Omega-3 EPA changes gut bacteria and lowers immune markers in aged normal and Alzheimer's mice

Female transgenic APP/PS1 (TG) mice (B6.Cg-Tg(APPswe,PSEN1dE9)85Dbo/Mmjax, Jackson Laboratory, RRID:MMRRC_034832-JAX, http://www.jax.org/strain/005864↗; access date 13 February 2020), 13–14 months old at the start of the experiment, and their female non-transgenic littermates, hereafter referred to as wild-type (WT) mice, were used. Mice were group-housed at the Paracelsus Medical University Salzburg under standard conditions, at a temperature of 22 °C and under a 12 h light/dark cycle, with ad libitum access to food and water. Animal care, breeding, handling, genotyping, and experiments were approved by local ethics committees (BMWFW, license numbers 2023-0.158.423 and 2021-0.009.986). The animals were bred and housed under specific pathogen-free conditions. One week before the start of the experiment, they were transferred to the conventional facility with individually ventilated cages.

EPA fish oil was extracted from commercially available ‘OMEGA 3 EPA’ capsules, kindly provided by the company Nahani (Errenteria, Spain), and mixed into standard chow (#3430, KLIBA NAFAG/Granovit, Kaiseraugst, Switzerland) to achieve a concentration of 0.3% EPA. Assuming a mouse chow consumption of 4 g per day (see Table 1) (resulting in 12 mg EPA per day), the EPA dose (461.5 mg/kg body weight; calculating with mouse body weight of 26 g) is equivalent to a human dose of 37.5 mg/kg, or approximately 2 g per day for a 60 kg person [27]. For the control diet, the same standard diet underwent the process of crushing and re-pelleting without adding any supplementation. TG and WT mice were randomly assigned cage-wise to either the control or EPA-supplemented diet for 3 weeks (n = 4 per group). Table 1 shows the weight of the animals before and after the experiment, the mean food consumption per day, and the increase in weight during the experiment.

At the end of the experiment, mice were anesthetized with a solution of ketamine (20.5 mg/mL, Ketamidor, Richter Pharma, Wels, Austria), xylazine (5.36 mg/mL, Chanazine, Chanelle, Loughrea, Ireland), and acepromazine (0.27 mg/mL, Vanastress, Vana GmbH, Vienna, Austria) in 1× PBS, resulting in 205 mg/kg (body weight) ketamine and 53.6 mg/kg (body weight) xylazine. Blood was collected by cardiac puncture using EDTA (0.1 M)-coated syringes and BD Microlance 23G × 1” needles. To further prevent blood coagulation, EDTA (0.1 M) was added to all samples at a 1:10 ratio. A volume of 50 µL of the blood was used for general blood analysis via an automated hematology analyzer (Sysmex pocH-100iV Diff; Sysmex Europe GmbH, Vienna, Austria), and 300 µL was centrifuged for plasma collection (2000× g, 4 °C, 10 min; followed by an additional 15 min centrifugation after separating the liquid phase). Washed platelets were prepared from the rest of the blood for the platelet activation assay. After blood collection, animals were transcardially perfused with 20 mL of ice-cold PBS (#14287072, Gibco/Thermo Fisher Scientific, Vienna, Austria).

Collection of the eyes was performed as described previously [28]. In short, the eyes were plucked from the eye sockets, and the left eye of each mouse was fixed by immersion in 4% paraformaldehyde (PFA) for 1 h at room temperature (RT), washed in 0.1 M phosphate buffer overnight at RT, and transferred into phosphate buffer containing 15% sucrose for 24 h. The eye was frozen at −80 °C by using 2-methylbutane (GPR RECTAPURE, VWR International, Vienna, Austria), and stored at −20 °C. The retina was dissected from the right eye, snap-frozen in liquid nitrogen, and stored at −80 °C until RNA extraction.

The brain was extracted and split into its hemispheres. The left hemisphere was fixed in 4% PFA for 4 h, washed in PBS, and cryopreserved in a 30% sucrose solution. It was then cut on dry ice with a Leica microtome into 40 μm thick sagittal sections, to divide the hemisphere into 10 representative parts, as described previously [29]. Sections were stored at −20 °C in cryoprotection solution (ethylene glycol, glycerol, 0.1 M sodium phosphate buffer pH 7.4, 1:1:2 by volume) until immunohistochemical staining.

