Akkermansia Muciniphila and Environmental Enrichment Reverse Cognitive Impairment Associated With High-fat High-cholesterol Consumption Through the Gut-brain Axis

Background: Non-alcoholic steatohepatitis (NASH) is one of the most prevalent diseases worldwide. A high-fat, high-cholesterol (HFHC) diet leads to an early NASH model. It has been suggested that gut microbiota mediates the effects of diet through the microbiota-gut-brain axis, modifying the host’s brain metabolism and disrupting cognition. Results: Here, we target NASH-induced cognitive damage by testing the impact of environmental enrichment (EE) and the administration of either Lacticaseibacillus rhamnosus GG (LGG) or Akkermansia muciniphila CIP107961 (AKK). EE and AKK, but not LGG, reverse the HFHC-induced cognitive dysfunction, including impaired spatial working memory and novel object recognition; however, whereas AKK restores brain metabolism, EE produces an overall decrease. Moreover, AKK and LGG did not induce major rearrangements in the intestinal microbiota, with only slight changes in bacterial composition and diversity, whereas EE entailed an increase in Firmicutes and Verrucomicrobia members. Conclusions: Our ndings illustrate the interplay between gut microbiota, the host’s brain energy metabolism, and cognition, and they suggest intervention strategies, such as the administration of AKK, for the management of the cognitive dysfunction related to NASH. e Spatial working memory test. Bar charts (mean±SEM)represent average latency on the sample and retention trials. Two-tailed paired t-tests (* comparison with its respective sample) and one-way ANOVA followed by Tukey’s test (# comparison between NC, NC + EE and HFHC + EE) were used. NC and NC + EE groups remembered the position of the platform in the retention trial; whereas HFHC was not able to remember it, HFHC + EE displayed recovered spatial working memory. (p < 0.001), anterodorsal thalamus (ADT; p < 0.001), basolateral amygdala (BLA; p = 0.006), dentate gryus (DG; p < 0.001), and CA3 (p = 0.001). Finally, HFHC + EE showed lower CCO values than NC + EE in the DG (p < 0.001). No differences were found between HFHC + EE and NC + EE in the prefrontal cortex, dorsal and ventral striatum, thalamus, amygdala, and perirhinal cortex. novel object in comparison with HFHC+PBS animals, but no differences were found between them. #p<0.05, **p ≤ 0.010. c Spatial working memory test. Bar charts (mean±SEM) represent the average latency on the sample and retention trials. Two-tailed paired t-tests (* comparison with its respective sample) and one-way ANOVA followed by Tukey’s test (# comparison between HFHC+PBS, HFHC+EE and HFHC+AKK) were used. HFHC+EE and HFHC+AKK displayed improved performance on the spatial working memory task, in comparison with HFHC+PBS, but no differences were found between them. *#p<0.05, **p ≤ 0.010.


Background
Non-alcoholic fatty liver disease (NAFLD) is characterised by hepatic fat accumulation, and it is closely associated with central obesity, diabetes, and other features of metabolic syndrome [1]. NAFLD encompasses a broad disease spectrum ranging from simple steatosis to non-alcoholic steatohepatitis (NASH), which may, in turn, develop into cirrhosis, end-stage liver disease, or hepatocarcinoma [2].
Although NASH has become a serious global health threat, therapeutic strategies that could prevent NAFLD-NASH progression have been overlooked.
We previously published [3] a study showing that consumption of a high-fat, high-cholesterol diet leads to an early NASH model that replicates the typical features of the human disease, including hepato-and splenomegaly, early NASH histopathology, hypercholesterolemia, increased serum liver enzymes, and increased pro-in ammatory cytokines [3,4]. Studies have reported that a high-fat, high-cholesterol diet and a high-fat diet produce changes in the gut microbial composition, as well as reductions in microbial diversity and changes in speci c bacterial taxa [3,5,6]. In this line, a recent study by Hoyles et al. [7] showed that faecal transplantation from human donors with hepatic steatosis triggered rapid development of hepatic steatosis in mice, which highlights the contributing role of microbiota in NASH development.
Gut microbiota has also been found to play an important role in the central nervous system (CNS) through the microbiota-gut-brain axis, and CNS dysfunction has been linked to several psychiatric and non-psychiatric disorders [8]. The involvement of gut microbiota dysbiosis and altered short-chain fatty acid (SCFAs) production has been proposed in neurodegenerative disorders such as Parkinson's disease [9]. Recent studies have also reported altered microbiota in autism spectrum disorder, anxiety, or myasthenia gravis, among other neurological conditions [8,[10][11][12]. Moreover, in a previous study we observed gut microbiota dysbiosis and microbial metabolite alterations linked to cognitive impairment in a NASH rat model [3]. These ndings highlight the functional connection between the liver-gut microbiota and cognition. Thus, gut microbiota could be a target for improving cognitive de ciencies associated with NASH, with probiotics as a promising tool for this purpose. However, the impact of probiotic administration on cognitive function is not well understood.
Furthermore, an inappropriate diet and physical inactivity often co-exist in these patients. Exercise is able to modulate the gut microbiota. Clarke et al. [13] described that athletes showed a higher diversity of gut microorganisms, an observation also supported by other studies [14,15]. In the brain, studies in elderly people have shown that exercise increases hippocampal volume and enhances hippocampus-dependent learning and memory [16]. The effect of physical exercise in rodents is modelled using environmental enrichment (EE), which induces a situation of increased motor stimulation and sensorial and cognitive enhancement [17]. EE has been shown to be associated with synaptic function and cellular plasticity changes, resulting in cognitive enhancement [17][18][19]. Therefore, EE could mitigate the negative impact of a high-fat, high-cholesterol diet on the brain and cognition.
In the present study, we target high-fat, high-cholesterol (HFHC) induced cognitive disturbances through the microbiota-gut-brain axis using two different approaches. First, we assess the effects of EE as a cognitive enhancer, and, second, we evaluate the impact of two speci cally selected strains of probiotics, Lacticaseibacillus rhamnosus GG (LGG) and Akkermansia muciniphila CIP107961 (AKK), on NASH pathology. In both cases, we compare their effects on microbiota changes, brain metabolism, and, ultimately, cognitive improvement.
