A diet rich in docosahexaenoic acid enhances reactive astrogliosis and ramified microglia morphology in apolipoprotein E epsilon 4-targeted replacement mice

Abstract

Docosahexaenoic acid (DHA) consumption reduces spatial memory impairment in mice carrying the human apolipoprotein E ε4 (APOE4) allele. The current study evaluated whether astrocyte and microglia morphology contribute to the mechanism of this result. APOE3 and APOE4 mice were fed either a DHA-enriched diet or a control diet from 4 to 12 months of age. Coronal brain sections were immunostained for GFAP, Iba1, and NeuN. Astrocytes from APOE4 mice exhibited signs of reactive astrogliosis compared to APOE3 mice. Consumption of DHA exacerbated reactive astrocyte morphology in APOE4 carriers. Microglia from APOE4-control mice exhibited characteristics of amoeboid morphology and other characteristics of ramified morphology (more processes, greater process complexity, and greater distance between neighboring microglia). DHA enhanced ramified microglia morphology in APOE4 mice. In addition, APOE4 mice fed the DHA diet had lower hippocampal concentrations of interleukin-7, lipopolysaccharide-induced CXC chemokine and monocyte chemoattractant protein 1, and higher concentration of interferon-gamma compared to APOE4-control mice. Our results indicate that a diet rich in DHA enhances reactive astrogliosis and ramified microglia morphology in APOE4 mice.

Introduction

Alzheimer’s disease (AD) is a neurodegenerative disease that affects cognitive functions such as thinking, learning, and memory. It is estimated that the number of individuals with dementia will be 74.7 million in 2030 and 131.5 million in 2050 [1]. Carriers of the ε4 allele of the APOE gene, which encodes for apolipoprotein E, have a higher risk of developing late-onset AD (LOAD) and develop LOAD at a younger age compared to carriers of the two other polymorphisms of the APOE gene: ε2 and ε3 [2]. One potential explanation of their higher risk is related to higher neuroinflammation [3], [4], [5], [6], [7], [8], [9]. Two of the key players in neuroinflammation are astrocytes and microglia [10]. In AD, astrocytes become activated and enter a state known as reactive astrogliosis which is characterized by increased expression of glial fibrillary acidic protein (GFAP), increased proliferation, hypertrophy, increased soma size, increased thickness of primary processes, and process interdigitation [10], [11]. In the absence of inflammatory stimuli, microglia exhibit ramified morphology which is characterized by a small soma with many processes that are long in length and have a high degree of branching complexity [9]. In AD, microglia transition to an activated amoeboid morphology which consists of a large soma area and retracted processes with little branching complexity [9]. These morphology changes are critical in maintaining brain health and function, but in chronic conditions such as AD, the constant activation of astrocytes and microglia and the resulting neuroinflammation can ultimately contribute to the disease progression [9].

Despite extensive research, no treatment or cure is currently available for LOAD. However, diet is an environmental factor that can modulate the risk of LOAD. For instance, the consumption of fatty fish, which contains long-chain omega-3 polyunsaturated fatty acids (n-3 PUFA) such as docosahexaenoic acid (DHA), has been associated with a lower risk of age-related cognitive decline [12]. About 50% of the brain is made up of fatty acids and DHA accounts for about 20% of them. DHA is especially concentrated in the gray matter where neuronal cell bodies, glial cells synapses, and capillaries are located. DHA is also a molecule giving rise to some resolvins which are anti-inflammatory and inflammation resolving. Higher blood levels of DHA are associated with a lower risk of developing dementia [13]. However, the link between fatty fish consumption and a lower risk of cognitive decline is not present in carriers of the APOE4 allele [14], [15]. Our group previously demonstrated that human APOE4 carriers metabolize DHA differently and consequently, APOE4 carriers exhibit a lower postprandial incorporation of DHA in plasma total lipids and have an attenuated plasma response to n-3 PUFA supplementation compared to non-carriers [16], [17]. We also showed that brain DHA uptake was lower in 13-month-old APOE4 mice compared to APOE2 mice [18]. However, a diet enriched with DHA prevented spatial reference memory decline in 12-month-old APOE4-targeted replacement mice [19]. Together these results suggest that despite their apparent lower brain uptake of DHA, APOE4 mice could benefit from long-term supplementation with DHA to prevent behavioral deficits. Unfortunately, in this specific study, we were unable to date to capture the mechanism explaining this important prevention result.

For this reason, we examined whether the consumption of a diet rich in DHA, could lower hippocampus cytokine levels together with lower activation of astrocytes and microglia in 12-month-old APOE4-targeted replacement mice, as measured by the morphology and quantity of astrocytes and microglia in the hippocampus. Through examining the effect of DHA on cytokine levels and the activation of astrocytes and microglia, we can infer whether DHA may help to reduce neuroinflammation in APOE4 carriers.

