Introduction

Dietary fibers are among the most important factors affecting the gut microbiota. Colonic bacteria have enzymes to break down dietary fibers converting these to beneficial compounds such as acetic, lactic, propionic and butyric acids. The benefits of dietary fibers to gut health have been described in many reviews1,2. However, long-term data on the effects of specific dietary fibers on the microbiota of the same person are scarce. In the era of personal medicine, this could be the focus of health-related studies. Gut microbiota is a complex ecosystem dynamically changing over time depending on the lifestyle. Only a few studies have investigated the long-term fluctuations of the gut microbiota3,4. Thus, David et al.3 showed in their year-long study that the gut microbiota was stable except during periods of substantial diet change or illness. Johnson et al.4 demonstrated significant changes in the gut microbiomes of people with a high abundance of intestinal Prevotella within two weeks on a habitual diet.

Researchers aim to systematize and find similarities between diverse individual gut microbiotas to better interpret their interactions with the whole organism and start to apply this knowledge in personalized medicine and nutrition. The understanding of the healthy microbiota state is another topic to discuss. Several attempts to move in this direction came out with different outputs. Thus, the widely acceptable grouping into three major enterotypes has been proposed5. However, additional research has indicated that there may be two to four gradual shifts in enterotypes6 depending on the geographical region. Lately, Bulygin et al.6 demonstrated the absence of distinct and stable clusters in microbiomes while Tap et al.7 showed up to 30 distinctive key genera in different microbiota types. In general, Bacteroides to Prevotella ratio is discriminative in many microbiome clustering analyses7,8 with some microbiome types more prone to fluctuations than others4,9. Walker et al.10 demonstrated that the consumption of dietary fibers can specifically alter the colonic microbiota with consequences for human health. As an example, the resistant starch-enriched diet increased the abundance of starch-degrading acetate-producing ruminococci, which through cross-feeding could support the growth of butyrate-producing bacteria during the same intervention period10. At the same time, the effects of dietary fibers common to Nordic foods such as rye bread and oat flakes on the gut microbiota have not been studied thoroughly11,12,13. It has been shown that high-fiber rye product consumption influences weight loss and levels of inflammatory markers in serum but not in gut microbiota11,12,13. Oat and barley beta-glucans and pectins offer numerous health benefits, including the reduction of blood cholesterol and postprandial glucose response14,15. The improvement of health characteristics might be related to the changes in gut microbiota, which has not been studied in detail, hence, being the focus of the current study.

Intestinal and fecal pH are other important factors to consider that directly correlate with gut microbiota composition. The stability of colonic pH is crucial to ensure a healthy intestinal environment and to support the proper functioning and diversity of the gut microbiota. The pH of the large intestine ranges from 6.0 to 7.0 on average, being lower in the ascending and transverse colon and higher in the descending and rectosigmoid colon16. The pH level in the large intestine is primarily influenced by diet, as well as by the metabolism and absorption of fatty and bile acids17,18. Although there are conflicting studies on fecal pH, a high-fat diet and alcohol consumption tend to be linked with elevated intestinal and fecal pH. On the other hand, a low-fat diet rich in vegetables promotes the production of short-chain fatty acids by gut microbiota and is associated with lower intestinal and fecal pH19. In any case, changes in the gut microbiota affect the digestion of food and the production of metabolites, which in turn has an influence on gut luminal pH.

Despite the recent collection of a large amount of data on the human intestinal microbiome, the individual resilience of gut microbiota in response to consecutive consumption of various fibers during long-term intervention has not been thoroughly studied yet. We conducted a study to analyze the stability of distinct microbiotas during the baseline, daily for a week while consuming beta-glucan, and during the wash-out period after the high fiber supplementation. Additionally, we investigated the impact of rye bran and two blended dietary fibers on the fecal microbiome. This design provides a detailed daily overview of changes in gut microbiota during a long 21-weeks observation period and is suitable for assessing the rate of response of gut microbiota to dietary changes and the persistence of its outcome.

