The ASO chains, in general, and the GR sequences considered in this study, take a rigid conformation as linear polyanion strands with the phosphate groups exposed to the outer helix surface (Uppuladinne et al., 2019). As a construction principle, we invoke the concept of counterion Manning-like condensation around the polynucleotide chains, behaving as rigid rods at a high anionic charge (Kebbekus et al., 1995). Figure 2 shows the supramolecular catanionic assembly composed of cationic DChol as condensed into the GR polyanion (Kim et al., 2008). The structure of the GR-polyanion is described in Figure 2A, and the DChol component as the condensation counterpart is described in Figure 2B. A lamellar condensed (GR/DChol bilayer) structure resulted from electrostatic interactions, as modulated in solvents with a variable dielectric permittivity (see Figure 2C). The Bligh–Dyer (BD) solvent was used to provide a tunable permittivity medium, as composed by ternary mixtures of water, methanol, and chloroform (see the phase diagram in Figure 3B). Electrostatics and hydrophobicity are both counterbalanced by tuning BD permittivity, hence resulting in catanionic GR:DChol₂ lipoplexes stoichiometrically formed upon a thermodynamic philic–phobic trade-off (Bligh and Dyer, 1959).

The GR components were chosen by their biological functionality for gene silencing and their chemical affinity as strong acids ( p K a ≈ 2 ) for the electrostatic nano-assembly. Figure 2A shows the GRs containing a central block of single-stranded deoxyribonucleotides for silencing L-PGDS-mRNA under RNase H cleavage. This ssDNA is flanked by 2′-O-methyl-modified ribonucleotide wings (LNA™), which shield the internal block from nuclease degradation (Monia et al., 1993). Coarse-graining as linear polyanions results in rigid GR rods with approximate cylindrical dimensions (for n = 16 bases; width D ≈ 1 n m , length L = n l 0 ≈ 5.5 n m , and l 0 ≈ 0.34   n m the average size of a single base) (Alberts et al., 2015). Due to the large persistence of the ribonucleotide chains ca. 72 n m ≫ L (Kebbekus et al., 1995), they stretch out as charged cylinders with a uniform surface density of charge σ G R ≈ − n e / π L D ≈ − e / n m 2 (being − e the charge of the electron (Uppuladinne et al., 2019)); see caption of Figure 2A for details. The basic DC-cholesterol (DChol) was chosen for cationic lipid formulation under Coulombic attraction and chemical affinity for anionic nucleic acids. Furthermore, DChol entails colloidal amphiphilicity and biocompatibility superior to other cationic lipids used in drug delivery (Chen et al., 2019). Figure 2B draws its chemical formula showing the hydrophilic modification terminated by a quaternary ammonium cation head group with a cross-sectional area sized ca. 0.5 nm². This is bound to the hydrophobic steroid moiety through a flexible hydrophilic linker. The head-to-tail length of this amphiphile is estimated to be ca. 2 nm, which assigns a cylindrical aspect ratio ( 1 / 2 ≪ α ≈ 0.9 ), thus being susceptible for the lamellar assembly through liquid crystal interactions as planar bilayers (Oswald and Pieranski, 2005; Israelachvili, 2011). The chemical affinity of DChol by the single nucleotides stands up on their high basicity in aqueous solutions ( p K a ≈ 9.3 ), accomplishing the acid–base complexation reaction G R + 2   D C h o l   ⇌   G R: D C h o l 2

at 1:2 lipoplex stoichiometry underlying structural bilayer relationships, i.e., a G R = a D C h o l and n G R = 2 n D C h o l (assuming each single nucleotide reacting as an equivalent GR-unit with the two molecules of DChol that constitute the bilayer). The electrostatic neutrality condition follows as n G R a G R + n D C h o l a D C h o l = 0 ; see caption in Figure 2 for details.

We exploited the Bligh–Dyer (BD) concept for hydrophobic extraction in modulated solvents, the gold standard for natural lipid extraction (Kebbekus et al., 1995). As described in Figure 3, our synthesis approach was implemented into a multistep BD-schema at modulated control of dielectric permittivity ( ϵ ). Figure 3A depicts the adaptation of the BD concept for optimized catanionic complexation, including the reaction, further compaction, and final extraction in homogeneous phases as composed by BD solvents. Briefly, catanionic GR–DChol interactions have become highly condensing under low dielectric permittivity in such a way that mixing double-excess DChol with GR yields catanionic GR:DChol₂ lipoplexes. The resulting hydrophobic homogenate is transferred into a biphasic BD system with a modulated phobic/philic trade-off under permittivity control; however, the lowest permittivity BD phase concentrates the GR:DChol₂ lipoplexes with the highest hydrophobicity, and the non-aggregated GR polyanions can be recovered in the high-permittivity phase. As composed by ternary mixtures of water (H₂O), methanol (MeOH), and chloroform (CHCl₃), the BD solvents were modulated under thermodynamic control; see the phase diagram in Figure 3B. The proposed BD procedures are fast, efficient, reproducible, and easy to implement (Kebbekus et al., 1995). By using greener solvents, the proposed catanionic complexation in BD solvents is potentially scalable up to pharmacologically enhanced standards translatable to human medicine (green organics, fluorinated oils, eutectic solvents, etc. (Breil et al., 2017)). As a featured piece of physicochemical understanding on the proposed synthetic procedures for catanionic lipoplexing, we further discuss the control mechanisms for solvation, reaction, and condensing interactions generically involved in BD solvents as follows.

The spatial strength of the catanionic lipoplex assembly was evaluated with respect to disentangling thermal fluctuations. In particular, the Bjerrum length ( λ B ) determines the distance above which Coulomb’s interactions weaken with respect to thermal energy ( k B T ). In BD media with a tunable dielectric permittivity ( ϵ B D ), the Bjerrum length is given as follows: λ B ≡ e 2 ϵ B D k B T ,

which is defined relative to the vacuum value ( ϵ 0 ; in electrostatic units 4 π ϵ 0 ≡ 1 ), where k B is the Boltzmann constant and T is the absolute temperature.