Materials and methods

Animals and sample preparation

Male and female APOE-targeted replacement C57BL6 mice were purchased from Taconic (Hudson, NY, USA) as previously described [19]. A total of 20 mice were used for this study (n = 10 for APOE3 (1 males and 9 female) and n = 10 for APOE4;(n = 4 males and n = 6 females). Breakdown of males and females in each group is provided in supplementary file. The mice were fed a commercial chow diet (Teklad 2018; Harlan Laboratories, Indianapolis, IN, USA) from weaning to 4 months of age and from 4 months to approximately 12 months of age, the mice were fed either a 0.7 g DHA / 100 g diet (DHA diet, n = 6 for APOE3 and n = 5 for APOE4) or a purified control diet (n = 4 for APOE3 and n = 5 for APOE4) (Research diet Inc, New Brundswick, NJ, USA) as described in Table 1 and in [19]. The mice were sacrificed at 12 months of age, as previously described [19]. Briefly, the mice were anesthetized with ketamine/xylazine and subsequently underwent transcardial perfusion with 50 mL 0.1 phosphate buffer saline solution containing phosphatases (1 mM sodium pyrophosphate and 50 mM sodium fluoride) and protease inhibitors (SigmaFast protease inhibitor tablets, Sigma-Aldrich, St-Louis, MO, USA). Following the sacrifice, the brains of the mice were dissected and halved along the midsagittal plane. One half of brain was fixed in 4% paraformaldehyde and embedded in paraffin for immunostaining. Only half of the animals had their brain embedded in paraffin in the paper published by [19] explaining why we had access only to 4–6 animals per diet per genotype for this study. The Comité d’éthique de la recherche du CHUQ-Centre hospitalier de l’Université Laval authorized the animal experimental protocol on August 2011.

Table 1.

Macronutrient, cholesterol, and fatty acid composition of the control and DHA diets (as previously described [19].

Diet composition	Control diet	DHA diet
Carbohydrate (%, w/w)	66.0	66.2
Fat (%, w/w)	5.0	5.0
Proteins (%, w/w)	20.3	20.1
Ingredients (g/kg)
Corn starch	150	150
Sucrose	500	500
Corn oil	30	10
Safflower oil	0	19
Soybean oil	10	0
Canola oil	10	0
DHA powder S170P100a	0	50
Casein	200	198
DL-methionine	3	3
Cholesterol, g/kg	0.6	0.6
Fatty acids, g/kg
16:0	5.13	5.06
18:0	1.32	0.58
18:1n-9	14.07	6.74
18:2n-6	20.64	17.23
18:3n-3	1.41	0.14
20:4n-6	0	0
20:5n-3	0	0.26
22:6n-3	0	7.15

Key: DHA, docosahexaenoic acid; 16:0, palmitic acid; 18:0, stearic acid; 18:1n-9, oleic acid; 18:2n-6, linoleic acid; 18:3n-3, alpha-linolenic acid; 20:4n-6, arachidonic acid; 20:5n-3, eicosapentaenoic acid; 22:6n-3, docosahexaenoic acid.

^aSupplied by DSM.

Immunohistofluorescence

Microglia, astrocytes, and neurons from paraffin-embedded brain sections that were available from a previous study [19] were revealed by immunohistofluorescence. Each slide contained 2 consecutive coronal sections of one brain hemisphere (4 µm thick) from the same individual. A total of 7 brain slices from 4 mice were considered for evaluation for APOE3-control, 8 brain slices from 4 mice were evaluated for APOE3-DHA, 10 brain slices from 5 mice were evaluated for APOE4-control, and 8 brain slices from 5 mice were evaluated for APOE4-DHA. Brain slices were incubated at 60 °C for 25 min and then washed in xylenes to remove the paraffin. The brain slices were rehydrated successively with 100%, 95%, and 70% ethanol solution followed by ultrapure water. Antigen retrieval was performed by boiling the samples in sodium citrate (10 mM, pH 6.0) for 20 min. Subsequently, the brain slices were washed in phosphate buffered saline (PBS) containing 0.1% (v/v) Triton X-100 and non-specific sites were blocked with PBS containing 0.1% (v/v) Triton X-100, 5% (v/v) goat serum and 5% (v/v) fetal bovine serum for 1 h at room temperature. Samples were then incubated overnight at 4 °C with either rabbit anti-Iba1 (#019–19741, Wako Chemicals USA Corporation, Richmond, USA, diluted 1:150) and mouse anti-GFAP (#3670 s, Cell Signaling Technology, Danvers, USA, diluted 1:50), or rabbit anti-NeuN monoclonal antibodies (#24307, Cell Signaling Technology, Danvers, USA, diluted 1:300). The slides were then washed twice with the washing buffer and incubated at room temperature for 1 h with anti-rabbit IgG (H + L) F(ab’)₂ – AlexaFluor® 488 (#4412, Cell Signaling Technology, Danvers, USA, diluted 1:50) and anti-Mouse IgG (H + L)F(ab’)₂ – AlexaFluor® 555 (#4409, Cell Signaling Technology, Danvers, USA, diluted 1:50) for the Iba1 and GFAP slides, or anti-rabbit IgG (H + L) F(ab’)₂ Fragment – AlexaFluor® 555 Conjugate (#4413, Cell Signaling Technology, Danvers, USA, diluted 1:50) for the NeuN slides. Brain slices were mounted with ProLong^TM Glass Antifade Mountant with NucBlue^TM (DAPI) stain (Life Technologies inc., Burlington, ON, CA) and a coverslip following the manufacturer’s instructions.

