Faecal microbiota transplant from aged donor mice into young recipients affects spatial learning and memory

Shifts in microbiota composition in ageing affect a variety of systems, including the central nervous system (CNS). We tested the hypothesis that faecal microbiota transplantation (FMT) from aged donor mice into adult recipients had a direct bearing on brain functions known to be affected during ageing. We found that FMT from aged donors led to impaired spatial learning and memory in adult recipients, whereas anxiety, explorative behaviour and locomotor activity remained unaffected. Label-free quantitative proteomics showed that FMT from aged donors modulated the expression of a variety of proteins implicated in maintenance of synaptic plasticity and neurotransmission in the hippocampus of adult recipients. Furthermore, microglia cells of the hippocampus fimbria acquired an ageing-like phenotype indicated by increased expression of F4/80 marker. We conclude that the age-associated alteration of the gut microbiota contributes to the decline of key functions of the CNS through the modulation of hippocampal synaptic plasticity-related proteins.


Introduction
Ageing is an inevitable process that starts immediately after birth and ultimately leads to the loss of functional capacity in several body systems, including the cardiovascular, skeletomuscular, osteoarticular and neuro-immune-endocrine, and is often associated with a decline in psychological wellbeing and cognitive function. The hallmark of ageing is a chronic low-grade inflammatory status termed "inflammageing" 1  Recently, the existence of bidirectional communication between the gut and the brain -the gut-brain axis -has emerged as an important player in shaping aspects of behaviour and cognitive function 6,7 . In particular, the gut microbiome has been reported to play an important role within this scenario. For example, a modest alteration in the composition of the gut microbiota, induced by either diet or antibiotics, is sufficient to cause changes in mouse brain chemistry and function 8,9 . In particular, the oral administration of antibiotics resulted in an increase in brain-derived neurotrophic factor (BDNF) in the hippocampus. Furthermore, reduced expression of the synaptic plasticity-related genes PSD-95 and synaptophysin, important in memory formation and maintenance, were also reported 10 . Furthermore, studies using germ-free (GF) mice showed that, in the absence of microbes, mice are less responsive to exposure to environmental stressors compared to conventionally reared mice, with conventionalisation of GF animals impacting significantly on brain development 11,12 . This observation, along with a growing body of evidence, has shown that the gut microbiota plays a major role in in the development and function of the CNS, affecting learning and memory via metabolic, neuroendocrine and immune pathways 13 .
Second, dysbiosis has been associated with a variety of neurological disturbances ranging from depression 14 to autism 15,16 , along with neurodegenerative diseases such as . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint Parkinson's disease and multiple sclerosis 17,18 . Not surprisingly, faecal microbiota transplantation (FMT) is being investigated as a therapeutic option not only for a variety of GI-tract-related diseases, including irritable bowel syndrome, inflammatory bowel disease and obesity 19,20,21 , but also for CNS disorders 22,23 . Thus, it is plausible to hypothesize that age-associated changes in the microbiota have a direct effect on the CNS, potentially contributing to the decline of cognitive function seen in the elderly.
Here, we tested the hypothesis that alteration of the profile of the gut microbiota observed in ageing directly affects aspects of cognitive and behavioural functions. In particular, we focused on events taking place in the hippocampus, an area of the CNS that controls aspects of cognition and behaviour and is heavily affected by the ageing process 24 . We observed that young/adult mice (henceforth identified as adult) colonized with the faecal microbiota from aged donors showed spatial learning and memory impairments while locomotor, explorative and anxiety-like behaviours remained unaffected. These changes were concomitant with the differential expression of an array of proteins in the hippocampus of adult recipients and with the appearance of phenotypic changes of microglia of the hippocampus fimbria typical of the ageing brain. Plasma levels of inflammatory cytokines and gut permeability remained unchanged. We interpreted these data as showing that alterations of the gut microbiome contribute to the deterioration of important functions of the CNS in later life.
. CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint

