Introduction

Neurodevelopmental disorders (NDD), such as Autism Spectrum Disorders (ASD), intellectual Disability or Attention Deficit Hyperactivity Disorder (ADHD), are a group of disabilities that arise from a disrupted development of the central nervous system (CNS) [1]. Individuals with NDD display cognitive deficits, associated with hippocampal dysfunction, among other symptoms [1]. Despite the identification of several genetic risk factors, they do not entirely explain NDDs. These disorders can also originate from complex interactions between various environmental factors [2]. Epidemiological studies reveal that maternal immune activation (MIA) or poor nutritional habits in the perinatal period are among the strongest of risk factors for NDD [3,4,5,6,7,8].

Long term structural and behavioral deficits relevant to NDD are commonly observed in the offspring of rodent models of MIA, including memory deficits and functional alterations in the hippocampus [6, 9,10,11,12,13,14,15]. Furthermore, some nutrients, including the polyunsaturated fatty acids (PUFAs), have a profound long-term influence on brain function [5, 16, 17]. PUFAs are essential fatty acids required for proper brain development and maturation [18,19,20,21,22,23,24,25]. Because they must be provided by alimentation, low dietary intake of these nutrients strongly affects neurodevelopment [8]. The principal forms of PUFAs in the CNS are the long chain (LC) arachidonic acid (AA, 20:4n-6) and the docosahexaenoic acid (DHA, 22:6n-3) [16]. Both clinical and preclinical studies show that low levels of LC n-3 PUFAs increase the risk of NDD and/or aggravate symptoms [8, 16, 26]. Recent clinical evidence suggests that n-3 PUFA homeostasis may be altered in ASD, either as a result of nutritional imbalance or genetic defect [27]. Finally, we and others already showed that decreasing n-3 PUFA dietary intake can affect the offspring’s susceptibility to MIA [8, 28]. However, the mechanisms linking the early-life nutritional status and inflammation to later life behavioral alterations are still unclear.

Microglia are essential for brain maturation, hence, interfering in their developmental activity gives rise to NDD-like symptoms in mice [9, 29,30,31]. In normal conditions, microglia guide axons, phagocyte apoptotic neurons, refine spines and synapses, in an activity-dependent manner [29, 30, 32,33,34,35,36,37,38,39,40,41,42,43]. Both MIA and nutritional imbalance have been shown to disrupt these processes, which might explain the long-term structural and behavioral defects [8, 9, 16, 44]. Myelination is also sensitive to early-life insults and a decrease in white matter integrity is a marker of most, if not all, NDDs [45, 46]. Both MIA and n-3 PUFAs can modulate myelination according to preclinical and clinical studies [47,48,49,50,51,52,53,54,55,56,57].

More recent literature highlights the role of the gut-brain axis in the occurrence of NDD [58,59,60,61]. Clinical studies have observed alterations in the composition of microbiota and associated metabolites in subjects with autism and schizophrenia [62,63,64,65,66,67,68]. The gut microbiome plays an important role in the modulation of microglial function and myelination and both PUFAs and MIA can alter its composition [69,70,71,72,73,74].

In the present study, we show that (1) n-3 PUFA deficiency reveals MIA-induced behavioral alterations in adult offspring but not in pups, (2) both n-3 PUFA deficiency and MIA affect brain lipid composition in adults, (3) this correlates to profound transcriptomic alterations in adulthood, especially for genes involved in CNS development, (4) n-3 PUFA deficiency worsens the impact of MIA on microglia-neuron crosstalk and alters the expression of oligodendrocyte and myelin markers in the developing brain, (5) n-3 PUFA deficiency worsens MIA-induced microbiota and gut alterations in pups, (6) microbial composition and gut inflammatory molecules correlate with neurobiological outcome during the post-natal period.

Material and methods

Animals

Animal husbandry and experimental procedures were in accordance with the EU Directive 2010/63/EU for animal experiments and approved by the Bioethical committee of our University (no. 50120186-A) and Région Aquitaine Veterinary Services (Direction Départementale de la Protection des Animaux, approval ID: A33-063-920). Every effort was made to minimize suffering and the number of animals used. All experiments were conducted in CD1 mice (Charles River, Arbresle, France). Mice were maintained under standard housing conditions in a temperature (23 ± 1 °C) and humidity (40–50%) controlled animal room with a 12h/12h light/dark cycle (07 h–19 h) and ad libitum access to food and water.

Diet and treatment

α-linolenic acid (ALA, 18:3 n-3) and linoleic acid (LA, 18:2 n-6) are the dietary precursors of omega-3 and omega-6 respectively. As soon as the male was introduced in the mating cage, female mice were fed isocaloric diets containing 5% fat with a high or low LA(n-6)/ALA(n-3) ratio (n-3 deficient group or “DEF” and n-3 sufficient group or “SUFF,” respectively) across gestation and lactation (i.e., from the first day of gestation until the post-natal day –PND-21) [28, 75,76,77,78,79]. These diets were custom-made and the pellets were prepared by the SAAJ-INRAE laboratory (Jouy-en-Josas, France).