Image acquisition and pre-processing

Brain tissue images were acquired using the epifluorescence microscope EVOS™ FL Auto Imaging System (Life Technologies inc.) since it is known to provide a good image resolution for morphological analysis of microglia and astrocytes from brain sections, especially in the hippocampal region, as reported by other groups [20], [21], [22], [23], [24]. Briefly, to cover the overall coronal section of each brain slice, 70 to 80 high resolution images (1.2 Mpx/images) per fluorescence channel were acquired with a 10x objective and stitched simultaneously using an automated “Scan-routine” functionality provided by the microscope software, thus resulting in an ultrahigh-definition images.

Iba1 and GFAP imaging

Pictures were simultaneously acquired in the blue channel (cell nuclei, DAPI; excitation, 357 ± 44 nm; emission, 447 ± 60 nm), green channel (Iba1; excitation, 470 ± 22 nm; emission, 525 ± 50 nm) and the red channel (GFAP; excitation, 531 ± 40 nm; emission, 593 ± 40 nm). Individual brain images were then pre-processed using an image analysis routine developed in our lab, based on OpenCV (Open Source Computer Vision Library [25]). Using Allen Brain Atlas, the hippocampal region was identified [26] on the full brain image acquired with the microscope.

After pre-treating the hippocampus images, the images were analyzed by an automated and unbiased images analysis routine. To be identified as microglia or astrocytes, hippocampus cells were required to satisfy the following two criteria: (1) the cell had to have a spreading area between 25 μm² and 500 μm² (microglia) or 150 μm² and 900 μm² (astrocyte), and (2) DAPI (cell nucleus) staining had to overlap with Iba1 (microglia) or GFAP (astrocyte) staining. For each hippocampus, all microglia or astrocytes meeting the criteria were systematically analysed. Depending on the experimental condition (genotype and diet), an average of 179 – 439 individual microglia and 40 – 110 individual astrocytes were analyzed for each half coronal brain slice over the whole hippocampal region.

The cell spreading area criterion was determined based on a thorough analysis of the cell soma spreading area as opposed to the area of representative artefacts. In addition, for every brain slice, a picture highlighting the cells to be considered for analysis was generated. These images were verified to ensure there were no false positives or false negatives and that there was no overlapping between two consecutive slides from the same individual. Cell spreading, number of cell processes, Sholl analysis [27], [28] and nearest neighbors analysis were performed for all astrocytes and microglia in the region of interest (hippocampus) of each brain slice. Cell spreading was determined using the Structural Analysis and Shape Descriptors functions from the OpenCV library and referred to the area covered by cell bodies. Each cell with their processes were thereafter reduced to binary objects to one-pixel-wide representations to represent the skeleton of each cell type. Processes analysis included quantification of their number and length for each soma. The number of cell processes was also calculated using the SKAN Python library [29] (Supplementary Fig. 2) and were classified into 3 classes: junctions between processes, process endpoints, or paths. Sholl analysis was also performed on the microglia and astrocyte skeletons [30], [31], [32], [33]. Finally, to calculate the distance of the nearest neighbors, the minimum Euclidian distance was measured between the centroid of each cell and that of all other cells within the hippocampus.

Fig. 2.

Morphology and quantification of microglia in APOE3 and APOE4 mice fed a control or DHA diet in the hippocampal region. A) Number of microglia processes. B) Microglia soma area. C) Microglia nearest neighbor’s analysis. D) Microglia Sholl analysis. E) Microglia density in the whole hippocampal region of the coronal section. All data is represented as mean ± SEM. Number of mice (N) and number of brain slices (n) for each group are as follows: N = 4, n = 7 for APOE3-control; N = 4, n = 8 for APOE3-DHA; N = 5, n = 10 for APOE4-control; and N = 5, n = 8 for APOE4-DHA.Two-way ANOVA were used to compare data. F-I) Representative images of Iba1 (green) and DAPI (blue) immunostaining in hippocampus of APOE3-control mice (F), APOE3-DHA mice (G), APOE4-control mice (H), and APOE4-DHA mice (I). Abbreviations: CTL, control diet; DHA, docosahexaenoic diet; APOE3, apolipoprotein E ε3; APOE4, apolipoprotein E ε4. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Cell density (microglia or astrocyte, cells/mm³) in the hippocampal region was determined by dividing the number of cells meeting the analysis criteria by the volume of the hippocampal region analysed (area of the hippocampal region multiply by the thickness of the coronal section).