The hippocampus of the right hemisphere was extracted, snap-frozen, and stored at −80 °C until eicosanoid analysis. The rest of the right hemisphere was stored in tissue storage solution (Miltenyi, Bergisch Gladbach, Germany) on ice until microglia isolation (maximum storage time: 2 h). Feces were collected out of the rectum, snap-frozen, and stored at −80 °C until microbiome analysis.

The whole procedure was performed at RT and adapted from a published protocol [30]. Blood (1:10 diluted with 0.1 M EDTA) was diluted by adding 2× volume of Tyrode’s buffer (134 mM NaCl, 12 mM NaHCO₃, 2.9 mM KCl, 1 mM MgCl₂, 0.34 mM Na₂HPO₄, 10 mM Hepes solution) and centrifuged (5 min, 300× g). The whole layer containing platelet rich plasma (PRP) was taken and centrifuged again (12 min, 200× g). After this, only the upper two-thirds were collected. PRP was diluted 1:20 in flow-buffer (phosphate-buffered saline (PBS) + 2% bovine serum albumin (BSA) + 2 mM EDTA), and antibodies were added for flow cytometry analysis. The following antibodies from BD Biosciences/Becton Dickinson Austria GmbH (Vienna, Austria) were used in 1:100 dilution: CD9-PE clone KMC.8, CD62P-FITC clone RB40.34, and isotype control FITC rat IgG1 λ. After 20 min incubation in the dark, reaction was stopped via 1:6 dilution with flow-buffer, and subsequently analyzed with a BD Accuri C6 Plus Flow Cytometer (BD Biosciences/Becton Dickinson Austria GmbH, Vienna, Austria). Platelets were defined based on their forward and sideward scattering patterns on a logarithmic scale and positive staining for CD9, a marker highly expressed on platelets. CD9+ cells were gated for CD62P+ expression, which is only expressed at the platelet surface upon activation [31].

An adult brain dissociation kit (Miltenyi) was used to dissociate the brain tissue, followed by magnetic bead separation (Miltenyi) of CD11b+ microglial cells. The whole procedure was performed following the protocol provided by Miltenyi [32]. In short, brain tissue was enzymatically dissociated at 37 °C on a gentleMACS Dissociator (Miltenyi) for 30 min. All the following steps were performed at 4 °C or on ice. After a debris removal and washing step, red blood cells were lysed. Following another washing, cells were solved in PBS + 0.5% BSA, and incubated 1:10 with CD11b MicroBeads (Miltenyi) for 15 min. Subsequently, cell suspension was loaded on an MS column (Miltenyi) which was placed in the magnetic field of an Octo MACS Separator (Miltentyi). Labeled cells captured in the column were eluted with PBS + 0.5% BSA. Cells were counted, and 2 × 10⁴ cells were seeded per well in 100 µL media (Advanced DMEM/F-12 with 5% fetal bovine serum, 1% pen/strep, 2.5 mM L-glutamine and 2% B27 supplement (Gibco)) in a poly-L-ornithine-coated 96-well plate, and kept at 37 °C and 5% CO₂. Approximately 24 h after seeding, half of the media was exchanged, and on the 3rd day, a phagocytosis assay was performed.

Beta-Amyloid (1-42), HiLyte™ Fluor 488-labeled (0.1 mg, Anaspec, Fremont, CA, USA), henceforth referred to as Aβ-488, was solved in 50 µL NH₄OH, diluted with PBS to a concentration of 100 µM, and stored in aliquots on −20 °C until usage. Adult primary microglia were incubated with 1 µM Aβ-488 for 4 h at 37 °C and 5% CO₂. Subsequently, the medium was removed, and after one washing step with PBS, pre-warmed Trypsin/EDTA was added for cell detachment. After 5 min incubation at 37 °C, 1× volume of medium was added, and cells were measured on a BD Accuri C6 Plus flow cytometer (BD Biosciences). Microglia were gated by forward/sideward scatter, and the threshold for phagocytosing cells was set based on the negative control without added Aβ-488.

