The Many Ages of Microbiome–Gut–Brain Axis

Frailty during aging is an increasing problem associated with locomotor and cognitive decline, implicated in poor quality of life and adverse health consequences. Considering the microbiome–gut–brain axis, we investigated, in a longitudinal study, whether and how physiological aging affects gut microbiome composition in wild-type male mice, and if and how cognitive frailty is related to gut microbiome composition. To assess these points, we monitored mice during aging at five selected experimental time points, from adulthood to senescence. At all selected experimental times, we monitored cognitive performance using novel object recognition and emergence tests and measured the corresponding Cognitive Frailty Index. Parallelly, murine fecal samples were collected and analyzed to determine the respective alpha and beta diversities, as well as the relative abundance of different bacterial taxa. We demonstrated that physiological aging significantly affected the overall gut microbiome composition, as well as the relative abundance of specific bacterial taxa, including Deferribacterota, Akkermansia, Muribaculaceae, Alistipes, and Clostridia VadinBB60. We also revealed that 218 amplicon sequence variants were significantly associated to the Cognitive Frailty Index. We speculated that some of them may guide the microbiome toward maladaptive and dysbiotic conditions, while others may compensate with changes toward adaptive and eubiotic conditions.


Introduction
Aging is a natural process affecting all living organisms, and it is characterized by a deterioration in physiological processes, leading to a higher probability to develop several disorders, i.e., cancer, metabolic, cardiovascular, and neurodegenerative diseases [1,2]. Indeed, aging is characterized by a decline in cognitive and locomotor functions [3]. Among different cognitive performances, the recognition memory has been described as a major component of mammalian and human personality [4][5][6], and it is early and heavily affected during aging.
In the elderly, frailty is defined as a clinical state with an increased vulnerability to stressors, thus exposing the organism to negative health-related outcomes, in the absence of recognized pathologies [7]. Frailty is a multisystem dysregulation leading to decreased physiological reserve [8,9]. Thus, frailty is related to aging, but it does not reflect chronologic age, showing big heterogeneity among subjects [10]. The age-related frailty is principally bacterial taxa. In parallel, we monitored the Cognitive Frailty Index during aging by using spontaneous behavioral tests for evaluating recognition memory. Finally, we recognized the specific changes in microbiome composition due to the cognitive frailty, and we tried to describe them from a functional point of view.

Mice and Behavioral Tests
Fourteen wild-type male mice (C57BL-6J) were obtained from Charles River, Italy. The mice were acclimated to their environmental conditions for one month before starting experiments, which were conducted in the Animal Care Facility at the University of Pavia. The mice were individually housed in plastic cages with an automatically controlled lightdark cycle: the dark period lasted from 07:00 to 19:00 and the light period was from 19:00 to 07:00. Water and food were furnished ad libitum. All experimental protocols and animal handling were carried out in strict conformity with European Council Directive 2010/63/EU and with the guidelines set by the Ethics Committee of Pavia University (Ministry of Health, License number 774/2016-PR).
The murine cognitive performance was studied accessing the knowledge component of recognition memory [6], tested through two spontaneous behavioral tests which consisted of (i) recognizing a new object (namely, novel object recognition (NOR) test) in an open arena or (ii) in exploring a new environment (namely, emergence test) in non-stressful conditions. Emergence and NOR tasks were carried out in accordance with Brandalise et al., 2017 [27]. Concerning the emergence test, the analyzed parameters were latency for the first exit(s), number of exits, and time of exploration(s). Regarding the NOR task, the parameters of interest were the discrimination capabilities (discrimination index, DI) for number and time of approaches. For all selected parameters, we obtained the corresponding Cognitive Frailty Index (FI), then the Cognitive FI related to the test (emergence or NOR), and finally the overall Cognitive FI, as previously reported [3].
In detail, spontaneous behavioral tests in mice were performed at five experimental time points: T0 = 11, T1 = 14, T2 = 17, T3 = 20, and T4 = 21.5 months of age. Notably, T0 and T1 belonged to adulthood, T2 to reproductive senescence, while T3 and T4 occurred during senescence. At the same experimental times, animals were weighed, and the fecal stool samples were collected and stored at −80 • C. The experimental design is summarized here below in Figure 1.  5 months of mice age, corresponding to T0, T1, T2, T3, and T4, respectively. It has to be noted that T0 and T1 belonged to adulthood, T2 to reproductive senescence, and T3 and T4 to senescence. At each time point, behavioral tests were performed and stool samples were collected).

