• AGEING
• COLONIC MICROFLORA
• GLUCOSE METABOLISM
• LIPID METABOLISM

We read with interest the recent article by Ng et al. (Gut, 2022; 71:910–8), who reported inhibition of the gut microbiota trajectory in patients with autism spectrum disorder.1 Similarly, the human gut microbiota ages in adults.2 3 Transplantation of gut microbes from elderly hosts, compared with their younger counterparts, deteriorates recipients’ age-related metabolic alternations.4 5 However, the gut microbiota in humans could differ in its pace of ageing, namely, accelerated or delayed microbiota ageing, even among those with similar chronological ages (figure 1A); this process is analogous to biological ageing.6 Therefore, we wondered whether the gut microbiota ageing trajectory could be used as a biomarker for metabolic diseases in adults. We analysed the gut microbiota compositions of 6376 participants of a population-level survey7 with Quantitative Insights Into Microbial Ecology (QIIME),8 among whom the median chronological age of the healthy participants is 43 (IQR, 31–45, online supplemental table S1). A random forest algorithm was applied to the data of 1083 healthy individuals to model gut microbiota ageing in relation to chronological age. The model was then applied to subjects with metabolic diseases, matching healthy individuals by chronological age. See the online supplemental file 1 for additional information.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Gut microbiome ageing patterns in participants with metabolic diseases and their healthy counterparts. (A) Three different patterns of ageing trajectories of the gut microbiota. (B–D) Gut microbiota ageing trajectory in patients with hypertension, diabetes and obesity versus healthy individuals. The random forest model was trained with the microbiota features of healthy individuals and then applied to predict the microbiota age of participants with hypertension, diabetes and obesity in relation to their chronological their age separately.

At the modelling phase, the microbiota age of the 1083 healthy individuals showed a significant and positive correlation with chronological age (online supplemental figure S1). The microbiota age in the healthy individuals was not significantly associated with gender or residing sites (developed vs less developed), but is significantly associated with grain, fruits and rice wine consumption (online supplemental table S2 and online supplemental figure S2). When the model was applied to individuals with metabolic diseases, the ageing trajectory of their microbiota was neither accelerated nor delayed but intersected with that of healthy individuals at nearly 50 years old (figure 1B–D). In individuals with metabolic disease, the microbiota age was older than that in healthy individuals at younger chronological ages, comparable with that in healthy individuals at middle chronological ages, and younger than that in healthy individuals at older chronological ages (online supplemental figure S3). Metabolic diseases (including hypertension, diabetes and metabolic syndrome) patients who took antibiotics or medication showed significantly lower microbiota age than the treatment-naive patients (online supplemental figure S4). We removed the medicated patients and the above intersecting patterns could still be reproduced (online supplemental figure S5).

To dissect the random forest ageing model, we analysed the ageing trajectories of the top 20 model-contributing taxa (figure 2A). The heatmap clearly showed that the relative taxonomic abundances correlated positively (upper half) or negatively (lower half) with chronological ageing in the healthy individuals, but the trajectories were disordered in hypertensive individuals. Typical microbial examples include Clostridium and Parabacteroides distasonis, with obvious intersecting patterns (figure 2B,C). Similar disordered taxonomic ageing trajectories were also observed in participants with diabetes, obesity, metabolic syndrome, hypercholesterolemia and hypertriglyceridaemia (online supplemental figure S6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Significant intersecting trajectories of taxonomic relative abundances in healthy and hypertensive participants. (A) Heatmap of the top 20 important amplicon sequence variant from the random forest regression model plotted against chronological age in healthy and hypertensive individuals. Significant intersecting patterns of the relative abundances of (B) seq93_Clostridium and (C) seq15333_Parabacteroides distasonis in healthy individuals and hypertensive participants. The interaction p indicates the significance of interaction term in the multivariable linear regression model.

In conclusion, the results of the current and Ng’s study suggest that the ageing trajectory of the gut microbiota could be a potential biomarker for both paediatric and adult chronic diseases. However, accelerated microbiota ageing should not be simply considered a risk factor for metabolic diseases in adults or vice versa, as it might be chronological-age dependent. Our finding corresponds to Wilmanski’s observation that elderly individuals who carried younger microbiota signatures could have lower survival rates,9 but analyses to understand covariates (lifestyle, diet, medication, etc), subpopulation differences and longitudinal disease risks of gut microbiota ageing are warranted. Moreover, whether a chronological age-dependent bacterial function could be observed and understood in mechanism studies is worth investigating. Interestingly, a common practice in faecal microbiota transplantation (FMT) is to choose young donors for health reasons. Analysing whether a chronological age-matched healthy donor for FMT could promote additional health benefits in the recipient is worth further investigation.