The maze was formed by a wooden structure in the shape of a cross with a central platform and four 35 × 6 cm arms raised 100 cm above the ground. The north and south arms were open, the east and west arms were enclosed by walls 20 cm high. During a 5-min trial the mouse was placed in the central platform and allowed to freely explore the maze. Since mice avoid open areas by confining movements to enclosed spaces or to the edges of a bounded space, a typical mouse tends to spend the majority of trial time in the closed arms. The entire apparatus was cleaned after each trial to remove olfactory cues. Trials were recorded by a ceiling-mounted camera and analyzed by a video analyzer (Ethovision XT, Noldus, The Netherlands). To evaluate anxiety level the following EPM parameters were measured: total entries and total time spent in the open and closed arms, defecations (Ruehle et al., 2013).

The Y-maze apparatus made of gray Plexiglas consisted of three identical arms (8 × 30 × 15 cm) with a 120° angle between adjacent arms. The three arms were designated as: start arm (always open), in which the mouse started to explore the maze; familiar arm (always open); novel arm, blocked during the first trial and opened during the second trial (Ma et al., 2007). Several visual cues were placed on the inner walls of the maze. Y-maze test was performed in a dimly lighted room and consisted of two trials separated by a 1 h inter-trial interval (ITI). During the first trial (training phase) lasting 10 min the mouse was allowed to explore two arms (the start arm and familiar arm) with the novel arm being blocked. During the second trial (retention phase) lasting 5 min the three arms were opened and the mouse placed in the start arm was allowed to freely explore all three arms. Maze floor and walls were cleaned after each trial to remove olfactory cues. Trials were recorded by a ceiling-mounted camera and analyzed by a video analyzer (Ethovision XT, Noldus, The Netherlands). To evaluate the preference for the novel arm (novelty) the following parameters of retention phase were analyzed: first entry in the novel arm, total entries, total distances and time spent in the familiar vs. novel arm.

The mouse was placed in a circular white pool (diameter 140 cm) filled with 24°C water made opaque by the addition of atoxic acrylic white color (Giotto, Italy) (Carrié et al., 2002). An escape platform (diameter 5 cm) with a rough surface was placed in the middle of the NW quadrant 20 cm from the side walls. It was submerged 0.5 cm the water level. The pool located in a room uniformly lighted by four lamps (25 W each) was surrounded by several extra-maze cues. The water maze was surmounted by a video camera whose signal was relayed to a monitor and to the image analyzer (Ethovision XT, Noldus, The Netherlands). The protocol consisted of two phases. During the Place phase, mice were submitted to five consecutive sessions of three 60 s-trials, with 30 s-ITI for a total of 15 trials. Inter-session interval was 15 min during which mice were put in their home cages. At the beginning of each trial, mice were gently released into the water from pseudo-randomly varied starting points and were allowed to swim around to find the hidden platform. Mice that did not locate the platform within 60 s, were guided there by the experimenter. After the animals climbed to the platform, they were allowed to remain on it for 30 s. After 24 h, mice were submitted to the Probe phase, in which the platform was removed and the mice were allowed to search for it for 30 s.

To evaluate localizatory memory the following MWM parameters were analyzed: latency to reach the platform, mean swimming velocity, percentage of time spent in each of the four quadrants during Probe phase.

The apparatus consisted in a parallelepiped chamber made of transparent Plexiglas (56 × 42 × 21 cm). The task consisted of three trials, habituation, pre-test and test (De Bruin and Pouzet, 2006; Arsenault et al., 2011). During habituation, mice were introduced in the empty chamber and left to freely explore it for 10 min. Afterwards, they were put in their home cages for 5 min during which two identical objects (A1 and A2: 8 cm-diameter white Plexiglas disk vertically fixed to a white rectangular base) were placed in the chamber. During the pre-test, mice were allowed exploring the objects for 5 min. After a 1 h-ITI, animals were placed again for 5 min in the chamber, where one of the original objects was replaced by a novel one (B: 8 cm-diameter gray plastic sphere mounted on a squared transparent base). A video camera connected to a monitor and to the image analyzer (Ethovision XT, Noldus, The Netherlands) was placed on top of the apparatus. To evaluate the spontaneous tendency to contact the novel object (novelty) the following ORT parameters were analyzed: contact time with objects, distance traveled in the arena, mean velocity, rearings, grooming, defecations. A discrimination index was calculated: contact time with the novel object (T_no) minus contact time with the familiar one (T_fo)/total contact time with objects, Tno−TfoTno+Tfo.