Bacterial DNA Extraction and 16s rRNA Sequencing
Total microbiota genomic DNA in the mice stools was extracted by using a QIAamp DNA stool mini kit (Qiagen, Dusseldorf, Germany) in accordance with the manufacturer's instructions. DNA quantification was performed by using Qubit FluorometerTM (Invitrogen, Molecular Probes, USA). Libraries for 16S rDNA amplicons sequencing of the V3 and V4 regions were obtained by using PCR primers containing a barcode (V3 Forward: 5 -CCTACGGGNGGCWGCAG-3 ; V4 Reverse: 5 -GACTACHVGGGTATCTAATCC-3 ).
Furthermore, for preparing the amplicons for the sequencing by MiSeq Illumina, we used the specific forward (V3 F: 5 -TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG-3') and reverse (V4 R: 5'-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG-3') adapters. Finally, we performed the PCR analysis by using Bio-Rad MJ Mini Personal Thermal Cycle, and PCR amplicons were sequenced by MiSeq Illumina, relying on the BMR Genomics SRL of Padova.

Illumina Data Processing and Microbiota Characterization
The raw paired-end FASTQ reads were imported into the Quantitative Insights Into Microbial Ecology 2 program (QIIME2, ver. 2020.2.01) [28] and demultiplexed using the native plugin. The Divisive Amplicon Denoising Algorithm 2 (DADA2) [29] was used to quality filter, denoise, and mergepair the data and remove chimeric sequences.
The resulting amplicon sequence variants (ASVs) with less than a 50× coverage were discarded from further analyses. The classification of the obtained ASVs was run using the feature-classifier plugin [30], implemented in QIIME2 against the SILVA SSU non-redundant database (138 release) [31], adopting a consensus confidence threshold of 0.7.

Data Analysis and Statistics
All data are expressed as mean ± SEM. The analysis on the bacterial diversity, and the corresponding figures, was performed using the phyloseq R package [32]. Microbiota diversity was described in terms of within (alpha) and between (beta) sample diversities. The Shannon diversity index (SDI) and Faith's phylogenetic distance (PD) [33] alpha diversity metrics were calculated to estimate the variation of bacterial diversity at the different time points. Values were compared using the pairwise Wilcoxon rank-sum test (WRST) [34] was used pairwise to determine whether the value of a specific metric (Shannon index and Faith's PD) changed significantly between different time points.
Beta diversity was estimated with quantitative distance metrics using the diversity function in the phyloseq R package. We estimated the Bray-Curtis dissimilarity indices by sampling 10,000 reads per sample [32] based on estimated rarefaction curves (see Supplementary Figure S1).
Variance partitioning and significant Cognitive FI variation over time were determined by performing linear mixed effects (LME) models with the lme4 R package [34].
Variance partitioning and significance of bacterial communities' diversity variation over time and with respect to the Cognitive FI were determined by performing PER-MANOVA test using the Adonis function implemented in the vegan R package. To overcome repeated measures, the variable that describes the mice was defined as strata.
Canonical analysis of principal coordinates (CAP) was computed using the capscale function from the vegan R Package [35,36]. To highlight the relationship between changes in community tolerance and shifts in community composition, a constrained ordination was performed by distance-based redundancy analysis with time, with Cognitive FI as the constraint variable. In addition, we used the unconstrained ordination method (principal coordinate analysis) to visualize patterns in bacterial community compositions. The significance of constraint variables was tested with a permutation test (number of permutations = 10,000) using the ANOVA function implemented in the vegan R package.
The structure of microbial communities was explored by the non-multidimensional scaling (NMDS) ordination approach [37].
The DESeq2 R package [38] was used to identify the bacteria with the most significant changes in ASV differential abundance considering the Cognitive FI. Improvements to the stability and dispersion of the counts (variance) were required before it was possible to calculate the differential abundances for the different species present in the samples being compared. To this end, we used the estimated size factors function in DESeq2 to transform the stabilization of the variance. The differential abundances were measured with the log2foldchange value, and the different conditions were compared using the Wald test with the Benjamini-Hochberg correction (Q parameter = 0.1, FDR < 10%). The differential abundance measurements were statistically significant if the adjusted p-value was <0.05.
The metacoder R package was used to generate tree plots [39]. All the graphical plots were generated by the ggplot2 R package [40]. The method for calculating ellipses is implemented in the ggplot2 package in the stat ellipse function based on John Fox and Sanford Weisberg (2011) [41].
The sequences generated in this study were deposited into the EMBL-EBI database in the PRJEB54046 project.

