Functional microbiome deficits associated with ageing: Chronological age threshold

Abstract Composition of the gut microbiota changes during ageing, but questions remain about whether age is also associated with deficits in microbiome function and whether these changes occur sharply or progressively. The ability to define these deficits in populations of different ages may help determine a chronological age threshold at which deficits occur and subsequently identify innovative dietary strategies for active and healthy ageing. Here, active gut microbiota and associated metabolic functions were evaluated using shotgun proteomics in three well‐defined age groups consisting of 30 healthy volunteers, namely, ten infants, ten adults and ten elderly individuals. Samples from each volunteer at intervals of up to 6 months (n = 83 samples) were used for validation. Ageing gradually increases the diversity of gut bacteria that actively synthesize proteins, that is by 1.4‐fold from infants to elderly individuals. An analysis of functional deficits consistently identifies a relationship between tryptophan and indole metabolism and ageing (p < 2.8e−8). Indeed, the synthesis of proteins involved in tryptophan and indole production and the faecal concentrations of these metabolites are directly correlated (r 2 > .987) and progressively decrease with age (r 2 > .948). An age threshold for a 50% decrease is observed ca. 11–31 years old, and a greater than 90% reduction is observed from the ages of 34–54 years. Based on recent investigations linking tryptophan with abundance of indole and other “healthy” longevity molecules and on the results from this small cohort study, dietary interventions aimed at manipulating tryptophan deficits since a relatively “young” age of 34 and, particularly, in the elderly are recommended.


| INTRODUC TI ON
The microbiota is now considered an additional organ in our body (Moya & Ferrer, 2016). Therefore, it undergoes changes throughout development, similar to other organs. Moreover, its physiological status, whether healthy or dysbiotic, influences the general health of individuals, although the directionality of this effect is not completely clear and is sometimes confusing (Kundu, Blacher, Elinav, & Pettersson, 2017). The numerous factors to which our body is exposed are also reflected by changes in the microbiota (Rojo et al., 2017). Various natural physiological changes are among the agents involved in modifying the microbial structure, both temporary (pregnancy or lactation) and permanent (the ageing process) (O'Toole & Jeffery, 2015). One of these changes is caused by chronological age.
Humans are not free of microbes at birth, and the microbial community is continuously enriched and diversified with ageing (Odamaki et al., 2016). Indeed, it is generally accepted that a healthy and stable microbiota is established at the age of 3 years that remains similar until the adult stage, despite periodic fluctuations of various types (Odamaki et al., 2016). However, this conclusion should be re-examined because the microbiota can be healthy throughout development but changes dynamically with age (Martí et al., 2017).
We would such as focus on the existing difference between the total and active microbiota (Moya & Ferrer, 2016). The total amount of bacteria at a given moment is different from the active working fraction, as this fraction has a functional role and is more relevant to the human health (Mills et al., 2019). In other words, although an examination of the temporal changes in the total (active and inactive) microbiota is interesting at any time scale, an evaluation of changes in the active members is more important in a broad sense (Moya & Ferrer, 2016) or in relation to ageing, life expectancy and age-related diseases (Zierer, Menni, Kastenmüller, & Spector, 2015).
As example, in their study of ageing and the microbiota, Sonowal et al. (2017) found that indoles produced from commensal active microbiota do not affect the fitness of young individuals but extend the healthspan of older individuals in diverse organisms such as Caenorhabditis elegans, Drosophila melanogaster and mice. During ageing, indoles induce the expression of host genes that promote healthy ageing. Thus, an assessment of microbial functions associated with active components of the human microbiota in well-defined age groups is necessary. However, these investigations are still rare.
Our goal in the present manuscript was to identify the association between the functional gut microbiome and ageing and to identify potential functional deficits associated with ageing. Studies using proteomics and metabolomics provide direct valuable insights into these deficits compared to other "omics" techniques, but due to technicalities, studies in ageing research are limited to a few examples and small (n = 12) sample sizes (Gelfi et al., 2006;Zierer et al., 2015). In the present study, total proteins from bacterial cells isolated from the faecal material of three well-defined age groups (n = 30) were subjected to shotgun proteomics; this approach allowed us to define the active fraction of the microbiota that synthesizes proteins. Subsequently, a functional analysis of the identified proteins was performed to assess presumptive age-dependent functional deficits. Finally, using liquid chromatography coupled with mass spectrometry, functional deficits were experimentally validated in an extended set of replicate samples collected over time (n = 83). The combined analysis identified a reproducible microbiome biomarker associated with ageing, namely, a link between an elderly age and tryptophan and indole deficits. The relevance of tryptophan and indole to healthy ageing (Sonowal et al., 2017) and the results reported in this study will provide opportunities for the development of putative and innovative dietary strategies for healthy ageing. at manipulating tryptophan deficits since a relatively "young" age of 34 and, particularly, in the elderly are recommended.