Microbiota of adult recipients acquires an aged phenotype following FMT from aged donors
The microbiota is known to change as we age 25 . In order to evaluate and prepare our samples for FMT, we investigated in the first instance the faecal microbiotas of adult (3 months) and aged mice (24 months) (Fig. 1). Measures of alpha diversity showed only a significant (P = 0.0107) difference in species richness between the two cohorts, with the faecal microbiota of aged mice harbouring more different amplicon sequence variants (ASVs) than adult mice (Fig. 1a). Beta diversity (Bray Curtis) analysis showed clear separation of the two groups of mice based on their faecal microbiotas (Fig. 1b) (Fig. 1c). Integration of faecal microbiomic and metabolomic data showed an association between lower levels of faecal short-chain fatty acids (SCFAs) and decreased representation of obligate anaerobes such as the Lachnospiraceae and Ruminococcaceae (Supplementary Fig. S1 and S2).
Based on this preliminary analysis, faeces from three of the adults and two of the aged mice included in the above-mentioned initial assessment were pooled and used as donors for FMT (Fig. 2). Adult mice received FMT either from adult donors (FMT adult) or aged donors (FMT aged). Measures of alpha diversity showed no significant differences between the two mouse sub-groups (Fig. 2a). Bray Curtis analysis showed clear separation of the pre-and post-FMT groups, with the post-FMT samples clustering with their respective inoculants (Fig.   2b). It is notable that antibiotic treatment of pre-FMT mouse 1 (Ms1) and mouse 2 (Ms2) was . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint not completely successful as they clustered more closely with the post-FMT adult group and adult pooled sample. However, after these mice had been inoculated with the adult pooled sample, they clustered with the aged post-FMT group. The differences in the faecal microbiotas between the adult and aged mice post-FMT were not as pronounced as the differences seen between the adult and aged groups in our initial study; as such a less stringent Benjamini-Hochberg cut-off (P < 0.1) was used when analysing these data (Fig. 1).
This less-pronounced change is unsurprising as the adult and aged mice in the initial study had not been subject to any intervention, but it is becoming clear that even subtle changes in the gut microbiome are associated with phenotypic changes in the host 26 and the same may be true of cognitive traits. Only four genera (Prevotellaceae, Faecalibaculum, Lachnospiraceae and Ruminococcaceae) were found to be significantly differentially abundant in the faeces of the post-FMT adult and post-FMT aged animals (Fig. 2c). Few significant associations were seen between the faecal microbiomic and metabolomic data ( Fig. 2d).

FMT from aged mice to adult recipients results in decreased spatial learning and memory but does not affect motor activity or anxiety-like behaviour
In the first set of experiments, we investigated the impact of FMT from aged mice to adult recipients to a series of spatial learning and memory tests. First, the Barnes maze test was used to assess memory and spatial learning. Following the training trials, a retention test was conducted in which the escape tunnel was removed and the latency before moving to the position of the former escape tunnel for the first time was measured. We observed that the average primary latency was significantly higher for FMT-adult recipients of microbiota from aged donors (FMT-aged) than the other control groups either left untreated or colonized with microbiota from adult, age-matched donors (FMT-adult) (41.7 ± 3.5 s; 30.5 ± 3.9 s and 23.5 ± 6.7 s, respectively; P<0.05) (Fig. 3a). Yet, during the retention test, FMT-aged mice spent less time in the target quadrant (28.9 ± 2.6 s in comparison to control 46.5 ± 6.3 s, and . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint FMT-adult 51.8 ± 10.2 s; P<0.05) that contained the escape tunnel compared to control group ( Fig. 3b; heat map in Fig. 3c) confirming a direct impact of the gut microbiota on memory and spatial learning. Learning and memory was further evaluated by the novel object recognition test. Significant differences between groups were observed in the time spent exploring two different objects (Fig. 3d). In particular, the time spent exploring novel and familiar objects was monitored; the object recognition was assessed by a defined discrimination index. Control groups, either untreated or FMT-adult mice, preferred the novel object more than the familiar one, whereas FMT-aged mice showed a significant reduction in the time spent exploring the novel object compared to control (0.23 ± 0.04 and 0.46 ± 0.05, respectively; P<0.01) suggesting reduced discrimination as a consequence of impaired memory capabilities. The decline of the discrimination capability between a novel and a familiar object by FMT-aged mice is further evidenced by the heat maps (Fig. 3e).
Given that the microbiota has been reported to affect locomotor activity and anxiety-like behaviour in animal models 10, 27 , we assessed whether FMT from aged animals to adult recipients could also affect these behavioural manifestations. For this purpose, we employed two validated tests: the open field and the elevated plus maze. In the open field task, no significant difference in the distance travelled by both FMT-aged and FMT-adult was observed when compared to the untreated control group indicating no significant motor impairments of the animals (Fig. 4a). It is worth noting that FMT-aged mice displayed a tendency to prefer the periphery or the corners of the arena instead of the centre (Fig. 4b).
The representative tracks of movement patterns are depicted in Fig. 4c-e (ANY-maze software).
In the elevated plus maze, FMT-aged mice did not display significant differences in time spent in either arms of the maze compared to control groups ( Fig. 4f- The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint

