small-angle X-ray scattering (SAXS) was employed to investigate the internal structure of the GPNPs and assess structural changes associated with increasing drug loading. Fig. 3A shows the SAXS curves for blank GPNP 1 : 1 : 2 : 10 and IRX4204-GPNPs 1 : 1 : 2 : 10 with 0.1, 0.25, 0.5, 1 and 2 mg mL⁻¹ drug and additional compositions such as blank GPNP 0 : 2 : 2 : 10 and IRX4204-GPNPs 0 : 2 : 2 : 10 with 0.1 and 0.5 mg mL⁻¹ drug and blank GPNP 0 : 10 : 10 : 50 and IRX4204-GPNPs 0 : 10 : 10 : 50 with 5 mg mL⁻¹ drug (Fig. 3B).

Fig. 3. SAXS evaluation of blank and IRX4204-GPNPs with alternative compositions and increasing drug concentrations. (A) SAXS profiles of GPNPs 1 : 1 : 2 : 10 with IRX4204 concentration ranging from 0–2 mg mL⁻¹. (B) SAXS profiles of GPNPs 0 : 2 : 2 : 10 and 0 : 10 : 10 : 50 with IRX4204 concentration ranging from 0–5 mg mL⁻¹. (C). Schematic illustration of the relationship between GPNP composition, micellar/vesicular structure and SAXS scattering behavior. Top row: composition symbols for GPNPs prepared with GS_RGD + GS or GS only (low and high-concentration). Middle row: representative illustrations of the structural information indicating that the GPNPs 1 : 1 : 2 : 10 heterogeneous interfacial packing, resulting in flexible, anisotropic micelles with increased structural adaptability, whereas GPNPs 0 : 2 : 2:10 display more uniform packing and form more rigid, less adaptable micelles and the GPNPs 0 : 10 : 10 : 50 show strong interparticle interactions and more tightly packed micelle/vesicular system. Bottom row: the corresponding conceptual SAXS schematic curves shown here indicate structural features that differentiate intra-particle and inter-particle effects. At lower concentrations and in GPNPs 1 : 1 : 2 : 10, a broad shoulder is observed, consistent with anisotropic or flattened micellar geometry. In contrast, at higher concentrations, the emergence of a correlation peak indicates the development of interparticle interactions and spatial organization. These features reflect a transition from form factor-dominated scattering to structure factor contributions associated with particle–particle correlations. GPNPs 1 : 1 : 2 : 10 show smooth decay with a subtle shoulder corresponding to micelle flattening, GPNPs 0 : 2 : 2:10 a smoother SAXS profile with minimal structural features and GPNPs 0 : 10 : 10 : 50 show strong interparticle interactions, leading to more tightly packed particles and the emergence of a correlation peak in the SAXS profile, indicative of short-range ordering between particles. Overall, increasing packing uniformity and intermolecular interactions reduce structural adaptability, shifting the system from flexible, drug-accommodating micelles toward interaction-dominated assemblies. A shoulder in the SAXS profile reflects anisotropic micelle shape (form factor), whereas the emergence of a correlation peak indicates interparticle interactions (structure factor). (D) Structural parameters of micellar aggregates derived from SAXS analysis. Effective headgroup area (a₀) and packing parameter (P) were calculated using cross-sectional dimensions obtained from elliptical cylinder fits. ΔP represents the change in packing parameter relative to the corresponding drug-free formulation within each system. The packing parameter (or critical packing parameter) is a dimensionless quantity that describes the preferred self-assembly of amphiphilic molecules. It captures the balance between headgroup repulsion and hydrophobic tail packing, thereby determining the curvature and resulting morphology of the assembly. Parameters are not reported for GPNPs 0 : 10 : 10 : 50 due to strong interparticle interactions, multimodal structure, and deviation from single-particle scattering behavior.