Endocannabinoids, including endocannabinoid-like amides, and free oxylipins were quantified in the hippocampal tissue of the right hemisphere at MetaToul-Lipidomique Core Facility [33] (I2MC, Inserm 1297, Toulouse, France), MetaboHUB-ANR-11-INBS-0010; doi.org/10.15454/VRJK-KY76. All tissues were snap-frozen with liquid nitrogen immediately after collection, and stored at −80 °C until extraction. For extraction, each frozen tissue was crushed with a FastPrep^®-24 Instrument (MP Biomedical, Illkirch Cedex, France) in 400 µL of methanol, and 7.5 µL of endocannabinoid internal standard and 40 µL of oxylipin internal standard (Deuterium labeled compounds: LxA4-d5, LTB4-d4, 5-HETE-d8, OEA-d4) were added to homogenates of samples. After 2 crush cycles (6.5 m/s, 30 s), a volume equivalent to 0.5 mg of sample was withdrawn for protein quantification. A 260 µL volume of cold methanol was added to the supernatants. After centrifugation at 5000× g for 15 min at 4 °C, the supernatants were transferred into 2 mL 96-well deep plates, and diluted in H₂O to 2 mL. Samples were then submitted to solid phase extraction (SPE) using an OASIS HLB 96-well plate (30 mg/well, Waters, Saint-Quentin-en-Yveline, France) pretreated with ethyl acetate (0.75 mL), methanol (0.75 mL), and water (0.75 mL). After sample application, extraction plate was washed with 10% MeOH (1 mL). After drying under aspiration, lipid mediators were eluted with 0.3 mL of acetonitrile, 0.4 mL of MeOH, and 0.3 mL of ethyl acetate. Before LC-MS/MS analysis, samples were evaporated under nitrogen gas and reconstituted in 15 µL of MeOH. For LC-MS/MS analyses, lipid mediators were separated on a ZorBAX SB-C18 column (2.1 mm, 50 mm, 1.8 µm) (Agilent Technologies, Santa Clara, CA, USA) using an Agilent 1290 Infinity HPLC system (Agilent Technologies) coupled to an ESI-triple quadruple G6460 mass spectrometer (Agilent Technologies). Data were acquired in Multiple Reaction Monitoring (MRM) mode with optimized conditions (ion optics and collision energy). Peak detection, integration, and quantitative analysis were performed using Mass Hunter Quantitative analysis software (version B.09.00, Agilent Technologies) based on calibration lines built with commercially available eicosanoids standards (Cayman Chemicals, Ann Arbor, MI, USA). From the target molecules (6-keto-prostaglandin (PG) F1a, resolvin (Rv) E1, thromboxane (TX) B2, 11B-PGF2a, PGE3, PGF2a, PGE2, RvD3, lipoxin (LX) B4, PGD2, RvD2, LXA4, RVD1, 8isoPGA2, leukotriene (LT) B5, PGA1, maresin 1 (MaR1), protectin Dx (PDx), RvD5, LTB4, 18-hydroxyeicosapentaenoic acid (18-HEPE), 5,6-dihydroxy-eicosatetraenoic acid (5,6-DiHETE), 15dPGJ2, 13-hydroxyoctadecadienoic acid (13-HODE), 9-HODE, 17-hydroxydocosahexaenoic acid (17-HDoHE), 14-HDoHE, 8-hydroxyeicosatetraenoic acid (HETE), 15-HETE, 12-HETE, 5-HETE, 14,15-epoxyeicosatrienoic acid (14,15-EET), 5-oxoeicosatetraenoic acid (5-oxoETE), 11,12-EET, 8,9-EET, 5,6-EET, eicosapentaenoyl ethanolamide (EPEA), palmitoylethanolamide (PEA), docosahexaenoyl ethanolamide (DHEA), N-arachidonoylethanolamine (AEA), 2-arachidonoylglycerol (2-AG), oleoylethanolamide (OEA), and stearoylethanolamide (SEA)), 26 could be measured in the hippocampus (6kPGF1a, TXB2, PGF2a, PGE2, PGD2, 8isoPGA2, 15dPGJ2, 13-HODE, 9-HODE, 15-HETE, 17-HDoHE, 14-HDoHE, 8-HETE, 12-HETE, 5-HETE, 14,15-EET, 5oxoETE, 11,12-EET, 8,9-EET, 5,6-EET, PEA, DHEA, AEA, 2-AG, OEA, SEA) and 13 in the blood (13-HODE, 9-HODE, 15-HETE, 14-HDoHE, 8-HETE, 12-HETE, 5-HETE, PEA, DHEA, AEA, 2-AG, OEA, and SEA).