NeuN imaging

Once pre-treated, the hippocampus of the brain slices from NeuN/DAPI staining were further analyzed. The blue channel (DAPI) and red channel (NeuN) were treated separately. Each channel was first binarized (e.g pixel with a value of either “1” or “0”) using Otsu’s binarization algorithm so that only pixels corresponding to a cellular feature were considered regardless of the fluorescence intensity. Since NeuN is a nuclear marker only expressed in mature neurons, the analysis considered only the pixels in the binarized red channel image (NeuN) colocalized with the pixels in the binarized blue channel image (nucleus) using a “logical and” bitwise operation. The resulting pixels with a value of “1” (pixels associated with neurons within the hippocampus) were counted. This value was then normalized with respect to the total number of pixels having a value of “1” in the binarized blue channel image (overall nuclei in the hippocampus). This ratio gives an approximation of the fraction of the cells that are neurons. This analysis was performed for each brain slice.

Cytokine and chemokine quantification

Eighteen cytokines and chemokines were quantified in the hippocampus using a Discovery Assay® called the Mouse High Sensitivity T-Cell Discovery Array 18-Plex (Eve Technologies Corp, Calgary, AB, Canada). The multiplex assay was performed using the Bio-Plex™ 200 system (Bio-Rad Laboratories, Inc., Hercules, CA, USA), and a Milliplex Mouse High Sensitivity T-Cell panel (Millipore, St. Charles, MO, USA) according to the protocol developed by Eve Technologies. The cytokines and chemokines analyzed in this assay consisted of granulocyte macrophage colony-stimulating factor (GM-CSF), interferon-gamma (IFNɣ), interleukin-1⍺ (IL-1⍺), IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-10, IL-12 (p70), IL-13, IL-17A, keratinocytes-derived chemokine/chemokine (C-X-C) motif ligand 1 (KC/CXCL1), lipopolysaccharide-induced CXC chemokine (LIX), monocyte chemoattractant protein-1 (MCP-1), macrophage inflammatory protein-2 (MIP-2), and tumor necrosis factor-⍺ (TNF-⍺).

Statistical analysis

Statistical analyses were performed using Statsmodels Python library [34] and Microsoft Excel - Analysis ToolPak. The statistical analyses were divided into 2 parts: i) individual mouse results (Supplementary equations 1–2) and ii) group results (Supplementary equations 3–6). Each brain slice for a given mouse was analyzed individually (2 brain slices per mouse), whereas the group results (composed of 4 to 6 individual mice) represented the genotype/diet group of the slides. The individual and group results enabled the comparison of APOE3-control diet, APOE3-DHA diet, APOE4-control diet and APOE4-DHA diet together. The average number of microglia or astrocytes per group was calculated as a simple arithmetic mean and a standard deviation over the individual results.

Two-way or three-way ANOVA evaluating the interaction between independent variables were performed in order to compare the data from the 4 experimental conditions (genotype/diet groups). The Statsmodels Python library was used to extract an ANOVA table. For the nearest neighbors analysis, the Sholl analysis and the number of process intersections, three-way ANOVA (genotype; diet; radius or class for Sholl and nearest neighbors; genotype × diet interaction) were performed. In addition, several two-way ANOVA (genotype; diet; genotype × diet interaction) were performed on several specific radii or classes to analyze these more precisely. For cell spreading, number of processes, cell density, and cytokine and chemokine concentration, two-way (genotype; diet; genotype × diet interaction) ANOVA were calculated. The null hypothesis was the studied hypothesis for every ANOVA. Only differences with p < 0.05 were considered statistically significant.

Results

Astrocyte number of cell processes, cell spreading, nearest neighbors analysis, Sholl analysis, and quantification

Astrocyte morphology and quantification are presented in Fig. 1A-I. There was a genotype effect for the number of astrocyte processes and astrocyte cell spreading where APOE4 mice had approximately 14% more processes or larger astrocyte cell bodies than APOE3 mice (Fig. 1A and B). There was also a diet effect (P < 0.01) (Fig. 1A) where mice who consumed the DHA diet had approximately 7% more processes than mice who consumed the control diet.

Fig. 1.