Sequencing Data Results
We performed a longitudinal study on fourteen mice sampled five times (from T0 to time T4). The gut microbiome of those 70 samples was examined by sequencing of the bacterial 16S rRNA gene. After quality control analysis, sequences of 70 libraries resulted in 1,748,748 million sequence reads, ranging from 564 to 43,360 reads, with a median of 25,801 for a sample (see Supplementary Figure S2). No reads were reported in negative controls. A total of 1458 ASVs were identified.

Aging Affects Overall Gut Microbiome Composition
Stool profiling revealed large microbiome composition differences over time during the mice lifespans. The alpha diversity (microbial diversity within the sample) changed significantly during physiological aging, but Shannon diversity index (SDI) and Faith's phylogenetic distance (PD) exhibited opposite behavior ( Figure 2). There was a significant increase in terms of biodiversity observed with Shannon index between time T1 vs. T2 (Wilcoxon rank-sum test (WSRT), p = 0.007) and between T2 vs. T3 (WSRT, p = 0.011); on the contrary, any significant variation was determined between T0 and T1, as well as between T3 and T4 (Tables 1 and S1). Supplementary Figure S3 reports the Simpson index variation during aging, which showed the same trend as the Shannon index. Concerning Faith's PD, which considers phylogenetic information, we observed a decreasing trend, even though a significant variation was perceived only between T3 and T4 (WSRT, p = 0.0002). As regards the other time points, any significant change was measured (Tables 1 and S1).
To obtain a complete picture of the variation over time of the measured diversity, we estimated the magnitude of change among successive time points by considering the beta diversity, a measure of similarity or dissimilarity of different microbiome communities. As shown in Figure 3 (NMDS based on Bray-Curtis distance matrix; stress value = 0.13), the murine microbiomes were similar at each experimental time but changed over the mice lifespan. This fact determined the identification of five consequential clusters specific for each time ( Figure 3). This longitudinal effect of beta diversity modification increased further and further over time, creating a large gap between the T0 and the T4 group (see results in Tables 2 and S2).