Females were exposed to males for 48 h. Mating was further confirmed by the observation of a vaginal plug and by measuring weight gain across gestation. At embryonic day 17 (E17), pregnant females were given an intraperitoneal (i.p.) injection of lipopolysaccharide (LPS, E. Coli LPS0127:B8, Sigma Inc, St. Louis, MO, USA; 0.12 µg/g mouse/100 µl). We assessed the development of a sickness behavior in dams by comparing their body weight at E17 and E18 (24 h after the administration of LPS). As expected, we found a significant decrease in body weight in both n-3 sufficient and n-3 deficient mice (Supplementary Fig. 1). We previously showed that this dose of LPS was sufficient to induce an inflammatory response in fetuses’ brain, which was exacerbated by n-3 PUFA deficiency [28]. The administration of the corresponding volume of saline solution (NaCl 0.9%, “Saline”) was used as a control [28, 80, 81]. Sickness behavior and weight were monitored at E17 and E18 to confirm that dams treated with LPS displayed a significantly lower weight gain than saline-treated dams (see MIA checklist, Table S1; [82]).

At birth, litter size was limited to 10 pups. At PND21, the weaned males were housed in groups of 3–6 without mixing the litters. We generated between 2 and 5 litters per condition. Only 1–2 pups from each litter were tested, to avoid any litter effect.

Cohorts were organised as follows (Fig. 1a): (1) One cohort for neonates’ behavior (PND4-8) (Fig. 1); (2) One cohort to assess microglial density/phenotype, microglia-neuron crosstalk and neuronal morphology between PND14 and PND28 (Fig. 3); (3) One cohort dedicated to gut assessment (intestinal permeability, microbiota composition). The brains of these mice were used to conduct immunohistochemistry studies (Iba-1, myelin markers) and perform correlation studies (Figs. 4, 5); (4) 2 cohorts were used for the behavioral assessment of the adult offspring (one for locomotion and anxiety and another for memory testing by the Morris Water Maze) (Fig. 1); (5) one cohort for lipid and transcriptomic profiling in the hippocampus of adults (Fig. 2).

Fig. 1: Effect of n-3 PUFA deficiency on MIA-induced behavioral deficits in neonates and in adult offspring.
figure 1

All graphs show Means ± SEM. a Experimental setup. b Average time spent by pups to achieve the Fox battery tests (negative geotaxis and righting reflex; 3 trials per day from PND4 to PND6). N = 14–19. Two-way ANOVA: MIA effect, F(1,62) = 11.67, p = 0.0011. c Average vocalization time (15-min sessions at PND7-8). N = 14–19. Kruskal–Wallis test followed by Mann–Whitney comparison; n-3 sufficient-Saline vs n-3 sufficient-LPS, **p < 0.01. d Neonate average locomotion measured as the distance traveled (in cm/min) during 1 min-session from PND5 to PND8. N = 14–19. Kruskal–Wallis test followed by Mann–Whitney comparison; n-3 sufficient-Saline vs n-3 deficient-saline, **p = 0.009, n-3 deficient-Saline vs n-3 deficient-LPS, ***p < 0.001. e Time course of locomotor activity of newborns from PND5 to PND8. N = 14–19. Two-way ANOVA on repeated measures followed by Bonferroni’s multiple comparisons test: n-3 deficient-Saline vs n-3 deficient-LPS, *p = 0.02. f Time course of the distance traveled in the Morris Water Maze during the learning phase (in cm). N = 10. Two-way ANOVA on repeated measures: diet effect, F(1,36) = 9.22, p = 0.004. g Percentage of time spent in the target quadrant. N = 10. One-sample t test; n-3 sufficient-Saline, ***p < 0.001; n-3 sufficient-LPS, **p = 0.004; n-3 deficient-saline, ***p < 0.001; n-3 deficient-LPS, p = 0.11. h. Time course of the distance traveled in the Morris Water Maze during the reversal learning phase (in cm). N = 10. Two-way ANOVA on repeated measures: MIA effect, F(1,36) = 3.27, p = 0.008; time effect, F(1,36) = 19, p < 0.001. i Basal locomotor activity (in cm). N = 12. Kruskal–Wallis test followed by Mann–Whitney comparison; n-3 deficient-Saline vs n-3 deficient-LPS, **p = 0.008. j Time spent in the light box (anxiogenic area) of the light-dark test. N = 11. Kruskal–Wallis test followed by Mann–Whitney comparison. k Percentage of time spent in the center of the open-field arena (anxiogenic area). N = 8–11. Kruskal–Wallis test followed by Mann–Whitney comparison.

Fig. 2: Dietary n-3 PUFA deficiency exacerbates MIA-induced alterations of the hippocampal lipid and transcriptional profiles in adulthood.
figure 2