The Bjerrum length determines the size of a systemic “binding cage” in which catanionic bonds dominate over thermal forces (see Figure 2C; left panel). Coulombic interactions are expected to be short-ranged in hydrophilic BD solvents ( ϵ B D ≈ ϵ w ≈ 80 ϵ 0 ; thus, λ w ≈ 0.01 λ 0 ). However, the Bjerrum cage is much smaller, thus weakly condensing in water ( λ w ≈ 0.7   n m ) than extremely condensing in vacuum ( λ 0 ≈ 60   n m ), or in organic-rich BD solvents ( ϵ o r g ≈ 5 ϵ 0 ≈ 0.05 ϵ w ; thus, λ o r g ≈ 20 λ w ). Hence, the dielectric BD compatibilizer (methanol) appears as the key order modulator regulating the Bjerrum’s cage for strengthening the catanionic assembly in organic-rich phases (because ϵ M e O H ≈ 30 ϵ 0 , intermediate between water ϵ w ≈ 80 ϵ 0 , and chloroform ϵ o r g ≈ 5 ϵ 0 , then λ w ≪ λ M e O H < λ o r g ≈ λ 0 ). Consequently, catanionic condensation becomes optimal in the lowest permittivity BD solvent (maximum chloroform), which imparts the highest electrostatic ordering inside the largest Bjerrum cage ( λ o r g ≈ 15 n m ≫ λ w ). On the contrary, electrostatic condensation weakens in water-rich BD solvents ( λ w ≈ 0.7 n m for ϵ w ≈ 80 ϵ 0 ).

The GR polyanions are initially diluted in water, where the Bjerrum length is significantly smaller than the distance between the neighboring nucleotides ( λ w > l 0 ≈ 0.34 n m ); here, they repel each other, thus making a conformationally persistent strand (the Kuhn length l p ≈ 70 n m ≫ λ w > l 0 ) (Kebbekus et al., 1995). Because Coulomb interactions strengthen the considered GR polyelectrolytes Γ ≡ λ w / l 0 ≈ 2.1 (Manning, 1969), we thus predicted a GR stretching out, followed by favored counterion condensation as dissolved in aqueous media; for a single GR polyanion in the presence of cationic lipid counterparts in water, we expect a condensation fraction θ ≈ 1 − Γ − 1 ≈ 0.5 (Manning, 1969). In the absence of cationic lipid (DChol), however, Manning’s condensation is extremely low ( θ ≈ 0 ), becoming only partial at a low electrolyte concentration below a critical value ( σ < σ c at Γ > 1 ) (Hayes et al., 2015). For the GR:DChol₂ lipoplex catanions considered at high dilution, we expect θ ≥ 0.53 ( Γ ≈ 2.1 ), implying that at least 53% of phosphate groups are condensed within the Bjerrum length. At moderate environmental hydrophobicity (e.g., in the BD monophase ϵ B D ≤ ϵ w ), a single counterion layer of DChol cations is expected to condense inside the Bjerrum cage, with a modulated size compatible with each single GR chain ( l 0 ≪ λ B D ≈ 16 l 0 ≈ 5 n m ), akin an inverse micellar aggregate. In more concentrated GR systems, however, higher electrostatic screening is needed with increasing ionic strength, which requests higher environmental hydrophobicity for Manning condensation to happen (Dobrynin and Rubinstein, 2005). Mean-field theoretical approaches predict higher condensation than estimated beyond the Manning threshold (Schiessel and Pincus, 1998; Schiessel, 1999), provided the freely mobile small counterions are considered within the Debye–Hückel approximation (Na⁺, K⁺, etc.); these theories predict a vanishing density of charge ( σ c → 0 ) (Stigter, 1995), in total condensation conditions ( θ → 1 ) (Hansen et al., 2001). Even at a high polyelectrolyte density but at low enough permittivity media, attractive counterion-mediated interactions are known to induce complete condensation (Ha and Liu, 1998; Podgornik and Parsegian, 1998).

The molecular ordering Manning interactions are expected to drive liquid crystal-like DNA condensation (Koltover et al., 2000; Muñoz-Ubeda et al., 2010). In concentrated GR/DChol phases, catanionic condensation could eventually happen in the “maximum chloroform” BD solvent (at ϵ o r g ≪ ϵ w , thus λ o r g ≈ 15 n m ≫ λ o r g ). The rigid GR polyanions could, thus, condense into catanionic GR:DChol₂ lipoplexes, with a planar rigid conformation and large size in a BD medium of low enough dielectric permittivity (see Figure 3). In addition, DChol is highly mesogenic (Oswald and Pieranski, 2005), with a high propensity for liquid crystalline catanionic condensation within GR into multilamellar arrangements with a hydrophobic steroidal core and exposed cationic surfaces at a high density of ammonium GR-binding groups (see Figure 2C for a molecular depiction; later contrasted with experiments). Because mesogenicity favors charge polarization on the outer DChol-surfaces, straightforward surface-guided Manning condensation should be able to reverse the interactions between polyanion chains from being repulsive to attractive (Ha and Liu, 1998; Podgornik and Parsegian, 1998). Those mesogenic Manning-like condensations induce a transition into a lamellar phase with the neutralizing 1:2 stoichiometry (under the chemical formula GR:DChol₂). These expectations correspond to the regime of high density and high (physiological) ionic strength, thereby holding charge neutralization in sandwiched catanionic layers; here, the resultant neutralized charge density holds σ G R a G R + σ D C h o l a D C h o l = σ a → 0 (having a ≈ 0.5 n m 2 the GR-to-DChol condensation area). In the pursue for optimal catanionic condensation into neutral lipoplexes (GR:DChol₂), which occurred under the lamellar self-assembly into lipid nanoparticles (GR@LNPs), we expect matching electrostatic neutralization leading to a vanishing zeta potential ( ζ → 0 ).

The lipid nanoparticles (@LNPs), as carriers of nucleic acids, particularly ASOs (GRs), are the most advanced delivery systems used so far in genetic medicines (Thi et al., 2015; McKay et al., 2020). In this work, we designed the synthesis procedure of the GR@LNPs to fulfill three principal terms: 1) to protect the GR payload at high compaction; 2) assist with cell uptake through the plasma membrane; and 3) release the GR payload into the cytosol at the highest transfection efficiency and the lowest toxicity as possible. Based on the aforementioned specifications, our formulation should include the following: a) ASO(GR), as an active pharmacological agent; b) cationic DChol lipid for ASO(GR) compaction under Manning condensation at 2:1 stoichiometry; c) unsaturated phospholipid POPC, a bilayer former as a lipid stabilizer; d) lipid helper cholesterol, as a hydrophobic filler and stability enhancer; e) PEGylated phospholipid DSPE-PEG, as a surface cell adhesion additive (Noiri et al., 2019); f) functional phospholipid DSPE-PEG-MAN for specific anchoring to the mannose receptor of BAMs (MANr), selectively from associated microglia that do not express MANr (Galea et al., 2005). The wanted lamellar phase represents the densest state achievable inside the solid-like @LNP-core with a stoichiometry, corresponding to two lipid counterions (DChol) per base pair (i.e., GR:DChol₂). The GR@LNPs were formulated with a functional shell for specific and selective adhesion to BAMs, as depicted in Figure 1B. The chemical formulation that embodies the engineered GR@LNPs is summarized in Table 1.