Detailed Microbiome Composition and Aging Distribution
A total of 1458 ASVs were identified throughout the whole experimental time. Of these, 76% was assigned at least at the phylum level, and specifically, we found 13 phyla, 20 classes, 54 orders, 76, families, and 145 genera.
We recorded a continuum in the phyla taxonomic level during aging. During the entire aging process, the most represented phylum was Bacteroidota followed by Firmicutes ( Figure 4A). The greatest changes during aging were the disappearance of the Verrucomicrobiota phylum from 13.84% at T0 to less than 0.01% at T4 and the decrease in Deferribacterota from 4.9% at T1 to 1.2% at T4. At the family and genus taxonomic levels, the greatest variations in microbiome composition were observed during aging ( Figure 4B,C). The most represented families during the entire physiological aging process were Tannerellaceae and Lachnospiraceae, of Bacteroidota and Firmicutes phyla, respectively. Some families underwent a strong variation during physiological aging. In particular, the Akkermansiaceae (Verrucomicrobiota phylum) decreased from 13.86% at T0 to 3.7% at T2 and almost disappeared at T3 and T4. On the contrary, the Muribaculaceae (Bacteroidota phylum) appeared at T3 at a percentage of 5.3% and increased even more to 14.4% at T4.
The most represented genera during the mice lifespan were Parabacteroides (Bacteroidota phylum; Tannerellaceae family) and Lachnospiraceae_NK4A136_group (Firmicutes phylum; Lachnospiraceae family) which did not undergo statistical fluctuations ( Figure 4C). The genera Akkermansia (Verrucomicrobiota phylum; Akkermansiaceae family) disappeared between adulthood and senescence ( Figures 4C and 5A, respectively), mirroring the variation at the phylum and family level. During senescence (at T3), we witnessed a statistically significant increase in the genus Clostridia_vandinBB60 (Firmicutes phylum; Clostridia family) (from 0.043% at T0 to 4.2% at T4) ( Figures 4C and 5B, respectively), paralleled by a similar sized increase of the genus Alistipes (from 4.6% at T0 to 9.8% at T4) ( Figures 4C and 5C, respectively), representing the entire Rikenellaceae family (Bacteroidota phylum). Furthermore, Muribaculaceae genera (Bacteroidota phylum, Muribaculaceae family) appeared in senescence (at T3 5.17% and at T4 13.57%, Figures 4C and 5D, respectively). Finally, we perceived a slight and continuing non-significant trend of growth during aging for Colidextribacter (Firmicutes phylum) and Clostridia UCG-014 genera (Firmicutes phylum, Clostridia family) ( Figures 4C and 5E,F, respectively). To investigate in more detail the community structure, we displayed our data as a heat tree map, where it is possible to follow all the taxonomic levels from phyla to species. Furthermore, in the heat tree map, the proportions were subjected to statistical testing. Based on the Wilcoxon rank-sum test, Figure 6 shows the heat tree map describing the different taxa at each experimental time during aging. The size and color of nodes and edges correlate with the abundance of bacteria at each experimental time. The heat tree map describes both the relative changes in sequential time and the whole change between adulthood (T0) and senescence (T4) (Figure 6). Figure 6. A differential heat tree based on the Wilcoxon rank-sum test, indicating which taxa were more abundant in each experimental time. Phyla, classes, orders, families, and genera are represented. Node label is the taxon name, node size is the number of ASVs, and node color is the abundances of the indicated phylum, class, order, family, or genus. A taxon colored brown was more abundant in the time points colored in brown and a taxon colored in green was more abundant in the time points colored in green, as reported in the legend. The tree differential plots were generated using the metacoder R package. (Number of animals: T1 n = 14, T2 n = 12, T3 n =14, and T4 n =13).