FMT from aged donors alters protein expression in the hippocampus of adult recipients
The observation that FMT from aged donors to adult recipients could alter behavioural patterns of transplanted mice prompted us to investigate the molecular mechanisms underlying such changes. To this extent, a one-shot label-free quantitative proteomics approach was employed, targeting the hippocampus, a brain region known to play an important role in learning and memory and to undergo several changes with the ageing process 28 . The analyses resulted in the quantitation of 2180 protein groups and 16083 unique peptides (Supplementary Table S1). The Pearson correlation coefficients between samples and technical replicates were 0.99, indicating a high reproducibility and confidence of the resulting data ( Supplementary Fig. S3A). The volcano plot was obtained by two-sided t test of the two groups: FMT aged faeces into adult mice related to the adult FMT into adult mice ( Supplementary Fig. S3B). 140 proteins were differentially regulated, 52 overexpressed and 88 down-expressed, of which 47 proteins were differentially regulated with a fold change higher than 1.5 (Table 1, Supplementary Table S1).
For quantitative western blotting confirmation of label-free proteomics data, we selected two candidate proteins implicated in the behavioural phenotype of transplanted mice Protein network analyses highlighted several differentially expressed proteins in the hippocampus of adult mice that received microbiota from aged donors compared to adult mice that received microbiota from age-matched adult donors. Table 1 summarises the main enriched categories. Cellular signalling during nervous-system development and related to synaptic transmission was described by 35 molecules, differentially expressed, indicating a downregulation during the transplant (z score -1.7). Neurotransmission was also found . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint downregulated by 41 differentially expressed molecules. Remarkably the changes in the expression of a total of 87 learning-related proteins and cognition tests in FMT-aged mice pointed to a decline of brain functions as seen during the physiological ageing process (Fig.   5). In addition, in FMT-aged mice several pathways related to lipid metabolism were downregulated ( Supplementary Fig. S6). This single label-free run approach resulted in an in-depth hippocampus protein characterization and above all in the identification of differentially regulated proteins involved in key pathways of ageing-related processes.

FMT from aged donors into adult recipients triggers phenotypic changes in glia cells of the hippocampus fimbria without affecting gut permeability or circulating cytokine levels
It has been suggested that FMT from aged donors triggers systemic inflammageing in adult The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint

Discussion
The role of the gut microbiome in health and disease 30 has become evident in the past few years and it is generally accepted that alterations of the gut microbiota have an impact on the ageing process 25 . However, the extent of the impact of the microbiome on specific functions of the CNS in ageing remains to be determined. The decline of cognitive and behavioural functions in ageing is one of the most detrimental effects of ageing 31 . In particular, a deficit of hippocampal-dependent spatial memory is a distinctive trait of ageing 24 . Here we report that FMT from aged donors into young adult recipients affects memory and spatial learning in transplanted mice, a phenomenon led through a differential expression of proteins involved in maintenance of synaptic plasticity and neurotransmission in the hippocampus area.
First, we observed that mice originated from the same colony and subjected to the same environmental conditions (i.e. housing and food) changed their microbiota profile with the passing of time. This observation suggested that the age-associated change in microbiota profile is initially triggered by changes taking place in the host due to the intrinsic, environment-independent ageing process. However, when the faecal microbiota was transferred to young adult mice, recipients of FMT from aged donors displayed impaired spatial memory and learning capacity, a typical trait of the ageing process. On the contrary, in agreement with a recent report that investigated the effect of FMT from aged donors in germ-free mice 32 we observed that anxiety, explorative prowess and locomotor activity were not affected by the FMT procedure. Although the presence of specific microbial families and genera have been associated with cognitive decline, anxiety behaviours and affective disorders 33 , to our knowledge this is the first evidence showing that FMT significantly affects important functions of the CNS in ageing. Such observations prompted us to investigate the . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint potential underlying mechanisms and whether the observed changes in cognition and behaviour may be triggered through the modification of proteins relevant for CNS function.
To this end, we employed a label-free quantitative proteomics approach.
Significant modifications of protein expression were observed in the hippocampus area. In particular, the variation of the expression of proteins, such as D-dopachrome decarboxylase (D-DT), neuronal membrane glycoprotein M6-a (Gmp6A) and M6-b and Ras/Rap GTPaseactivating protein SynGAP was of significance. Indeed, alterations in the expression of these proteins in the CNS have been implicated in important CNS functions and disturbances, ranging from filopodium formation, synaptogenesis, learning disability, behavioural anomalies and neuronal plasticity to neurodegenerative disorders 34, 35, 36, 37, 38 . In addition to these observations, it has been suggested that changes in the microbiota are associated with increased gut permeability and systemic inflammation that may play an important role in age-associated alteration of cognitive and behaviours in ageing 39 . With regard to inflammation it has been described that FMT from aged donors to young germ-free recipients triggered systemic inflammageing including enhanced CD4+ T cell differentiation and distribution of several Th1 and Treg subsets, in particular in the systemic compartment 4 . However, the latter report fell short of measuring circulating pro-inflammatory cytokines and cognition/behaviour in the recipients. In our experiments, a small non-significant increase was observed for IL-1β and TNF-α. Consistent with the lack of a significant inflammatory response is our observation that levels of GFAP did not increase in various areas of the hippocampus. However, a direct effect of pro-inflammatory cytokines on the protein expression and function of the hippocampus cannot be ruled out. It is plausible that a drastic increase of systemic levels of cytokines is not required and that only a very minor increase of specific systemic cytokines may suffice to trigger the changes observed here.
The latter hypothesis is supported by the phenotypic change of the microglia, resident immunocompetent cells in the CNS that respond to a plethora of inflammatory stimuli 40 .
Indeed, young adult recipients that had received FMT from aged donors showed a significant . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint increase in the expression of the marker F4/80, a specific trait of the ageing brain 29 .
Currently, whether the FMT-mediated phenotypic change of the microglia is paralleled by its activation remains to be determined.
Soluble factors of microbial origin other than host-derived molecules/cytokines might be at work. Indeed, integration of faecal microbiomic and metabolomic data showed a clear association between lower levels of SCFAs and decreased representation of obligate anaerobes such as the Lachnospiraceae and Ruminococcaceae that have been associated with production of SCFAs by the human faecal microbiota (Supplementary Fig. S1 and Supplementary Fig. S2). It should, however, be noted the biological relevance of increased levels of faecal propionate, butyrate and isobutyrate being significantly correlated with increased relative abundance of lactic acid bacteria such as Staphylococcus, Jeotgalicoccus and Facklamia is currently unknown (Supplementary Fig. S1 and Supplementary Fig. S2), as our knowledge as to the ability of these species to produce SCFAs other than acetate or lactate from fermentation of carbohydrate or protein sources is poor as is our knowledge on cross-feeding potential of the mouse gut microbiota.
In addition to these aforementioned genera, a strong decrease in Prevotellaceae was also observed following FMT of aged faeces in young adult recipients. We previously reported higher levels of Prevotellaceae and Ruminococcaceae in APOE3/E3 and APOE2/E3 genotype carriers, respectively, relative to APOE4 carriers 41 , one of the strongest prevalent risk factors for neuropathology and Alzheimer's disease 42,43 . Whilst the underrepresentation of Prevotellaceae has been previously reported to diminish the levels of health-promoting neuroactive SCFAs in humans 44 and the biosynthesis of thiamine and folate, two vitamins decreased in Parkinson's disease 45,46 , depletion of Ruminococcaceae has been associated with Alzheimer's disease 47 . Interestingly, a previous study showed that SCFAs may regulate host serotonin biosynthesis 48 , a multifaceted neurotransmitter modulating cognition, learning and memory, along with numerous physiological processes 49 .
Finally, the role of the vagus nerve on microbiota-mediated alterations of the hippocampus 1 3 area cannot be rule out. This cranial nerve exerts an important role in gut-brain communication 50 and it has been shown that it mediates the effects of orally delivered probiotic strains on aspects of behaviour 51 . Future experiments of selective vagotomy will help to elucidate the role of the vagus nerve in the microbiota-mediated decline of spatial and learning memory.
Taken together these results point to a direct effect of age-associated shifts in microbiota composition on the decline of key functions of the CNS. This notion, along with recently collected evidence that correcting age-associated shift of microbiota profile is beneficial for health and life expectancy 52 , lends support to the hypothesis that microbe-based approaches that aim to restore a young-like microbiota might improve cognitive function and in so doing the quality of life of the elderly, an ever-increasing demographic segment of modern societies.
. CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint Environmental temperature was kept at 23±1 °C with a 12 h light/dark cycle. All efforts were made to minimize animal suffering and to reduce the number of animals used.