Bioactive payloads: L-PGDS gene-silencing GRs. The active payload of the GR@LNPs was made up of catanionic GR:DChol₂ lipoplexes; two types of single-stranded antisense oligo-GRs were chosen with a L-PGDS silencing activity: GR1) 14–16pbs (5′-3′-TAC​TCT​TGA​ATG​CAC​T) and GR2) 14–16pbs (5′-3′-AGT​TAC​ATA​ATT​GCC​A). As a negative control (no gene interference), we used a random no-sense sequence: GRc) 14–16pbs (5′-3′-AAC​ACG​TCT​ATA​CGC). These GRs are labeled at the 5′ end with a fluorescein-derived isomer being emitted at 488 nm (6-FAM and 6-carboxyfluorescein).

@LNP formers: structural lipids. Cationic DNA-complexing DChol was chosen as a biocompatible lipid as it can pack the nucleic acid payload into lipoplexes (Muñoz-Ubeda et al., 2010). By mimicking membrane-forming components (Alberts et al., 2015), the rest lipid formula for the vehicle assembly is as follows: a) structural POPC (60%); b) rigidizing cholesterol (30%); c) surface adhesion lipid DSPE-PEG (1%), and d) functional lipid DSPE-PEG-MAN (9%) for specific adhesion to MAN-receptor of BAMs (indicated as molar percentages). Our colloidal synthesis plan takes advantage of this chemical formulation within the nano-assembly concepts discussed earlier; these are as follows: a) the catanionic (acid–base) stoichiometric reaction modulated by dielectric permittivity in BD solvents and b) tunable Manning-driven aggregation for controlled lipid extraction and compatibilization.

For assessing the targeting activity of the GR@LNP nanovectors to deliver GRs in the BAMs adjacent to the intracerebroventricular (ICV) spaces, an ICV injection was chosen as a validated administration setting. The ICV injection of mannosylated liposomes is a well-established procedure to specifically target the mannose receptor (CD206)-expressing BAMs without affecting resident microglial numbers in both rats (Polfliet et al., 2001a; Newman et al., 2005; Sayd et al., 2020) and mice (Galea et al., 2005; Hawkes and McLaurin, 2009). These studies demonstrated that effective targeting of BAMs required the mannosylated liposomes to be administered directly into the cerebrospinal fluid (CSF). Given the unidirectional CSF flow, the maximal exposure of the brain surface to GR@LNPs in the CSF is necessarily accomplished by the ICV injection, as described in Galea et al. (2005). The GR@LNP uptake efficiency was evaluated into BAMs with respect to the reference control constituted by clodronate-containing liposomes (CLO#GR@Ls), as based on the commercial CLO preparation, previously validated to reach BAMs using the ICV injection (Clodrosome^® macrophage depletion kit dissolved in PBS at 5 mg/ml) (Sayd et al., 2020). For this purpose, six rats were anesthetized with an intraperitoneal injection of ketamine and xylazine (2.5:1 mg/kg, SC). They were then mounted on the stereotaxic ICV frame. We used a 26-GA needle guided through the right lateral ventricle (1 mm posterior to Bregma, 0.25 mm lateral to the midline, and 3.5 mm ventral to the surface of the skull), with a motorized stereotaxic ICV injector (Stoelting^® 53,311). The GR-based preparations (either commercial CLO 5 mg/ml or PBS buffer at pH 7.4) were respectively infused to three animals in a volume of 25 μL over 10 min using a Hamilton syringe (Bonaduz, AG Switzerland^®). CLO-induced BAM clearance was evaluated under CD163+ depletion, as detected by immunofluorescence. As expected, there was a lack of immunosignals for the cellular marker CD163⁺ (ED2) in parenchymal vasculature and meninges in CLO animals compared to PBS-treated rats; see previous results in Sayd et al. (2020).

Adult male Wistar Hannover rats were divided into three groups (N = 9; three per group): a) the control group, receiving LNPs without GR (only the vehicle); b) the negative control group, receiving GR@LNPs with no-sense GR3 (non-complementary to the RNA fragment coding for L-PGDS); and c) the positive control group, receiving GR@LNPs with antisense GR1 and GR2 (complementary to RNA fragments coding for L-PGDS). The animals were anesthetized and placed on a stereotaxic frame, and different GR@LNPs were administered by ICV. The GR@LPN preparations were slowly diffused into the right lateral ventricle of the rat brain over 10 min via a motorized stereotaxic injector at a dose of 25 μL GR@LNP preparation (the total GR concentration estimated at 0.15 nmol/μL). One week after injection, the animals were euthanized with an overdose of pentobarbital, and immediately after the brain was removed, it was frozen on dry ice and stored at –80°C for subsequent histological preparation of tissue samples. To identify the BAM route and evaluate the silencing activity of the GR@LNPs, a histological characterization was performed as follows.