Methods

Study design

The study was conducted from January to June 2022 by the AS TFTAK and Elsavie OÜ in Estonia. Participants were selected according to a questionnaire and definite criteria. The questions concerned the participants’ eating habits, description of their health status (incl. gut health), and data on their general lifestyle. The exclusion criteria were as follows: antibiotic administration within 3 months before the study and any severe or chronic diseases (e.g., cancer, Crohn’s disease, ulcerative colitis), specific diets such as vegan, reported as mindful consumption of low carbohydrate (30–40%E) or high-fat (> 40%E), travel to subtropic or tropic area within the previous three months. Participants were asked to follow their usual eating habits throughout the study. Subjects who did not follow the study protocol were excluded from the study. In total, 19 volunteers filled in the questionnaire and 15 of them were chosen for the study. Finally, 12 individuals aged from 31 to 50 years (10 females and 2 males) completed the study. According to the questionnaires, none of the participants took pro- and prebiotic supplements during the entire study with one exception (Fu1) when the antibiotic was prescribed. The median BMI of the participants was 22.9 kg/m2. The study consisted of nine periods (Fig. 1). The one-week base period was followed by four three-week test periods with alternating two-week washout periods. The test periods included the consumption of oat beta-glucan, rye bran, and two fiber mixtures (“Red” and “Green”, one per period) in addition to the standard menu. The additional doses of dietary fibers were 6.8, 6.8, 8.7 and 8.5 g/day, respectively. “Red” fiber mix refers to the Elsavie fiber mixture “Calm your rumbly tummy” and a daily amount for participants consisted of 1.6 g oat fiber (40% dietary fiber of which 75% is beta-glucan), 4.7 g citrus fiber (90% dietary fiber) and 3.8 g psyllium husks (80% dietary fiber). “Green” fiber mix indicates the Elsavie fiber mixture “Feel good inside” and a daily amount for participants consisted of 2.6 g oat fiber, 1.2 g citrus fiber, 2.7 g psyllium husks, 3.2 g polydextrose (90% dietary fiber), 1.2 g inulin (85% dietary fiber) and 2.1 g rye bran (40% dietary fiber). The fibers were mixed in smoothies, porridge, yogurt or similar products according to participants’ preferences and consumed in portions during the day. Throughout the study (Fig. 1), participants kept detailed food diaries (Nutridata V10) and provided stool samples at the end of each period. Additionally, daily (seven days) stool samplings were collected within one week of the baseline period, during the first week of the beta-glucan supplement period, and one week after the beta-glucan consumption period. In total, 23–29 fecal samples for 16S rRNA gene sequencing, 9 fecal samples for metabolite analyses, and 147 food diaries were collected per participant. The study was approved by the Ethics Committee of Human Studies of The National Institute for Health Development in Estonia (No 1002, 13/01/2022) and registered at www.clinicaltrials.gov (NCT05424640, 21/06/2022) confirming that all experiments were performed in accordance with relevant guidelines and regulations. All participants signed written informed consent forms before the study.

Fig. 1
figure 1

Study design. The study lasted 21 weeks and consisted of nine periods. The one-week base period was followed by four three-week test periods with alternating two-week washout (WO) periods. The test periods included oat beta-glucan, rye bran, and two fiber mixtures “Red” and “Green” (one per period) in addition to the standard menu and the additional dose of dietary fibers of 6.8–8.7 g/day. In total, 12 participants completed the study. P1-P31 indicates the timepoints for fecal sample collection for microbiota analyses.

