At the time of the study, the male athlete was 18 years old and was studied over the course of a 31-week period during the 2019 race and training season preceding the world championship competition in U19 category of academic rowing. By this time, the subject had been undertaking rowing for 4 years and had not previously sought any conditioning or dietary advice. Furthermore, the athlete was not on any prescribed medication and was a non-smoker, and his usual diet was previously supplemented with whey protein [SiS (Science in Sport) Limited products] only. In the 7 months prior to the supplement intake, the subject held a normal diet that was alike on a daily basis, comprising mostly of meals that were high in carbohydrate and protein, medium in fat, and modest in dietary fiber. The athlete was fully informed of the study aims and confirmed participation in the study by signing a consent form, understanding that the parameters of health were not associated with the study and that the subject was not physically harmed by the study. This study was approved by Protobios (1-05/2019).

The primary goals of the support provided were to (a) get insights into the subject's intestinal microbial community during periods of high exercise and (b) examine the effects of dietary fiber intake to the bacterial compositions associated with energy production.

The participant was informed to maintain the usual dietary habits throughout the study. Body composition estimates were made preseason (sample #1) and post study (sample #3) with no substantial change (BMI 23.3 ± 0.2, fat percentage of 8.4 ± 0.0) according to the medical sports health survey records. Based on diet recall, the usual macronutrient intake was assessed using image-based dietary assessment software (NutriData, National Institute for Health Development, Estonia). We used this tool to ascertain the usual eating patterns of the subject including type, frequency, and amount of foods consumed. Foods consumed were matched to the nutritional analysis for the specific menu items that had been coded in NutriData. If not consumed from the menu, the item was coded against the most appropriate matching food. On average, across the study period, the athlete consumed 2,560 ± 750 kcal/day, and the estimated macronutrient intake was 23 ± 7.12% protein, 52 ± 19.1% carbohydrate, 26 ± 19.3% fat, and 15 ± 4.5 g fiber. The consumption of the nutrient supplement (Food, not only for thought, Elsavie, Estonia) started on week 27 and continued on a daily basis until week 31 (altogether 30 days), with intake immediately after breakfast, and the recommended daily intake [1.5 tablespoons (1.5 tbsp/20 g) mixed with water] was not exceeded. The dietary supplement provided to the participant as the prebiotic mix (20 g) included dietary fiber (8.79 g), consisting of resistant starch (2.25 g), arabinoxylan (2.05 g), citrus fiber (2 g), beta-glucans (1.03 g), inulin (1.03 g), and rye fiber (0.57 g). The athlete kept the training diary in Sportlyzer (Sportlyzer, Estonia). The 31-week high-training and intensive competition program is presented in Figure 1. The mean physical activity (8 months duration) of the subject during the study was 472 min/week of water rowing, 354 min/week of indoor rowing, and 60 min/week of stretching exercises (Figure 1). Since the time the first microbial sample (#1) was taken, the subject was participating in a series of national and international competition activities including five international and 11 local competitions (53.25 km total of total race distance) following the planned program as in Figure 1. The second sample (#2) was taken at week 27, by the end of the water season and at early weeks of the indoor rowing season. Thereafter, the subject began to take the dietary fiber supplement as recommended for 30 days. A third sample (#3) was taken at week 31. The seasonality of sampling was spring (#1), autumn (#2), and winter (#3), respectively.

Microbiome composition was determined on three occasions (week 1, week 27, and week 31, Figure 1). Samples were self-collected from morning stool samples by using the commercially available kit (INTEST.pro, BIOMES NGS GmbH, Germany) in accordance with the specifications laid out by the manufacturer. The first two samples were taken at normal nutrition (at the start of and after the active competition period), followed by 1 month of dietary supplement intake to investigate the dynamics of the intestinal microbiome and the effects of the fiber supplement on the microbiota. Collected samples were transported to the lab according to the service provider's instructions where the microbiome composition was analyzed via 16S rRNA gene amplification and sequencing by BIOMES NGS GmbH (Germany). In brief, microbial genomic DNA from fecal material was extracted by the bead-beating technique, the V3–V4 region of the 16S rRNA gene was amplified, and sequencing was performed on the Illumina MiSeq platform using a 2 × 300-bp paired-end protocol (Illumina, San Diego, CA, USA). These DNA sequencing techniques were then used to generate data outputs that provided a comprehensive bacterial taxonomic profile of the subject (28) in comparison with the average microbiome of the European population as the evidence indicates that microbiome may vary by geography (29).