| General characteristics of the study population and study design
Faeces from ten infants (I), ten adults (A) and ten elderly individuals (E) were collected and analysed with a proteomic approach using a pooling strategy. Generally, sample pooling results in some unforeseen methodological and statistical bias. A recent proteomic study using individual and pooled serum samples from controls and patients with Creutzfeldt-Jakob disease (CJD) revealed that compared to the analysis of individual samples, sample pooling affected the coefficients of variation of the minimum, maximum and mean values in both the control and CJD groups. However, the authors were able to identify biomarkers that significantly differed among groups (Molinari et al., 2018), which were then subjected to an in-depth analysis using independent samples. This strategy was used in the present study because our main objective was to identify biomarkers of protein and functional deficits that substantially differed among the three well-defined age groups (see Table 1).
Briefly, pooling reduced the number of samples to analyse from 10 individual samples to two pools of five individuals for each of the three well-defined age groups. Figure S1 summarizes the experimental groups analysed in this study. The idea was to establish a proteomic analysis with low resource and time requirements that would allow us to detect substantial differences among the three groups, which would be further validated using individual samples.

| An elderly status is associated with elevated levels of active bacteria
In total, 64,313 quality-filtered nonredundant proteins were obtained from the faecal samples of all six pools, with a median value of 14,892 ± 1,875 proteins per pool (Table S1). Significant differences in the number of proteins were not observed among all six pools. This number is consistent with the average values reported in previous proteomic studies of the gut microbiota (Deusch et al., 2018;Mills et al., 2019;Serrano-Villar et al., 2016). We calculated the alpha diversity parameters of active bacteria, namely, microbial richness (d), Pielou's evenness (J′) and Shannon index (H), from raw proteomic data using a previously described procedure (Deusch et al., 2018). As shown in Table 2, the richness of active species measured using d values was slightly increased with age, and the value of the E group was ca. 1.4fold higher than the I group. The number of species, as measured by calculating J′ and H, was also slightly increased with age. Thus, ageing is associated with a slightly greater protein diversity and richness of the corresponding active bacteria. Importantly, a statistical analysis was not performed because only two pools per group were analysed. However, the standard deviations among pools were very low and thus the differences observed were considered significant.