The scattering profiles show a broad peak generally centered around q = 0.1 Å⁻¹ in each sample characteristic of micellar structures or disordered systems. After excluding the spherical micelle model and core–shell particle model, analysis showed that the cylindrical micelle model also failed to adequately describe the data (χ² values >40). The resulting fits produced unstable parameters and unrealistically large micelle lengths, reflecting the limited sensitivity to the longitudinal dimension of anisotropic particles. Similarly, lamellar fits yielded thicknesses outside of physically reasonable range, suggesting that the system is better represented as finite anisotropic aggregates with bilayer-like tendencies rather than ideal cylinders or extended lamellae. Among the models evaluated, the elliptical cylinder model provided the overall best description of the scattering data, although it still should be regarded as an approximation. The elevated χ² values observed in some samples suggest that additional factors, such as polydispersity, particles flexibility, or weak interparticle interactions, are not fully captured. Consequently, the uncertainty associated with the aggregate dimensions propagates into the calculated N_agg values, therefore, these values are considered comparative metrics rather than absolute quantities (Fig. 3D). Similarly, the assumption of a homogeneous elliptical geometry may not fully account for local structural heterogeneity within the aggregates. The combined SAXS, TEM and DLS results suggest the coexistence of multiple structural populations. TEM showed both smaller anisotropic aggregates (about 10 × 50 nm) and larger spherical particles (about 60–80 nm), while DLS showed peaks corresponding to these two populations, with a dominant peak at around 60 nm and a secondary population near 10 nm (Fig. 1 and 2). These results indicate that the particle population may be bimodal, consisting of flattened micelles coexisting with larger vesicular structures. Peak position shifts likely reflect changes in interparticle spacing or disruption of weakly organized domains. With increasing drug concentration (GPNP 1 : 1 : 2 : 10 series), peak intensity decreased, indicating progressive disorder due to drug-induced rearrangement and altered molecular packing.

Structural evolution across formulations is primarily driven by drug concentration. In GPNP 1 : 1 : 2 : 10, increasing drug loading leads to larger and more anisotropic micelles with decreasing effective headgroup area and increasing packing parameter (Fig. 3C and D). This suggests drug partitioning at the interface, increasing effective hydrophobic volume and reducing curvature, driving a transition from compact micelles to flattened, low-curvature aggregates.

Beyond about 1 mg mL⁻¹ drug loading, SAXS profiles and micellar dimensions plateau, indicating saturation of drug incorporation and suggesting that excess drug partitions into a secondary phase. This is supported by DLS and TEM, which show larger, heterogeneous structures consistent with aggregation or phase separation that contribute weakly to the SAXS signal. Overall, stable flattened micelles coexist with larger aggregates at high loading, indicating limited micellar capacity and greater structural heterogeneity, with the four-component system (1 : 1 : 2 : 10) showing higher loading tolerance than the three-component (0 : 2 : 2 : 10) or more concentrated (0 : 10 : 10 : 50) formulations (Fig. 3C and D).

In GPNP 0 : 2 : 2 : 10 and 0 : 10 : 10 : 50 formulations drug-induced structural changes were observed, however, the presence of GS_RGD in GPNP 1 : 1 : 2 : 10 enhanced these effects, producing greater anisotropy and larger aggregates suggesting a modulatory role in interfacial packing.

The system approaches a curvature-driven transition regime where both micellar and bilayer-derived structures are accessible. The presence of vesicular aggregates together with increasing packing parameter supports a transition toward low-curvature organization, although the absence of clear lamellar features indicates that extended bilayers do not dominate. Instead, the system exhibits coexistence of flattened micelles and vesicles (Fig. 3C and D).

Analysis of SAXS-derived cross-sectional dimensions yielded headgroup areas within a narrow range (about 25–30 Ų), indicating similar geometric packing constraints across formulations. However, differences in scattering behavior, and structural adaptability indicate that simple geometric parameters do not fully capture interfacial organization. The mixed GS_RGD + GS system (1 : 1 : 2 : 10) showed greater packing heterogeneity and structural flexibility, suggesting more dynamic and heterogeneous interfacial environments.