The rat brain was cut by using a cryotome into 15-μm thick sections. In order to detect the BAM marker CD163, the sections on coverslips were washed three times for 5 min with 0.02 M KPBS and incubated in a blocking solution (10% bovine serum albumin and 0.1% Triton X-100 in 0.02 M KPBS) for 60 min at room temperature. Once removed from the blocking solution, the sections were incubated with antisera for a mouse monoclonal anti-CD163 (ED-2) antibody (sc58965, 1:200; Santa Cruz Biotechnology^®) during 2 h at room temperature. Subsequently, the sections were washed with KPBS five times during 5 min each and then incubated for 2 h at room temperature with Alexa Fluor^® 555-conjugated donkey antimouse IgG (h + l) highly cross-adsorbed secondary antibody (A-31570, 1:1000; Life Technologies^®). The sections for confocal studies were washed in KPBS five times for 5 min each and then blocked for 30 min with 10% BSA and 0.1% Triton X-100 in KPBS. Then, the sections were incubated 1 h at room temperature with antisera for a mouse antirat RECA-1 antibody (MCA970R, clone HIS52, 1:1000; BioRad^®). Subsequently, the sections were incubated for 0.5 h at room temperature with IRDye^® 680R goat antimouse 926-68070 (D10512-15, 1:1000, Li-Cor). The sections were washed in 0.02 M KPBS five times for 5 min. For microglial immunofluorescence, antigen retrieval was performed over a second set of sections with a solution of sodium citrate at pH 6.0 during 40 min, ranging from 40°C to 65°C. The sections were washed three times for 5 min with 0.02 M KPBS and incubated in the blocking solution (10% bovine serum albumin and 0.1% Triton X-100 in 0.02 M KPBS) for 60 min at room temperature. Then, they were incubated with antisera for a rabbit polyclonal anti-ionized calcium-binding adapter molecule 1 (Iba1) (ab108539, 1:100; Abcam^®) for 48 h at 4°C. Subsequently, the sections were washed with KPBS five times for 5 min, and then incubated for 1.5 h at room temperature with Alexa Fluor^® 555-conjugated donkey antirabbit IgG (H + L) highly cross-adsorbed secondary antibody (A-31572, 1:1000; Life Technologies^®). The sections were washed in 0.02 M KPBS five times for 5 min. Finally, 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) containing Fluoroshield (Sigma-Aldrich) mounting medium was added to the slides. The sections were cover slipped and frozen at −20°C or immediately visualized using a high-performance fluorescence microscope. The confocal images were obtained by using the confocal Olympus microscope FV1200, from CAI-UCM “Centro de Citometría y Microscopía de Fluorescencia.” The images were processed for adjusting brightness, contrast, and merging images in the processing package “Fiji.” We have analyzed the brain prefrontal cortex of the three animals of the group, and two images for three consecutive brain sections of the animals have been used. The images shown are a qualitative representation of the 6-FAM immunosignal in the different groups studied. To verify whether there was non-specific binding of fluorescent secondary antibodies, negative controls were performed for each of the antibodies using the same incubation protocol without the addition of the primary antibody.

Finally, a third set of sections were incubated with a 1:100 solution of the primary monoclonal rabbit antireceptor mannose antibody (CD206) (ab64693, Abcam, United Kingdom), which was maintained at 37 °C overnight. Subsequently, the sections were washed with KPBS five times during 5 min each and then incubated with a 1:300 solution of the secondary antibody goat antirabbit IgG bound to Alexa Fluor^® 555 (A31572, Thermo Fisher Scientific, United States) for 1 h at room temperature. Over these sections, some images were obtained using a TI-FL Epi-fl Illuminator^® Nikon Eclipse Ti Series epifluorescence microscope equipped with ultraviolet, blue, and green exciting filter sets and a Nikon DS-Qi1Mc monochrome camera. The following filters were used to visualize the fluorescence signal of probes: UV for DAPI (exc/em: 358/461 nm), blue (exc/em: 495/517 nm) for 6-FAM, and green (exc/em: 555/580 nm) for Alexa Fluor^® 555. The photographs were processed with ImageJ®x software.

Coronal brain sections containing the prefrontal cortex (15 μm-thick) were obtained and processed, as described in Ferres-Coy et al. (2016). The oligodeoxyribonucleotide probe sequence, complementary to bases rat PGDS mRNA, was CTC ACC TGT GTT TAC TCT TGA ATG CAC TTA TCC GGT TGG GGC AGG (GenBank accession NM 013015.2 obtained from IBIAN Technology, Zaragoza, Spain). The oligonucleotide probe was individually labeled (2 pmol) at the 3′ end with [³³P]-dATP (>2500 Ci/mmol; DuPont-NEN) using terminal deoxynucleotidyl transferase (TdT, Calbiochem). For hybridization, the radioactively labeled probe was diluted in a solution containing 50% formamide, 4x standard saline citrate, 1x Denhardt’s solution, 10% dextran sulfate, 1% sarkosyl, 20 mM phosphate buffer, pH 7.0, 250 μg/ml yeast tRNA, and 500 μg/ml salmon sperm DNA. The final concentration of radioactive probes in the hybridization buffer was in the same range (∼1.5 nM). The tissue sections were covered with the hybridization solution containing the labeled probes, overlaid with Parafilm coverslips, and incubated overnight at 42 °C in humid boxes. The sections were washed four times (45 min each) in a buffer containing 0.6 M NaCl and 10 mM Tris-HCl (pH 7.5) at 60 °C. Hybridized sections were exposed to a Biomax MR film (Kodak, Sigma-Aldrich, Madrid, Spain) for 24–72 h at –80°C with intensifying screens. For specificity control, adjacent sections were incubated with an excess (50x) of unlabeled probes. The films were analyzed, and relative optical densities were evaluated in three adjacent sections of the prefrontal cortex including four measurements in each section (two measurements in medial and lateral edges, respectively) as duplicate for each rat and averaged to obtain individual values using a computer-assisted image analyzer (MCID, Mering). The MCID system was also used to acquire black and white images. Image management was performed using Adobe Photoshop software (Adobe software). Contrast and brightness of images were the only variables that were digitally adjusted under expert guidance.

All values are expressed as the mean ± SEM. Statistical comparisons were performed by GraphPad Prism 9.0 (GraphPad software, Inc., San Diego, CA, United States) using the appropriate statistical tests, as indicated in the figure legend. The outlier values were identified by the Grubbs’ test (i.e., extreme studentized deviate, ESD method) by GraphPad Prism software and excluded from the analysis when applicable. Differences among means were analyzed by two‐way analysis of variance (ANOVA), followed by Tukey’s multiple comparison test. Differences were considered significant at p < 0.05.

Prior in vivo evaluation of the L-PGDS gene-silencing activity, the structure and contents of the outcome @LNP nanovectors were characterized by reference to a CLO-based @liposome formulation. The synthesis procedure was completed with three independent batches. Characterizations were performed at least in triplicate for samples on the obtained batches (experimental errors correspond to standard deviations; t-Student 95% confidence level; p < 0.05 ). Our characterization analytics was performed as follows.