| Functional deficits associated with ageing
We performed the procedure described in the Section 4 to optimally evaluate the proteome of each of the two pools of the three welldefined groups. In this experiment, only those proteins that were expressed in both pools from each of the three groups were considered to exhibit protein-level changes due to ageing that were not due to inter-individual variability. Three thousand four hundred seventy-five of the 64,313 quality-filtered nonredundant proteins met the following criterion: they were present in both pools, regardless of the relative abundance level. Those proteins were subsequently compared. Relative protein abundances were obtained using the procedure described in the Section 4. As shown in Figure 1, an exponential distribution of relative protein abundances was observed in all the three age groups, with a minor fraction of proteins (less than 1%) defined as super-abundant. The abundance data thus suggest that the diversity of proteins being synthesized and expressed was not dominated by a particular type of protein or highly similar clusters of proteins, but consists of diverse proteins with similar abundances in the three well-defined age groups investigated.
A subsequent examination of the 3,475 proteins revealed that only ca. 18% of the proteins were shared among the three well-defined age groups ( Figure S2). A sub-set of ca. 14% was only expressed in infants, ca. 21% in adults and ca. 25% in the elderly, suggesting that ageing slightly and progressively induced the synthesis of proteins that are produced at levels below the detection limit in early life. Rather than examining the individual differentially expressed proteins among groups, which may be biased because of the biological variation after sample pooling, we performed a functional analysis based on the assumption of a representative functional deficit in the individual groups or samples; sample pooling has been shown suitable for these purposes (Molinari et al., 2018). The identified proteins were assigned to Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthology (KO) pathways to assess the effects of ageing on different metabolic pathways defined by the KEGG. After calculating the relative abundance of each KO (see Section 4), no statistically significant differences were observed for most of the KO functions identified ( Figure 2). Among the different KO annotations, we focused on those that discriminated among extreme age states, namely, pathways present in the I group and in the E group, or vice versa. By using this strict criterion, only the levels of proteins assigned to KO1667 (TnaA; tryptophanase) and KO1696 (TrpB; tryptophan synthase) were significantly increased in infants compared with adults and were below the detection limit in the elderly individuals ( Figure 3). Most of these proteins belong to bacteria of the phylum Firmicutes. TrpB and TnaA are enzymes that catalyse the final steps in the biosynthesis of tryptophan and its further metabolism into indole ( Figure 4). The observation that both proteins were undetectable in both pools of elderly individuals suggests that the capacity of an "aged" microbiota to produce both tryptophan and indole may be significantly reduced compared to the "infant" or "adult" microbiota.

| Age-dependent changes in tryptophan and indole metabolism
We validated the functional deficits detected using the pooling strategy by subsequently quantifying the relative abundance of TrpB (essential for the biosynthesis of tryptophan) and TnaA ( Undoubtedly, the elderly presents a microbiota that is unable to generate indole and tryptophan, or produces these metabolites at very low levels, and those signatures are among the best active biomarkers of chronological age identified to date.