Full size image

Fecal sampling, analyses of pH and organic acids

Fecal samples were self-collected by the participants using swab and spoon kits and FecesCatcher from TagHemi (Zeijen, The Netherlands). Samples collected with the swab kit were used for 16S rRNA sequencing. These DNA/RNA Shield™ Collection Tubes with Swabs (Zymo Research, Irvine, CA, USA) stabilize the samples at the time of collection, inactivate pathogens and enable sample storage at ambient temperature. Fecal samples were stored at home at 4 °C until transport to the laboratory (within 24 h after sampling). Samples collected with the spoon kit were used for pH and organic acid measurements. For that, one to two grams of stool were collected immediately after defecation into a collection tube without any preservation reagent by an integrated scoop (Sarstedt, Germany). The tube with the sample was kept at − 20 °C until the frozen transportation in a cold container (Sarstedt, Germany) to the laboratory for no longer than two days (transportation time was less than 8 h). In the laboratory, the swab samples were stored at − 20 °C before the DNA extraction, while spoon samples were stored at − 80 °C for further analysis.

For pH and organic acid analyses, the fecal sample was diluted five times with deionized water. The fecal suspension was mixed vigorously and centrifuged at 10,000 g for 10 min to remove the undissolved particles. pH of the supernatant was measured at room temperature using a pH meter (Mettler Toledo, MP125, Switzerland), and electrode InLab Pro (Mettler Toledo, Switzerland, calibrated with pH buffers 4.01 and 7.00).

Concentrations of organic acids were determined from the fecal supernatant prepared as above and filtered through AmiconR Ultra-3K filters (cut-off 3 kDa) according to the manufacturer’s instructions (Millipore, USA) at 4 °C. The prepared samples were analyzed by high-performance liquid chromatography (HPLC, Alliance 2795 system, Waters, Milford, MA, USA), using BioRad HPX-87H column (Hercules, CA, USA) with isocratic elution of 0.005 M H2SO4 at a flow rate of 0.5 mL/min and at 35 °C. For quantification of the substances, refractive index (RI) (model 2414; Waters, USA) and UV (210 nm; model 2487; Waters, USA) detectors were used. The detection limit for the HPLC method was 0.1 mM. Analytical grade standards of acetate, butyrate, formate, isobutyrate, isovalerate, lactate, propionate, succinate and valerate were used for quantification.

Gut microbiota sequencing

Bacterial profiles in feces were measured using 16S rRNA gene amplicon sequencing technology (Illumina). For that, microbial DNA was extracted by ZymoBIOMICS DNA Miniprep Kit (Zymo Research, Irvine, CA, USA), sequencing library was prepared using 16S rRNA v4 region and Nextera XT Index Kit by MiSeq v2 kit and 2 × 150 bp read length (Illumina) as described previously20. DNA sequence data was analyzed by BION-meta software according to the authors’ instructions21 as described previously using SILVA reference 16S rRNA gene database (v138) for alignment of consensus reads22. All mapped taxa with relative abundance < 0.001 were discarded and considered as potential noise. The raw sequences obtained were demultiplexed and uploaded to the European Nucleotide Archive (ENA, www.ebi.ac.uk/ena) under the accession number PRJNA1033718.

Nutrition data processing

This study focuses on the personal-level associations between the consumption of dietary fibers and fecal characteristics: pH, organic acids, and microbiota composition. The nutrition data was analyzed using weight in grams of food products consumed per day recorded by a program Nutridata (Version 10, tap.nutridata.ee, National Institute for the Health Development, Estonia). The foods were categorized into 15 main groups based on the fiber composition and/or food type (berries, fruits, fats/oils, nuts/seeds, sauces, sweets, vegetables, dairy, egg, fish, meat, refined grain products, whole grain cereal products, alcoholic and non-alcoholic drinks) as described in Adamberg et al.22. The main food groups were divided into 47 sub-groups as shown in Supplementary Table S1. By grouping the foods, very general and heterogeneous groups such as porridge cereals or dried fruits were avoided. Instead, the foods were grouped based on the amounts of potential substrates for the gut bacteria. In cereal products, the types of dietary fibers such as beta-glucans in oat and barley or arabinoxylans in rye and wheat were considered. Foods containing polyols or lactose formed separate groups. Hence, the meals or food products containing several fiber types belonged to several groups.