Aging and Cognitive Frailty Index on Microbiome Composition
We investigated the knowledge component of the recognition memory, which significantly worsened during aging, as previously reported [3]. In particular, by using specific spontaneous behavioral tests (emergence and NOR tasks), we evaluated the mice's capabilities in distinguishing an environment (emergence task) or an object (NOR task) as familiar or novel, during the aging process. Altogether, the overall Cognitive FI (used for evaluating the cognitive impairment) values significantly increased during physiological aging in wild-type mice (Figure 7), indicating that physiological aging was accompanied by cognitive decline. Remarkably, there was a significant increasing effect of time on Cognitive FI (See Supplementary Table S3 for LME model details), confirming the direct correlation between the Cognitive FI and the aging process. The next question to address was the possible identification of a frailty microbiome signature during aging. Therefore, we investigated the relationship between the Cognitive Frailty Index and the alpha diversity, evaluated by Shannon (see supplementary Figure S4) and Faith's PD indices (Figure 8). Faith's PD index, but not Shannon index, showed a statistically significant relationship with the Cognitive FI in the performed linear models. In particular, in relation to an increase in the Cognitive FI value (coeff = 6.345, Figure 8, see Table S4 and Figure S4 in Supplementary Materials), a negative correlation in the estimated biodiversity occurred based on Faith's PD index. This result was also supported by partial canonical analysis of principal coordinates (CAP analysis, Figure 9). The CAP scale revealed that an advanced state of senescence and a higher Cognitive FI value explained the largest proportion (18.2%, p < 0.001) of the variation in beta diversity (Figure 9). However, the estimated LME models did not report a significant effect (p value > 0.05) of the Cognitive FI variation on the beta diversity of the microbiomes (see Table S5 in Supplementary Material), probably due to the low sample number.
Furthermore, we applied the DESeq2 algorithm to test for differential abundance in bacterial groups that changed with respect to the Cognitive FI. We found 218 ASVs significantly linked to the variation in the Cognitive FI. Interestingly, the heat map returned different clusters (T0 and T1 versus T2, T3, and T4) in microbiome composition during the mice lifespan, confirming the beta diversity data ( Figure 10). Figure 10. The heat map shows the distribution of the abundances of 218 ASVs whose variation was statistically significant in relation to the variation of the Cognitive FI. The analysis was performed using Deseq2 R package and the heat map was generated using phetamap R package. The variations in terms of abundance are indicated using the coloring scale in the legend. A green-white color scale is used to indicate the variation of Cognitive FI.
These two identified clusters corresponding to microbiome composition in adulthood and senescence were specifically composed of some bacterial ASVs which were decreasing or that were increasing with physiological aging and frailty. In the discussion section, we address this point with specific comments about those ASVs specifically linked to the Cognitive FI change during the aging process.