Environmental enrichment. The NC + EE and HFHC + EE groups were continuously submitted to EE from week 8 to week 12 (Fig. 1a). During this time, each experimental group was separately housed in groups Faecal sample collection and processing. Fresh faecal pellets from each animal were collected and kept at -80 °C until DNA extraction. Before that, samples were weighed, diluted in PBS (1:5 w/v), and homogenized (3 min, full speed) in a stomacher (LabBlender 400). From 1 mL of the homogenate, faecal supernatant and cellular pellets were separated by centrifugation (10.000 rpm, 15 min) and stored at -20 °C. The supernatant was used for SCFA and BCFA analysis, and the pellets were used for DNA extraction with the QIAamp DNA stool kit (Qiagen, GmbH, Germany), as described elsewhere [27].
Analysis of faecal microbial groups by 16S rRNA gene pro ling and quantitative PCR. Extracted Isolated DNA was used as a template for ampli cation of the V3 region from partial 16S rRNA gene sequences by PCR, using the primers and conditions described by Milani and coworkers [28]. The amplicons obtained were then sequenced by using the MiSeq (Illumina) platform) at GenProbio srl (Italy). The individual reads obtained were ltered, trimmed, and processed [29]. 16S rRNA Operational Taxonomic Units were de ned at ≥ 97% sequence homology using the uclust tool developed by Edgar [30]. All reads were classi ed in the lowest possible taxonomic rank using QIIME 2 and a reference dataset from the SILVA database [31]. Akkermansia and Lactobacillus genus were also determined by quantitative PCR using the previously described primers and conditions [32].
Determination of short-chain and branched-chain fatty acids in faeces. SCFA and BCFA levels were determined in the faecal supernatants by means of gas chromatography, as described by Moris et al. [11]. Brie y, 250 µl of cell free-supernatants were mixed with 100 µl methanol, 50 µl internal standard solution (2-ethylbutyric 1.05 mg/ml), and 50 µl of 20% v/v formic acid. The mix was then centrifuged, and the supernatant was injected into a system composed of a 6890NGC injection module (Agilent Technologies Inc., USA) connected to a ame injection detector (FID) and a mass spectrometry detector (MS, 5973N) (Agilent) for quanti cation of both SCFA and BCFA.
Locomotor activity assessment. The study of locomotor activity was carried out by using Rotarod 7750 for rats, which consists of a motor-driven rotating rod (Ugo Basile Biological Research Apparatus) [33].
After habituation for 1 min at a constant speed of 2 rpm, the rats were evaluated for 5 min, during which the speed increased constantly until 20 rpm. Rotation speed was recorded and used as a measure of locomotor function.
Novel object recognition. The object recognition test was carried out in an open eld (66 × 46 × 45 cm) made of grey breglass, with an open roof. The open eld was situated in a room with two diffuse white lights on its sides, providing an illumination density of 50 lux at its centre. The animal's behaviour was recorded by a video camera (Sony V88E) connected to a computer equipped with a computerised videotracking system (EthoVision Pro, Noldus Information Technologies, The Netherlands). After each trial, the apparatus was thoroughly cleaned with a 75% ethanol solution. The objects used were constructed from a combination of plastic pieces of different colours and shapes. The animals were habituated for two days before the test. On the rst day, the animals were situated rst in groups and then individually in the open eld for two trials lasting 6 minutes each. On the second day, each animal was placed in the eld, with two equal objects in the centre, for 3 trials lasting 6 minutes each, separated by 20 minutes.
The test was composed of two phases. In the rst one, in the open eld, each subject was exposed for 4 minutes to two copies of a new object (2 x object A), located in opposite corners at 10 cm from the walls.
After a 50-minute delay, each animal carried out the second trial, which was similar to the previous one, except that a copy of the previously-encountered object was changed to a novel one (object A + object B).
Exploration of an object was de ned as directing the nose towards the object at less than 2 cm and exploring it (i.e. sni ng and or interacting with the object). For each animal, time spent exploring the objects in both trials was measured. The proportion of exploration in the rst (e1) and second (e2) phases of the test was considered as the ratio between the exploration of all objects in each phase and the total time available for exploration. The discrimination of the objects was assessed through the discrimination index (d1), the difference in time spent exploring the two objects in the second phase (novel minus familiar), and the discrimination ratio (d2), the difference in time spent exploring the two objects in the second phase divided by the total time exploring both objects [34].
Spatial working memory. Spatial working memory was measured in the Morris water maze (MWM), which is a cylindrical breglass tank, virtually divided into four quadrants, measuring 150 cm in diameter with a 40-cm high wall [35]. The water level was 30 cm, and its temperature was 22 ± 2 °C. The pool contained a cylindrical platform that was 10 cm in diameter and 28 cm high, of which 2 cm was below the surface and used by the animals to escape. The MWM was in the centre of a 16 m 2 lit room (two halogen lamps of 4000 lx), surrounded by panels on which several extra-maze clues were placed. The animal's behaviour was recorded by a video camera (Sony V88E) connected to a computer equipped with a computerised video-tracking system (EthoVision Pro, Noldus Information Technologies, The Netherlands). One day before the test, the animals were habituated to the task in 3 trials with the platform, using different starting positions in a small square water tank (47 × 75 × 38 cm).
The spatial working memory test is a paired sample task carried out for ve days. Each session consisted of two trials (sample and retention). Within the same session, the platform, which was invisible to the animals because it was hidden underwater, was situated in one of the four quadrants and the animal started facing the wall of another quadrant. The trial ended once the animal had found the platform, or when 60 s had elapsed. If the animal had not reached the hidden platform after this time, it was placed on the platform for 15 s. During the intertrial interval, the animals were placed in a bucket for 5 s. The quadrant containing the platform and the animal's starting point changed between sessions. Latencies to reach the escape platform were recorded and used as a measure of task acquisition. The learning criterion was when the animals spent signi cantly less time in the retention trial than in the sample trial in each session [36].