Different packages of MS Excel (based on MS Excel 2011) and licensed MedCalc (version 19.1.6) statistical analysis programs were used for taxonomic, functional analysis, and visualization of bacterial composition data obtained from microbiome sample study reports. Spearman's correlation coefficients r_s were calculated for comparing abundances of genera at two time points. Statistical significance of differences of Shannon's indices (α-diversity) across time points was assessed using Hutcheson's modified t-test with a significance level of p < 0.05. Statistical significance of differences between the athlete's characteristics and the control group (general population) was assessed using the single mean t-test with a significance level of p < 0.05. Relative abundance values of bacteria on genus and species level at different time points were compared using non-parametric Wilcoxon rank sum tests (also named the Mann–Whitney U-test or Mann–Whitney–Wilcoxon test) with a significance level of p < 0.05.

The Shannon index indicating the diversity of bacterial families present in the samples of the subject ranged from 1.67 to 2.11 (Figure 2). At the beginning and end of the training period, the Shannon indices were similar (2.11 and 2.08, respectively), concluding that the microbiome diversity did not change significantly at times of high competition (Hutcheson's modified t-test, p-value > 0.05). After the dietary fiber mix intake, the Shannon index dropped to 1.67 with a significant decrease in community diversity (Hutcheson's modified t-test, p-value < 0.05). The Shannon index of the control group was 1.63 (data not shown).

The evenness of the distribution of species in communities showed that at high exercise, microbiome uniformity indices were similar (J1 = 0.54 vs. J2 = 0.53), whereas, after the fiber intake, J showed a substantial drop (to J3 = 0.42). Although, also the evenness values of the microbiota were the lowest after the fiber mix diet (changed from 0.53 to 0.42), the statistical significance of the reduction could not be concluded. The uniformity index for the control group was J = 0.41. These data allowed us to conclude that at the family level, the athlete's microbiome at high training and competition loads (samples #1 and #2) were more diverse and more balanced (even) than after the dietary fiber consumption (sample #3). Intense exercise accompanied by a high intake of dietary fiber did not lead to the increased diversity of gut microbiota as was initially expected.

As for the phylogenic composition, a total of nine phyla were present in all the samples with the dominance of Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria (Figure 3). The Firmicutes [67.8 ± 6.2 (mean abundance across three time points ± SD)] outranked Bacteroidetes (18.0 ± 1.8), Actinobacteria (8.8 ± 1.7), Proteobacteria (2.7 ± 1.6), Verrucomicrobia (1.6 ± 1.2), and Cyanobacteria (1.0 ± 1.4) phyla. Dietary fiber consumption had a positive effect on the abundance of Firmicutes (+20%), whereas, it showed drastic negative effects on Verrucomicrobia and Cyanobacteria, whose drop in abundance was almost 100%, and on Proteobacteria, Bacteroidetes, and Actinobacteria that declined by 75, 18, and 13%, respectively (Figure 3B). Analysis of Firmicutes/Bacteroidetes (F/B) ratio values showed relative stability during the study, where during the intense competition period, the reduction in Firmicutes [changing the F/B ratio by 17.9% (from 3.78 to 3.11)] was rescued upon fiber consumption by increases in abundance by 20% and resulting in the F/B ratio of 4.55 (Figure 3C). Compared with the control group, the F/B ratio of the athlete was significantly higher in all time points (t-test of one mean, p-value < 0.05, Figure 3D). Altogether, these results showed that dietary fiber along with high exercise loads affected the phylogenic composition by the microbiome of the athlete becoming relatively poorer at the phylum level. Overall, these results were in good agreement with data showing that training promoted relative increases in Firmicutes (30, 31) and that F/B ratio correlated significantly with cardiorespiratory fitness (32).

The relative abundance of 77 genera, of which 29 were shared across all samples, was significantly different between samples (p-values < 0.05, Wilcoxon rank sum test, Figure 4). At the genus level, Prevotella [11.7 ± 2.1 (mean abundance across three time points ± SD)], Faecalibacter (5.1 ± 1.8), Blautia (5.4 ± 1.2), Ruminococcus (3.8 ± 2.8), and Bifidobacterium (5.0 ± 3.8) were the most abundant genera (Figure 4). The predominance of Prevotella compared with the families of Bacteroides and Ruminococcus indicated that the subject had Prevotella-predominant enterotype, e.g., enterotype II (Figure 4A). Prevotella's abundance was associated with long-term fiber intake (33).Similar trends were noted also in the current study whereupon dietary fiber intake resulted in enhanced abundance of Prevotella that became 41.7% more abundant as compared with the previous time point (Figure 4A). It is known that extreme dietary changes can lead to wide-ranging shifts in the gut bacterial community (34). Herein, the relative abundance of the acetate-producing bacteria (e.g., Blautia, Bifidobacterium, Sutterella groups) and the lactate-producing bacteria (e.g., Bifidobacterium, Streptococcus, Lactococcus groups) increased during high training and competition period (sample #1 vs. #2), but showed decreasing patterns (except Blautia) by the end of the dietary supplement intake period when acetate- and lactate-consuming and butyrate-producing genera including Faecalibacterium increased significantly (sample #2 vs. #3, p-value < 0.05, Wilcoxon rank sum test, Figures 4B,C). Interestingly, the propionate-producing genera (35) showed differential patterns. Bacteroides and Acidaminococcus showed increasing trends in contrast to Phascolarctobacterium and Veillonella that decreased upon competition, with only the levels of Veillonella slightly rescued upon fiber consumption (Figures 4B,C).