The nanoscopical morphology and the internal structure of the synthetized GR@LNPs were both studied by ultrastructural TEM analysis, inferred from the population catalog supplied as Supplementary Materials (Supplementary Figures S2A, B). We consider two principal types of @LNPs: Type A) the GR-containing nanoparticles (GR@LNPs), as obtained from the pelleted fraction of the reaction outcome, were further identified at different degrees of payload compaction (Type A1: compact and Type A2: core–shell stratified); B) the hollow nanoparticles not containing GR (hollow@LNP), as recuperated from the supernatant fraction from the reaction outcome. Figure 7 shows the main results exemplified with representative specimens for the homotypical populations stratified from the general textural characterization specified earlier (in Figure 5; see Methods). As revealed by uranyl staining (revealing negative charges under electrodensity contrast), the high-resolution TEM images provided the localization of the electron-dense GR polyanions inside the GR@LNPs (Figure 5), which were analyzed in terms of a lamellar structure factor (Supplementary Figure S3). Figure 7 shows quantitative results from the ultrastructural analysis of the lamellar structure factor (performed in terms of radial profiles; right panels). The quantitative results from such TEM characterization were obtained for representative homotypic specimens (as classified in Supplementary Figure S3, with best-fit parameters reported in Supplementary Table T1); by the LNP type, they are classified as follows:

Type A1) Compact GR@LNP archetypes, as found in the pelleted part of the synthesis outcome: The lipid nanoparticle shows an electron-dense core with a visible lamellarity (Figure 4A; left panel). Detailed profiling analysis showed a repetitive multilamellar structure (Figure 4A; right panel). The radial fittings to the core–shell structure factor revealed a uniform lamellarity characterized by a repeating spacing D A = 5.4 ± 0.5   n m (associated with the natural radial decay k A ≪ R 0 − 1 ≈ 0 and almost zero lamellar dilation δ A ≈ 0 ; see Supplementary Figure S2 for details). The repeating lamellar structure is characterized by a high compaction factor ( α ≈ 8 ), which determines the narrow thickness of the electron-dense layer as d A = 0.9 ± 0.4   n m ≪ D A . This ultrastructural lamellarity is compatible with the molecular periodicity of catanionic lipoplexes, ideally designed in Figure 2C,for repeating compact layers of GR polyanions ca. 1nm width ( ≈ d ≈ D G R ), intercalated by cationic DChol bilayers ca. 4 nm thick ( ≈ D b i l ); this results in the nominal lamellar spacing D = D G R + D b i l = 5   n m , in agreement with the experimental value ( D A ≈ 5   n m ), whereas the uranyl-stained GR component of the lamellar structure appeared to be highly electron-dense ( d ≈ D G R ≈ 1 n m ≪ D ), and the thicker interlayers corresponding to lipids (and other electron-poor species) appear much lighter (near zero TEM-contrast). This particularly asymmetric lamellarity suggests the heaviest GR compaction together with metallic electrolytes and minimal interstitial water into the thinnest polyanion layers as possible ( D G R ≈ 1 n m ), which alternate with much thicker lipid bilayers hydrophobically organized at the practical absence of electrolytes and/or interstitial water ( D b i l ≈ 4   n m ).

Type A2) Core–shell stratified GR@LNPs: The considered GR@LNP specimens appeared sectioned in the equatorial plane at which a swollen corona was distinguishable around the more compact core (Figure 7B; left panel). The radial decreasing contrast indicated lesser peripheric GR than found in the previous Type A1. The profiling analysis for Type A2 GR@LNP’s revealed lamellar periodicity again at D B = 7.3 ± 0.6   n m (Figure 7B; right panel). Such a lighter prototype appears thicker than observed for compact specimens (Type A1), being characterized by strong radial decay ( k B = 0.05 ± 0.02   n m − 1 ≫ D B − 1 ) and non-negligible lamellar dilatation along the radial direction ( δ B ≫ 0 ); see Supplementary Figure S3 for details. These “swollen” Type A2 nanovector objects would represent less compact GR@LNP specimens, probably containing more interstitial water and ions, thus allowing the outer layers for expanding as including a lighter corona of PEGylated lipids (both mannosylated and neutral as included to favor, respectively, BAM-specificity and surface adhesion within the GR@LNPs).

These TEM-classified subtypes A1 and A2 correspond to the extremal status of lamellar packing within a same class of GR-containing nanovectors. Nevertheless, both subtypes appeared mixed within the pelleted fraction of each synthesis outcome and indistinguishable in terms of size (as determined by DLS) and surface electrostatic charge (as determined by the zeta potential). Hereinafter, we will generically refer to as GR-containing Type A lipid nanoparticles, or simply GR@LNPs; these are completely different to the hollow nano-objects found in the supernatant (aka hollow @LNPs).

Type B) Hollow lipid nanoparticles devoid of GapmeR: these hollow @LNPs are systematically found in the supernatant fraction of the synthesis outcome. Yet armored with a lamellar ultrastructure, the hollow@LNP specimens of this GR-devoid subtype appeared comparatively less dense than the GR-containing relatives found in the pelleted fraction (see Figure 7C; left panel). They appeared as white objects identifiable with a higher lipid content (see also Supplementary Figure S2). The practical absence of uranyl-stained GapmeR within these lamellar structures induced smooth periodic compacity ( α = 1 ), without any sign of core–shell stratification ( k C ≈ δ C ≈ 0 ). Quantitative profiling revealed lipid periodicity D C = 5.2 ± 0.6   n m (Figure 7C; right panel), which confirms sterol mesogenicity being capable to induce lamellarity even if no GR is retained inside the hollow Type B @LNPs ( D C ≈ D b i l ; see Figure 2C).

Experimental design: synthesis outcomes and nanovectors, as based on the aforementioned protocol for @LNP synthesis (see Figure 4); we performed several reaction rounds using different GR payloads (considered under chemical formulation in Table 1; all of them under lipid mannosylation). Each nanovector construct was synthetized at least in triplicate and subjected to fractioning either as GR@LNPs (Type A collected in the pelleted fraction) or as hollow @LNPs (Type B recuperated in supernatants). The synthesis outcomes previously stratified in physicochemical terms were batched by reference to the different experimental constructs designed for further in vivo evaluation (see Table 4; Reaction outcomes for evaluation of GR nanovectorization in vivo). Additionally, the commercial clodronate prepared was considered for reference, as evaluated in previous studies (Kida et al., 1993; Polfliet et al., 2001b; Williams et al., 2001; Schiltz and Sawchenko, 2002; Hawkes and McLaurin, 2009; Steel et al., 2010; Vasilache et al., 2015; Pedragosa et al., 2018; Sayd et al., 2020); the corresponding CLO-containing liposomes were identified as CLO#GR@L (Type C preparation). We designed a rationale for in vivo evaluation of the nanovectorization activity, as considering three relevant cell types in the neurovascular unit: PVMs, MGMs, and microglia. The experimental design is summarized in Table 4 with reference to the structural nomenclature established earlier (from Figures 5, 7).