Brain metabolic activity. Ninety minutes after the last session of the spatial working memory task, the animals were decapitated, their brains were frozen and subsequently sliced in 30 µm-thick brain coronal sections. The protocol followed was previously described by our group [37], and staining variability across different baths was controlled through sets of brain tissue homogenate standards of known CCO activity from rat brains cut at different thicknesses (10, 30, 50, and 70 µm) [38].
The CCO histochemical staining intensity was quanti ed by densitometric analysis, using a computerassisted image analysis workstation (MCID, Interfocus Imaging Ltd). The mean optical density (OD) of each region was measured in twelve readings that were averaged to obtain one mean per region for each animal. These OD values were then converted to CCO activity units (µmol of cytochrome c oxidised/min/g tissue wet weight), determined by the enzymatic activity of the standards measured spectrophotometrically.
Statistical analysis. Data derived from microbiota were analysed using the IBM SPSS Statistics Version 24.0 (IBM Corp., USA) software. Comparisons of the groups following normal distributions (alphadiversity) were performed using one-way ANOVAs, followed by Tukey's post-hoc analyses. For nonparametric distributions (qPCR, SCFAs and BCFAs data), the Kruskal-Wallis test was performed, followed by Dunn's post-hoc test. Differences in bacteria abundances were calculated using the linear discriminant analysis effect size (LEfSe) method [40].
Body weight, behavioural data, and CCO values were analysed with the SigmaStat 3.2 program (Systat, USA). Body weight was analysed through a two-way ANOVA (Group x Week), followed by a post-hoc Tukey test when appropriate. Maximum speed spent on the Rotarod was compared between groups by using a Kruskal-Wallis test. On the novel object recognition test, two-tailed paired t-tests were used to analyse the exploration between the two trials (e1 and e2), and Mann-Whitney U tests for independent samples were performed on d1 and d2 to nd out whether the mean scores differed from zero (with chance performance indicating equal exploration of the two objects). Additionally, d2 values were compared between the groups using a one-way ANOVA. On the spatial working memory test, sample and retention trial latencies were compared within each group with a two-tailed paired t-test, and between groups with a one-way ANOVA. Brain metabolic activity was evaluated between the groups through oneway ANOVA. When normality or equal group variances failed, Kruskal-Wallis one-way analysis of variance on ranks was performed. Post-hoc multiple comparison analyses were carried out, when allowed, using the Tukey method. The results were considered statistically signi cant if p < 0.05.

Results
Environmental enrichment restores cognitive de cits caused by the HFHC diet. No statistically signi cant differences were found in the body weight between groups throughout the 14 weeks of administration of the diet (Fig. 1a) (p = 0.306). As expected, signi cant differences between weeks were revealed (p < 0.001), with normal weight gain according to age in both experimental groups (Fig. 1b). The normal increase in weight was accompanied by the absence of statistically signi cant locomotor de cits in all groups (p = 0.151; Fig. 1c).
When the experimental groups performed the novel object recognition task, there were no statistically signi cant differences between e1 and e2 in any of the groups (NC: p = 0.153; NC + EE: p = 0.157; HFHC: p = 0.895; HFHC + EE: p = 0.106), which means that these animals spent a similar amount of time exploring in the rst and second phases of the test. However, when we compared the discrimination ratios, we found that the NC group was able to discriminate between a previously-encountered object and a novel object because d1 and d2 were statistically higher than zero (d1: p < 0.001; d2: p < 0.001), whereas the HFHC group was not able to distinguish the new object (d1: p = 0.442; d2: p = 0.442). After the implementation of EE, the NC + EE group maintained its recognition ability (d1: p < 0.001; d2: p < 0.001), and we found that the HFHC + EE group was now able to recognise the novel object (d1: p = 0.010; d2: p = 0.010). When we compared the discrimination ratio between the experimental groups, we found that the NC, NC + EE, and HFHC + EE groups showed statistically signi cant higher d2 than the HFHC group (p = 0.003). The HFHC group initially displayed an object recognition impairment that was reversed by the introduction of EE (Fig. 1d).
To explore whether or not the EE improvement only occurred in a particular network, the experimental groups were tested on a spatial working memory task. Whereas the NC group was able to remember the position of the platform, showing a statistically signi cant lower latency in the retention trial compared to the sample trial (p = 0.003), the HFHC group did not show statistically signi cant differences between the sample and retention trials (p = 0.333). After EE, the NC + EE maintained their capacity to remember the position of the platform (p = 0.002), and HFHC + EE were now able to remember it as well (p = 0.002). Our results showed that the HFHC impaired spatial working memory was ameliorated by the EE. When comparing the mean latencies in each trial between groups, we observed that, although the sample latencies were statistically similar, (p = 0.941), the retention latencies differed (p = 0.008) because they were signi cantly higher in HFHC than in NC, NC + EE, and HFHC + EE (Fig. 1e).
Each group was fed their respective diet for 14 weeks. From week 8 to 12, NC+EE and HFHC+EE were subjected to EE, whereas HFHC+LGG and HFHC+AKK were given their respective probiotic daily, and HFHC+PBS received phosphate buffered saline. From weeks 12 to 14, the cognitive evaluation took place, at the end of which the animals were sacri ced and the samples were collected. b Body weight across weeks. Two-way ANOVA (Group x Week) was used to assess weight gain. No changes in body weight between groups were found throughout the 14 weeks of the administration of the diet, and signi cant differences across the weeks were revealed. c Locomotor function evaluation measured on the Rotarodaccelerod test. Bar charts (mean±SEM) represent the maximum speed (rpm) of the animals on the rod, compared with the Kruskal-Wallis test. There were no statistically signi cant differences between groups. d Novel object recognition test. Bar charts (mean±SEM) represent the discrimination ratio (d2) between the new object and the one previously observed. Mann-Whitney's U test for independent samples (* comparison with zero) and one-way ANOVA followed by Tukey's test (# comparison of NC, NC+EE, HFHC and HFHC+EE d2 value) were used. NC and NC+EE groups were able to recognise the new object; whereas HFHC was not able to discriminate it, HFHC+EE showed a recovered novel object recognition ability. #p<0.05, **p≤0.010, ***p<0.001. e Spatial working memory test. Bar charts (mean±SEM)represent the average latency on the sample and retention trials. Two-tailed paired t-tests (* comparison with its respective sample) and one-way ANOVA followed by Tukey's test (# comparison between NC, NC + EE and HFHC + EE) were used. NC and NC + EE groups remembered the position of the platform in the retention trial; whereas HFHC was not able to remember it, HFHC + EE displayed recovered spatial working memory. #p < 0.05, **p ≤ 0.010.