Morphology and quantification of astrocytes in APOE3 and APOE4 mice fed a control or DHA diet in the hippocampal region. A) Number of astrocyte processes. B) Astrocyte soma area. C) Astrocyte nearest neighbor’s analysis. D) Astrocyte Sholl analysis. E) Astrocyte density in the whole hippocampal region of the coronal section. All data is represented as mean ± SEM. Number of mice (N) and number of brain slices (n) for each group are as follows: N = 4, n = 7 for APOE3-control; N = 4, n = 8 for APOE3-DHA; N = 5, n = 10 for APOE4-control; and N = 5, n = 8 for APOE4-DHA. Two-way ANOVA were used to compare data. F-I) Representative images of GFAP (red) and DAPI (blue) immunostaining in hippocampus of APOE3-control mice (F), APOE3-DHA mice (G), APOE4-control mice (H), and APOE4-DHA mice (I). Abbreviations: CTL, control diet; DHA, docosahexaenoic diet; APOE3, apolipoprotein E ε3; APOE4, apolipoprotein E ε4. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

For the nearest neighbors analysis (Fig. 1C), there was a genotype × diet interaction for the relative percentage of astrocytes at a distance of 0 – 100 µm and 100 – 400 µm from each other (P < 0.0001). In APOE4 mice, the relative percentage of nearest astrocytes at a distance of 0 – 100 µm was ∼ 17% higher than APOE3 mice. In addition, APOE4 mice who consumed the DHA diet had a 9% higher proportion of nearest neighbor astrocytes at a distance of 0 – 100 µm compared to APOE4 mice who consumed the control diet (Fig. 1C). Conversely, the relative percentage of nearest astrocytes at a distance of 100 – 400 µm was ∼ 36% lower in APOE4 mice compared to APOE3 mice.

For the Sholl analysis, there was a genotype × diet interaction for the number of intersections at 13, 15, 18, 21, 23, 26, and 29 µm from the astrocyte soma (P < 0.0001 for all) (Fig. 1D). Overall, the number of intersections at 13 and 15 µm was on average 18% higher in APOE4 mice compared to APOE3 mice whereas the number of intersections at 18, 21, 23, and 26 µm from the astrocyte soma was 41–53% higher in APOE4 mice compared to APOE3 mice. In contrast, for APOE4 mice fed the DHA diet, the number of intersections was 7–50% higher at 13–29 µm from the astrocyte soma compared to APOE4 mice fed the control diet.

For the astrocyte density analysis, there was a genotype effect (P < 0.01) (Fig. 1E) for the number of astrocytes where APOE4 mice had twice more astrocytes compared to APOE3 mice.

Microglia number of cell processes, cell spreading, nearest neighbors analysis, Sholl analysis, and quantification

Microglia morphology and quantification are presented at (Fig. 2A-I). In the number of microglia processes analysis, there was a genotype × diet interaction (P = 0.0031) (Fig. 2A). Overall, the number of microglia processes was ∼ 117% higher in APOE4 mice compared to APOE3 mice. In APOE3 mice fed the DHA diet, the number of microglia cell processes was 43% higher than APOE3 mice fed the control diet (0.51 ± 0.03 processes for APOE3-DHA vs. 0.35 ± 0.03 processes for APOE3-control, P = 0.015).

For the area analysis, there was a genotype effect (P < 0.0001) (Fig. 2B) for the microglia soma area where APOE4 mice had ∼ 25% larger microglia soma area compared to APOE3 mice. In addition, there was a significant diet effect (P = 0.004) (Fig. 2B) where mice who consumed the DHA diet had ∼ 5% larger microglia soma area compared to mice who consumed the control diet.

For the nearest neighbors analysis, there was a genotype × diet interaction for the relative percentage of nearest microglia at a distance of 0 – 50 µm, 50 – 100 µm, and 100 – 200 µm from each other (P < 0.0001) (Fig. 2C). For APOE4 mice, the proportion of nearest microglia at a distance of 0 – 50 µm was 28% lower compared to APOE3 mice but they had 88–171% higher proportion of nearest microglia at a distance higher than 50 µm. With regards to diet, the relative percentage of nearest microglia was opposite with respect to the genotype of the mice, with higher percentage in APOE3 mice who consumed the DHA diet compared to APOE3 mice who consumed the control diet (80.7 ± 0.1% of microglia in APOE3-DHA vs. 70.7 ± 0.4% of microglia in APOE3-control, P < 0.0001) in the 0 – 50 µm distance. In contrast, in APOE4 mice who consumed the DHA diet compared to APOE4 mice who consumed the control diet the relative percentage of nearest microglia within 0 – 50 µm was 12% lower. The opposite result was also observed in the 50 – 100 µm range and in the 100 – 200 µm range.

For the Sholl analysis, there was a genotype × diet interaction for the number of intersections at 4–13 µm from the microglia soma (P < 0.0001) (Fig. 2D). Overall, the number of intersections at 4–13 µm from the microglia soma was 61–97% higher in APOE4 mice compared to APOE3 mice. At 4–10 µm from the microglia soma, the number of intersections was higher in both APOE3 and APOE4 mice who consumed the DHA diet compared to their control diet counterparts. At 13 µm from the microglia soma, the number of intersections was 30% higher in APOE3 mice fed the DHA versus the control diet and 4% lower in APOE4 mice fed the DHA versus the control diet.