Packing parameters (P approximately 0.7–1.0) indicate all systems lie within a low-curvature regime (Fig. 3D). The mixed GS_RGD + GS formulation shows the largest increase in P with increasing drug concentration, approaching values near unity and indicating transition toward bilayer-like organization. In contrast, the three-component systems show smaller changes and reduced structural adaptability. The high-concentration system (0 : 10 : 10 : 50) exhibits interaction-dominated behavior with stronger interparticle correlations and constrained structure.

These results indicate that drug incorporation depends on the ability of the system to reorganize. The mixed GS_RGD + GS system exhibits greater packing heterogeneity and interfacial flexibility, enabling continuous structural adaptation, while GS only systems display more uniform packing, limiting reorganization and aggregation at higher concentrations.

In addition to the broad peak observed in all SAXS curves typical of micellar systems, all SAXS profiles also show small superimposed sharper peaks in both drug containing and blank GPNPs indicating a secondary structural feature or coexisting phase such as a minor ordered domain or short-range packing or a crystalline feature that could be related to excipients in the formulation.

Based on these physicochemical and structural analyses, IRX4204-GPNPs 1 : 1 : 2 : 10, 0.1 mg mL⁻¹ was selected for in vivo evaluation due to complete (or near-complete) drug encapsulation (as estimated from the DLS profile), structural stability over extended storage, favorable particle size and surface charge and absence of drug crystallization.

Enhanced CNS delivery with intranasal IRX4204-GPNPs

Changing the route of administration of IRX4204 from oral to intranasal delivery was hypothesized to provide improved targeting to the CNS while reducing systemic exposure. This is important since IRX4204, like other RXR agonists, is known to elevate plasma lipid levels and can cause hypothyroidism, neutropenia and hepatomegaly (the latter in mice), although to a lesser extent than first-generation RXR agonists.⁹

Therefore, to evaluate whether intranasal delivery of IRX4204 using GPNPs improves delivery efficiency and safety relative to oral administration of unencapsulated IRX4204, an in vivo study was conducted comparing tissue drug exposure following both dosing routes. Nanoparticles in the 50–200 nm range with a moderately positive zeta potential (around +30 mV) have been reported to facilitate transport across nasal pathways, including mucosal penetration and brain uptake,²⁰ although effective delivery can also depend on the material-specific properties of the NPs.²¹

Intranasal delivery of IRX4204 in GPNPs (24 µL total dose, 0.12 mg kg⁻¹) was compared to oral administration of IRX4204 suspension (12 mg kg⁻¹). Despite the 100-fold lower administered dose, intranasal delivery produced comparable IRX4204 concentrations in the spinal cord and hindbrain at 2 h and markedly higher concentrations in the striatum at this time point.

At later time points (6 and 24 h) intranasal delivery resulted in consistently higher concentrations across all CNS regions (Fig. 4A). By 12 h, IRX4204 was undetectable in the CNS after oral administration, whereas measurable levels persisted following intranasal delivery at 24 h.