Feces samples were sent to GVG Diagnostics GmbH (Leipzig, Germany) for microbiome analysis. DNA was extracted from feces samples using an optimized column DNA purification kit (NEBNext Microbiome DNA Enrichment kit (New England Biolabs, Ipswitch, MA, USA), according to the manufacturer’s protocol. The DNA concentration was determined by fluorometric measurement using a Qubit Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA) during all preparatory steps. In addition, the integrity of the isolated DNA was checked by using a Genomic DNA ScreenTape assay (Agilent Technologies), which provided an objective assessment of the sample integrity during all steps. A microbiome DNA enrichment kit was employed to remove large amounts of host DNA. Surplus mouse DNA was extracted from the samples by selective binding and removing CpG-methylated host DNA. The high-quality DNA obtained was subsequently used for enzymatic fragmentation and library preparation, with the quantities normalized to optimal loading concentrations and sequenced (2× 150 bp) using an Illumina NextSeq (Illumina, San Diego, CA, USA). After the shotgun sequencing process, the sequence data obtained were de-multiplexed according to their nucleotide bar codes and employed for taxonomic profiling. Data analysis and taxonomic profiling was performed with CLC Genomics Workbench 20.0.03, including the Microbial Genomics module (Qiagen, Hilden, Germany) and the GVG Bioinformatic Toolbox 1.5 (GVG Diagnostics GmbH).

Retinal RNA isolation and gene expression analysis were performed as previously described [34]. Snap-frozen retinas were homogenized in 1 mL of TRI reagent (Sigma-Aldrich, Vienna, Austria) and centrifuged (10 min, 12,000× g, 4 °C). The supernatant was collected and mixed with 200 μL of chloroform. After a centrifugation step (15 min, 12,000× g, 4 °C), the aqueous phase containing RNA was collected and mixed with 96% isopropanol (1/2 volume). For the following steps, the High Pure RNA Isolation Kit (Roche, Vienna, Austria) was used, following the manufacturers’ instructions. cDNA was synthesized from 600 ng RNA using the iscript™cDNA synthesis Kit (Bio-Rad, Vienna, Austria). Real-time quantitative PCR (RT-qPCR) was performed in a 15 μL reaction in duplicates, each containing 24 ng of cDNA template. GoTaq Probe Mastermix (Promega, Walldorf, Germany) and the following primers were used from Integrated DNA Technologies (IDT, Leuven, Belgium) (Alox5 Mm.PT.56a.30176779, Alox5ap Mm.PT.58.5140995, Il6 Mm.PT.56a.10005566, Il10 Mm.PT.58.13531087, Tspo Mm.PT.58.43313736, Nlrp3 Mm.PT.58.13974318, Lamp2 Mm.PT.58.13168833, Trem2 Mm.PT.58.7992121) and Thermofisher (H2-Aa Mm00439211_m1, TNF Mm00443258_m1). The RT-qPCR was performed on a thermal cycler (CFX96 Real-Time System, Bio-Rad) in a two-step cycling protocol (95 °C for 15 s, 60 °C for 30 s, 39 cycles). Expression of the target genes was normalized to the mean of three reference genes (Sdha Mm01352366_m1 (Thermo Fisher Scientific, Vienna, Austria), Psmd4 Mm.PT.56.13046188 (IDT), Gusb Mm.PT.39a.22214848 (IDT)).