For the microglia density analysis, there was no statistically significant difference (Fig. 2E).

Neuron quantification

The number of neurons was also analyzed in this study (Fig. 3A-E). There was no statistically significant difference (Fig. 3A). Neurons accounted for 47–53% of all cells in the hippocampus, which is consistent with the literature as neuron density in mouse hippocampus accounts for ∼ 54% of the cell population [35].

Fig. 3.

Quantification of neurons in APOE3 and APOE4 mice fed a control or DHA diet. A) Number of neurons. Quantifications of neurons is presented as a ratio between the number of red pixels (NeuN) and blue pixels (DAPI) over the whole hippocampal region. All data is represented as mean ± SEM. Number of mice (N) and number of brain slices (n) for each group are as follows: N = 4, n = 7 for APOE3-control; N = 5, n = 8 for APOE3-DHA; N = 5, n = 9 for APOE4-control; and N = 4, n = 7 for APOE4-DHA. Two-way ANOVA were used to compare data. B-E) Representative images of Neu-N (red) and DAPI (blue) immunostaining in hippocampus of APOE3-control mice (B), APOE3-DHA mice (C), APOE4-control mice (D), and APOE4-DHA mice (E). Abbreviations: CTL, control diet; DHA, docosahexaenoic diet; APOE3, apolipoprotein E ε3; APOE4, apolipoprotein E ε4. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Cytokine and chemokine quantification

There was a genotype × diet interaction for the concentration of LIX and MCP-1 (P = 0.045 and 0.002, respectively) (Fig. 4A and 4B, respectively). For LIX, the effect of diet was opposite, depending on the genotype. For MCP-1, APOE4 and APOE3 mice fed the control diet had similar MCP-1 concentrations, whereas APOE4 mice fed the DHA diet had 33% lower MCP-1 concentration compared to APOE3 mice. There was also a significant diet effect for IL-7 where APOE3 and APOE4 mice fed the DHA diet had ∼ 34% lower IL-7 concentration compared to APOE3 and APOE4 mice fed the control diet (P = 0.008) (Fig. 4C). For IFNɣ, there was a diet effect where APOE3 and APOE4 mice fed the DHA diet had ∼ 18-fold higher IFNɣ concentration compared to APOE3 and APOE4 mice fed the control diet (P = 0.023) (Fig. 4D). For the other cytokines there were no significant differences (Supplementary Fig. 1).

Fig. 4.

Quantification of cytokines and chemokines in the hippocampus of APOE3 and APOE4 mice. A) LIX concentration. B) MCP-1 concentration. C) IL-7 concentration. D) IFNɣ concentration. E) IL-2 concentration. F) IL-4 concentration. G) IL-6 concentration. H) KC concentration. I) GM-CSF concentration. All data is represented as mean ± SEM. The number of mice (N) per group is as follows: N = 11 for APOE3-control, N = 10 for APOE3-DHA, N = 13 for APOE4-control, and N = 11 for APOE4-DHA. Abbreviations: APOE3, apolipoprotein E ε3; APOE4, apolipoprotein E ε4; CTL, control; DHA, docosahexaenoic acid; LIX, lipopolysaccharide-induced CXC chemokine; MCP-1, monocyte chemoattractant protein-1; IL, interleukin; IFNɣ, interferon-gamma; KC, keratinocytes-derived chemokine; GM-CSF, granulocyte macrophage colony-stimulating factor.