Quantification of the levels of total PUFAs (a), n-6 PUFAs (b), DTA n-6 (c) or DPAn-6 (d) in the hippocampus of adult mice, expressed as the percentage of total fatty acids. All graphs show Means ± SEM. N = 6. Two-way ANOVA followed by Bonferroni’s multiple comparisons test: Total PUFAs: n-3 sufficient-Saline vs n-3 deficient-Saline, ***p = 0.0001; n-3 deficient-Saline vs n-3 deficient-LPS, *p = 0.0156; n-3 deficient-Saline vs n-3 sufficient-LPS, ***p = 0.0003. Total n-6 PUFAs: n-3 sufficient-Saline vs n-3 deficient-Saline, ***p < 0.0001; n-3 deficient-Saline vs n-3 deficient-LPS, **p = 0.0055; n-3 deficient-Saline vs n-3 sufficient-LPS, ***p < 0.0001; n-3 sufficient-Saline vs n-3 deficient LPS, ***p < 0.0001; n-3 deficient-LPS vs n-3 sufficient-LPS, ***p < 0.0001. DTA n-6: n-3 sufficient-Saline vs n-3 deficient-Saline, ***p < 0.0001; n-3 deficient-Saline vs n-3 deficient-LPS, **p = 0.0013; n-3 deficient-Saline vs n-3 sufficient-LPS, ***p < 0.0001; n-3 sufficient-Saline vs n-3 deficient LPS, ***p = 0.0004; n-3 deficient-LPS vs n-3 sufficient-LPS, *p = 0.0158. DPA n-6: n-3 sufficient-Saline vs n-3 deficient-Saline, ***p < 0.0001; n-3 deficient-Saline vs n-3 deficient-LPS, **p = 0.0045; n-3 deficient-Saline vs n-3 sufficient-LPS, ***p < 0.0001; n-3 sufficient-Saline vs n-3 deficient LPS, ***p < 0.0001; n-3 deficient-LPS vs n-3 sufficient-LPS, ***p < 0.0001. e Venn diagram highlighting the number of genes that were modulated by MIA in the hippocampi of adult n-3 sufficient (blue) or n-3 deficient (red) mice. Lower panel: Number of genes that were up- or down-regulated in n-3 sufficient and n-3 deficient mice. Representation of the 20 most significantly dysregulated genes in n-3 sufficient (f) and n-3 deficient (g) mice. Genes that appear in both n-3 sufficient and n-3 deficient mice are bold. h PCA analysis of MIA-induced differentially expressed genes (DEG) in both dietary groups. Confidence ellipses appear around each group. i, j Gene Ontology analysis of DEGs (light red and blue: up-regulated genes; dark red and blue: down-regulated genes).

All methodological details appear as supplementary information.

Statistical analysis

All data are expressed as means ± SEM. Normality and homoscedasticity of distributions were assessed by Shapiro-Wilk test and Brown–Forsythe test respectively. Data were then analysed using parametric two-way analysis of variance (ANOVA) (diet and prenatal treatment as factors) followed by Bonferroni post-hoc test when applicable. If normality and/or homoscedasticity tests failed, data were analyzed with a non-parametric Kruskal–Wallis test followed by Mann–Whitney U test when applicable. To analyse recognition index results in the probe test of the Morris Water Maze, we performed a one sample t tests (comparison to chance level set at 25%). To calculate the inflammation Z-score, all cytokines were converted to Z-scores (each cytokine value subtracted by the mean of the group and divided by the SD), and the Z-scores were added. Spearman correlations were used to explore associations between microbial composition, the gut immune cells reactivity and markers of myelin and oligodendrocytes. All analyses were conducted with GraphPad Prism 7 (GraphPadSotfware, San Diego, USA) except for repeated measures analysis (learning and reversal phase in Morris Water Maze task; pups’ exploratory behavior), where we used Statistica 6.0 (StatSoft, Tulsa, USA) and for correlation matrix where we used R version 3.5.2. For all results, statistical significance was set at p < 0.05. Details for all statistical results are presented in Tables S2 and S3.

Results

Low maternal n-3 PUFA dietary intake results in MIA-induced locomotor and spatial memory deficits in adult offspring

We assessed short and long-term consequences of MIA and dietary manipulation on offspring’s behavior. Regardless of the diet, MIA negatively affected the psychomotor score of pups between PND4 and PND6, measured by the increase of time needed to perform the 2 tests of the Fox battery. In addition, MIA decreased the whistles time of vocalizing PND7-8 pups in response to maternal separation, independent of maternal diet. (Fig. 1b, c). The exploratory behavior of neonates was significantly decreased in n-3 deficient mice (Fig. 1d, e). Moreover, MIA significantly increased the mean exploratory behavior of n-3 deficient animals.

In adulthood, n-3 deficient mice displayed learning deficits during the training phase of the Morris Water Maze task (Fig. 1f). During the probe test, MIA-exposed n-3 deficient mice spent the same amount of time in all 4 quadrants of the maze, a marker of spatial memory deficits (Fig. 1g). When assessing locomotor activity, MIA-exposed n-3 deficient mice traveled significantly longer distances than n-3 deficient-Saline or n-3 sufficient mice during the 1 h-recording period (Fig. 1i).

Finally, we could not find any effect of MIA or diet on the reversal task in the Morris Water Maze, or on the anxiety levels as assessed in the light-dark box and the open-field (data not shown) (Fig. 1h–k).

Overall, our data showed that MIA induced spatial memory deficits and hyper-locomotion only in n-3 deficient adult offspring.