The apparatus consisted of a conditioning chamber (49 × 21 × 21 cm; Mod. 7532, Ugo Basile, Italy) with walls made of gray Plexiglas and ceiling made of transparent Plexiglas to allow video-recording. The grid-floor (steels spaced by 1.5 cm) was connected to a shock generator scrambler (conditioner 7531, Ugo Basile, Italy) (Cutuli et al., 2013).

CFC encompassed three sessions: baseline, training and context test. After a 3-min acclimation to the conditioning apparatus (baseline), the mouse received three foot-shocks (0.5 mA; 2 s) delivered with 60 s-ITI (training). After 24 h, it was placed again in the chamber for 6 min receiving no foot-shock (context test) (Crawley, 1999). Animals' behavior was recorded by a ceiling-mounted video camera and then analyzed through an image analyzer (Ethovision XT, Noldus, The Netherlands). To evaluate aversive learning the following CFC parameters were analyzed: freezing duration (considered as behavioral immobility, except for respiration movements) and defecations.

Four mice per group received daily i.p. injections of BrdU at a dose of 50 mg/kg BrdU dissolved in saline (0.9% NaCl adjusted to pH 7.2 with NaOH) for 6 days. One day after the final injection the animals were sacrificed and transcardially perfused, under deep anesthesia with chloral hydrate with 4% paraformaldehyde in 0.1 M phosphate buffer (PBS). The brains were removed and kept overnight in 4% paraformaldehyde (PFA). Afterwards, brains were equilibrated in 30% sucrose and cryopreserved at −80°C.

The hippocampus from brains embedded in Tissue-Tek OCT (Sakura, USA) was cut by cryostat at −25°C in 40 μm coronal serial free-floating sections. To perform BrdU detection, DNA was denaturated with 2N HCl for 40 min at 37°C to facilitate antibody access, followed by 0.1 M borate buffer pH 8.5 for 20 min. Sections were incubated overnight at 4°C with a primary antibody rat anti-BrdU (AbD Serotec Cat# MCA2060 RRID:AB_323427) diluted 1:300 in TBS containing 0.1% Triton, 0.1% Tween 20 and 3% normal donkey serum (blocking solution). For immunofluorescence analysis, sections were then stained for multiple labeling by using fluorescent methods. After permeabilization with 0.3% Triton X-100 in PBS, the sections were incubated with 3% normal donkey serum in PBS for 16–18 h with the following primary antibodies: 1:200 goat polyclonal antibodies against Doublecortin (DCX) (Santa Cruz Biotechnology, Inc. Cat# sc-8066 RRID:AB_2088494), 1:300 rabbit polyclonal antibodies against Glial Fibrillary Acidic Protein (Promega Cat# G5601 RRID:AB_430855), 1:150 rabbit monoclonal antibody against Ki67 (Lab Vision Cat# RM-9106-S RRID:AB_149707); 1:100 rabbit polyclonal antibody against Iba-1 (Wako Chemicals USA, Inc. Cat# 019-19741 RRID:AB_839504), 1:100 goat polyclonal antibody against Iba-1 (Abcam Cat# ab5076 RRID:AB_2224402); 1:100 rabbit polyclonal antibody against cleaved Caspase-3 (Cell Signaling Technology Cat# 9661S RRID:AB_331440). Secondary antibodies used to visualize the antigen were 1:200 donkey anti-rat Cy3-conjugated (Jackson ImmunoResearch; BrdU) 1:100, donkey anti-rabbit Cy2-conjugated (Jackson ImmunoResearch; Ki67, Iba-1), 1:100 donkey anti-goat Cy2-conjugated (Jackson ImmunoResearch Cat# 705-225-147 RRID:AB_2307341; DCX, Iba-1), 1:200 anti-rabbit Cy3-conjugated (Jackson ImmunoResearch; GFAP, Caspase-3).