The CCO values are expressed as mean ± SEM. The studied regions included the prefrontal cortex (prelimbic (PrL), infralimbic (IL) and cingulate (Cg) cortex), the dorsal striatum (dST), the ventral striatum (accumbens core (AcbC) and shell (AcbSh)), the thalamus (anteromedial nucleus (AMT), anterodorsal nucleus (ADT), and anteroventral nucleus (AVT)), the amygdala (central (CeA), basolateral (BLA) and lateral (LaA)), the dorsal hippocampus (dentate gyrus (DG), CA1 and CA3 areas), and the perirhinal (PRh) and entorhinal (Ent) cortices. Data were analysed through one-way ANOVA followed by Environmental enrichment has an effect on microbiota composition and bacterial metabolism. To assess the effect of the environmental enrichment (EE) implementation on the gut microbiota, we studied the gut microbiome pro le of the NC + EE and HFHC + EE groups, by comparing them with the pro les of the control groups (NC and HFHC) and with each other. We rst studied how EE affects microbial diversity within the communities by calculating the Chao1 (richness estimator) and Shannon's index (richness and evenness estimator). We con rmed signi cantly (p < 0.000) less bacterial diversity in the HFHC group compared to the NC group, as previously described in the characterization of the NASH animal model used [3], and we observed an increase in bacterial diversity in the NC + EE group compared to the NC group (p < 0.01). However, EE did not affect the bacterial diversity in the HFHC + EE group, which was signi cantly lower (also for evenness) than in the NC + EE group (p < 0.000) (Fig. 2a).
Next, we assessed the gut microbiota composition in each group of rats. When comparing the NC and HFHC groups, the pattern was substantially different (Additional le 1: Supplementary Figure S1), in agreement with previous studies [3]. The main differences were found in the greater abundance of Lactobacillaceae and Ruminococacceae in the NC groups compared to the HFHC groups, which harboured a greater abundance of Enterobacteriaceae, Bacteroidaceae, or Peptostreptococcaceae. Then, in order to explore how EE could affect the different phylotypes on the microbiota, we applied a linear discriminant analysis effect size (LEfSe) method at the family level to investigate the taxa most likely to explain differences in abundances across the groups. When we compared the four animal groups: NC, NC + EE, HFHC, HFHC + EE, the results identi ed statistically signi cant increased abundance of Micrococcaceae, Christensenellaceae, and Ruminococcaceae in the NC + EE group compared to the rest of the groups, and increased abundance of different families belonging to the Firmicutes phyla and Akkermansiaceae as the most differential microorganisms in the HFHC + EE when compared with the other three groups (Fig. 2b).
To explore the metabolic implications of the microbial differences observed after EE implementation, we analysed the main SCFAs and branched-chain fatty acids (BCFAs) derived from the bacterial metabolism. First, when comparing the HFHC group and the NC group, we con rmed the previous observations [3], where the main SCFAs (acetate, p < 0.01; propionate, p < 0.05; and butyrate, p < 0.01) showed statistically signi cant lower concentrations in the HFHC group. Then, the comparisons of NC + EE and HFHC + EE with their control groups (NC and HFHC, respectively) did not show any signi cant differences in acetate, propionate, butyrate, or valerate, even though a decreasing tendency was observed after EE. On the other hand, the concentrations of the BCFAs iso-butyric (p < 0.05) and iso-valeric (n.s.) were lower in NC + EE and HFHC + EE, compared to NC and HFHC, respectively (Fig. 2c).
Box-and-whiskers (median and IRQ range) represent comparisons of alpha-diversity of gut microbiota using Chao1 and Shannon indexes among the groups studied, compared using a one-way ANOVA followed by Tukey's test (* comparison with all the groups; # comparison with HFHC+EE group; $ comparison with HFHC and HFHC+EE groups). *#$p<0.01. b Gut microbiota composition. Taxonomic cladogram obtained from LEfSe analysis (LDA scores > 2 and signi cance of p<0.05 as determined by Wilcoxon's signed-rank test) showing bacterial taxa with differential abundance among the groups studied. Red indicates differential abundances in the HFHC group; green indicates differential abundances in the HFHC+EE group; blue indicates differential abundances in the NC group; purple indicates differential abundances in the NC+EE group. c SCFA and BCFAs. Bar charts (mean±SEM) represent comparison of the SCFA and BCFA levels (mM) compared using the Kruskal-Wallis test followed by Dunn's analysis (* comparison with HFHC and HFHC+EE groups; # comparison with NC+EE group; $ comparison with NC+EE and HFHC+EE groups). *#$p<0.05.