PFA-fixed retinas were incubated in 10% citrate buffer in a heat steamer (Braun, Neu-Isenburg, Germany) at 65 °C for one hour for antigen retrieval. Following extensive washing in tris-buffered saline (TBS), retinas were incubated overnight in blocking medium (0.2% fish skin gelatine (Sigma-Aldrich), 0.05% Triton X-100 (Merck, Vienna, Austria) and 1% bovine serum albumin (Roth, Karlsruhe, Germany) in TBS). Incubation with primary antibodies (goat collagen IV (ColIV), 1:300, Merck #AB769; rabbit NG2, 1:600, Merck #AB5320) was performed at 4 °C for 4 days. After three washing steps, the retinas were incubated with secondary antibodies from Thermo Fisher Scientific (dilution 1:1000) (donkey anti rabbit Alexa 647, donkey anti goat Alexa 555) and DAPI in blocking medium overnight at RT. After final washing, the retinas were mounted onto glass slides and coverslipped with ProLong Gold Antifade (molecular probes). The procedure was adapted from published protocols [34].

Immunohistochemistry on free-floating slices was performed as described previously [29]. Each staining was performed on a representative tenth of the left hemisphere. After washing off the cryoprotection solution in PBS, antigen retrieval was performed in 10% citrate buffer in a heat steamer at 100 °C for 15–20 min. For staining of MHCII, the antigen retrieval step was exchanged for incubation in 50 mM glycine for 1 h. After blocking unspecific binding sites in blocking buffer (1% BSA, 0.2% fish skin gelatin, 0.1% triton X-100 in 1× PBS), sections were incubated with primary antibodies (rabbit ColIV, 1:500 dilution, Abcam (Amsterdam, Netherlands) #236640; goat PDGFRb, 1:300 dilution, R&D Systems/Biotechne (Dublin, Ireland) #AF1042; rabbit CD68, 1:500 dilution, Abcam # 125212; goat Iba1, 1:500 dilution, Abcam #ab5076; rabbit Lamp1, 1:300 dilution, Abcam #24170; rat MHCII, 1:100 dilution, Thermofisher #14532182; rabbit Iba1, 1:500 dilution, FUJIFILM Wako Chemicals Europe GmbH (Neuss, Germany) #019-19741) overnight in blocking buffer. On the following day, the sections were extensively washed and incubated with secondary antibodies from Thermofisher (1:1000 dilution) (donkey anti rat Alexa 488, donkey anti rabbit Alexa 647, donkey anti goat Alexa 568, donkey anti rabbit Alexa 488) in blocking media for 3–4 h. For Aβ plaque staining, AmyTracker 520 (Ebba Biotech, Solna, Sweden) was added in a dilution of 1:1200 with the secondary antibodies. Lipid droplets were stained with Bodipy (Thermofisher; stock solution 1 mg/mL in DMSO) 1:1000 in PBS for 15 min [35]. For lipid droplet staining, no antigen retrieval was performed, and all solutions were used without any detergent. After final washing steps, brain slices were mounted on glass slides and coverslipped with ProLong Gold Antifade mounting media (molecular probes).

For brain sections, pictures were taken of at least three slices per mouse, containing a dorsal hippocampal region with a triangle-shaped dentate gyrus. From each hippocampus, z-stacks of two different fields of view were captured to image the whole dentate gyrus. From each retina, z-stacks from three different fields of view, each covering superficial, intermediate, and deep plexus, were taken from the medial area. All imaging was performed on a confocal laser scanning microscope (LSM700 and LSM710, Zeiss, Vienna, Austria) using a 20× objective, except for lipid droplet imaging, where a 63x objective was used. For analysis of plaque size distribution, the whole hippocampus was imaged with a VS120 Virtual-Slide-Microscope (Olympus, Hamburg, Germany) with a 20× objective.

ColIV-stained vessels (hippocampus and retina) with a diameter below 15 μm were included in the following analysis. Vessel length was measured via the software Imaris (Bitplane, Zurich, Switzerland, version 10.0.0), using the filament tracer tool to determine the total vessel length per field of view. Pericytes (PDGFRb+ cells in brain tissue and NG2+ cells in retinal tissue) were counted manually with the software ImageJ (version 1.54f) and the plugin ‘cell counter’, based on the presence of PDGFRb+ or NG2+ signal co-localizing with relatively round nuclei (DAPI) located on vessels with a diameter below 15 μm.