Discussion

In this study, we hypothesized that long-term consumption of a diet rich in DHA would decrease astrocyte and microglial activation in 12-month-old APOE4-targeted replacement mice. Our results indicate that the morphology and quantity of astrocytes in APOE4 mice fed a control diet is more consistent with reactive astrogliosis compared to astrocytes in APOE3 mice. The consumption of a DHA diet seemed to exacerbate some of the reactive characteristics of astrocytes in APOE4 mice. This unexpected result hence supports that long-term supplementation with DHA enhances astrocyte reactivity in APOE4 carriers. Reactive astrogliosis is an inflammatory astrocytic response to brain injury or disease which has been observed in AD patients and in animal models [11]. The increased number of astrocytes in APOE4 mice is however not associated with any difference in the number of neurons, further supporting the conclusion that APOE4 mice in the current studies are exhibiting signs of reactive astrogliosis. Other studies have demonstrated that when inflicted with acute injury to the cortex and hippocampus or to the entorhinal cortex, APOE4 mice had greater GFAP immunoreactivity compared to APOE3 mice [36], [37]. Moreover, it was also reported that APOE4 mice had higher GFAP immunoreactivity in vascular mural cells surrounding arterioles compared to APOE3 mice [38]. In addition, in post-mortem studies, the brains of human APOE4 carriers exhibited increased relative proportions of astrocytes and GFAP expression compared to non-carriers [39], [40]. Therefore, our results on astrogliosis reflect the findings of previous studies in mice and humans. In the current study, we showed that unlike our hypothesis, APOE4 mice who consumed a DHA diet had higher indications of reactive astrogliosis compared to those who consumed control diet. More precisely, astrocytes in APOE4-DHA mice were closer together with a greater number of processes and greater process branching complexity compared to APOE4-control mice. Another group reported in rats that a DHA diet induced proliferation of cultured rat astrocytes [39], [41]. Hopperton and colleagues measured astrocyte activation in response to intracerebroventricular infusion of amyloid-β 1–40 in 12-week-old wild-type mice fed a fish oil diet or safflower oil diet [42]. In line with the findings of the current study, they reported that ten days post-amyloid-β infusion, mice fed the fish oil diet exhibited increased GFAP expression, hypertrophic astrocytes, increased number of astrocyte processes and increased process length compared to wild-type mice who consumed the safflower oil diet [42]. A similar phenomenon was observed in 22-month-old mice, where mice who consumed an omega-3 diet (containing DHA and eicosapentaenoic acid) for two months had increased astrocyte process length in the CA1 and CA3 regions of the hippocampus compared to aged mice fed the control diet and compared to 3-month-old mice fed either a control or omega-3 diet [43]. These previous studies reported that short-term DHA supplementation in rodents increased astrocyte activation, especially when exposed to short-term stress such as amyloid-β 1–40 infusion, or long-term stress such as aging. The current study supports the findings of these previous studies since there was also higher reactive astrocyte morphology. Several studies have found a link between IFNɣ and astrocyte activation and proliferation [44], [45], [46], [47], [48]. In the current study, mice fed the DHA diet had a greater IFNɣ concentration compared to mice who consumed the control diet. This wasn’t because of oxidated DHA because the diet composition has been measured before and after the experiments and the levels of DHA was similar at both times. The increase in reactive astrocyte morphology largely seemed to be driven by the APOE4 genotype rather than the DHA diet. One potential explanation of this effect could be related to a compensatory mechanism in which astrocyte processes proliferate to compensate for the increased BBB permeability that is characteristic in APOE4 carriers [3], [4], [5], [6], [11], [7], [8]. Moreover, in APOE4 mice, we previously reported lower Vglut1 level in the hippocampus [19] which could be involved in lower supportive function of astrocytes. Hence, astrocytes number might be enhanced here to cope for its lower functions [49]. This could also support why the number of neurons remain similar between APOE4 and APOE3 mice. However, whether this potential coping mechanism is sustainable at older ages in APOE4 mice remains to be established in another study.

Another aspect of neuroinflammation investigated in the current study was microglia morphology. In contrast to astrocytes, there seemed to be a greater impact of both APOE genotype and DHA diet on microglia morphology. Indeed, for microglia, results were mixed in terms of showing reactive microglia morphology in APOE4 mice compared to APOE3 mice. For APOE4 mice fed the DHA diet, the decrease in number of processes and slight increase in the soma area is consistent with a reactive morphology, whereas the decrease in proportion of neighbouring microglia within 0 – 50 µm and the increase in branching complexity is more consistent with the ramified morphology. While the larger soma area is consistent with the active amoeboid morphology, the greater number of processes, greater process complexity, and greater distance between neighboring microglia is more consistent with the inactive ramified morphology [50], [51]. Previous studies reported that microglia from APOE4 carriers are larger in size, greater in number, and display an amoeboid morphology consistent with microglia activation compared to non-carriers [36], [52], [53], [54]. In the current study, while APOE4 microglia exhibit a larger soma area compared to APOE3 microglia, they do not demonstrate any of the other characteristics observed in the aforementioned studies. The number of studies which report the impact of DHA supplementation on microglia morphology in APOE4 carriers is limited. Previous studies indicate that DHA consumption attenuates microglial activation in response to a variety of neuroinflammatory stimuli [51], [55], [56], [57], [58]. Moreover, Hopperton and colleagues measured microglia activation in response to intracerebroventricular infusion of amyloid-β 1–40 in 12-week-old wild-type mice fed a fish oil diet or safflower oil diet [51]. The authors found that ten days post-amyloid-β infusion, mice fed the fish oil diet exhibited smaller microglia process length, higher microglia branching complexity, and no significant difference in the process endpoints per cell in the CA1 and CA3 regions of the hippocampus and the dentate gyrus [51]. The results of the current study also report some evidence of a greater degree of inactive microglia morphology in APOE4 mice fed the DHA diet compared to those who consumed the control diet. In particular, the microglia of APOE4 mice fed the DHA diet were further apart from one another and had greater branching complexity compared to microglia from APOE4 mice fed the control diet. The anti-inflammatory effect of DHA on microglia morphology is thought to be neuroprotective since it may prevent microglia from being chronically activated in chronic neuroinflammatory conditions. Chronic activation of microglia in conditions like AD can lead to neurotoxicity and subsequently neuronal death [59]. Therefore, a decrease in microglia activation, especially under exposure to chronic stressors like aging and AD, can be a protective mechanism. One study found that in rat microglial cells co-cultured with organotypic hippocampal slice culture, treated with IFN-ɣ and IL-4 induced a protective phenotype which improved neural survival in the hippocampal slice culture [60]. Interestingly, the current study demonstrated greater IFN-ɣ concentration in mice fed the DHA diet. This may possibly provide one explanation for some of the less activated microglial phenotype characteristics observed in mice fed the DHA diet. It is also possible that the less activated microglia phenotype and the more reactive astrocyte phenotype seen in the current study could be the result of crosstalk between astrocytes and microglia, whereby the more activated astrocytes decrease microglial activation [61]. For instance, in a retinal degeneration mouse model, DHA has been shown to decrease retinal gene expression of CCL2 or MCP-1, a pro-inflammatory chemokine secreted by astrocytes which promotes microglia activation and motility [56], [61], [62]. Our results also suggest that APOE4 fed the DHA had a significant lower production in MCP-1, an interesting result that remained significant after correcting for multiple cytokines statistical testing while other cytokines were no longer significant after this correction. In line with this, DHA also reduced the migration of BV-2 microglial cells in vitro [56]. The current study showed a lower concentration of MCP-1 in APOE4 mice who consumed the DHA diet. In addition, the current study indicates that a greater proportion of microglia nearest neighbors were further away in APOE4 mice fed the DHA diet. It is possible that the decreased levels of MCP-1 in APOE4 mice fed the DHA diet could be a contributing factor to the increased distance between microglia nearest neighbours in this group. However, this would need to be verified in further studies.