Images of the immunostained sections were obtained by laser scanning confocal microscopy by using a TCS SP5 microscope (Leica Microsystem; Germany).

Analyses were performed in sequential scanning mode to rule out cross-bleeding between channels.

Quantitative analysis of hippocampal cell populations was performed by means of design-based (assumption-free, unbiased) stereology. Slices were collected using systematic random sampling. The brain was coronally sliced in rostro-caudal direction, thus including the entire hippocampus. Approximately 40 coronal sections of 40 μm were obtained from each brain; about 1-in-6 series of sections (each slice thus spaced 240 μm apart from the next) were analyzed by confocal microscopy and (by unbiased stereological method) used to count the number of cells expressing the indicated markers throughout the rostro-caudal extent of the whole hippocampus. The total estimated number of cells positive for each of the indicated markers within the dentate gyrus (DG), CA1 and CA3 areas was obtained by multiplying the average number of positive cells per section by the total number of 40 μm sections comprising the entire DG, CA1 and CA3 (spaced 240 μm) (Jessberger et al., 2005; Kee et al., 2007; Farioli-Vecchioli et al., 2008).

Cells with lipofuscin deposits were recognized by their characteristic appearance and fluorescence of similar intensity in all three channels investigated (Kempermann et al., 2002). We counted only heavily loaded cells, in which the deposits obscured more than half of the nucleus.

An investigator blind to the experimental specimen performed cell count and proportional analyses. Cell number quantification was performed in at least 3 animals per group.

Volumes of the DG, Ammon horn CA1 and CA3 and whole hippocampus were estimated by quantitative light microscopy using the Cavalieri's method (Pakkenberg and Gundersen, 1997). In brief, rostro-caudal sections from hippocampus of each animal (taking every sixth serial section) were mounted onto glass slide and stained with 4′,6-diamidino-2-phenylindole (DAPI) for 1 min. Stained sections were viewed at low magnification using Olympus BX53 digital photomicroscope. Digital images were then captured electronically and displayed on a computer screen. For each animal, DG, Ammon horn and whole hippocampus volumes were subsequently derived by multiplying the calculated mean surface area by the section thickness (40 μm) and the total actual number of sections in which the hippocampus was present.

Areal neuronal density values (number of neurons/mm²) were calculated within each of the hippocampal subfields (DG, CA3, and CA1) in the same tissue sections (stained with DAPI) that underwent the quantitative image analysis described elsewhere (Tandrup, 2004). Cells were counted within a rectangular area on a computer monitor, ranging in size from 1989 μm² (60 × 35.15 μm) to 2270 μm² (58.4 × 38.9 μm), and the rectangular area was superimposed onto the pyramidal cells of CA1 and CA3 areas and onto the granule cells of DG. The results were expressed as the cell density per mm².

Dendritic analysis of DCX⁺ neurons was performed by acquiring z-series of 15–25 optical sections at 1–1.5 lm of interval with a 40X oil lens, with the confocal system TCS SP5 (Leica Microsystem). Two-dimensional projections at maximum intensity of each z-series were generated with the LAS AF software platform (Leica Microsystem) in the TIFF format, and files were imported in the I.A.S. software (Delta Systems) to measure dendritic length. The number of branching points was counted manually in the same images. For each data point, 20–30 cells from 3 mice were analyzed (Farioli-Vecchioli et al., 2008).