Discussion
Declarative memory is one of the main features of the human personality, which is the capability to remember an object, a person, or a place previously encountered. The maintenance of the recognition memory capability of declarative memory has a great impact on the elderly social life. Animal and human studies have evidenced the so-called gut-brain axis, in which a crucial role is played by the gut microbiome through a dynamic, complex bidirectional relationship with the host, thus identifying the microbiota-gut-brain axis [15,42].
Elderly people display a different gut microbiome profile compared to healthy adults for multifactorial factors such as lifestyle, nutrition, lower motor capabilities, and reduced immune system functions [15]. Indeed, the gut microbiome is highly sensitive to environ-mental stimuli and its composition changes during the host lifespan. Animal models allow to minimize most of the differences in epigenetic factors such as, food, light, temperature, humidity, etc. A change in the gut microbiome composition during aging could be due both to a dysbiotic condition and/or to an adaptive mechanism trying to compensate for the dysbiosis. This longitudinal study aimed at investigating (i) the modifications in gut microbiome composition during aging; (ii) the relationship between gut microbiome composition and cognitive decline of recognition memory; and (iii) functional speculation about eubiotic and adaptive or dysbiotic and maladaptive changes during aging.
To achieve those goals, we monitored the gut microbiota composition and the Cognitive Frailty Index (FI) during aging [43]. Previously, we tuned an FI in mice for describing the aging decline of recognition memory by means of a battery of spontaneous behavioral tests that discriminate between the knowledge component and the recall component of the recognition memory [3,6].
The current longitudinal study demonstrated that the overall gut microbiome composition changed during the aging process. Nevertheless, as time went on, Cognitive FI increased, making it extremely difficult to isolate the effects of the two covariants (time and Cognitive FI) on overall gut microbiome composition (alpha and beta diversity levels). Despite this, we described a slight correlation between Cognitive FI and microbiomes (in terms of their taxonomic composition in the different mice examined), and this consideration was partly supported by the results of the CAP scale analysis. The not-significant effect of the LME model used for testing beta diversity in our experimental condition highlighted that it is not possible to use this variable as a predictor of total bacterial diversity. However, in a deeper taxonomic microbiome analysis, it was possible to identify some ASVs that specifically increased or decreased with Cognitive FI variation.
During aging, gut microbiome alpha diversity measures gave opposite results: the Shannon index increased, whereas Faith's PD index decreased, in senescence. Shannon alpha diversity is sensitive to both the richness and the evenness (the total number of species and their relative abundance in the fecal sample). These data are in agreement with a recent paper [44], describing changes in human gut microbiome composition during aging also associated with healthy or unhealthy conditions. Faith's PD index represents the number of phylogenetic tree units and their distance within a sample. Therefore, our results suggest that during the aging process, we were witnessing an increase in the number of ASVs that were phylogenetically more similar in senescence compared to adulthood. We may speculate that, despite the high number of species in the aged mice, the similarity to the taxa level accounts for a lower flexibility from a functional point of view compared to young animals. Confirmation that aging significantly affected the overall microbiota composition also came from the inter-individual variation, estimated through the beta diversity, by a non-multidimensional scaling (NMDS) ordination approach. In fact, we identified five clusters, sequentially expressed and each one specific for each experimental time, thus creating a large gap between the adult and senescent mice microbiome.
The core phyla during aging were Bacteoroidota and Firmicutes: their relative abundance and ratio remained relatively stable throughout the lifespan. Regarding subcore phyla during aging, Verrucomicrobia disappeared between 20-21.5 months, matched by total disappearance of the Akkermansiaceae family and Akkermansia genus. A change in Akkermansia's relative abundance was associated with metabolic diseases (i.e., diabetes, inflammatory bowel disease, and obesity) [45][46][47] but also with neurodegenerative diseases (i.e., Alzheimer's and Parkinson's diseases) [48][49][50]. At the gut level, Akkermansia, regulating the host immune response and reducing local inflammation, helps maintain the integrity of the gut barrier. Furthermore, Akkermansia influences the fat and sugar metabolism [47]. Recently, Ou et al. (2020) [51] and Higarza et al. (2021) [52] demonstrated that Akkermansia could improve cognitive performance in two different preclinical models. In particular, Ou et al. (2020) [51] found that the Akkermansia gavage, in addition to regulating the inflammation and sugar metabolism, significantly reduced the Aβ 40-42 levels in the cerebral cortex and improved the spatial and recognition memory of APP/PS1 mice, a model of Alzheimer's disease (AD) [51]. Higarza et al. (2021) [52] demonstrated that Akkermansia gavage restored cognitive impairment related to nonalcoholic steatohepatitis in rats [52]. In parallel with the disappearance of Akkermansiaciae, we recorded the appearance of Muribaculaceae (previously known as S24-7). The Muribaculaceae family is one of the major utilizers of mucus-derived monosaccharides in the gut, contributing to SCFA production [53], in particular, propionate [54,55]. The SCFAs have a pivotal role in the host's homeostasis and physiology, and SCFA concentrations are predictive of the host's lifespan [54]. Indeed, the Muribaculaceae bacterial family was found as the dominant bacterial taxa in the gut of Spalax leucodon, an exceptional animal model of longevity [53]. Previous works have demonstrated a lower abundance of the Muribaculaceae family in aged mice compared to young animals, and the anti-aging intervention has been shown to partially reverse the gut microbiota composition of elderly mice, also increasing the relative abundance of the Muribaculacea [56,57]. We hypothesized that the increased Muribaculaceae relative abundance during physiological aging suggests a possible adaptive mechanism for the disappearance of the Akkermansia, trying to prevent a possible leaky gut. Next, we demonstrated a significant increase in the Rikenellaceae family during aging. Notably, in our samples, the Alistipes genus represented the entire Rikenellaceae family. Regarding this genus, conflicting data are reported [58], from a pathogenic role in cancer [59] to a preventative role in mental health, such as anxiety and depression [60].
Furthermore, we found a significant increase in Clostridia vadinBB60 between adulthood and senescence. Recently, Juckel and colleagues (2021) [61], in patients with schizophrenia, suggested that an increase in the Clostridia vadinBB60 group could be responsible for the neuroplasticity reduction in the central nervous system (CNS). Therefore, we supposed that the increase in Clostridia vadinBB60 could exert a detrimental effect on cognitive performance during aging.
The relative abundance of some ASVs was significantly associated with Cognitive FI changes. In particular, ASVs belonging to Parabacteroides, Clostridia UCG-014, Oscillibacter, Mucispirillum, Coprostanoligenes, Flavonifractor, Blautia, Monoglobus, Sedis, Lachnoclostridium, Oscillospiraceae_UCG_005, Marvinbryantia, Erysipelotrichaceae, Clostridium sensu stricto 1, Halomonas, Cutibacterium, Ruminococcaceae Candidatus Soleaferrea, Ruminococcaceae incertae sedis, Ruminococcaceae Paludicola, Ruminococcaceae Phocea, and Lachnospiraceae ASF356 significantly decreased in relation with the Cognitive FI increase. On the other hand, ASVs belonging to Ruminococcus, Colidextribacter, Anaerotruncus, Anaeroplasma, Peptococcaceae, Lactobacillus, and Ruminococcaceae incertae sedis increased in parallel to the Cognitive FI increment. We speculated that the increase in health-promoting or the decrease in pathobiont bacteria associated with the Cognitive FI increase could have an adaptive or compensatory role; on the contrary, the increase in unhealthy bacteria and a decrease in healthy bacteria with the FI increase could exert a dysbiotic role (Table 3). Table 3. Proposed dysbiotic or adaptive roles of gut bacteria that significantly changed with Cognitive Frailty Index. We searched in the literature the involvement of selected bacteria on host health, and in particular, on CNS.

