This post compares two similar studies assessing the influence of exercise on the GI microbiome.    The pediatric study took place in León, Spain. The adult study originated out of the University of Hong Kong. The original thought was to ask if these papers say the same thing. The Hong Kong study, all 42 pages of it, turned into a productive rabbit hole with links to other posts on this site. Sometimes a person finds a Wonder Land down a rabbit hole. This study took a lot of time to analyze, but is was worth it!

The pediatric exercise study [1]

This is the protocol of the pediatric study. [1]

1. The Proteobacteria phylum, of E coli fame

Proteobacteria include Escherichia coli.

Fig. 1 Comparison of the phylotypes between healthy control children (Hc) and obese patients assigned randomly to a control (Oc) ora training (Oe) group at the beginning (t = 0) and at the end of the study (t = 12 weeks) at the phylum level. a Bar graphs representthe relative abundance of the total population. b Box plot summarizing significant differences using the Kruskal–Wallis test followed by the Mann–Whitney U test (p < 0.05). *Hc vs Oc (t = 0), aHc vs Oe (t = 0), bHc vs Oc (t = 12 weeks), dOc (t = 0) vs Oe (t = 12 weeks), #Oe (t = 0) vs Oe (t = 12 weeks).

3. The Gammaproteobacter class was the only significant change

This figure was modified from a PDF file so that it might actually be readable. Enterbacterales are a family in phylogenetic class.

4. Some little known genera really change in ways that don’t make sence

An attempt was made to establish the phylogenetic classification of these genera of bacteria that change between the healthy controls and the before and after exercise obese groups. There really is not that much information on say, Paraprevotella. To make matters worse, the before “no exercise” group is more like the healthy controls than the other treatment groups.

Fig. 4 Box plots represent the differences at the genus level among heathy control children (Hc), obese control patients (Oc) and trained obese patients (Oe) at the beginning (t = 0) and at the end of the study (t = 12 weeks) using the Kruskal–Wallis test followed by the Mann–Whitney U test (p < 0.05). *Hc vs Oc (t = 0), aHc vs Oe (t = 0), bHc vs Oc (t = 12 weeks), cHc vs Oe (t = 12 weeks).

Note, except for Akkermansia, genera that are not significantly different have been edited out of the Figure 4 of this publication. [1]

5. Metabolites, what they are making, is more important they who is there

A total of 30 fecal metabolites were detected in the feces of the participants.  These included bile acids, short-chain fatty acids (SCFAs), free fatty acids, amino acids, carbohydrates, nucleotides, and organic acids. A PLS- discriminant analysis (DA)  method was performed to discriminate between the healthy controls and obese before the start of the intervention, Figure 5a, and shows a clear cluster formed with metabolites from all obese patients at the beginning of the study, while healthy children were dispersed.

Fig. 5 Relationship between fecal microbiota composition and metabolic profile. a Partial least squares-discriminant analysis (PLS-DA) of metabolites from healthy control children (Hc) and obese patients (O) at the beginning of the study. b PLS-DA showing the exercise performance effect on the metabolic profile in pediatric obese patients. Colored ellipses represent the 95% confidence range for the indicated experimental group. The explained variance of each component is shown in parentheses on the corresponding axis. Figure 5b compares the before and after exercise small molecule profile of the feces.  The two are not as separate as healthy versus obese at the start of the study, but some changes in the fecal small molecules are dispersing before versus after.  These factors are branched-chain amino acids such as isoleucine and leucine (reduction) and formate and alanine (moderate reduction).  Sugars xylose, glucose, and galactose were decreased after exercise.

Fig. 5 Relationship between fecal microbiota composition and metabolic profile.  This figure has been rotated to the right from the original publication.  c Heat map of the correlations between fecal bacterial populations at the genus and metabolite levels considering the results obtained longitudinally in all groups. Each square represents the Spearman’s correlation coefficient (p < 0.05). Red and blue cells specify positive and negative correlations. p values are corrected for multiple comparisons based on the false discovery rate (FDR).

Some theoretical regression graphs have been added just to make the point that a Spearman’s correlation coefficient of 1.0 or -1.0 represents an extremely tight direct, or indirect, correlation.

6. A look at protein makers of host inflammation.

• NLRP3 is a subunit in the interleukin producing inflammasome.
• CASP1, caspace 1 is a protease and the last effector of the inflammasome complex.  It is the active caspase 1 that was detected. The authors proposed that the inflammasome detects danger signals associated with obesity.
• OPN, osteopontin, is a protein that the authors claimed to be known to maintain the inflammatory state in obesity.  The authors cited references claiming a role of osteopontin in enabling adipose tissue inflammation.
• TLR4, or the Toll like receptor 4, is a receptor for lipopolysaccharide, a pro inflammatory component of the cell wall of Gram negative bacteria.  TLR4, the receptor for LPS, protein levels in blood revealed no significant decrease in the obese exercise gorup  after the 12-week training program, while the opposite pattern was observed in the Oc
• group at the end of the experimental period.

Note that these changes happened in the absence of any changes in the LPS content of the blood plasma. It is questionable if any of these changes are functionally significant.