Fig. 4. Biodistribution and pharmacodynamic measurements of IRX4204 following a single oral administration of IRX4204 suspension or intranasal GPNP administration. (A) IRX4204 concentrations in the plasma, lungs, liver, spinal cord, olfactory bulb, hindbrain, striatum, and cortex of C57Bl/6 mice at varying times after oral IRX4204 suspension (A; 12 mg kg⁻¹), and, (B) intranasal IRX4204-GPNPs (B; 0.12 mg kg⁻¹; n = 3–6; mean ± SEM). Selected pair-wise comparisons between treatment groups at individual time points using Welch's two-tailed t-tests showed higher plasma and liver levels after oral administration at 2 h (p < 0.05–0.01), while intranasal delivery produced significantly higher CNS concentrations at 24 h (p < 0.01; p < 0.001 in striatum). (C) SCD1 gene expression normalized relative to reference gene expression 2, 6, or 24 h after a single intranasal dose of empty-NPs or IRX4204-NPs (0.12 mg kg⁻¹) in the hindbrain of C57Bl/6 mice (n = 2–4; ordinary one-way ANOVA, Dunnett's multiple comparisons test, mean ± SEM). *p < 0.05, **p < 0.01 and ***p < 0.001 relative to 0 h. (D) LXR1 gene expression normalized relative to reference gene expression 2, 6, or 24 h after a single intranasal dose of empty GPNPs or IRX4204-GPNPs (0.12 mg kg⁻¹) in hindbrain of C57Bl/6 mice (n = 2–4; ordinary one-way ANOVA, Dunnett's multiple comparisons test, mean ± SEM). *p < 0.05 and **p < 0.01 relative to 0 h. (E) Comparative tissue distribution, absolute drug amounts, and delivery efficiency following intranasal and oral administration of IRX4204. Mean concentrations (nM), calculated tissue drug amounts (µg), and percentage of administered dose are shown for peripheral tissues and CNS regions as means of measured concentration across the sampling period (24 h). Calculations incorporate corrected tissue weights for adult mice. Intranasal administration (0.12 mg kg⁻¹) resulted in 1.4-fold higher absolute CNS exposure and 136-fold greater delivery efficiency compared to oral dosing (12 mg kg⁻¹), while reducing systemic and hepatic exposure. Fold changes are expressed relative to oral administration. (F) Calculated parameters to compare the improvement after intranasal versus oral administration of IRX4204. DE represents the percentage of the dose reaching the CNS, DTE shows the selectivity/preferential delivery to the CNS, DTP estimates the fraction of the drug reaching the CNS directly via nose-to-brain pathways and K_p represents brain penetration/partitioning across the BBB.

CNS concentrations after intranasal administration of IRX4204 GPNPs exceeded the biologically effective threshold (about 10 nM) required for inducing oligodendrocyte progenitor cell (OPC) differentiation and suppressing inflammation and dopaminergic neuron loss in preclinical models.²² In contrast, oral administration resulted in more transient exposure that maybe insufficient for sustained pharmacological activity. In addition to enhanced CNS exposure, intranasal delivery significantly reduced systemic drug levels. Peak plasma (130.6 nM; 2 h) and liver (422.2 nM, 2 h) concentrations following oral administration were 5-fold and -18-fold higher, respectively, than those measured after intranasal IRX4204 GPNP administration (24.7 nM; 24 h and 23.2 nM; 24 h, respectively) (Fig. 4A). Quantitative analysis confirmed that intranasal administration delivered approximately 1.5 times as much drug to CNS tissues (0.0045 vs. 0.0033 μg), despite the lower dose (Fig. 4E and F panel 1). When normalized to dose, this corresponds to 0.19% of the intranasal dose reaching the CNS, compared to about 0.0014% following oral administration, representing an approximately 136-fold improvement in delivery efficiency (Fig. 4E). The improved CNS targeting is reflected in the brain-to-plasma ratio (K_p), which increased from about 0.83 (oral) to about 3.11 (intranasal), corresponding to an about 4-fold enhancement (Fig. 4F).

Pharmacodynamic confirmation of CNS delivery

To determine whether IRX4204 delivery using GPNPs was released in the brain at levels sufficient to activate RXRs, pharmacodynamic responses were assessed.

Initial experiments with oral IRX4204 (12 mg kg⁻¹) identified several RXR-responsive genes (including Delta-like non-canonical Notch ligand 1 (DLK1), RXRγ, Stearoyl-CoA desaturase 1 (SCD1), PGC1α, liver X receptor (LXRβ) and Acetyl-CoA carboxylase α (ACACA)) that were upregulated in hindbrain tissue.^23,24 Among these, SCD1 expression was found to be consistently elevated following both acute and repeated dosing and therefore was selected as a robust marker of RXR activation.

Evaluation of gene expression following intranasal administration demonstrated sustained pharmacodynamic activation. SCD1 mRNA levels in the hindbrain increased steadily over 24 h following intranasal IRX4204 GPNPs (1 : 1 : 2 : 10) (0.1 mg mL⁻¹) administration, whereas oral administration resulted in a delayed peak at 12 h followed by a decline (Fig. 4C).