Whole blood taken from the tail was collected on filter paper as dried blood spot (DBS), which is particularly suitable for small volume samples. The determination of amino acids (proline, valine, leucine/isoleucine/hydroxyproline, ornithine, methionine, phenylalanine, arginine, citrulline, tyrosine, glycine, alanine, serine, threonine, asparagine, aspartate, lysine/glicine, glutamate, histidine) and carnitines (free carnitine, acetylcarnitine, propionylcarnitine, butyrylcarnitine, isovalerylcarnitine, 3-hydroxybutyrylcarnitine/malonylcarnitine, 3-hydroxyisovalerylcarnitine/methylmalonylcarnitine, glutarylcarnitine/3-hydroxyhexanoylcarnitine, adipylcarnitine, tetra-decenoylcarnitine, myristoylcarnitine, hexadecenoylcarnitine, palmitoylcarnitine, linoleylcarnitine, oleylcarnitine, stearoylcarnitine) was performed on filter paper card DBS samples by direct infusion mass spectrometry (DIMS) following a standardized high-throughput screening method as described elsewhere (Chace et al., 2002; Rossi et al., 2011). The analysis was performed in DBS samples by adding isotopically labeled internal standards, according to the principle of isotope dilution internal standardization. Briefly, filter paper disks containing approximately 3–3.2 μL of whole blood, were punched out from DBS samples and from the quality controls (QCs) using an automatic puncher, into a polypropylene microliter plate. 100 μL of the extraction solution containing the internal standards were added to each paper disk, and the plate was shaken in a thermo mixer (700 rpm, 45°C, 50 min). The internal standards as well as the extraction solution and the QCs were from the NeoBase Non-derivatized MSMS Kit (Perkin Elmer Life and Analytical Sciences, Finland) (Ostrup and Wallac, 1994; Food and Drug Administration, 2004). 75 μL from the content of each well were transferred to a new microliter plate. The plate was placed in the autosampler for analysis. Low and high blood spot QCs were run in replicate in each plate, before and after the real samples. The DIMS analysis for the evaluation of metabolite profile in DBS samples was performed using a Liquid Chromatography-tandem Mass Spectrometry (LC/MS/MS) system consisting of an Alliance HT 2795 HPLC Separation Module coupled to a Quattro Ultima Pt ESI tandem quadrupole mass spectrometer (Waters Corporation, USA) (Sirolli et al., 2012; Bonomini et al., 2013; Rizza et al., 2014). The instrument was operated in positive electrospray ionization mode, using MassLynx V4.0 Software (Waters) with auto data processing by NeoLynx (Waters Corporation, USA). The 30 μL injections were made directly into the ion source through a narrow peek tube with a total run time (injection-to-injection) of 1.8 min. The mass spectrometer ionization source settings were optimized for maximum ion yields for every analyte. The capillary voltage was set to 3.25 kV, the source and desolvation temperature were 120°C and 350°C respectively, and the collision cell gas pressure was 3–3.5 e-3 mbar Argon (Sirolli et al., 2012; Bonomini et al., 2013; Rizza et al., 2014).

Fatty acids were extracted using the method reported by Folch (Folch et al., 1957) with slight modifications. Briefly, brains were homogenized in CHCl₃/MeOH (2:1 v/v) to a final dilution of 20-fold of the original sample volume, assuming that the tissue has the same specific gravity of water. Heptadecanoic acid was used as internal standard. The resulting organic phase was evaporated to dryness in a speed-vac at room temperature and then derivatized with BSTFA-TMCS 99:1 v/v (Sigma-Aldrich, Italy) for 1 h at 60°C. Derivatized samples were transferred in the injection vial and added with 20% v/v of Acetone. GC/MS analyses were performed using a Focus GC (Thermo Scientific, USA) equipped with 30 m × 0.25 mm fused silica capillary column SLB™-5MS (Supelco) and connected to a PolarisQ mass spectrometer (Thermo Scientific, USA). 2 μL of samples were injected in split mode (1:10 ratio), the injector temperature was set at 200°C; the carrier gas was Helium and the flow rate was maintained constant at 1 ml/min. The initial oven temperature of 100°C was held for 1 min and then raised to 250° C at 10°C/min and maintained for 6 min. After then the oven temperature was increased up to 310°C at 20°C/min and held for 5 min. Mass transfer line was maintained at 280°C and the ion source at 200°C. Analyses were performed in Selected Ion Monitoring (SIM) mode and fatty acids were identified by comparison with commercial standards.