A pause before proceeding…

Many of the figures in the Quiroga publication were modified to make them readable online. Perhaps the most reproducible changes are in the short chain fatty acid and branched chain amino acid profiles.

Of mice and pre-diabetic, adult Chinese men…[2]

The FMT into mice part of this study will just have to be addressed in a continuation post.

Liu and coworkers started out with the assumption that exercise is a good way of preventing pre-dieabetic, overweight, sedentary individuals. Not everyone who starts and exercise plant actually sees improvement in type 2 diabetes markers. Liu and coworkers hypothesized that those that don’t respond to exercise intervention have the wrong microbiome. To make things less complicated, the authors only considered males in their study, all Chinese at that. After the 12 weeks of strength, endurance, and flexibility training, they looked at type 2 diabetes for non-responders,

This image is from figure 1 of the Liu study. [1]. These authors must have been montioring HOMeostatic Assessment of Insulin Reistance along with fasting blood glucose, and so on. They might have noticed that most guys were losing weight… Probably, by the end of the study, it became apparent that fasting insulin levels and the body’s response to insulin, HOMA-IR were not improving for some guys.

Curious, when these guys were defined as “non responders” they authors saw that they improved as well as the other guys in terms of weight loss, percent body fat, lean mass, and expectantly, fasting glucose.

2. A tiny snap shot of big data phylogenetic analysis

This post is making no attempt to really explain the entirety of the large amount of data in Liu et al Figure 2. Sometimes things get lost in Big Data. There are some small points in Figure 2E worthy of mentioning.

• (2D) The blue oval encompasses data from responders whereas the red oval data points of the non responders. There are responder data points that do not intersect the non responder oval, but not vice versa.
• (2E) Significantly altered species (p < 0.05) caused by exercise intervention in R and NR, respectively. Fold change was defined as the ratio of relative microbial abundance after exercise to those at baseline. Note that Akkermansia munciphila numbers decrease nearly 10x in the exercise responders but don’t change in the exercise non-responders.. Moving down the list, Lachnospiraceae bacterium6 _1_63FAA increases 3.2x (translating from log₁₀ units) in response to exercise. Nothing happens to this species in the responders. These data are presented in a slightly different manner in this post.

3. Relating everything to species of bacteria

The authors deserve a lot of respect for the amount of data they analyzed. As a general note, this figure has been rotated 90o for presentation and to add greater emphasis on the dependent variables such as HOMA-IE and the seated row exercise, and so on.

Figure 3. Exercise-Induced Alterations of Microbial Species Are Closely Associated with Improvements of Clinical Indices Independent of Body Weight and Adiposity Heatmap of the Spearman’s correlation coefficients between changes in different clinical indices and taxonomic alterations caused by exercise intervention after adjustment for body weight, fat mass, and waist-to-hip ratio. +p < 0.05, *FDR < 0.1, and **FDR < 0.05. Yellow and purple headers indicate increased and decreased relative abundance, respectively. Only species with significant correlations (at least one based on FDR or two based on raw p value) were shown.

4. What the microbiome is doing, or not

This post will not get into the differential expressed gene sets in the feces of responders and non responders. Most notably absent from this list is genes for short chain fatty acid synthesis. [2] Panel 4aC of the Liu publication [2] has been highly edited to make it easier to follow. Yellow symbols have been added to the pathway map and the Fold Change box plot. Note that the expression of sulfur metabolism enzymes increases in the non responders. A arrow has been added to the change in the K08738 enzyme for liberation of sulfate from sulfur containing amino acids. Sulfate metabolism by Akkremansia and Desulfovibrio and Akkermansia metabolic switching. have been addressed in the links given.

This panel really requires some thoughtful study. We see shifts to GABA production in responders and glutamate production in non responders. Both of these are neurotransmitters. How is the enteric nervous system affected? How does GI transit time affect the microbiome?

D) Heatmap showing the microbial metabolites in fecal samples from R and NR before and after exercise intervention. The colors changing from blue to red indicate higher abundance

5. But the FMT mice are not exercising

The cartoon in Panel 5A showing the FMT protocol is not shown. The panels showing no change in body mass, CVO₂ max and exercise related parameters that did not change are not shown. What is shown are all of the glucose handling parameters that did change. The reader will have to consult the public access publication and enlarge it because the font is very difficult to read as it.

The trend seems to be that only the FMT from the 12 week exercise conditioned responders improves the pre-diabetic markers.

References

1. Quiroga R, Nistal E, Estébanez B, Porras D, Juárez-Fernández M, Martínez-Flórez S, García-Mediavilla MV, de Paz JA, González-Gallego J, Sánchez-Campos S, Cuevas MJ. Exercise training modulates the gut microbiota profile and impairs inflammatory signaling pathways in obese children. Exp Mol Med. 2020 Jul;52(7):1048-1061. PMC free article
2. Liu Y, Wang Y, Ni Y, Cheung CKY, Lam KSL, Wang Y, Xia Z, Ye D, Guo J, Tse MA, Panagiotou G, Xu A. Gut Microbiome Fermentation Determines the Efficacy of Exercise for Diabetes Prevention. Cell Metab. 2020 Jan 7;31(1):77-91.e5. free article