MIA-exposed n-3 deficient adult mice display alterations in the metabolism of n-6 PUFAs in the hippocampus

N-3 PUFA dietary deficiency significantly altered fatty acid composition of the hippocampus in the adult offspring (Fig. 2a–d, Table 1, Table S3), including an increase in the total amount of n-6 PUFAs, Saturated Fatty Acids and MonoUnsaturated Fatty Acids, and a decrease in n-3 PUFAs. MIA alone also increased SFA concentration and decreased total PUFAs, total n-6 PUFAs, DocosaTetraenoic Acid (DTA), and DPA levels (Table 1 and Table S3). Post-hoc analysis revealed that these effects were more pronounced in MIA-exposed n-3 deficient mice (Fig. 2a–d).

Table 1 Total fatty acid composition of adult hippocampus.

Dietary n-3 PUFA deficiency exacerbates MIA-mediated alterations of gene expression in the adult hippocampus

Next, we used transcriptomics to assess gene regulations sustaining the behavioral effects of n-3 PUFA deficiency and MIA in adults. We focused on the hippocampus, a critical structure for learning and memory abilities [83]. When comparing MIA-exposed mice to saline-treated individuals, 337 genes were significantly differentially expressed (DEGs) in n-3 sufficient mice vs 610 DEGs in n-3 deficient animals, among which 120 genes were common to both dietary groups (Tables 2, 3). This suggests that MIA had a greater impact on gene expression under low n-3 PUFA intake (Fig. 2e). Figure 2f, g illustrates the most significantly dysregulated genes in n-3 sufficient (Fig. 2f) and n-3 deficient mice (Fig. 2g). Principal component analysis (PCA), using DEGs between n-3 sufficient-LPS vs n-3 deficient-LPS mice, showed separation between dietary groups (Fig. 2h).

Table 2 List of differentially expressed genes (DEGs) in all experimental groups (comparison saline vs LPS in bofth dietary groups).
Table 3 Common differentially expressed genes (DEGs) between n-3 sufficient and n-3 deficient group.

We then used the MGI database [84] to retrieve the non-redundant Gene-Ontology annotations of interest that are related to the set of DEGs identified above. We computed the number of genes for each selected GO-term, in every group of DEGs identified for each diet type (Fig. 2i, j). The analysis revealed that the majority of these genes belong to “Cell communication” (e.g., synaptic activity, neurotransmitter release), “Immune/defense response” (e.g., leukocyte activation/migration/homeostasis), “CNS development” (e.g., myelination, gliogenesis, neurogenesis) or “Lipid metabolism” (e.g., fatty acid metabolism, membrane lipid metabolic process) pathways (Fig. 2i, j). Based on these transcriptomic data, we further analysed the impact of low n-3 PUFA intake and exposure to MIA on: (1) microglia-neuron crosstalk in the developing hippocampus; (2) neuronal phenotype; (3) myelination.

MIA induces persistent neuronal morphology alterations in n-3 deficient mice, likely by modifying microglia-neuron interactions

During the first post-natal weeks, microglia prune nonfunctional and immature synapses as a maturation mechanism for neuronal networks [29, 30, 85]. We quantified the level of colocalization between microglia (Iba-1) and dendritic spines (PSD95) at PND14, when synaptic refinement peaks in the hippocampus [86]. Our data show that n-3 PUFA deficiency significantly increased the level of Iba-1/PSD95 colocalization at PND14, suggesting greater synaptic pruning activity (Fig. 3a). Prenatal exposure to LPS did not show any effect. However, MIA differentially affected the expression of c3, cd47, cx3cr1, cx3cl1 according to n-3 PUFA intake, all these genes being involved in microglia-mediated synaptic pruning in the developing brain (Fig. 3b). These effects could not be attributed to modifications of microglial density (quantification of Iba-1 immunoreactivity) and/or phenotype (flow cytometry analysis of phenotypic markers on sorted microglia) (Supplementary Fig. 2a, b). We could only observe a slight, yet significant, increase in the proportion of CD86+ microglia in MIA-exposed n-3 deficient mice at PND14, suggesting that the cells may be skewed towards a proinflammatory phenotype (Supplementary Fig. 2b) [87]. Prenatal LPS exposure also did not affect cytokine mRNA production in the hippocampus of the offspring (Supplementary Fig. 2c).

Fig. 3: Effect of n-3 PUFA deficiency and MIA on microglia-neuron crosstalk pathways, spine density, oligodendrocyte and myelin protein expression.
figure 3

All graphs show Means ± SEM. a Colocalization of Iba-1 and PSD95 proteins immunoreactivity in the CA1 region of the hippocampus of PND14 pups. Representative confocal image of Iba-1 (green) PSD95 (red) costaining (Top panel: scale bar = 10 µm) and Imaris 3D reconstruction (Bottom panel, scale bar = 1 µm). N = 72–122. Kruskal–Wallis test followed by Mann–Whitney comparisons; n-3 deficient-Saline vs n-3 sufficient-Saline, ***p < 0.0001; n-3 deficient-LPS vs n-3 sufficient-LPS, ***p < 0.0001. b qRT-PCR quantification of microglia-neuron interaction mRNA markers in the hippocampus of PND14 mice (data normalized to the saline group, dotted line). N = 4–6. Kruskal–Wallis test followed by Mann–Whitney comparisons; *p < 0.05, **p < 0.01 (all comparisons in Table S2). c Quantification and representative images of Golgi staining spine density in the CA1 region of the hippocampus at PND28. N = 8–21. Kruskal–Wallis test followed by Mann–Whitney comparisons; n-3 deficient-Saline vs n-3 deficient-LPS, ***p < 0.0001; n-3 deficient-Saline vs n-3 sufficient-Saline, **p = 0.001; n-3 deficient-LPS vs n-3 sufficient-LPS, *p = 0.025. d Western blot-based quantification and representative images of PSD95 protein expression in the hippocampus of PND28 mice. N = 4–8. Kruskal–Wallis test followed by Mann–Whitney comparisons; n-3 deficient-Saline vs n-3 deficient-LPS, **p = 0.004; n-3 deficient-LPS vs n-3 sufficient-LPS, *p = 0.03. Quantification of Olig2 (e), PLP (f), APC (g), MAG (h) and MBP (i) immunoreactivity in the hippocampus of PND14 mice. N = 4–7. Two-way ANOVA. Olig2: diet effect, F(1,20) = 3.48, p = 0.08; MIA effect, F(1,20) = 4.78, p = 0.041. PLP: MIA effect, F(1,22) = 5.01, p = 0.036. APC: diet effect, F(1,17) = 4.96, p = 0.0397.