All data were tested for normality (Shapiro-Wilk's test) and homoscedasticity (Levene's test) and presented as mean ± s.e.m. Behavioral data and metabolic correlates were analyzed by using one- and Two-Way ANOVAs (with group as between-factor and arm/session/quadrant as within-factor) followed by Newman-Keuls's tests. When parametric assumptions were not fully met, non-parametric analyses of variance (χ² and Mann-Whitney U-tests) were used. Morphological data were analyzed by using Student's T-test. Values of p < 0.05 were considered significant (Statistica 7, Statsoft).

As indicated by One-Way ANOVAs on entry frequencies [F_{(2, 34)} = 0.41, p = 0.67], time spent in the close vs. open arms [F_{(2, 34)} = 3.08, p = 0.06] and defecations [F_{(2, 34)} = 1.02, p = 0.37], the anxiety levels resulted similar for the three experimental groups, regardless of the food supplementation type.

When allowed to choose between the novel and familiar arm, n-3 PUFA mice showed better discrimination abilities of spatial novelty as indicated by the higher number of animals first entering the novel arm and by the total entries in the novel arm. No motivational differences were observed among groups, given their similar total entries. Namely, analyses on frequency of first entry in the novel arm revealed that only n-3 PUFA mice showed a significant preference for it (χ²₁ = 8.33, p = 0.004), while OLIVE OIL (χ²₁ = 0.07, p = 0.78) and naïve (χ²₁ = 0.69, p = 0.41) mice chose an arm by chance (Figure 2A). A Two-Way ANOVA (group x arm) on total entries revealed significant arm effect [F_{(1, 35)} = 32.05, p < 0.0001] and interaction [F_{(2, 35)} = 4.82, p = 0.01], while group effect was not significant [F_{(2, 35)} = 0.14, p = 0.87]. Post-hoc comparisons on interaction demonstrated that only n-3 PUFA group entered more the novel arm (p = 0.0001), while OLIVE OIL (p = 0.13) and NAÏVE (p = 0.09) groups equally entered the novel and the familiar arm (Figure 2B). No significant differences among groups were found on time spent in each arm [group: F_{(2, 35)} = 2.79, p = 0.07; arm: F_{(1, 35)} = 0.44, p = 0.51; interaction: F_{(2, 35)} = 2.56, p = 0.09] and total distances [group: F_{(2, 35)} = 1.04, p = 0.36; arm: F_{(1, 35)} = 11.07, p = 0.002; interaction: F_{(2, 35)} = 2.43, p = 0.10].

Regardless of the experimental conditions, all mice learned to localize the hidden platform as the Place sessions went by, as revealed by a Two-Way ANOVA (group × session) on latency values [group: F_{(2, 34)} = 2.68, p = 0.08; session: F_{(4, 136)} = 10.76, p < 0.0001; interaction: F_{(8, 136)} = 0.94, p = 0.49] (Figure 2C). Furthermore, no differences in mean swimming velocity among groups were found [F_{(2, 34)} = 0.28, p = 0.76]. A Two-Way ANOVA (group × quadrant) performed on time spent in the four quadrants during the Probe trial (24 h after Place phase) revealed significant quadrant effect [F_{(3, 102)} = 11.05, p < 0.0001] and interaction [F_{(6, 102)} = 7.07, p < 0.0001], while group effect [F_{(2, 34)} = 0.26, p = 0.77] was not significant. Interestingly, post-hoc comparisons revealed that only n-3 PUFA group swam significantly longer in the quadrant previously rewarded by the presence of the platform (n-3 PUFA vs. OLIVE OIL: p = 0.01; n-3 PUFA vs. NAÏVE: p = 0.001; OLIVE OIL vs. NAÏVE: p = 0.15) (Figure 2D). Time spent in the previously rewarded quadrant was significantly higher than the time spent in the other quadrants only in n-3 PUFA group (always p < 0.001).