The abundance of the shared largest 29 genera was better correlated in samples #1 and #2 (Spearman's correlation coefficient r_s = 0.87) as compared with that in samples #2 and #3 (rs = 0.76, Figure 4D). However, the strongest correlation in abundance of these major genera was observed between baseline and endpoint (samples #1 and #3, r_s = 0.92). This result was somewhat surprising. Similar to other studies (36), we also noted strong patterns of individuality of the response to exercise and diet, with a possible explanation that these activities supported the original (primary, “keystone”) bacterial community dynamics of the subject.

Taken together, high exercise along with dietary fiber intake resulted in dynamic shifts in genera composition especially in the balance of lactate- and acetate/butyrate-producing bacteria.

Next, we compared the mean relative abundance of 32 individual species to identify those that were the most affected by the dietary switch. Among the studied species, the six most abundant bacterial species (Prevotella copri, Faecalibacterium prausnitzii, Akkermansia muciniphila, Bifidobacterium adolescentis, Coprococcus eutactus, Collinsella aerofaciens) accounted for 92.5% on average of the abundance of the top species. For most of these analyzed species, a notable variation was associated both with the intense exercise loads and the dietary fiber consumption (Figure 5). Interestingly, A. muciniphila (Verrucomicrobia) that produces both propionate and acetate (37, 38) showed decreased proportions upon fiber consumption and was replaced by the abundance of the butyrate producer C. aerofaciens (Figure 5A). Another major shift upon fiber consumption was noted in the abundance of C. eutactus with known beneficial effects on butyrate production (39). However, the abundance of 12 species attributed with a protective role on the intestinal mucosa reduced significantly after dietary fiber intake (p-value < 0.05, Wilcoxon rank sum test), from on average 50.4 to 33.5% of abundance of the detected species (Figure 5B). Herein, two of the five species which were among the most significantly affected were the Bacteroidetes spp. Also, the species of Akkermansia, Bacteroides, Bifidobacterium, and Ruminococcus with protective functions on the intestinal mucosa showed a significant decrease upon fiber intake, whereas, F. prausnitzii, one of the major manufacturers of butyrate, showed increased abundance upon dietary fiber (Figure 5B). Veillonella dispar species were specifically monitored because of their potential impact on performance enhancement as a lactic acid-utilizing community (6). Although, changes in V. dispar abundance during periods of high exercise load and dietary fiber intake were detected, the statistical significance of these findings could not be determined (Figure 5C). It is noteworthy that the number of V. dispar in the microbiome of the young athlete was significantly higher than of the control group (Figures 5C,D, single mean t-test, p-value < 0.05). The observed changes in the abundance of bacteria producing SCFAs associated with energy consumption of the skeletal muscle (40) supported the initial work hypothesis that dietary fiber intake could facilitate athletic endurance by favorable shifts in microbial composition.

The major strength of the study was that it was conducted in “real-life” scenario as the temporal dynamics of the athlete's microbiota was explored by combining high-endurance exercise specifically with the athlete-designed dietary fiber supplement. The benefits of this study may lead to new insights into the cumulative effects a particular physiological interference has on the gut microbiome, that is on the role of the host's enterotype (with defined keystone species) has on the covariation of microbial communities upon dietary shifts and at high exercise loads. Finally, this is an individual athlete's study, and as such, it does not allow to draw solid and supportive conclusions. The value of the study lies in the aspect that overall variability in the physiological response of athletes to training and nutrition has not yet been adequately explored.

Firstly, it was not possible to fully control dietary intake, although, the participant was instructed to maintain normal habits. Secondly, the study did not examine the causal relationship between exercise performance and the gut microbiota of the athlete, although, a high number of intestinal bacteria of the Veillonella family that would be of advantage to the athlete were observed. Also, recording of metrics inflammation and immunosuppression could have helped to examine the microbial gut stress levels of the athlete. Finally, given that microbiome sequencing had a limited capacity to resolve taxa to the species level, therefore, the study focused on the proportionally largest reservoir of multiple species' diversity and functionality.