We further studied the consequences of altered microglia-neuron interactions on neuronal morphology at PND28, when most of the synaptic refinement is complete in the hippocampus. N-3 PUFA deficiency significantly decreased spine density (Fig. 3c). We also observed that MIA-exposed n-3 deficient mice had greater number of spines on pyramidal neuron and expressed higher amount of PSD95 protein vs n-3 deficient-Saline animals (Fig. 3c, d). MAP2 protein expression was similar across all experimental groups (Supplementary Fig. 2d).

Overall, our data suggest that n-3 PUFA deficiency affected the amplitude of response to MIA in terms of dendritic spine density and microglia-neuron communication during the post-natal period.

No interaction between MIA and n-3 PUFA deficiency on myelin and oligodendrocyte protein expression

According to the transcriptomic analysis, “gliogenesis” and “myelination” processes are likely to be altered by MIA and n-3 PUFA deficiency (Fig. 2j). MIA significantly reduced the expression of Olig2 (marker of all oligodendrocytes) and PLP (mature oligodendrocytes marker) at PND14, the peak of oligodendrocyte maturation and myelination in the developing brain (Fig. 3e, f). APC protein (for Adenomatous Polyposis Coli, a marker for mature oligodendrocytes) levels were significantly increased by n-3 PUFA deficiency while the expression of the myelin proteins MAG and MBP were not altered (Fig. 3g–i). Overall, we did not observe any interaction between MIA exposure and low n-3 PUFA dietary intake on oligodendrocyte and myelin protein expression.

MIA differentially alters microbiota composition in both dietary groups and reveals gut alterations under n-3 PUFA deficiency

A growing body of evidence suggests microbial modulation of brain functions, especially when occurring during the post-natal developmental phase [12, 59, 61, 69, 70, 74, 88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107]. We therefore assessed the gut microbiota at PND14 and PND21, i.e., during a critical time window for gut physiology and gut-brain communication. Alpha diversity was significantly higher in n-3 sufficient-Saline group than in any other group, at both time points, (Fig. 4a–e; Tables 4, 5). While n-3 PUFA deficiency decreased diversity at both time points, at PND21 MIA restored microbial diversity up to control levels in n-3 deficient mice (Fig. 4e; Table 5). A strong effect of diet and MIA was observed at phyla, family and genus levels (Fig. 4b, c–f, g; Tables 4, 5). At both ages, n-3 PUFA deficiency increased the amount of Proteobacteria and decreased the proportion of Lachnospiraceae. It also decreased the amount of Prevotellaceae and increased Akkermansiacaceae and Entoreobacteriaceae at PND14, and decreased the Lactobacillaceae and increased the Tannerellaceae and Rikenellaceae at PND21 (Fig. 4c–g). All groups clustered together by PCA analysis at PND14, while they clustered distinctly from controls (n-3 sufficient-Saline) at PND21 (Fig. 4d–h). Our data revealed a robust effect of MIA and low n-3 PUFA intake, alone and combined, on gut microbiota composition, in an age-dependent manner.

Fig. 4: Effect of n-3 PUFA deficiency and MIA on gut microbiota composition at PND14 and PND21.
figure 4

All graphs show Means ± SEM. a 16S rRNA-sequencing-based alpha diversity analysis of the microbiota, measured by Shannon index in PND14 mice. N = 8–12. Two-way ANOVA: diet effect, F(1,39) = 12.76, p < 0.001; MIA effect, F(1,39) = 5.39, p = 0.026. Bacteria phyla (b) and family (c) observed in all experimental groups at PND14. d PCA of all subjects at PND14. Confidence ellipses appear around each group. e 16S rRNA-sequencing-based alpha diversity analysis, measured by Shannon index in PND21 mice. N = 8–13. Two-way ANOVA: n-3 sufficient-Saline vs n-3 sufficient-LPS, ***p = 0.0005; n-3 sufficient-LPS vs n-3 deficient-LPS, *p = 0.019; n-3 sufficient-Saline vs n-3 deficient-Saline, **p = 0.0078. Bacteria phyla (f) and family (g) observed in all experimental groups at PND21. h PCA of all subjects at PND14. Confidence ellipses appear around each group. Quantification of MLN lymphocytes cytokine release measured by ELISA at PND14 (i) and PND21 (j). N = 6–15; Kruskal–Wallis test followed by Mann–Whitney comparisons; *p < 0.05, **p < 0.01, ***p < 0.001 (all comparisons in Table S2). Z-score of T cells inflammatory reactivity in PND14 (k) and PND21 (l) mice. N = 7–15. Kruskal–Wallis test followed by Mann–Whitney comparisons; PND14: n-3 deficient-Saline vs n-3 deficient-LPS, ***p < 0.001. PND21: n-3 sufficient-Saline vs n-3 sufficient-LPS, **p = 0.0011, n-3 sufficient-LPS vs n-3 deficient-LPS, ***p = <0.0004.