The Iba1+ and CD68+ immune reactive area was determined via manual threshold setting in ImageJ and analyzed via the ‘analyze particles’ function. MHCII+ cells were manually counted with ImageJ cell counter. For analysis of microglial lipid droplets, colocalization of Iba1 and Bodipy staining was performed with Imaris, and the surface tool was used for 3D modeling of the colocalization channel to obtain the volume and number of lipid droplets. Aβ-plaques and Lamp1+ dystrophic neurites were masked with the surface tool from Imaris to obtain the count and volume [29]. Plaque size distribution was performed in ImageJ with the segmentation tool ‘Robust Automatic Threshold Selection’ and ‘analyze particles’. All analyses were conducted in a blinded manner.

All statistical analyses were performed using GraphPad Prism software version 9 or 10. Normal distribution of data points was verified with the Shapiro–Wilk test. Because of the small sample size, we decided to use a two-way ANOVA to detect the main genotype- and diet-specific effects. Multiple comparisons were added in the case of a significant ANOVA test, and corrected with Šídák’s multiple comparison test. A 95% confidence interval was used, and p-values less than 0.05 were considered significant. Graphs are presented as the mean ± standard deviation, with all data points displayed.

Transgenic APP/PS1 (TG) mice and their non-transgenic littermates (WT) received an EPA-supplemented or control diet for three weeks to analyze the effects on several parameters of neuroinflammation. The gut microbiota are at the forefront of processing dietary intake, can adapt rapidly, and their influence on inflammatory processes in the brain is increasingly recognized. In light of this, fecal samples were analyzed for microbiome composition. On average, 334 (±46) operational taxonomic units (OTUs) were identified in each sample. In general, all samples had a high similarity, indicated by a low value of beta-diversity (0.0796). The Shannon diversity index, a measure for the diversity of the microbiota, revealed no change in the alpha diversity between treatment groups (Figure 1A). For further comparisons, we focused only on the OTUs identified in all samples and with an average abundance of at least 5%.

The most abundant bacteria phyla were Firmicutes (68.4% ± 8.6) and Bacteroidetes (17.2% ± 5.9), with no difference between WT and TG animals. EPA supplementation significantly increased the abundance of Firmicutes and reduced the abundance of Bacteroidetes, leading to a modest, but significant, increase in the ratio of Firmicutes to Bacteroidetes (F:B ratio) (Figure 1B–D). EPA supplementation effects on the gut microbiota were also observed at hierarchical lower taxonomic levels. Animals supplemented with EPA showed a higher abundance of bacteria of the class Clostridia, and a lower abundance of class Bacteroidia, while the class Bacilli stayed unchanged (Figure 1E–G). Bacteria of the order Clostridiales were elevated, while bacteria of the order Bacteroidales were present at lower levels with EPA supplementation (Figure 1H,J). Lactobacillales remained unaffected by diet (Figure 1I). At the family level, the EPA diet led to a decreased abundance of Bacteroidaceae, while Lachnospiraceae and Lactobacillaceae were not significantly changed (Figure 1K–M). All dietary effects on the microbiota were visible in WT and TG animals. We did not observe genotype-specific differences in the microbiome.

In short, EPA supplementation increased bacteria of the Firmicutes phylum, while Bacteroidetes were decreased, leading to an elevated Firmicutes:Bacteroidetes ratio, which is known to decline with age [36].