Similarly, LXRα mRNA levels also showed rapid 2-fold increase within 2 h after intranasal treatment, that was maintained up to 24 h, while expression declined after oral dosing (Fig. 4D). Mice treated intranasally with IRX4204 suspension did not show increased hindbrain SCD1 expression relative to the PBS control mice (result not shown), indicating that NP encapsulation is important for efficient CNS delivery and pharmacodynamic activity. The hindbrain was selected for pharmacodynamic assessment due to its well-established involvement in demyelination and neurodegeneration in experimental autoimmune encephalomyelitis (EAE) and multiple sclerosis (MS), as well as its relevance to other CNS disorders.^25–27

Comparative CNS targeting and delivery efficiency of intranasal versus oral IRX4204

These results suggest that the intranasal delivery of IRX4204 to the CNS using GPNPs provides advantages over oral administration. Intranasal IRX4204-GPNPs achieved sustained, pharmacologically relevant drug concentrations in the brain and spinal cord, reduced systemic and hepatic exposure, and significantly enhanced delivery efficiency.

Fig. 4F summarizes the calculated parameters reflecting comparative aspects of intranasal versus oral administration of IRX4204. Intranasal delivery results in 136-fold increase of IRX4204 concentration in CNS regions compared to oral, and when normalized to administered dose, CNS exposure, delivery efficiency (DE) is enhanced by about 136-fold relative to oral delivery. Drug Targeting Efficiency (DTE) was estimated to be 380%, indicating a 4-fold enhancement in CNS targeting relative to oral administration. Direct Transport Percentage (DTP; here defined as CNS delivery attributable to the intranasal route relative to oral administration or the fraction of CNS exposure that cannot be explained by systemic absorption alone) was estimated to be about 73.5%, suggesting that the majority of CNS exposure following intranasal administration occurs via direct nose-to-brain transport rather than systemic circulation. Region-specific DTP ranged from 62% in the olfactory bulb to 94.1% in the striatum, with most CNS regions exhibiting values between 70–82%, confirming that intranasal delivery is dominated by direct nose-to-brain transport rather than systemic exposure. Intranasal administration significantly enhanced brain partitioning (K_p) across all regions, with 3–17-fold increases relative to oral delivery, confirming efficient BBB bypass and preferential CNS targeting (Fig. 4F).

DTE, DTP and K_p were calculated using concentration ratios at the 6 h time point because this represented a common sampling point with quantifiable drug concentrations in both plasma and CNS tissues. This timepoint-specific analysis demonstrated maximal and most interpretable brain-targeting efficiency at 6 h (DTE = 377%; log DTE% = 2.58; DTP = 73.5%). Values at 2 h and 24 h were strongly influenced by very low plasma concentrations and resulted in unreliable values. The striatum exhibited the earliest and most pronounced increase in drug exposure following intranasal administration, suggesting preferential regional delivery. Its proximity to the ventricular system may facilitate access via cerebrospinal fluid-associated transport pathways, although this remains to be confirmed experimentally.

Intranasal nanoparticle-based nose-to-brain delivery systems reported in the literature generally demonstrated DTE in the range of approximately 150–400% and DTP of 40–80%, depending on formulation type and experimental conditions. For example, sumatriptan-loaded nanostructured lipid carriers exhibited DTE of about 258% and DTP of 61%,²⁸ while other lipid-based (solid lipid nanoparticles, nanostructured lipid carriers, liposomes) and nanoemulsion systems typically show DTE values around 200–300% and DTP values in the range of 60–70%.^29–32 In this context, the present dataset, with a DTE of 380%, a DTP of 73.5%, and a 4-fold increase in brain-to-plasma partitioning, is consistent with the upper end of values reported for optimized nanoparticle systems. Nguyen and Maeng demonstrated that intranasal nanoparticle systems typically increase log DTE% above the baseline value of 2 (corresponding to DTE = 100%), with most formulations distributed between about 2.2 and 2.6. The present system, with a log DTE% of 2.58, lies within the upper range of these reported values, indicating efficient brain targeting consistent with optimized lipid nanoparticle platforms.³³