Table 4 16S rRNA-sequencing-based bacteria genus observed in all experimental groups at PND14.
Table 5 16S rRNA-sequencing-based bacteria genus observed in all experimental groups at PND21.

At PND14, n-3 PUFA deficiency significantly shortened colon length and increased its permeability. MIA displayed no effect on these parameters (Supplementary Fig. 3a, b). At PND21, n-3 deficient-Saline mice still exhibited shorter colon but lower permeability compared with n-3 sufficient group. MIA increased both parameters in n-3 deficient mice while it had no effect in mice fed n-3 PUFA sufficient diet (Supplementary Fig. 3c, d).

We then assessed mesenteric lymph node T-cells cytokine secretion as a marker of gut immune reactivity [59, 108]. T cells sorted from PND14 n-3 deficient-LPS mice released greater levels of TNFα, IFN-γ, and IL-17 upon anti-CD3/anti-CD28 stimulation compared with n-3 sufficient-LPS group, while production of the anti-inflammatory cytokine IL-10 was hampered (Fig. 4i). We calculated a z-score that summarizes the cytokine profile of mice and confirmed that the effect of MIA on cytokine release was more pronounced in n-3 deficient mice (Fig. 4k). At PND21, MIA increased cytokine production more drastically in stimulated T cells sorted from n-3 sufficient mice (Fig. 4j, l). At histological levels, inflammation (Ameho score) was not different between groups (Supplementary Fig. 3e, f).

Hence, n-3 PUFA deficiency enhanced gut immune reactivity during the post-natal period, shifting the MIA-mediated proinflammatory response towards the earlier time point.

Microbial modifications correlate with gut inflammatory reactivity and neurobiological alterations

Spearman’s correlation matrix between gut immune cells reactivity and bacterial genera show that the microbial composition was tightly correlated with gut inflammatory reactivity both at PND14 and PND21 (Fig. 5a–c). Our data also revealed correlations between microbial composition and markers of myelin (olig2, MAP2, PLP) and microglial cells (Iba-1) (Fig. 5b–d).

Fig. 5: Correlations between microbial modifications, gut inflammation, and neurobiological parameters.
figure 5

a Spearman’s correlation matrix between gut immune cells reactivity (e.g cytokine release after T-cells stimulation) and bacterial genera in PND14 mice (*p = 0.05). b Spearman’s correlation matrix between neurobiological measurements (PLP, Olig2, Iba-1 and MAP2) and bacterial genera in PND14 mice (*p = 0.05). c Spearman’s correlation matrix between gut immune cells reactivity (e.g cytokine release after T-cells stimulation) and bacterial genera in PND21 mice (*p = 0.05). d Spearman’s correlation matrix between neurobiological measurements (PLP, Olig2, Iba-1, and MAP2) and bacterial genera in PND21 mice (*p = 0.05). e Spearman’s correlation between gut immune cells reactivity (e.g released cytokines after stimulation) and neurobiological parameters in PND21 mice (*p = 0.05). N = 20–22. Escherichia-Shig: Escherichia-Shigella; Eubacterium copro: Eubacterium coprostanoligenes group; Lachno NK4A136: Lachnospiraceae NK4A136 group; Lachno UCG-008: Lachnospiraceae UCG-008; Prevo UCG-001: Prevotellaceae UCG-001; Rikenellaceae RC9: Rikenellaceae RC9 gut group; Ruminococcus gg: Ruminococcus gnavus group. f Schematic summarizing the main findings. Exposure of n-3 PUFA deficient dams to MIA alters the gut microbiota composition and increases the inflammatory reactivity of the gut T-lymphocytes in the offspring during the post-natal period. This is correlated with an impairment in microglia-neuron crosstalk during this phase, with consequences on hippocampus function and memory abilities later in life.

We further assessed potential correlations between gut inflammatory reactivity and neurobiological features at PND21. Spearman correlation analysis revealed a strong correlation between Iba-1 and T cells-induced cytokines (Fig. 5e). We also observed significant correlation between Olig2 expression and the production of TNF-α and IL-17 (Fig. 5e). This suggests that alterations of gut physiology are linked to neurobiological defects in the developing brain.