Nanoengineered gene delivery is prominent as exploiting lipid nanoparticles (Frazier, 2015; Novak et al., 2018; Hou et al., 2021). Many BAM-vectorized gene drugs based on ASO/GRs are currently under scrutiny for the maximal gene interference efficiency, minimal off-target binding, and non-toxicity (Frazier, 2015; Huggett and Paisner, 2017; Novak et al., 2018). Despite the variety of gene interferential pathways and expression knockouts available for ASO/GR-based therapeutic intervention nanovectorized into BAMs, a current and growing controversy is on the rise because most of the revealed circuitries were explored under harmful clodronate depletion (Polfliet et al., 2001a; Faraco et al., 2016; Pedragosa et al., 2018; Sayd et al., 2020; Zhang et al., 2021). We have proposed an interferential gene therapy platform based on mannosylated GR@LNPs, as loaded into non-suicidal BAMs. For the sake of validation, we used Wistar rats, in which the targeted BAMs were specifically accessed from the ventricular brain spaces by the ICV route of administration. Immunofluorescence experiments revealed the presence of the mannosylated GR@LNPs in a very specific localization of the perivascular and meningeal macrophages. When delivering antisense GRs into BAMs, their interferential ability to inhibit L-PGDS gene expression was verified by in situ hybridization localized in the macrophages of the rat prefrontal cortex. However, no interferential activity was detected with the nonsense GR construction used as a negative control of L-PGDS gene interference. As compared to the conventional clodronate agent formerly used as suicidal GR vehicles toward BAMs (Sayd et al., 2020), an important reduction of L-PGDS mRNA expression was found in the cortical brain areas of rats receiving antisense GR@LNPs.

The mannosylation of the GR@LNPs, as particularly considered in this study with the hydrophobic mannose derivative 4-aminophenyl alpha-D-mannopyranoside, directs the nanovectors to the phagocyte cells of the innate immune system presenting the mannose receptor CD206, including not only the brain BAMs but also other peripheral macrophages (Marichal, 2020; Nielsen et al., 2020). Preclinical checking has evidenced CD206 specificity to improve incorporation of mannosylated liposomes across the BBB (Umezawa and Eto, 1988). Nevertheless, CD206 is also expressed by other non-CNS vascular cells, for instance, liver Kupffer macrophages (Nielsen et al., 2020) or lung interstitial macrophages (Marichal, 2020). In our preclinical setting with model Wistar rats, we opted for testing the mannosylated GR@LNPs, as directed into the rat brain by an ICV injection. Such a rationale assures unambiguous evaluation of interferential gene activity, specifically into BAMs at circumventing off-target binding. Although systemic delivery could have altered L-PGDS expression in multiple rat organs with uncontrolled consequences in the immunity status and indirectly into BAM status, the directed use of GR@LNPs in the brain by the ICV injection follows a more controlled delivery into targeted BAMs. Classical studies have demonstrated indeed that ICV-delivered liposomes containing aminophenyl mannoside were efficiently incorporated into the mouse brain (Umezawa and Eto, 1988). How would mannosylated GR@LNPs be specifically targeted to brain macrophages from systemic delivery is a very challenging question that requests for more extensive and advanced research in the preclinical setting. However, the ICV delivery of the non-suicidal platform in this study developed as GR@LNPs could ultimately offer a chance for better BAM-targeting control than currently achieved with the BAM-depleting clodronate agent. Indeed, the putative specificity of BAM depletion over other cell populations is controversial as currently elicited with mannosylated clodronate-based formulations. In this vein, some authors have stated that resting and inflammatory microglia express CD206 in the rodent brain and are also suitable to be depleted by clodronate-containing liposomes (Marzolo et al., 1999; Zimmer et al., 2003). Other authors conclude that only BAM expresses CD206 in child-like brains and chronic neurodegenerative disease and not microglia (Galea et al., 2005). Our research group and others have found that administration of mannosylated clodronate liposomes did not affect the number of microglial cells in different brain areas of rodents subjected to pathological conditions (Polfliet et al., 2001a; Faraco et al., 2016; Pedragosa et al., 2018; Sayd et al., 2020; Zhang et al., 2021). However, ICV administration of CLO@liposomes with/without mannosylation has been used as an efficient tool to deplete microglia in different experimental conditions, both in vitro and in vivo (Graykowski and Cudaback, 2021; Yao et al., 2021). Using our mannosylated GR@LNPs (devoid of CLO under the current nanotechnological design), no evidence of GR-immunosignal was, however, found in any parenchymal cellular type other than microglia. Since microglial cells do not need to present an activated profile to phagocyte our GR@LNP carriers, they represent a realistic possibility to further explore neuroinflammatory responses. Therefore, the targeting specificity gained with the mannosylation of GR@LNPs could be further exploited in more advanced/enhanced (pre)clinical scenarios in considering ASO-assisted interventions. Although mannosylation is classically considered to be the most efficient ligand cell-targeting strategy (Umezawa and Eto, 1988), however, the functional role of the CD206 mannose receptor presented by different phagocytic cells might be further assessed by functionally comparing the cellular uptake of mannosylated GR@LNP nanovectors differentially from incorporating the naïve (non-mannosylated) nanoparticles. Different ASO/GR backbones and/or other @GR ligand-conjugations (mannosylated, PEGylated, etc.) could also be exploited for BAM-targeted interferential gene delivery.

Current advances in gene delivery systems based on lipid nanoparticles have been developed in a variety of chemical synthesis scenarios, including: 1) variable backbone chemistries based on native and/or artificial nucleic acids (Kheirolomoom et al., 2015; Horejs, 2021; Hou et al., 2021; Zhu et al., 2022); 2) tunable self-assembly electrostatics with ionizable lipid moieties (Kheirolomoom et al., 2015; Han et al., 2021); and 3) practicable synthesis solvents with a modulated dielectric permittivity (Duong et al., 2020), including those based on the “green” Bligh and Dyer concept (Breil et al., 2017). Colloidal engineering proposed in this study has provided a synthesis concept for GR@LNP fabrication under modulated dielectric permittivity in conventional BD solvents. Our @LNPs build upon polyanionic ASO(GR) backbones, which become strongly compacted by cationic cholesterol under mesogenic Manning condensation. The catanionic lipid/GR formulation includes additional mannosylated and PEGylated lipids, resulting in a hybrid multifunctional nanovector with an enhancing influence on the physicochemical and biological properties of the pre-structured GRs. A similar B and D-guided synthesis concept was previously exploited to encapsulate prototypical microRNAs for interferential gene delivery toward peripheric vascular endothelia (Kheirolomoom et al., 2015). However, those colloidal RNA nanovectors resulted with a much lighter liposomal vesicle-like structure as mostly composed by an aqueous lumen enclosed by a lipid bilayer coating. A post-insertion methodology was additionally developed to functionalize the outer vesicle leaflet with a PEGylated lipid functionalized with an endothelial cell-targeting peptide. As revealed by TEM, only a few luminal RNA strands were found complexed with the cationic lipid in the inner bilayer leaflet (Kheirolomoom et al., 2015). Such lipid–RNA surface construction is bulky liquid-like as dominated by a water core mostly empty of nucleic acid. This is radically different to our current GR@LNPs with a solid catanionic core supported by a mesogenic skeleton made of nucleic acids complexed into lipid multilayers. The former RNA@liposomal constructs are rather (Kheirolomoom et al., 2015): 1) swollen and comparatively lighter (because of containing more water and less nucleic acid); 2) electrically charged (as no effective Coulombic condensation happens in the aqueous core); and 3) potentially more unstable (because of containing more water). Our GR@LNPs have, however, resulted to be highly compacting, electrically neutral, and internally dry, better protecting for the nucleic acid payload from water degradation. As compared to the previous synthesis method in Kheirolomoom et al. (2015), our nano-colloidal aggregates have been claimed with a high loading efficiency and high biochemical stability (patented procedure ES2698565B2).