Faecal material preparation and FMT regime
Faecal material was collected from adult and aged mice and placed into Eppendorf tubes containing 500 ml of freezing solution (sterile saline solution with 12.5% glycerol), tubes were left on ice for 60 min and later homogenized for 2 min on ice using a hand-held electric pellet pestle device with sterile, re-usable pestles. When fully homogenized, the suspended pellets were frozen and stored at -80°C until utilized for the FMT procedure. To this end, mice were randomized into the following treatment groups (n=12 per group): control (no antibiotic treatment, no FMT); FMT-adult (antibiotic treatment followed by FMT from young age-matched donors), FMT-aged (antibiotic treatments followed by FMT from aged donors). The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint differences between different diversity measures was determined using Wilcoxon rank sum test. ALDEx2 was used to determine statistically significant differences (Welch's t test, Wilcoxon) between mouse groups. The 16S rRNA gene sequence data have been deposited in the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/) under accession number PRJNA524024.

Metabolomic analyses
Metabolites were analysed and quantified by 1 H-NMR analysis as previously described 41 .
Briefly, 20 mg of frozen faeces were thoroughly mixed on a vortex with 1 mL of saline

Cognition and behavioural tests
To evaluate the effects of FMT on cognition and behaviour the following tests were conducted: Barnes maze, novel object recognition, elevate plus maze, open field test.
Detailed methods for these and other analysis and the associated references can be found in Supplementary Materials and Methods.

Statistical analyses
. CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint Microbiomic data (genus level) were correlated (Pearson) with metabolomic data using aldex.corr() within the Bioconductor package ALDEx2 53 . For the MS study all data were evaluated by Perseus statistical software. The protein expression fold change variation between two groups was analysed by two sides t test, setting FDR less than 0.015 and s0 of 0.1. The differentially expressed proteins, with a significant ratio FMT-aged/FMT-adult (P value < 0.05), were analysed by Ingenuity Pathway Analyses (Qiagen) and only the differentially regulated pathway with a significant z score, after Bonferroni test, were    Barnes Maze test (Fig 3a-c): mice were trained to find the cage for 4 consecutive days (twice daily; 2 trials). The average primary latency (Fig 3a) was significantly higher for adult recipients of microbiota from aged donors (FMT-aged) than the other control groups either left untreated or colonized with microbiota from adult, age-matched donors (FMT-adult).
Furthermore, during the retention test, FMT-aged mice spent less time in the target quadrant . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint that contained the escape tunnel compared to control groups ( Fig. 3b; heat map in Fig. 3c).
The values represent the mean ± SEM for each group (n = 10-12 mice/group). *P<0.05 vs control animals and FMT-adult. The novel object recognition test (Fig. 3d-e): on day 1, mice were exposed to two similar objects (A+A); on day 2, animals were re-exposed to the testing area containing one novel object (A+B). The time spent by the animals exploring each object was recorded. The discrimination index, calculated as (TB-TA)/(TB+TA), was used to assess the preference for the novel object. Control groups, either untreated or FMT-adult mice, preferred the novel object more than the familiar one, whereas FMT-aged mice showed a significant reduction in the time spent exploring the novel object (heat map in 3e) suggesting reduced discrimination as a consequence of impaired memory capabilities. The values represent the mean ± SEM for each group (n = 10-12 mice/group). **P<0.01 vs control animals. Open Field Test did not show significant difference in the distance travelled by both FMTaged and FMT-adult was observed when compared to the untreated control group indicating no significant motor impairments of the animals (Fig. 4a). However, FMT-aged mice displayed a tendency to prefer the periphery or the corners of the arena instead of the center (Fig. 4b). The representative tracks of movement patterns are depicted in Fig. 4c-e (ANYmaze software Furthermore in the elevated plus maze, FMT-aged mice did not display significant differences in time spent in either arms of the maze compared to control groups (Fig. 4f-g).
The values represent the mean ± SEM for each group (n = 10-12 mice/group).
. CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/866459 doi: bioRxiv preprint 1 Supplementary Fig S2. Pearson correlation of faecal microbiomic and metabolomic data for adult and aged mice. ALDEx2 was used to correlate the datasets. +, P < 0.05 (Benjamini-Hochberg). All data are shown: significant results were used to generate Supplementary