| D ISCUSS I ON
In the present study, we comprehensively assessed the age-associated changes in the gut microbiota, particularly the changes related F I G U R E 1 Distribution of relative protein abundances values in I, A and E groups. Represented is the number of proteins (absolute frequency) being expressed at different relative abundances (given as nemPAI or "ng protein per total µg of protein") in the two pools per each of the three groups; the percentage of proteins per each expression group is given above the bars. Relative abundances represent the mean values of two pools of five individuals each for each of the three well-defined age groups. The figure was obtained using R script to protein synthesis and eventually were able to distinguish among infants, adults and elderly individuals. Our study was performed using samples from three well-defined age groups by analysing faecal shotgun proteomes and was further validated by classical analytical methods for metabolite quantification. Although ageing-specific microbial species biomarkers have been identified (see Section 1), the identification of microbial deficits associated with ageing has not experienced the same increase in knowledge. This information is important to obtain an understanding of the roles of whole-microbiome characteristics in ageing. An analysis of the proteome presents a distinct opportunity for studying the ageing-associated microbiota by providing a functional perspective. Moreover, the identification of reproducible potential biomarkers of protein or functional deficits associated with ageing may facilitate the design of therapeutic or nutritional interventions.
Here, we have compared the proteins associated with the active fraction of the microbiota in infants, adults and elderly individuals.
Ageing is associated with the progressive activation of gut bacteria, possibly because bacteria must react to increasing number of factors associated with preserving the health status in response to exposure to an increasing number of environmental conditions that are distinct and greater than the conditions experienced during early life stages.
Most importantly, we identified a link between ageing and the microbial pathway associated with tryptophan and indole production and metabolism by the commensal microbiota. The key proteins involved in tryptophan-to-indole metabolism, TnaA and TrpB are both more abundant and expressed in the gut microbiota of infants. Both were expressed at significantly lower levels in adults and at even lower levels or below the detection limit in elderly individuals.
As shown in a recent study, indoles from commensal bacteria extend the healthspan of geriatric worms, flies and mice, and indoles may represent a class of therapeutics that improve the way we age, but not how long we live (Sonowal et al., 2017). The essential amino acid tryptophan, which is the least abundant in terms of its use in proteins, is provided by the diet or produced by gut bacteria, can cross the blood-brain barrier under the influence of the gut microbiota (Sandgren & Brummer, 2018) and  our study suggests a threshold or age at which the microbiota-based metabolism of tryptophan and indole begin to be significantly reduced, which may have health-related consequences on ageing if not treated accordingly. Indeed, based on our results, from the age of 11 years, the human gut microbiota may exhibit a decreased capacity to produce these metabolites, and from the age of 34 years, this capacity may be reduced by more than 90% compared to childhood.
The results of this study reinforce the hypothesis that dietary supplementation with indole (Sonowal et al., 2017) and tryptophan exert a beneficial effect on elderly individuals because their gut bacteria exhibit an impaired capacity to produce these molecules required for extending the healthspan. This supplement can be administered beginning at the age of 11 years, at which time a 50% decrease in the production of these metabolites occurs, and particularly beginning at the age of 34 years, when a greater than 90% reduction occurs ( Figure 5).
Notably, this study has two major limitations. The first one related to the pooling strategy applied, which as discussed also by Our results actually demonstrated that ageing significantly decrease tryptophan content in our gut environment. It is noteworthy that a 90% reduction occurs at a still "young" age of 34 years.
A justification of this threshold reduction remains unclear and needs further investigation. As well, age-thresholds, if any, for other microbiota products need to be established and reasoned.
Waiting for these issues to be further evaluated, a recent investigation demonstrated that some bacteria are important in predicting age, as the abundance and loss of certain bacteria associated with person's lifetime (Galkin et al., 2018). We thus think that the tryptophan deficiency from a certain age may be associated to the co-occurring changes in microbial composition and ecological interactions within the gut. Correlation analyses from 16S rRNA and tryptophan composition in faeces will allow, in the future, detecting networks of co-occurring bacteria and tryptophan level at different ages.
F I G U R E 3 Distribution of total relative abundance of proteins (in ng/μg total protein) assigned to tryptophan metabolism in I, A and E groups. Only the nonredundant proteins assigned to tryptophan metabolism found to be expressed in the two pools of five individuals each for each of the three well-defined age groups were considered. Data are the mean values of the two pools, with standard deviation shown. The figure was obtained using R script F I G U R E 4 Microbiota-based metabolism of tryptophan and indole. In brief, bacterial members of the microbiota synthesize tryptophan through TrpB tryptophan synthase; this metabolite is further degraded into the healthspan-related indole by the action of TnaA tryptophanase Whatever the case, tryptophan is known to play a fundamental role in health and neuroprotection (Fang, 2019;Platten, Nollen, Röhrig, Fallarino, & Opitz, 2019), as the kynurenine pathways use tryptophan to produce nicotinamide adenine dinucleotide (NAD + ).
This coenzyme is a longevity molecule, which can be metabolized F I G U R E 5 TnaA and TrpB relative abundances (in ng/μg total protein) and indole and tryptophan quantitative concentration (in mg/L faecal fluid) as a function of age. The data correspond to samples at time 0, 3 and 6 months of each of the 10 individuals and a total number of 83 samples-for details see Section 4. The fit was obtained using R script and the "Im" function, to extract a polynomial regression. Correlation and r 2 correspond to a polynomial regression model, whereas the grey zone represents the confidence value of 95% (meaning that the data within the grey are fit to the model with a confidence of 95%). Based on the fitting, the ages at which 50% and 90% reduction of initial values is reached are specifically shown F I G U R E 6 The relationship between TnaA and indole as well as TrpB and tryptophan abundances are responsible for the lower amount of both metabolites in elders. The samples considered and the data analysis (correlation and r 2 ) were as described in Figure 5. In the left panels, the fit for relationships between proteins and metabolites abundances is shown. In the right panels, box plot of the abundance of indole and tryptophan in the three well-defined age groups of individuals examined (extracted from Figure 5) is shown. Statistical evaluation of the differences between groups was carried out by a Mann-Whitney U test Based on the results presented and the importance of tryptophan and their side products, we suggest that nutritional intervention based on tryptophan supplementation may be required for a "healthy" longevity. The data suggest this intervention not being essential during infancy and early adulthood, as these essential molecules are produced by the corresponding gut microbiota at those ages, albeit at significantly lower level in adults. However, the "aged" microbiota in elderly individuals does not produce these molecules.