By exploiting ionizable lipids (other than cationic cholesterol) and greener BD solvents, for instance, newer and optimized GR@LNPs could be designed under colloidal Manning-modulated synthesis. Modulating these physicochemical factors could enhance biocompatibility, tissue penetration, intracellular targeting, and dosing, resulting in fewer toxicities. In nanotechnological engineering terms, a modulable GR-backbone compaction adapted to a programmed pharmacokinetics, and a hidden cationic lipid moiety facilitating lysosomal escape would render into a gene delivery improvement toward the targeted cells. The novel nanomedicines will require compact, stable, safe, and effective @LNPs that protect the nucleic acid from degradation, allowing specific and selective cellular uptake and efficient intracellular release for genetic interference. Our gene delivery approach with ASO/GR into compact @LNP nanovectors can easily be applied to other therapeutics such as bare RNA, DNA, plasmids, and drugs.

As a main outcome of our nanotechnological proof-of-concept, a preliminary validation was obtained on the interferential activity of antisense GR payloads into vectorized BAMs as directly accessed from intracerebroventricular spaces. Likewise, other clinically translatable gene therapies were based on interferential ASOs (Novak et al., 2018; Quemener et al., 2020); our GR@LNP nanovectors have been, hence, shown with a pharmacological potential for CNS-targeted ASO delivery. Regarding the route of administration while the ICV injection is occasionally used in the clinics, it is the most often considered suboptimal since it has a high rate of complications and requires extensive medical intervention (Atkinson, 2017; Cohen-Pfeffer et al., 2017). Other more practical routes, such as IV or IP, must be assessed in future translational studies. Our novel nanocolloidal strategy gains biomedical importance, considering that the BAM cellular targets are a bridge between the periphery and brain parenchyma. The proposed nanotechnological approach ought to be chemically bio-orthogonal as it occurs inside the living system without interfering and avoiding chemical injury of brain cells (Silva, 2006; Nguyen et al., 2021). Furthermore, genetic BAM modulation by GR@LNPs can produce changes in other cellular types conforming the neurovascular complexes and probably, in the adjacent parenchymal cells, such as neurons and microglia. Although the present study therapeutically focused BAMs to inhibit the expression of the anti-inflammatory molecule L-PGDS, however, our methodology can be further extended to modulate the expression of any gene that is over activated in other neuroinflammatory/neurodegenerative pathological conditions. At this point, it is worth discussing the unexpected lack of the 6-FAM immunosignal in the positive control group receiving GR@LNPs with nonsense GR3 (non-complementary to the RNA fragment coding for L-PGDS). The reasons for this lack of signal are unknown and could be related to a faster degradation of the nonsense sequence. Indeed, further time-course experiments after ICV administration of our different liposome preparations are needed to shed light into the interpretation of this result. Another important limitation to our immunofluorescence methodology is that it does not allow discriminating BAMs between PVMs and MGMs beyond their different anatomical localization (infiltrating vs. surrounding brain parenchyma, respectively). The discrimination between perivascular and meningeal macrophages is an open and controversial issue in the field.

Recent studies using single-cell sequencing technologies have revealed that MGMs and PVMs are considered as genetically homogenous populations in the homeostatic state, sharing most of their known phenotypic markers and only distinguishable by their specific localization at the CNS interfaces (Kierdorf et al., 2019). Another important limitation to our methodology to be generalized with other ASOs is that we cannot completely rule out the other cellular types in proximity with BAMs in the brain vasculature, such as endothelium, pericytes, or leptomeningeal cells, which are able to incorporate our mannosylated @LNP preparations. This limitation is also extended to our in situ hybridization results and could be solved in the future carrying out dual-labeling experiments combining in the same preparations immunolocalization and in situ hybridization. Further and complementary studies using high-resolution confocal microscopy would be needed to identify the precise intracellular localization and biological activities of the novel ASO@LNP nanovectors not only inside the targeted BAMs but also in other neurocellular components, both in vitro and in vivo.

Although the new nanotechnological developments provide invigorated interest in ASO/GR interventions, important pharmacological challenges and toxicity issues remain to be addressed with the proposed GR@LNPs to optimize target specificity reducing off-target adverse effects, including 1) the finest route of administration providing better cell-targeting specificity; 2) optimal cell uptake and endosomal trafficking; 3) lysosomal release for enhanced gene interference; 4) minimized proinflammatory effects (vasculitis/inflammatory infiltrates); and 5) negligible (or absent) neurotoxicity under minimal systemic toxicities (mainly nephrotoxicity and hepatotoxicity related to lysosomal accumulation). Further preclinical research requests to elucidate mechanisms for these issues, allowing a better understanding of the clinical relevance of the developed GR@LNPs and pharmacological implications of their toxicities. Pharmacologists, toxicologists, pathologists, and regulatory reviewers need to be familiar with the new ASO/GR@LNP development and their implications. To undergo preclinical toxicity testing, a greater number of ASO/GRs and cationic lipid formulations need to be screened with respect to several cell types targeted from several routes of pharmacological administration. In the therapy context of neuro-inflammation, several authors have tried to selectively deplete BAMs with the intravenous (IV) or intraperitoneal (IP) injections of clodronate-loaded liposomes, and the efficiency was found low, probably because in the systemic circulation, the liposomes have multiple opportunities to be either phagocytosed or extravasated through fenestrated vessels, resulting in a short half-life and rapid degradation in circulation before reaching the CNS (Huitinga et al., 1990; Bauer et al., 1995). The highest BAM-targeting efficiency resulting in a complete depletion of BAM was found indeed with a single ICV injection of clodronate liposomes (Polfliet et al., 2001a; Galea et al., 2005). Nevertheless, the CLO-based liposomes that were not phagocytosed through the ventricular and cisternal spaces of the brain could be retrogradely leaked into the venous blood; consequently, their lost cargoes can eventually affect other cell populations including peripheral macrophages. Therefore, a directed access to CNS phagocytes required the CLO liposomes to be administered directly into the cerebrospinal fluid (Galea et al., 2005; Sayd et al., 2020). Here, we chose the ICV administration as a logical first step to test the success of our nanoformulations to reach BAM targets in the brain, but certainly, the current experimental design cannot discern the differential impact of our GR@LNP nanovectors in different classes of systemic macrophages. However, using ICV administration instead of the IV or IP routes, we have mostly directed the impact of our formulations into BAMs, thus reducing retrograde effects in peripheral macrophages. Therefore, not only the ICV route but also other (systemic and intranasal) routes must be further tested in future preclinical work.