This is one of the first studies to examine the effect of long-term DHA supplementation on astrocyte and microglia morphology and their quantification in APOE4 mice. The main strength of the study is the method of analysis. In our approach, instead of analysing only a subset of a ROI from which the data are further extrapolated to the entire ROI, we simultaneously analysed all the cells meeting the criteria in the overall hippocampus. We also used thin coronal slices (4 µm) and cross validate that there was no overlapping within a same individual. Indeed, an automatic computer program was developed and used in this study which is extremely robust and limit bias compared to the eye of a human. In addition, the computer program analyzed all cells in the whole hippocampus region that fit the criteria rather than selecting an arbitrary smaller number of cells in a small ROI and extrapolate for the rest of the hippocampus. One important limitation of the current study is that there were more female than male in our groups and only one male in the APOE3 group hence, there is a potential sex bias in our results since sex differences have previously been reported with regards to number of astrocytes and microglia in mouse hippocampus [63]. Although this is an important limitation, in each group there were more female than male supporting that our results seem to be more influenced by a genotype by diet interaction rather than a sex by diet interaction. Another limitation was the small number of mice in each diet and genotype group and hence, the study could have benefited from more mice in each genotype and dietary groups.

Moreover, only the hippocampus was examined. It may be useful for future studies to examine the impact of DHA on astrocyte and microglia morphology in other parts of the brain. Another limitation is the fact that 2D image analyses were performed on thin slides (4 µm) instead of a stereological approach base on 3D reconstructed z-stack images from thicker sections (40–50 µm) as reported by other groups [42], [64]. In future studies, the systematic morphological approach that we have developed could be further transposed to 3D reconstructed images to increase even more the accuracy of such analysis. Finally, one last limitation of this study is the approach that was used to estimate the number of neurons. We used an indirect strategy based on a nuclear neuronal marker (NeuN) and DAPI co-localization analysis. Despite being in accordance with the literature [35], the precision of those results could be further improved with the used of stereological counting.

In conclusion, we reported here reactive astrogliosis in APOE4 mice compared to APOE3 mice and no diet effect. In contrast, microglia in APOE4 mice seemed to have mixed results regarding whether they were exhibiting ramified or active morphology. The consumption of the DHA diet seemed to ameliorate some characteristics of active morphology so that the APOE4 microglia were more similar to the ramified morphology.

Data statement

Upon reasonable request, data and materials supporting the results or analyses presented in this paper will be available freely.

Disclaimer statements

Contributors: MP and M-AL designed the research; M-AP conducted the immunofluorescence experiments; RCW conducted the dissections and preparation of the brain samples; FC provided the mice and technical expertise; MV was responsible for the mouse colony; M-AP and M-AL analyzed the data and HCM wrote the manuscript. All authors read and approved the final version of the manuscript.

Disclosure statement

MP has received funding from Neptune Wellness Solutions for conducting clinical trials unrelated to the current project. No other potential competing interest was reported by the authors.

Funding

This work was supported by the Canadian Institutes of Health Research between 2012 and 2017 under Grant MOP119454; and the Natural Sciences and Engineering Research Council of Canada under Grant RGPIN-2019-06055. M-A. Poulin was supported by an Undergraduate Student Research Award from the Natural Sciences and Engineering Research Council of Canada.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.