GPNPs differ from conventional lipid-based carriers, including liposomes, emulsions, and SLNs/NLCs, in both internal organization and interfacial behavior, which may, at least partially, explain their improved in vivo performance. The twin-head/twin-tail architecture reduces packing constraints, conferring greater membrane fluidity and elasticity. GSs enhance mucosal performance through their amphiphilic dicationic structure and low critical micellar concentration (CMC), enabling strong mucosal interactions, sustained structural integrity and efficient membrane permeation. This enables a balance between mucoadhesion and diffusion that is difficult to achieve with more rigid lamellar or crystalline lipid systems, which tend to favor adhesion at the expense of penetration. Consistent with these properties, IRX4204-GPNPs produce sustained CNS exposure with reduced systemic distribution and efficient brain targeting, along with a substantial contribution of direct nose-to-brain transport, supporting a mechanistic link between the flexible, high-curvature GPNP structure and improved in vivo behavior.

Overall, the physicochemical parameters of IRX4204-GPNP 1 : 1 : 2 : 10, 0.1 mg mL⁻¹ formulation, such as particle size range 60–70 nm, positive charge, stability and structural features, indicate favorable properties for intranasal CNS delivery, as small, sub-100 nm non-aggregating nanoparticles exhibit enhanced diffusion through the mucus barrier and improved transport along olfactory pathways, whereas larger or interacting aggregates are sterically hindered and rapidly cleared.³² This behavior is consistent with literature reports, indicating nanocarriers improve brain targeting by overcoming BBB via olfactory and trigeminal pathways, but only when particle size and interaction state are optimized.^20,32,34

While these results demonstrate clear advantages of intranasal delivery, several limitations should be considered. Tissue distribution estimates rely on time-averaged concentration data, which may introduce some uncertainties in absolute values. However, these assumptions were applied consistently across both delivery routes, ensuring robust relative comparisons. Future studies incorporating full pharmacokinetic modeling (e.g., AUC-based analysis) would further refine these findings. Additionally, mechanistic studies are needed to more precisely delineate the relative contributions of olfactory, trigeminal, and systemic transport pathways. The regional distribution pattern, with enhanced delivery to hindbrain, spinal cord, and striatum supports a mixed transport mechanism involving both direct nose-to-brain pathways and systemic redistribution (noticeable from the increased lung IRX4204 concentration after intranasal delivery), which might be related to nanoparticle inhalation or swallowing part of the dose in the mouse model.

Conclusions

IRX4204-GPNPs exhibit well-defined physicochemical and structural properties that enable efficient intranasal delivery to the CNS. Optimized formulations provide stable drug encapsulation, resulting in enhanced CNS exposure, reduced systemic distribution, and approximately 136-fold improved delivery efficiency compared to oral administration. The combined pharmacokinetic and pharmacodynamic findings demonstrate that intranasal delivery of IRX4204-GPNPs provides a substantial advantage over oral administration. This approach enables higher and more sustained CNS exposure, reduces systemic and hepatic accumulation, and improves delivery efficiency by approximately two orders of magnitude. Importantly, these improvements are achieved using a substantially lower administered dose, highlighting the potential for reduced toxicity and improved therapeutic index. From a translational perspective, these findings are particularly relevant for the development of therapies targeting neurodegenerative diseases. Effective treatment of such conditions requires sustained drug exposure across multiple CNS regions, including the striatum, spinal cord and hindbrain. The ability of intranasal IRX4204-GPNPs to deliver therapeutically relevant concentrations to these regions, while minimizing peripheral exposure, suggests a promising strategy for improving both efficacy and safety.

Author contributions

All four authors designed the experiments and analyzed the results. A. H. (physicochemical, in vitro experiments), G. K. (in vivo experiments) carried out experimental work. All four authors wrote the manuscript and provided revisions. M. F. and G. R. acquired funding.

Conflicts of interest

All authors are inventors on a patent application related to this work.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper. Fig. 1–3, including the tables, provide the dataset for the physicochemical characterization and electron micrographs. Fig. 4A–D and table in Fig. 4E contain dataset on the in vivo biodistribution data.