After being processed by the gut microbiota, nutrients are absorbed through the walls of the small intestine into the blood, before reaching other organs such as the brain. Accordingly, our next step was to analyze various blood parameters, such as blood cell counts, fatty acid-derived signaling molecules (eicosanoids, endocannabinoids and endocannabinoid-like amides), and platelet activation, to obtain an indication of the systemic inflammatory state. A general hematological analysis using an automated analyzer showed no differences between the groups (Supplemental Table S1). As ω-3 fatty acids and their eicosanoids have been suggested to play a role in platelet function [37], we measured platelet activation via the expression of the activation marker CD62P. TG mice had a higher percentage of CD62P-positive platelets than WT animals, with no significant effect of EPA supplementation (Figure 2A and Supplemental Figure S1A–F). Quantification of fatty acid-derived lipid mediators in the blood (i.e., eicosanoids, endocannabinoids and endocannabinoid-like amides) revealed no genotype-specific differences in the measured eicosanoids (5-HETE, 8-HETE, 12-HETE, 15-HETE, 9-HODE, 13-HODE, 14-HDoHE) (Figure 2B and Supplemental Figure S1G). A significant genotype effect was visible in the endocannabinoid-like amide PEA, with slightly lower levels in TG mice, and in OEA, which tended to be present at lower levels in TG animals than in WT mice (Figure 2C). AEA, 2-AG, DHEA, and SEA did not show genotype-specific effects (Figure 2C). A significant diet effect was present in 5-HETE and AEA. EPA supplementation reduced 5-HETE and AEA levels in WT mice, but not in TG animals (Figure 2B,C). PEA tended to be reduced overall with EPA supplementation, but this did not reach significance.

In summary, TG animals had higher platelet activation and lower plasma endocannabinoid-like amides than WT animals, but did not show a higher inflammatory status in the measured plasma eicosanoids. EPA supplementation decreased the pro-inflammatory eicosanoid 5-HETE and the endocannabinoid AEA in WT animals, but did not significantly affect any blood parameters measured in the TG mice.

The primary objective of this study was to investigate the effect of EPA supplementation on neuroinflammation within the CNS. To this end, two distinct regions were analyzed: the hippocampus, a central structure involved in learning and memory functions, and the retina, which can be considered the window to the brain. The retina has been identified as a site of pathological manifestation in AD, and may serve as an indicator of brain function [26]. We performed retinal gene expression analysis of several inflammatory and AD-associated markers (Alox5, Alox5ap, H2-Aa, Il6, Il10, Tnf, Tspo, Nlrp3, Trem2, Lamp2) to identify potential targets for further investigation in brain tissue. However, the retinal gene expression levels of Alox5, Il6, Il10, Tnf, Tspo, and Nlrp3 were found to be too low to provide reliable results. Alox5ap and Lamp2 expression did not differ between genotypes (Supplemental Figure S2A,B), and Trem2 expression showed a tendency to be downregulated in TG animals, although this did not reach statistical significance (Figure 3A). Neither Alox5ap, Lamp2, nor Trem2 were affected by EPA supplementation. Interestingly, retinal H2-Aa, the gene encoding major histocompatibility complex class II (MHCII), was significantly reduced by the EPA-supplemented diet, although it was not differentially regulated between genotypes (Figure 3B).

Deteriorations of the neurovascular unit are strongly associated with AD and neuroinflammation [38,39]. Since the APP/PS1 mouse model features substantial retinal vascular deficiencies [40], and ω-3 fatty acids are supposed to act beneficially on vascular integrity [41], we stained for blood vessels and pericytes. In contrast to published data, total capillary vessel length and pericyte density were not decreased in the TG animals, and EPA supplementation had no effect on the parameters analyzed (Figure 3C–E).

Similarly, we observed no genotype- or diet-specific effects on capillary length or pericyte density in the hippocampus (). Supplemental Figure S3

Lipid mediators play an essential role in neuroinflammation during AD progression [6]. Therefore, we measured fatty acid-derived eicosanoids, endocannabinoids, and endocannabinoid-like amides in hippocampal tissue to determine the inflammatory state in the brain of the TG AD mouse model used in this study. Among the eicosanoids quantified in the hippocampus, the arachidonic acid-derived mediators PGD2, TXB2, 5,6-EET, 5-HETE, 15-HETE, and 5oxoETE were found to be genotype-specifically increased in TG mice (Figure 4A), whereas others (6kPGF1a, PDF2a, PGE2, 8isoPGA2, 15dPGJ2, 13-HODE, 9-HODE, 17-HDoHE, 14-HDoHE, 8-HETE, 12-HETE, 14,15-EET) did not differ between genotypes (Supplemental Figure S4). EPA supplementation did not affect any of them. Among the endocannabinoids and endocannabinoid-like amides measured, AEA and PEA showed only a tendency to have lower levels in TG mice, whereas DHEA was significantly decreased in TG animals (Figure 4B). We did not observe a genotype-specific effect in the levels of 2-AG, OEA, and SEA, and EPA supplementation did not affect any of the endocannabinoids or endocannabinoid-like amides measured.