| Study participants and sample preparation strategy
Thirty healthy volunteers from Valencian Community (Spain) were involved in this study, and ten infants (I) were recruited between 2 and 5 years (age average 3.9 ± 1.45), ten adults (A) at 25-45 years (age average 35.4 ± 6.59) and ten elderly ( Samples, hereafter referred to as "faecal suspension," were stored at −80°C in freezers until further processing. For this study, three groups (I, A and E) were analysed after pooling. Five individual samples from the same group were pooled to constitute a pool ( Figure S1). For that, 1-ml of the "faecal suspension" of each of the five individuals were pooled to yield a total of 5-ml.
(NOTE: only the initial samples (time 0) for each of the ten volunteers per groups were considered for pooling).

| Protein extraction
The 5-ml pool suspension obtained as above was lyophilized, and the freeze-dried material was dissolved with chaotropic lysis buffer (

| Proteomics data analysis and sequence search
Mass spectrometry data obtained were processed using PeakView v2.2 Software (SCIEX) and exported as mgf files which were searched using Mascot Server v2.5.1 (Matrix Science) against a high-quality reference microbiome protein sequence database (Li et al., 2014;Pasolli et al., 2019;Zou et al., 2019). Sequences and ID codes, and functional and taxonomic assignations are detailed elsewhere (https ://db.cngb.org/ micro biome/ genec atalo g/genec atalog_human/ ; Li et al., 2014;Pasolli et al., 2019;Zou et al., 2019). Data acquisition of the total number of identified peptide spectra matched for a given protein (referred to as peptide-to-spectrum matching (PSM)), and Exponentially Modified Protein Abundance Index (emPAI), were calculated for each of the proteins. The PSM and emPAI can be used as a relative quantitation score of the proteins in a complex mixture based on protein coverage by the peptide matches in a database search result. Here, the emPAI was used (Arike & Peil, 2014 and therefore the nemPAI value can resemble a relative protein abundance in "ng protein per total μg of protein" ).
Diversity parameters of active bacteria, namely, those actively synthesizing proteins, were calculated from raw proteomic data as described previously (Deusch et al., 2018). Parameters were calculated for each of the six separate pools (two for I group, two for A group and two from E group), and the results for each age-group were given as mean values and the standard deviation of each of the two pools.
When coarse-grained data were required for protein and functional biomarkers discovery, the following procedure was applied: relative protein abundances were calculated separately for each of the six pools (two for I group, two for A group and two from E group), the two-pools data were combined by only considering proteins that were present in both pools of the same age-group, and the average protein abundance and the standard deviation were calculated. These filtered data sets were the ones used for coarse-grained protein biomarker discovery analyses and for functional analysis of proteins, which was performed by assigning KO in order to analyse the varying abundance of specific proteins in certain pathways. KO assignations available elsewhere were used (Li et al., 2014;Pasolli et al., 2019;Zou et al., 2019).
Relative abundance of each KO was obtained by the sum of the relative abundances of all proteins assigned to each KO .
In case of individual proteomic analysis, paired two-sample t tests were used for pairwise comparisons of the relative abundance of proteins of interest (IBM SPSS Statistics, version 20.0. IBM Corp).

| Determination of indole and tryptophan in faecal fluid
The following reagents and standards have been used: acetonitrile (LC-MS grade, Sigma-Aldrich) and DL-tryptophan (Sigma-Aldrich). All solutions were prepared using MilliQ ® water (Millipore).
A total of 1 ml of each of the "faecal suspensions" per individual samples were sonicated using a pin Sonicator ® 3000 (Misonix) for a total time of 20 s (10 W) on ice three times (with 20 s stop between cycles) and centrifuged at 15,000 g for 15 min at 4°C and the supernatant was retained and used for indole and tryptophan quantification.
The corresponding supernatants were transferred to analytical vials.
Indole was analysed using the Kovács reagent (Sigma-Aldrich) as described elsewhere (Darkoh, Chappell, Gonzales, & Okhuysen, 2015)). For tryptophan determinations, a series of calibration samples were prepared from a solution of acetonitrile:water  in the design or conduct of the study, the analysis and interpretation of the results, the writing of the report, or the decision to publish.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are available from the authors upon reasonable request.