Akkermansia muciniphila, as a probiotic treatment, restores cognition de cits. Slight changes were found in the body weights of the groups throughout the 14 weeks of the administration of the diet because the HFHC + LGG group gained more weight than the NC group (p = 0.038); signi cant differences were observed across weeks (p < 0.001), as expected (Fig. 3a). The normal increase in weight was accompanied by the absence of locomotor de cits in all groups (p = 0.092; Fig. 3b). On the object recognition test, the experimental groups did not show signi cant differences between e1 and e2 (NC: p = 0.153; HFHC + PBS: p = 0.565; HFHC + LGG: p = 0.157; HFHC + AKK: p = 0.642) while performing the novel object recognition test. Indeed, the NC group was able to discriminate between a previously encountered object when d1 and d2 were statistically higher than zero (d1: p < 0.001; d2: p < 0.001). However, the HFHC + PBS group was not able to discriminate the new object (d1: p = 1.000; d2: p = 1.000). After the administration of probiotics (AKK or LGG), the HFHC + LGG group was still unable to recognise the new object (d1: p = 1.000; d2: p = 1.000), but the HFHC + AKK group was able to do so (d1: p = 0.010; d2: p = 0.010). When we compared the discrimination ratio between the groups, we found that NC and HFHC + AKK showed signi cantly higher d2 than HFHC + PBS and HFHC + LGG (p < 0.001). These results showed that HFHC + PBS displayed an object recognition impairment that was not recovered by the administration of LGG, but it was reversed by AKK (Fig. 3c).
Finally, when we evaluated the e cacy of probiotic administration on a relevant prefrontal dependent task such as spatial working memory, we found that the HFHC + LGG group was unable to execute the task (p = 0.486), as well as the HFHC + PBS group (p = 0.103). However, the HFHC + AKK group performed the task (p = 0.019) as correctly as the NC group, and they were able to remember the position of the platform because they showed a statistically signi cant lower retention latency compared to the sample trials (p = 0.003). When we compared the mean latencies in each trial between groups, we observed that, whereas the sample latencies were statistically similar, (p = 0.798), the retention latencies differed (p = 0.003) because they were signi cantly higher in HFHC + PBS and HFHC + LGG than in NC and HFHC + AKK (Fig. 3d). These results highlight the e cacy of A. muciniphila CIP107961 as a probiotic treatment to reverse impaired spatial working memory, compared to other probiotics such as L. rhamnosus GG.
Two-way ANOVA (Group x Week) was used to assess weight gain. Slight signi cant changes in body weight between groups were found throughout the 14 weeks of the administration of the diet because the HFHC+LGG group gained more weight than the NC group; signi cant differences across the weeks were also revealed. *p<0. represent the average latency on sample and retention trials. Two-tailed paired t-tests (* comparison with its respective sample) and one-way ANOVA followed by Tukey's test (# comparison between NC, HFHC+PBS, HFHC+LGG, and HFHC+AKK) were used. The NC group remembered the position of the platform on the retention trial; whereas HFHC+PBS and HFHC+LGG were not able to remember it, HFHC+AKK displayed recovered spatial working memory. *#p<0.05, **p≤0.010.
Akkermansia muciniphila restores brain metabolic activity to normal. When we explored the brain metabolic activity underlying these cognitive changes, we rst found that the HFHC + PBS group showed signi cantly lower levels of CCO activity than the NC group in the prefrontal cortex (p < 0.001), dorsal and ventral striatum (p < 0.001), amygdala nuclei such as CeA (p = 0.004), BLA (p = 0.014), hippocampus (p ≤ 0.007), and PRh (p < 0.001; Table 2).
When the animals were treated with A. muciniphila CIP107961, their brain metabolic activity equalled that of the NC group in most of the regions previously affected by HFHC, such as the prefrontal cortex, dorsal striatum, amygdala, hippocampus, and perirhinal cortex. In line with this, HFHC + AKK showed an increased CCO value in the dorsal striatum (dST, p < 0.001), central amygdala (CeA, p = 0.004), and basolateral amygdala (BLA, p = 0.014), compared to HFHC + PBS. However, the HFHC + AKK group also showed lower CCO levels than the NC group in the infralimbic cortex (IL, p < 0.001), ventral striatum (p < 0.001), and dentate gyrus (DG, p < 0.001).
When comparing the two probiotic treatments, we observed that the HFHC + AKK group showed statistically signi cant higher CCO values than the HFHC + LGG group in the prefrontal cortex (p < 0.001) and dST (p < 0.001). These results point out that the decreased CCO shown by HFHC + PBS cannot be recovered with LGG administration, whereas the opposite effect was found when applying AKK. More importantly, A. muciniphila CIP107961 was able to reverse the HFHC-associated decrease in CCO activity in most of the brain regions previously affected by the HFHC diet. Table 2 Brain oxidative metabolism in experimental groups subjected to PBS, LGG and AKK. The CCO values are expressed as mean ± SEM. The studied regions included the prefrontal cortex (prelimbic (PrL), infralimbic (IL) and cingulate (Cg) cortex), the dorsal striatum (dST), the ventral striatum (accumbens core (AcbC) and shell (AcbSh)), the thalamus (anteromedial nucleus (AMT), anterodorsal nucleus (ADT), and anteroventral nucleus (AVT)), the amygdala (central (CeA), basolateral (BLA) and lateral (LaA)), the dorsal hippocampus (dentate gyrus (DG), CA1 and CA3 areas), and the perirhinal (PRh) and entorhinal (Ent) cortices. Data were analysed through one-way ANOVA followed by Probiotics do not induce major rearrangements in the faecal microbiota. To assess the effect of two different probiotics on the NASH-associated cognitive disturbances, we rst analysed the effect of the administration of the strains by oral gavage on the gut microbiota in the gavage feeding groups (HFHC + AKK, HFHC + LGG, and HFHC + PBS [control group]), also including the NC group (no gavage) as external control. Miseq sequencing produced an average of ~ 63,000 ltered partial sequences per sample, and it showed the biggest differences at the diversity and compositional levels between the NC group and the other three NASH groups (Fig. 4a, Additional le 1: Supplementary Figure S2a), con rming the strong effect of diet (control vs. HFHC diet). The analyses of the SCFAs followed the same trend (Additional le 1: Supplementary Figure S2b). These results are in concordance with our previous observations in the NASH [3] animal model. Thus, by focusing more on the speci c effect of the two probiotics on the gut microbiota, we observed that the administration of both bacteria led to a unique change in the microbial composition at the phylum level, by signi cantly decreasing Bacteroidetes, compared to placebo administration (PBS) (p = 0.016 HFHC + LGG vs. HFHC + PBS; p = 0.018 HFHC + AKK vs. HFHC + PBS). At lower taxonomical levels, applying a linear discriminant analysis effect size (LEfSe) method, we observed that only a few families and genus suffered differential changes in their abundance depending on the probiotic administration.