As indicated by the discrimination index (Figure 2E), n-3 PUFA group better recognized object novelty in comparison to both NAÏVE (Mann-Whitney U, Z = −2.40, p = 0.01) and OLIVE OIL (Mann-Whitney U, Z = −2.81, p = 0.004) groups. No significant differences among groups were found in emotional (grooming and defecations) as well as in explorative (distance, velocity and rearing) indices.

n-3 PUFA supplementation enhanced associative memory between aversive stimulus and environmental context, as indicated by the higher freezing behavior of n-3 PUFA group, a readout of increased memory. Namely, a Two-Way ANOVA (group × session) on freezing duration revealed significant group [F_{(2, 34)} = 6.93, p = 0.003] and session [F_{(2, 68)} = 191.07, p < 0.0001] effects. Also the interaction [F_{(2, 68)} = 3.39, p = 0.01] was significant. Post-hoc comparisons on interaction revealed that while during baseline and training all animals exhibited similar freezing levels, during context test n-3 PUFA group exhibited significantly higher freezing levels in comparison to OLIVE OIL (p = 0.02) and NAÏVE (p = 0.0001) groups (Figure 2F). Defecations did not differ among groups throughout the test [F_{(2, 34)} = 0.69, p = 0.51].

n-3 PUFA mice showed a total hippocampal volume significantly increased in comparison to OLIVE OIL (p = 0.03) and NAÏVE (p = 0.04) groups (Figure 3). To thoroughly analyze n-3 PUFA effect on hippocampal subfields, the volumes of Ammon Horn (CA = CA1 + CA3) and dentate gyrus (DG) were measured. In n-3 PUFA group CA volume significantly increased in comparison to OLIVE OIL (p = 0.0004) and NAÏVE (p = 0.0004) groups. Even in DG n-3 PUFA group exhibited enhanced volume in comparison to OLIVE OIL (p = 0.04) and NAÏVE (p = 0.04) groups.

To evaluate whether the volumetric differences were related to changes in neuronal density DAPI-stained nuclei were counted in DG, CA3, and CA1 hippocampal subfields (Figure 3). In DG n-3 PUFA supplementation resulted in a significant increase of granule cell density in comparison to OLIVE OIL (p = 0.005) and NAÏVE (p = 0.0001) groups. Even in CA3 pyramidal neuron density significantly increased in n-3 PUFA group in comparison to OLIVE OIL (p = 0.03) and NAÏVE (p = 0.03) groups. Finally, in CA1 n-3 PUFA group showed a significant increment in the pyramidal neuron density in comparison to OLIVE OIL (p = 0.0007) and NAÏVE (p = 0.0008) groups. All together, these data indicate that the n-3 PUFA-induced increase of hippocampal volumes was due to a decreased neuronal loss, as revealed by the increased neuronal density observed in the n-3 PUFA-treated aged mice.

To verify whether the increased cell density observed in n-3 PUFA group was linked to a decrease of apoptotic cell death, the apoptotic marker activated caspase-3 was measured in DG, CA1, and CA3. The number of caspase-3⁺ cells was very low in the three hippocampal subfields, without any difference among groups. By contrast, several caspase-3⁺ cellular debris were scattered throughout DG, CA1 and especially CA3 subfields in all groups, with a marked reduction was observed in n-3 PUFA group (Figure 4A).