Genera Published Effects on Host Health
Changes with Cognitive Frailty Increase

Monoglobus
Negatively related to amyloid presence in the brain [84]. Decrease Dysbiotic

Marvinbryantia
Negatively related to amyloid presence in the brain [84]. Decrease Dysbiotic

Increase Adaptive
Furthermore, about the other ASVs which decreased or increased, not yet presently described in the present study, any previous literature data are available, to the best of our knowledge, regarding their relationship with both aging and cognitive frailty.
In summary, our data demonstrated that aging significantly affected the overall gut microbiome composition in a complex way, increasing the number of ASVs and decreasing the phylogenetic distance among them. Beta diversity confirmed the presence of different clusters between adult and senescent mice. Cognitive frailty was associated with some specific changes in ASVs. We proposed that some of those ASV modifications could be a challenge in the eubiotic and adaptive direction and the opposite in the dysbiotic and maladaptive direction. The inverse correlation between Faith's PD index and cognitive frailty suggests that cognitive decline is accompanied by a shrinkage of the gut microbiome functionality. Furthermore, this potential plasticity in the gut microbiome composition paves the way to the use of psychobiotics for prevention of cognitive frailty during aging [131].
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu14142937/s1, Figure S1: Rarefaction curve; Figure S2: Number of reads; Figure S3: Simpson index; Figure S4: correlation between Shannon index and Cognitive FI. Table S1: Statistical analysis of the effect of aging on alpha diversity based on Shannon diversity index (on the top) and Faith's phylogenetic distance (on the bottom), including the comparison among all time points; Table S2: Statistical analysis of the effect of aging on beta diversity based on Bray-Curtis distance matrix, including the comparison among all time points; Table S3: LME used to estimate the effect of time on Cognitive FI; Table S4: LME used to estimate the effect of time and Cognitive FI on Faith's PD index; Table S5: LME used to estimate the effect of time and Cognitive FI on the beta diversity, based on Bray-Curtis index.

Data Availability Statement:
The sequences that were generated in this study have been deposited into the EMBL-EBI database in PRJEB54046 project.