In summary, TG animals showed increased levels of inflammatory eicosanoids in the hippocampus, and reduced levels of the anti-inflammatory endocannabinoid DHEA. EPA supplementation did not affect the levels of the lipid mediators analyzed in the hippocampus.

As the quantification of eicosanoid levels revealed a pro-inflammatory state in the hippocampus of TG animals, we analyzed whether dietary EPA influenced microglial function. Since gene expression analysis of the inflammatory marker MHCII in the retina was significantly reduced upon EPA supplementation, this prompted us to investigate whether a similar effect could be observed in brain tissue. MHCII+ cells were detected in the hippocampal tissue of the TG mice, whereas no signal was found in the WT animals. In the TG mice, EPA supplementation significantly reduced the number of hippocampal MHCII+ cells, which mainly co-localized with Iba1 staining (Figure 5A,B). Based on this encouraging downregulatory effect on a marker of pro-inflammatory microglia, we further investigated microglial phenotypic characteristics, including phagocytosis, and amyloid plaque pathology.

Ex vivo microglial phagocytosis of fluorescence-labeled Aβ-488 peptide showed a genotype-specific effect, with a higher amount of phagocytosing cells, as well as a higher uptake of Aβ-488, in the TG animals (Figure 5C,D and Supplemental Figure S5A–F). In addition, lysosomal CD68, a marker often used for phagocytic microglia, was present at much higher levels in TG animals than in WT animals (Supplemental Figure S5G,I,J), which was expected because of the Aβ pathology in TG mice. The hippocampal area occupied by Iba1+ microglia did not differ between genotypes (Supplemental Figure S5G,H). Dietary EPA did not affect ex vivo microglial phagocytosis, CD68+, or Iba1+ area (Figure 5C,D and Supplemental Figure S5). Subsequent analysis of Aβ plaque pathology in the TG animals did not show any differences in total Aβ plaque volume, count, or plaque size distribution with dietary treatment (Supplemental Figure S6A–D). In addition, the amount of dystrophic neurites around the Aβ plaques was not affected by EPA supplementation (Supplemental Figure S6A,E).

The accumulation of lipid droplets, particularly in microglial cells, is considered an age-related process. The role of lipid droplets in neurodegenerative diseases such as AD is still under investigation. To determine whether EPA supplementation affects microglial lipid droplet formation, we stained lipid droplets and Iba1+ microglia, and quantified the co-localization volume. Surprisingly, the number and volume of microglial lipid droplets were lower in TG animals (Figure 5E–G). The mean lipid droplet size did not change between genotypes (Figure 5H). Dietary EPA supplementation had no observable effect on lipid droplet accumulation (Figure 5E–H).

Overall, microglia from TG mice displayed distinct phenotypic characteristics, with increased phagocytic activity and lysosomal CD68, in addition to remarkable MHCII expression compared to WT animals. In contrast, microglial lipid droplet formation was higher in WT mice. EPA supplementation reduced MHCII+ cells, but did not affect microglial phagocytosis, lipid droplet accumulation, or Aβ pathology.

The present study was designed as a pilot experiment to investigate the potential effects of EPA supplementation on neuroinflammation and gut microbiota, and to gain insight into the analysis that will be the focus of a future project. Therefore, the sample size was intentionally limited to minimize cost and effort, but we recognize that the study may be statistically underpowered. Although the positive effects of ω-3 fatty acids were found to be more beneficial with longer supplementation [13], we chose a shorter time period to investigate whether the effects could be observed in the brain before major changes in the microbiome occurred. The results showed that the gut microbiota had a rapid and substantial response, while the effects on the brain were modest. Therefore, we recommend that future studies of ω-3 supplementation take into account communication via the gut–brain axis. To reduce the number of animals used, only female subjects were included in this study, because females (mice and humans) are more affected by AD pathology [76,77], and thus have greater relevance. Whether an EPA diet has the same effects in male animals remains to be investigated.

The original contributions presented in the study are included in the article and, further inquiries can be directed to the corresponding authors. Supplementary Materials