In the HFHC + AKK group, we observed higher abundance mainly in Faecalibacterium, Prevotella 9, and the Ruminococcus UCG005 group (Fig. 4b). Changes in the concentration of the Lactobacillus and Akkermansia genus were validated by qPCR, con rming higher levels of the Lactobacillus genus in the HFHC + LGG group and no differences in the Akkermansia genus across the groups (Fig. 4c). On the other hand, the Chao1 index showed statistically higher alpha-diversity in the HFHC + AKK group when compared with the HFHC + PBS group (p < 0.05); however, the Shannon index did not show any signi cant differences among the groups (Fig. 4d). These results indicate that probiotics can affect microbial diversity, but they did not induce major rearrangements of the faecal microbiota.
Average relative abundance of gut microbiota at the family level from NC, HFHC+PBS, HFHC+LGG, and HFHC+AKK groups. Bacterial taxa representing less than 0.5% of the total abundance are included in Others. b Gut microbiota differences. Results of LEfSe analysis (LDA scores > 2 and signi cance of p<0.05 as determined by Wilcoxon's signed-rank test) showing signi cantly different taxa among the HFHC+PBS, HFHC+LGG, HFHC+AKK groups. Red indicates differential abundance in the HFHC+AKK group; green indicates differential abundance in the HFHC+LGG group; blue indicates differential abundance in the NC+PBS group. Bacterial taxa representing less than 0.5% of the total abundance are included in Others. c qPCR concentration. Bar charts (mean±SEM) represent comparison of Lactobacillus and Akkermansia genus concentration analysed by qPCR (log 10 cells/g faeces), compared using the Kruskal-Wallis test followed by Dunn's analysis (*p<0.05, comparison with all the groups). d Bacterial diversity. Box-and-whiskers (median and IRQ range) represent comparison of alpha-diversity of gut microbiota using Chao1 and Shannon indexes among groups, compared using a one-way ANOVA followed by Tukey's test (*p<0.05, comparison with HFHC+PBS group).
Finally, the levels of the main SCFAs did not show differences between the probiotic groups and the HFHC + PBS group, with the only exception of acetate, which was lower in the probiotic groups (p < 0.01, HFHC + LGG vs. HFHC + PBS; p < 0.05, HFHC + AKK vs. HFHC + PBS) (Additional le 1: Supplementary Figure S2c).
Cognitive recovery is equally improved by environmental enrichment and Akkermansia muciniphila. No locomotor dysfunction has been found in any experimental group (p = 0.310; Fig. 5a). Both treatments, environmental enrichment and the administration of A. muciniphila CIP107961, were able to improve novel object recognition in HFHC animals. When we compared the discrimination ratio (d2) between HFHC + PBS, HFHC + EE, and HFHC + AKK to observe if there was a differential effect, we found that HFHC + EE and HFHC + AKK presented a higher d2 value than HFHC + PBS (p = 0.013), but the d2 value did not differ signi cantly between the HFHC + EE and HFHC + AKK groups. Thus, environmental enrichment and AKK administration were equally successful in reversing the novel object recognition impairment caused by the HFHC diet (Fig. 5b).
Regarding spatial working memory, we previously showed that EE and AKK improved the performance of HFHC animals. When comparing the mean latencies of HFHC + PBS, HFHC + EE, and HFHC + AKK, we observed that, whereas there were no statistically signi cant differences in the sample trials (p = 0.647), both HFHC + EE and HFHC + AKK showed lower retention latencies than HFHC + PBS (p = 0.004). However, retention latencies between HFHC + EE and HFHC + AKK were not signi cantly different. Thus, EE and AKK administration were both equally effective in restoring spatial working memory in the HFHC condition (Fig. 5c).
Bar charts (mean±SEM) represent the maximum speed (rpm) of the animals on the rod, compared using the Kruskal-Wallis test. There were no statistically signi cant differences between groups. b Novel object recognition test. Bar charts (mean±SEM) represent the discrimination ratio (d2) between the new object and the one previously observed. Mann-Whitney's U test for independent samples (* comparison with zero) and the Kruskal-Wallis test followed by Tukey's analysis (# comparison of HFHC+PBS, HFHC+EE and HFHC+AKK for d2 value) were used. HFHC+EE and HFHC+AKK showed improved recognition of a novel object in comparison with HFHC+PBS animals, but no differences were found between them.
#p<0.05, **p≤0.010. c Spatial working memory test. Bar charts (mean±SEM) represent the average latency on the sample and retention trials. Two-tailed paired t-tests (* comparison with its respective sample) and one-way ANOVA followed by Tukey's test (# comparison between HFHC+PBS, HFHC+EE and HFHC+AKK) were used. HFHC+EE and HFHC+AKK displayed improved performance on the spatial working memory task, in comparison with HFHC+PBS, but no differences were found between them.
Environmental enrichment and Akkermansia muciniphila have differential effects on brain metabolic activity. Next, we compared CCO values in the HFHC + PBS, HFHC + EE, and HFHC + AKK groups in order to discover whether there was a differential treatment effect on brain metabolism. We rst observed that HFHC + EE displayed lower CCO levels than HFHC + PBS in the accumbens core (AcbC, p < 0.001), thalamus (p < 0.001), lateral amygdala (LaA, p = 0.002), DG (p < 0.001), and CA3 (p = 0.001). However, the opposite results were found in the HFHC + AKK group, where the CCO values were higher than in the HFHC + PBS group in the cingulate cortex (Cg, p = 0.002), dST (p < 0.001), and BLA (p = 0.003).

Region
HFHC + PBS, we observed that the probiotic did not induce a major community-wide compositional change, compared to the corresponding control group (HFHC + PBS). Only some previously mentioned changes were observed in the HFHC + AKK group.