Microglial cells are responsible for maintaining the extracellular environment of the brain, by regulating uptake and degradation of non-functional or apoptotic material. Reduced microglia self-renewal capacity and impaired clearance functions are reported in aging (Sierra et al., 2010; Gemma and Bachstetter, 2013). n-3 PUFA group displayed a significant increase in the number of microglial Iba1⁺ cells in comparison to both control groups in DG (n-3 PUFA vs. OLIVE OIL: p = 0.01; n-3 PUFA vs. NAÏVE: p = 0.0007), in CA3 (n-3 PUFA vs. OLIVE OIL: p < 0.0001; n-3 PUFA vs. NAÏVE: p = 0.0002), and in CA1 (n-3 PUFA vs. OLIVE OIL: p = 0.03; n-3 PUFA vs. NAÏVE: p = 0.007) (Figure 4A). More specifically, we observed in n-3 PUFA group an increase of the microglial cells phagocytosing apoptotic caspase-3⁺ neuronal debris in DG (n-3 PUFA vs. OLIVE OIL: p = 0.03; n-3 PUFA vs. NAÏVE: p = 0.04) and CA1 (n-3 PUFA vs. OLIVE OIL: p = 0.03; n-3 PUFA vs. NAÏVE: p = 0.01) when compared with the other experimental group (Supplementary Figure 1 and Figure 4A, square boxes). These results suggest that n-3 PUFA supplementation increases the phagocytosing microglial population, likely providing a protective mechanism during aging.

A typical biomarker of the aging brain is the astrocyte hypertrophy with increased GFAP expression (Salminen et al., 2011). In n-3 PUFA group, GFAP immunoreactivity was significantly reduced in comparison to both control groups in DG (n-3 PUFA vs. OLIVE OIL: p = 0.0002; n-3 PUFA vs. NAÏVE: p = 0.0002) and CA3 (n-3 PUFA vs. OLIVE OIL: p = 0.002; n-3 PUFA vs. NAÏVE: p = 0.001), but not in CA1 (n-3 PUFA vs. OLIVE OIL: p = 0.43; n-3 PUFA vs. NAÏVE: p = 0.70) (Figure 4B). These results indicate that n-3 PUFA supplementation markedly reduced astrocytosis.

The non-specific lipofuscin deposits in neurons are a prominent sign of aging and their low levels are considered a measure of “cellular health” (Kempermann et al., 2002). In n-3 PUFA group the number of lipofuscin autofluorescent cells was significantly reduced in all hippocampal subfields in comparison to both control groups (DG: n-3 PUFA vs. OLIVE OIL: p = 0.04; n-3 PUFA vs. NAÏVE: p = 0.01; CA3: n-3 PUFA vs. OLIVE OIL: p = 0.009; n-3 PUFA vs. NAÏVE: p = 0.0009; CA1: n-3 PUFA vs. OLIVE OIL: p = 0.0006; n-3 PUFA vs. NAÏVE: p = 0.0007) (Figure 4C). These data indicate that n-3 PUFA supplementation resulted in ameliorated neuronal status promoting a decreased cellular degeneration and consequent neuronal loss.

Cell proliferation. The subgranular zone (SGZ) of the hippocampal DG constitutes one of the two neurogenic niches of the adult brain along with the subventricular zone (SVZ) (Eriksson et al., 1998). In SGZ, newborn cells proliferate from neural stem cells and give rise to immature neurons which migrate into the inner part of granule cell layer, where they mature and integrate into the pre-existing circuitry by receiving inputs from the entorhinal cortex, and extending projections into the CA3 (Zainuddin and Thuret, 2012). In order to evaluate the effect of n-PUFA administration on the hippocampal neurogenesis of aged mice we investigated the proliferation and differentiation of newborn neurons in the three experimental groups. In the n-3 PUFA group the number of BrdU⁺ cells was significantly increased in comparison to both control groups (n-3 PUFA vs. OLIVE OIL: p = 0.004; n-3 PUFA vs. NAÏVE: p < 0.0001), while no significant difference was evident between OLIVE OIL and NAÏVE groups (Figure 5A). Similarly, a significantly higher number of Ki67⁺ cells (Scholzen and Gerdes, 2000) was found in n-3 PUFA group compared to both control groups (n-3 PUFA vs. OLIVE OIL: p < 0.0001; n-3 PUFA vs. NAÏVE: p < 0.0001). The number of Ki67⁺ cells found in OLIVE OIL and NAÏVE mice was not significantly different (Figure 5A). These data demonstrate that in the DG the n-3 PUFA supplementation was effective in strongly increasing the cellular proliferation rate of newly-generated neurons that was conversely barely detectable in NAÏVE and OLIVE OIL mice.