Discussion

We and others previously showed that maternal dietary n-3 PUFA intake influences neurobiological and behavioral outcome in mice exposed to prenatal inflammation [6, 9, 28, 109, 110]. Our study is the first to extensively explore the mechanisms underlying the cumulative effects of MIA and n-3 PUFA deficiency, along the gut-brain axis and at various ages (Fig. 5f). We show that (1) n-3 PUFA dietary deficiency reveals MIA-induced behavioral alterations in adult offspring but not in pups, (2) MIA-exposed n-3 deficient adult mice display alterations in the metabolism of n-6 PUFAs in the hippocampus. (3) This was correlated to profound transcriptomic alterations, especially for genes involved in CNS development processes, such as synaptic plasticity, myelination and inflammation. (4) N-3 PUFA deficiency worsened the impact of MIA on microglia-neuron interaction in pups. (5) N-3 PUFA deficiency worsened MIA-induced microbiota and gut alterations in the post-natal period. (6) Microbial composition and gut inflammatory molecules correlated with neurobiological outcome in pups. This study must now be extended to female mice, since most, if not all, the parameters studied are sensitive to sex, and to increase its clinical relevance [74, 111,112,113].

We show that MIA reveals spatial memory deficits and hyper-locomotion behavior only when combined with low n-3 PUFA intake. This is contradictory to previous studies that showed a significant effect of MIA alone on behavior of adult mice, including learning and memory abilities [8, 114,115,116,117,118,119,120,121,122,123,124]. We however confirmed a previous report from our group in which n-3 PUFA deficiency induces memory deficits in MIA-exposed mice [28]. As a plausible explanation for this discrepancy, our control animals are fed with an n-3 sufficient diet, whose composition is distinct from standard chow (less MUFA, more PUFA and higher n-6/n-3 ratio in the standard chow vs n-3 sufficient diet) [125]. Moreover, the type and the dose of MIA-inducing agent, as well as the embryonic age of exposure, are also key for long-term behavioral deficits [8].

One cannot rule out that MIA-induced maternal care defects is a plausible cause for the long term deficits observed in the offspring. Indeed, it was previously shown that the adoption of control neonates by surrogate rearing mothers, previously exposed to MIA during pregnancy, is sufficient to trigger cognitive deficits in the fostered offspring [13, 109, 126,127,128,129]. Nonetheless, unlike other studies using a similar approach to ours, i.e., one single i.p. administration of LPS at E17 in mice, we could not find any major effects of MIA in n-3 sufficient mice, especially on memory and anxiety [11, 80, 82, 118, 130, 131]. This suggests intrinsic differences between standard chow-fed (as used in previous studies) and n-3 PUFA sufficient-fed mice and questions the underlying mechanisms. We previously showed that the brain fatty acid composition of mice fed either a standard or n-3 sufficient diet is quite similar [125]. However, the two diets vary in total saturated fat, monounsaturated, PUFAs, LA/ALA ratio, proteins, carbohydrates, and total lipids. Hence, aspects other than brain fatty acid composition are likely to explain differences between both groups, such as cell energy metabolism, lipid composition at the cellular resolution or production of lipid derivatives. This remains to be addressed.

We evaluated the effect of dietary n-3 PUFA deficiency combined with MIA on gut inflammation and microbiota composition, as a plausible mechanism underlying behavioral alterations. α-divesity was decreased by both interventions at PND14 and PND21. Similar observations were made in patients with neurodevelopmental diseases such as ADHD [132]. Of note, bacteria from the Prevotellaceae family were less abundant in n-3 deficient mice while Akkermansia was increased. While we must be cautious in making comparisons between human and mice findings regarding gut microbiota composition [133], similar differences have been observed between autistic children and controls of similar age [134, 135]. Moreover, increased levels of Prevotella following a fecal material transfer in autistic patients have been associated with improvements of autistic symptoms [136], while some studies report that Akkermansia levels are elevated in multiple sclerosis (MS) or Parkinson’s Disease patients’ [137,138,139]. We also observe an elevation of Enterobacteriaceae in n-3 deficient mice, which was similarly found to be increased in autistic children [140] and decreased after n-3 PUFA supplementation [141]. Our work is also in line with two studies demonstrating that maternal n-3 PUFA deficiency induces compositional and functional disturbances to the gut microbiome, closely associated with long term behavioral consequences in the offspring [142, 144].

We show for the first time that low maternal n-3 PUFA intake alone or combined to MIA alters gut structure and physiology during post-natal development in a time-dependent manner. Previous study reported altered intestinal permeability in the offspring in a poly(I:C) model of MIA [59]. Several studies reveal the impact of n-3 PUFAs on gut health (intestinal permeability and gut innervation) [143, 145,146,147,148,149,150]. The gut immune response, tightly related to gut permeability, was exacerbated in n-3 deficient mice at PND14, while at PND21, it was enhanced in n-3 sufficient mice. More studies are needed to understand whether and how the delay of response in n-3 sufficient mice explains their protection in terms of long-term cognitive abilities. A recent study highlighted the prominent role of T helper 17 (TH17)-derived IL-17A in neurodevelopmental abnormalities in the offspring of pregnant mothers undergoing MIA [61]. In our study, we observed a general overactivation of the gut inflammatory response in MIA-exposed offspring during CNS development. Interestingly, clinical studies report defects in gut permeability and inflammatory response in ASD and schizophrenic patients [60, 107, 133], suggesting that the double environmental insult recapitulates some aspects of these diseases.