Average relative abundance of prevalent microbiota at the family level from HFHC+EE, HFHC+PBS, HFHC+AKK groups. Bacterial taxa representing less than 0.5% of the total abundance are included in Others.
Regarding the levels of the main SCFAs, only propionate showed different concentrations between HFHC + EE and the HFHC + AKK and HFHC + PBS groups; and in the BCFA, iso-butyrate reached higher concentrations in the HFHC + AKK group, compared to the placebo group and the HFHC + EE group (Additional le 1: Supplementary Figure S3).

Discussion
In this study, we described cognitive, brain metabolism, and microbiota alterations associated with highfat and high-cholesterol consumption. In addition, we clearly showed that environmental enrichment and A. muciniphila CIP107961 restore cognitive dysfunction. Furthermore, we revealed that cognitive improvement was due to differential effects of environmental enrichment and this strain of A. muciniphila on brain metabolism and gut microbiota. Finally, we discovered that restored cognitive function was associated with the administration of A. muciniphila CIP107961, but not L. rhamnosus GG, which may be clinically relevant when selecting probiotics for treating HFHC-derived pathologies.
Environmental enrichment (EE) has been described as motor and sensorial stimulation that is able to modulate the physical and cognitive status [18]. Moreover, EE has been found to have brain repairing therapeutic effects in several neuropsychiatric diseases [17][18][19]. Our results revealed that the HFHC group presented cognitive de cits. Speci cally, novel object recognition impairment was restored by EE, in accordance with de Souza et al. [41]. Our data extend these ndings demonstrating that spatial working memory alterations, which we have previously described [3], were also improved by the EE. The EE's effect on the HFHC group was accompanied by decreased levels of brain metabolism in the prefrontal cortex, striatum, and perirhinal cortex. Moreover, other brain regions such as the thalamus, the amygdala, and the hippocampus showed the same pattern. The prefrontal cortex sends projections to the perirhinal and entorhinal cortices [42] and is in reciprocal connection with the striatum [43] and the thalamus [44]. This network supports spatial working memory and novel object recognition abilities [34,42,44,45]. Therefore, any improvement in this network will be re ected in an improvement on both behavioural tasks, as we demonstrated.
One additional nding is the decrease in brain metabolic levels when introducing the EE, which was clearly observed not only in the HFHC + EE group, but also in the NC + EE group. The NC + EE group displayed lower CCO values than the NC group in the prefrontal cortex, striatum, thalamus, and hippocampus, whereas HFHC + EE showed lower levels than HFHC in the striatum, thalamus, amygdala, and hippocampus. These results are in line with previous data by Tong et al. [46] showing that exercise decreased the hippocampal expression of genes involved in mitochondrial metabolic processes, including cytochrome c-oxidase. In fact, physical activity and EE have been found to regulate metabolic and redox activity, resulting in a bene cial modulation of brain function [47]. Likewise, our results support the overall effect of EE on brain metabolism through the reduction in metabolic levels in both healthy and HFHC conditions.
When we assessed the gut microbiota composition of the HFHC and NC groups, we con rmed previous observations in the same HFHC animal model [3], such as lower diversity and important changes at the compositional level. At the phylum level, the HFHC group showed higher levels of Proteobacteria and lower levels of Firmicutes, alterations that were translated into lower taxonomical levels. Changes in the gastrointestinal microbial population correlate with speci c patterns of faecal metabolites [48], and we observed that the faecal metabolite pro le also showed alterations with respect to the control group, with lower concentrations of the main SCFAs in the faecal water of the HFHC group. Physical exercise appears to be able to modify microbiota [49]. In our experimental model, the EE did not induce a signi cant community-wide compositional change when applied to the NC group, although the bacterial diversity richness increased in the NC + EE group. However, EE induced more modi cations at the taxonomical level in the HFHC group, decreasing Proteobacteria and increasing Firmicutes. These changes were accompanied by decreased iso-butyrate levels, but without any change in the bacterial diversity. This is consistent with the ndings of Allen et al. [50], who showed that the effect of EE was dependent on the health status of their experimental groups because obese subjects showed a different response to exercise from that of healthy controls in terms of gut microbiota. In our case, EE did not affect the gut microbiota of the healthy control group; however, the HFHC + EE displayed a signi cant modi cation of the microbiota composition, illustrating the complex association between microbiota and health.
Gut microbiota changes provoke a dysregulation of the enteric nervous system, which leads to a breakage of the gut-brain axis and ultimately to neural de cits [10,51]. In addition, gut dysbiosis has been associated with mitochondrial dysfunction through the delivery of metabolites that can affect reactive oxygen species and ATP production [51], suggesting a potential mechanism for neural damage.
The NC, NC + EE, and HFHC + EE groups did not show the cognitive impairment displayed by the HFHC group; they manifested a better capability for novel object recognition and spatial working memory.
Hence, these results demonstrated that EE is able to target the gut-brain axis, leading to the modulation of brain oxidative metabolism, which has an impact on cognitive amelioration.
In a previous pilot study, we found that these gut microbiota changes in the HFHC + EE group were re ected in a signi cant increase in Akkermansia levels compared to the HFHC group. Moreover, the comparison of the NC and HFHC groups indicated a reduction in the relative proportions of lactobacilli in the HFHC group. These two observations prompted us to evaluate the impact on cognition of the administration of two strains of the cited microbial groups, L. rhamnosus GG and the potentially novel probiotic A. muciniphila CIP107961, for which bene cial effects have been recently reported [23][24][25].