Cell differentiation. In n-3 PUFA group an increased number of DCX⁺ cells in the SGZ of DG was found in comparison to OLIVE OIL (p = 0.001) and NAÏVE (p < 0.0001) groups, while no statistical difference was found between OLIVE OIL and NAÏVE groups (p = 0.48) (Figure 5A). To investigate whether n-3 PUFA supplementation resulted in altered morphology of early-differentiated DCX⁺ neuroblasts, in term of dendritic length and branching points, two-dimensional projections of confocal z-series obtained from DCX⁺ cells were analyzed. In n-3 PUFA group a significant increase of dendritic length (n-3 PUFA vs. OLIVE OIL: p = 0.001; n-3 PUFA vs. NAÏVE: p < 0.0001) and branching points (n-3 PUFA vs. OLIVE OIL: p = 0.001; n-3 PUFA vs. NAÏVE: p < 0.0001) was observed (Figure 5B). These data indicate that the n-3 PUFA supplementation significantly stimulated neuroblast differentiation leading to major dendritic complexity.

Blood levels of amino acids and carnitines are reported in Table 2. Supervised and Unsupervised Principal Component Analysis did not reveal any significant difference among the experimental groups. However, One-Way ANOVA on the acetylcarnitine (ALC) levels revealed a significant difference [F_{(2, 12)} = 5.52, p = 0.02] among groups. Post-hoc comparisons indicated that ALC blood levels were significantly higher in n-3 PUFA group in comparison to OLIVE OIL (p = 0.03) and NAÏVE (p = 0.02) (Figure 6A).

No significant differences were found for amino acids blood levels.

EPA, DHA, and DPA were measured as the major n-3 PUFA components of brain cell membranes. Moreover, the EPA+DHA+DPA/Arachidonic Acid (AA) ratio was assessed due to its role in cognitive dysfunction and neuroinflammation (Rao et al., 2007; Labrousse et al., 2012). EPA and DHA, but not DPA, levels increased in n-3 PUFA group in comparison to NAÏVE and OLIVE OIL groups, as revealed by one-way ANOVAs on EPA and DHA levels [EPA: F_{(2, 12)} = 26.77, p < 0.0001; DHA: F_{(2, 12)} = 6.21, p = 0.01; DPA: F_{(2, 12)} = 1.85, p = 0.20]. Post-hoc comparisons indicated that n-3 PUFA mice exhibited higher EPA (n-3 PUFA vs. NAÏVE p = 0.0001; n-3 PUFA vs. OLIVE OIL p < 0.0001) and DHA (n-3 PUFA vs. NAÏVE p = 0.04; n-3 PUFA vs. OLIVE OIL p = 0.02) levels (Figures 6B,C). No significant differences were found by comparing data of NAÏVE and OLIVE OIL mice. Moreover, One-Way ANOVA on the EPA+DHA+DPA/AA ratio revealed significant differences among groups [F_{(2, 12)} = 17.54, p = 0.0003]. Post-hoc comparisons indicated a significant increase of EPA+DHA+DPA/AA ratio in n-3 PUFA group in comparison to NAÏVE (p = 0.0005) and OLIVE OIL (p = 0.001) groups (Figure 6D).