At PND14, we found that levels of several members of the Lachnospiraceae family correlate with and myelin markers and markers of gut immune reactivity. Recent study highlighted, Ruminococcus Gnavus as a central player in Crohn’s disease. These bacteria have been shown to stimulate gut inflammation through the release of specific metabolites [150]. Lachnospiraceae is decreased in MS patients [151], who are known to exhibit an altered T-cell response together with alterations of the intestinal barrier permeability [152]. Contrary to our own data, Hsiao et al. showed that MIA increases Lachnospiraceae [59]. The discrepancy could result from different timelines of analysis: we analyzed microbiota composition in PND14-21 mice, while Hsiao et al. studied its composition after weaning. It could also be explained by the different MIA protocols used: bacterial at E17 vs viral mimetic at E12.5. At the genus level, we observed opposite variation of members of this family. Linking global Lachnospiraceae levels with biological outcomes appears to be inappropriate, as the family comprises of a large number of distinct bacteria. It could explain why a human study of associations between the levels of Lachnospiraceae and neurological diseases, such as depression, also found conflicting results [153].

N-3 PUFA deficiency exacerbated MIA-induced defects in microglia-neuron interactions, while microglial density, phenotype and inflammatory activity were unaffected in pups. This confirms most previous reports in which no effect of MIA on microglial density were ever observed [34, 116, 154,155,156,157,158]. We previously demonstrated that the proinflammatory cytokine expression is exacerbated in both maternal and embryonic brains of the n-3 deficient group [28], which is no longer the case at PND14. However, increased number of Iba1/PSD95 positive cells were measured in the hippocampus of n-3 deficient offspring at PND14, suggestive of synaptic pruning [29]. In line with these data, the hippocampal synaptic density was decreased at PND28. These findings reinforce our previous observations that early-life n-3 PUFA dietary deficiency alters post-natal microglia phenotype and phagocytic activity in the hippocampus [44, 77, 159]. However, while n-3 PUFA deficiency combined with MIA further increased Iba-1/PSD95 colocalization at PND14, at PND28 the spine density in this group was higher compared with that of saline-injected n-3 deficient mice. These data suggest that the regulation of spine density relies on microglia-independent mechanisms in the n-3 deficient/LPS mice. Microglial phagocytic capacity might be overridden in that context, or compensatory mechanisms are put in place between PND14 and PND28, leading to a significant increase in spine density. More experiments are required to test the functional state of the excess spines of n-3 deficient/LPS mice at PND28 and how it relates to behavioral deficits observed in these mice in adulthood. Our study also revealed an impact of n-3 PUFA deficiency and MIA on microglia-neuron crosstalk, confirming a previous report describing a correlation between spine density and CX3CR1 expression in the hippocampus of MIA-exposed mice [160]. The crucial pathways for this interaction are likely to be dysregulated in our experimental context (complement cascade, fractalkine pathway, CD47) [29, 30, 161]. This concurs with previous studies showing that n-3 PUFAs modulate spine density and microglial activity [20, 44, 77, 162,163,164].

In conclusion, our study is the first to examine potential mechanisms underlying the link between low n-3 PUFA intake and MIA, including interactions between enteric microbiota and the CNS. We uncover a correlative relationship between diet- and MIA-induced gut alterations and deleterious neurobiological outcomes. We also highlight the post-natal period as a vulnerable time window for perinatal dietary and immune stress. We finally demonstrate the long-lasting effect of these stressors, both at the transcriptional and behavioral levels. This work furthers our understanding of the link between MIA, poor nutritional habits and NDD, emphasising a potential role of the gut-brain axis.

Funding and disclosure

PG reports having received consulting fees from Chiesi Pharma AB. All other authors reported no biomedical financial interests or potential conflicts of interest. Funding for this research was provided by the Institut National pour la Recherche Agronomique, l’Alimentation et l’Environnement (INRAE), the Bordeaux Univ, the New Aquitania Region, NeurATRIS (Paris, France), the International Laboratory OptiNutriBrain, the Foundation for Medical Research (FRM) (DEQ20170336724), the French Foundation (FDF, #00070700). QL was supported by the Region Ile de France (PICRI, the Ceberal Palsy Foundation #13020605) and by the FRM, FD is supported by an Idex PhD grant, GB is supported by the French Ministry for Research and Higher Education, CL was supported by Idex grant, AQ and IV were supported by an Erasmus Mundus scholarship, ABB was supported by the Ecole Normale Supérieure (Lyon). TB, BM and PG were supported by Inserm, Université de Paris, Région Île-de-France (DIM Cerveau et Pensée, NeoPhen), Investissement d'Avenir (ANR-11-INBS-0011, NeurATRIS), ERA-NET Neuron (Micromet). The Western blot analyses were done in the Biochemistry and Biophysics Platform of the Bordeaux Neurocampus at the Bordeaux University funded by the LABEX BRAIN (ANR-10-LABX-43). This work benefited from the facilities and expertise of the imaging platform Imag'In (www.incia.u-bordeaux1.fr), which is supported by CNRS and Region Aquitaine and from the Bordeaux Imaging Center. The microscopy was done in the Bordeaux Imaging Center a service unit of the CNRS-INSERM and Bordeaux University, member of the national infrastructure France BioImaging supported by the French National Research Agency (ANR-10-INBS-04).