The administration of LGG was not effective in ameliorating the cognitive dysfunction in the HFHC group, showing similar results to those found in the control group (HFHC + PBS). These results are in line with a previous study by Kelly et al. [52] in healthy humans, where LGG was not able to improve cognitive function, measured as visuospatial memory performance, attention switching, rapid visual information processing, and emotion recognition. At the compositional level, the gut microbiota of the HFHC + LGG group suffered slight modi cations with regard to the placebo group (HFHC + PBS), with a considerable increase in lactobacilli and a lower concentration of acetate in the faecal water. In contrast to the results observed with LGG, the HFHC animals receiving A. muciniphila CIP107961 displayed correct performance on novel object recognition and spatial working memory, which means that this microorganism is effective in reversing HFHC-associated cognitive defects. Previous studies had found that A. muciniphila was effective in metabolic disorders derived from HFHC diets by reversing endotoxemia [53,54]. In this way, it has been observed that A. muciniphila can also lower plasma LPS levels [26] through the reinforcement of the gut barrier. Moreover, several studies have shown that the administration of this microorganism, either dead or alive in animal models, reduced levels of different hepatic enzymes considered to be markers of in ammation and hepatic steatosis [26,54,55]. These enzymes have been found to be increased in our HFHC model [56], along with the presence of hepatic steatosis [3]. In accordance with these physiological bene ts, for the rst time, our results show a potential role of A. muciniphila in improving cognitive function.
We found that A. muciniphila CIP107961 not only restored metabolic levels in the amygdala and the dorsal striatum and increased the CCO values compared to HFHC + LGG in this latter area, but higher CCO values were also found in the prefrontal cortex. Moreover, the group treated with A. muciniphila CIP107961 displayed similar values to the NC group in regions previously affected by HFHC, such as the prefrontal cortex, striatum, hippocampus, and perirhinal cortex. When we addressed the L. rhamnosus GG administration, no changes were found in the CCO values in any of the measured regions compared to HFHC + PBS, which is consistent with the inability of this group to improve their cognitive function.
Previous studies have suggested that A. muciniphila is associated with endocannabinoid regulation because its administration led to increased intestinal levels of endocannabinoids [53], and, in turn, the blockage of endocannabinoid receptor CB1 increased levels of this microorganism [57]. The endocannabinoid system can be modulated along the gut-brain axis through vagal control [58], and CB1 receptors, which are densely located in the striatum and hippocampus [59,60], have been suggested to be present in the mitochondrial membrane, so that they are able to control cellular respiration and energy production [61]. Therefore, this aforementioned endocannabinoid regulation of A. muciniphila in the intestine could be extending its effects along the gut-brain-axis, ultimately manifested as mitochondrial activity upregulation, which could be causing the cognitive improvement.
An interesting nding of our study is the decreased CCO levels when comparing the HFHC + PBS and NC groups in the prefrontal, hippocampus, amygdala and perirhinal cortices. These results could be re ecting a stress response [62,63] potentially induced by gavage administration [64]. The gut microbiota study after A. muciniphila CIP107961 administration did not show major rearrangements, with only slight modi cations and a modest increase in bacterial diversity when compared with the microbiota of animals receiving PBS or LGG. This suggests that major microbiota changes are not involved in the observed effects, which is in line with previous studies carried out in other animal models [24,54] or in humans [26] that reported bene cial effects of this microorganisms without observing major changes in the gut microbiota composition. However, it is important to note that more in-depth studies may be needed to fully elucidate this issue because in this study we rely on 16S rRNA gene-based microbiota pro ling, and, therefore, we do not have a complete overview of the total intestinal metagenome.
We nally wanted to compare the two successful treatments, environmental enrichment and A. muciniphila CIP107961 administration, in order to elucidate their differential effects along the gut-brain axis. We rst showed that each treatment produces an opposite effect on brain metabolism; whereas EE caused an overall brain metabolic decrease, AKK restored the activity. Second, in a similar manner, HFHC + EE and HFHC + AKK animals showed clear differences in their microbiota patterns. Whereas the HFHC + AKK group showed only minor differences with regard to the corresponding experimental control group (HFHC + PBS), the differences with the HFHC + EE group were larger, including higher levels of Bacteroidetes and Proteobacteria at the phylum level. However, as previously discussed, the existence of differences in the basal microbiota between these two groups keeps us from drawing rm conclusions.
Indeed, these differences suggest that the observed effects of both EE and A. muciniphila administration on the reversion of NASH-associated cognitive impairments are independent from a de ned overall microbiota composition. In this regard, some studies have identi ed speci c A. muciniphila proteins, such as the membrane protein Amuc_1100, as being able to mediate some of these bene cial effects [55].
Finally, we want to point out the observed changes in the SCFAs and their metabolites, which showed a decrease in iso-butyrate and acetate when the animals were exposed to environmental enrichment or treated with probiotics, respectively. In this regard, some microorganisms, such as bacteroides or Clostridia members, are able to produce BCFAs from the proteolytic metabolism of branched-chain amino acids. The higher levels of Bacteroidaceae or Peptostreptococcaceae observed in the HFHC group could be related to the higher levels of BCFAs observed, as found in other diseases related to the gut-brain axis, such as obesity [65], Rett syndrome [66], or anorexia nervosa [67]. On the other hand, there is evidence in mice that acetate can alter the levels of glutamate, glutamine, and GABA [68], and some studies have shown that SCFAs and their metabolites can stimulate vagus nerve signaling [69,70]. Moreover, other studies have correlated changes in SCFAs and iso-butyrate with positive effects on behaviour through the microbiota-gut-brain axis [8]. Hence, these results could be highlighting the contribution of SCFAs and BCFAs to NASH-cognitive improvement.

Conclusions
In conclusion, the microbiota and cognition are intimately connected through the gut-brain axis, and in HFHC pathologies they can be in uenced by environmental enrichment and A. muciniphila CIP107961 administration. Cognitive improvement was accompanied by changes in brain metabolic activity and gut microbial composition analysis, pointing to speci c microbiota targets for intervention in diet-induced pathologies. However, some mechanisms other than major changes in microbiota composition and the combined effect of environmental enrichment and A. muciniphila administration, which we identi ed in this study, may also be biologically relevant and will need to be investigated in future studies due to their relative contributions to the selection of effective treatments for patients.

Declarations
Ethics approval The procedures and manipulation of the animals used in this study were carried out according to the

Consent for publication
Not applicable Availability of data and materials The datasets generated during and analysed during the current study are available from the corresponding author on reasonable request.