The chemical structure of ASO/GRs is designed to increase the resistance to nuclease degradation and enhance in vivo stability. Despite their constitutional stability against endonuclease degradation as extensively determined in vitro (Duschmalé et al., 2020), naked ASO/GRs can be degraded in vivo into multiple and partially uncontrolled ways (Agrawal et al., 1995; Eder et al., 2009). We thus prevented the naked GRs from non-controllable use in the current proof-of-concept context. Comparatively, our compact GR@LNPs have been objectively shown to be more stable than the liposomal clodronate preparation, which is relatively more stable than conventional liposomes. As compared to naked ASO/GRs, the vesicle liposomes protect the ASO/GR cargoes into the luminal space, hence improving stability in the biological milieu, enhancing cellular uptake, easy escape from the endocytic pathway, and promoting drug distribution (Pakunlu et al., 2006; Zylberberg and Matosevic, 2016). Because the GRs considered in our work are likely to be phagocytosed by multiple cells (endothelial cells, neurons, microglia, and astrocytes), we designed a methodology to specifically reach the macrophages in perivascular and meningeal locations. With respect to previous approaches based on diluted cargoes by luminal carrier liposomes non-specifically targeted under surface-adherent PEGylation (Bunker et al., 2016; Cheng et al., 2022), the novel mannosylated GR@LNP nanovectors compactly charged toward phagocytic brain macrophages are strengthening therapeutic promise for the following reasons: 1) they specifically target the mannose receptor of BAMs; 2) their natural fate is phagocytosis; once internalized, they do not easily escape from the cell; 3) the intracellular release occurs along the endocytic pathway; once inserted in the BAM endocytic way, the lipid components are digested by lysosomal lipases; and 4) they accumulate in perinuclear proximity, favoring GR release into the ribosomes. Consequently, the compact GR-payload into cored @LNPs with a mannosylated corona assures optimal gene interference with respect to conventional liposomes under non-specific cell adherence and lighter loading.

As a main strength from the nanotechnological engineering side, our physicochemical characterization has shown possibilities for synthetic regulation of the GR-compaction status through a modulable formulation within the boundaries established under the conditions for the Manning condensation and the colloidal stability of the resulting catanionic aggregates. Specifically for the mannosylated GR@LNPs validated in this work, we have measured a size around 170 nm (as a hydrodynamic diameter) and an effectively zero electrostatic charge (as compatible with a catanionic maximal condensation). They resulted in an electrically neutral multilamellar aggregate self-assembly as a compact GR-containing nanovector. These very particular synthesis characteristics seem to be the most adequate for the artificial GR@LNPs to specifically target BAMs (without needing to cross the BBB). However, one can be wondering whether other characteristics by design are better (or worse) to this purpose. Specifically, nanoparticle sizes ranging 150–200 nm are expected optimal as far as they are easily endocytable by the targeted BAMs (Manzanares and Ceña, 2020) but refractive to the entry by the tight junctions in the BBB (Ceña and Játiva, 2018). A modulable charge could also play a relevant role in regulating the GR@LNP uptake (Asati et al., 2010; Billiet et al., 2012). In general, the cationic nanovectors are better internalized into the cells due to the cell surface negative charges operating over short distances. However, neutral or negatively charged nanovectors are considered less efficiently internalized (Manzanares and Ceña, 2020). We have shown that a nanoparticle electrostatics effectively neutral seems ideal not only as the consequence of an optimal catanionic condensation in the considered GR@NPLs but also in order to prevent undesired nanovector-BAM repulsions. It is worth mentioning that, independently of the net charges involved, electrostatic screening operates by necessarily unbinding non-specific uptakes because of the high ionic strength stressed by the physiological milieu over long distances. The fact that the neutral GR@LNPs appeared well internalized into BAMs but not into microglia seems quite a matter of specific adhesion interactions mediated by mannosylation and not by non-specific Coulombic interactions.

As a potential biological weakness of the proposed ASO@LNP constructs, mannosylation could become not sufficiently selective or not adequately adherent to phagocytic BAMs but efficiently interfering with other phagocytic cells. The ASOs leaked in the venous circulation could thus eventually interact with other components of the neurovascular unit and ultimately affect other cells by retrograde gene silencing in other unspecific cell populations. However, some research groups have proven that the use of mannosylated liposomes is the most effective way to reach phagocytic cells expressing the mannose receptor (Kong et al., 2012; Belogurov et al., 2013). To discern the differential impact of the GR@LNPs among the other classes of peripherical macrophages, our experimental rationale focuses, thus, intentionally and limitedly to BAMs, as accessed from the ICV spaces. Hence, in effect, to certainly circumscribe conclusions on the efficiency of L-PGDS gene interference on the CNS, we used ICV administration (instead of pharmacologically best-suited IV or -IP routes). Although being pharmacologically suboptimal, the ICV administration of GR@LNPs used in our proof-of-concept has the fundamental potential to modulate gene expression by the venous retrograde impact in several types of peripheral macrophages, expressing the mannose receptor (Kupffer macrophages, lung interstitial macrophages, and circulant monocytes). The degree of the effects will depend on the gene manipulated and the nature of the physiological/pathophysiological process studied (e.g., if BAM infiltration is seen or not seen in the brain parenchyma).