Statistical analyses

Bacterial abundances exceeding the identification limit (corresponds to relative abundance > 0.001) were considered for further analysis. Αlpha-diversity (Shannon index) analysis was based on raw read counts and interindividual (n = 12) changes between periods were analyzed using the signed-rank Wilcoxon paired test. Enterotypes of all samples were determined using PCoA algorithm published by Arumugam et al.5. The differences were considered significant at p-values < 0.05.

For the determination of the stability of alpha-diversity or bacterial abundances in fecal microbiota, individual fecal samples (n = 23–29 per participant) were used. The range of values of each bacterium in the base period of each individual was set as normal fluctuation. To determine the significant changes in bacterial composition after the intervention periods of dietary fibers outliers compared to the base period were determined using a z-score (< − 3 or > 3). To find relationships between diet, fecal pH, organic acids, and bacterial composition Pearson correlations were calculated between all variable pairs (food groups, microbiome data, pH, and organic acid data) of each participant. Pearson correlations were considered significant if the p-value was smaller than 0.05.

For data analysis and statistics, Microsoft 365 Excel and R version 4.3.1 programs were used. The main R packages used for statistics and data analyses were indicspecies, dplyr, GGally, ggpubr, reshape2, rstatix and tidyverse. Graphics were built using RStudio, and the ggplot2 package was used for the visualization.

Results

The participants’ food consumption patterns did not align with the national food recommendations

Although the self-administered questionnaires completed at study recruitment did not reveal any specific dietary preferences or changes in macronutrient distribution, the analysis of food diaries showed that some participants had an unbalanced nutritional status. Nutritional data analysis showed that the baseline average carbohydrate and fat intake of the study participants did not follow the national dietary guidelines (Table 1). Average fat intake was higher than recommended (39.2%E), while daily carbohydrate intake was lower (41.4%E). However, the distribution of saturated and unsaturated fatty acids in the menu of most participants was balanced, despite the higher fat intake. Mean protein intake was in line with national recommendations of 10–20% E. There was considerable variation in fiber intake among participants. Although the median fiber intake was sufficient, half of the participants consumed less than 25 g of fiber per day.

Table 1 Description of the diet of 12 subjects during the baseline period of 7 days.
Full size table

Considering the above, the pre-defined study criteria for the nutritional value of the food consumed before the intervention started were revised and released. Additionally, the data indicated that individual food preferences remained consistent throughout the study period, but the range and selection of foods were notably affected by seasonal availability, particularly for fruits, berries, and vegetables.

Resilience of gut microbiota

Our previous data on several cohort studies (n = 152) on the gut microbiota of the Estonian population9,22,23 by 16S rRNA sequencing based on an enterotype clustering algorithm5 identified three definite microbiota types (Supplementary Figures S1). These clusters were named by the most dominant and discriminating genus in the given cluster – Prevotella (Pr), Bacteroides (Ba), and Christensenellaceae (Ch). The current study participants’ fecal microbiota distribution was similar to the last more representative clustering with several outliers (Fig. 2). There were six participants of the Prevotella-type microbiota (Pr) and three of the Bacteroides-type microbiota (Ba). Additionally, we observed three individual microbiotas, which neither matched the Ba- nor the Pr-type microbiotas. The dominant genera characteristic of those consortia were either Christensenellaceae (Ch), Phascolarctobacterium (Ph), or Fusicatenibacter (Fu). Thus, the current study participants can be considered as representative of the Estonian cohort group.

Fig. 2
figure 2

Fecal microbiota enterotypes of 12 participants based on the baseline period samples. Participants’ samples are named by the most abundant genus.

Full size image

All baseline samples from the same person were typically grouped into the same microbiota cluster excluding Pr3 and Pr5 whose microbiota patterns belonged to two clusters. Each type of microbiota can be characterized by its unique microbial patterns. In the Pr-type microbiota, the accompanying discriminating genera were Collinsella and Olsenella, while the Ba-type microbiota comprised higher abundances of Roseburia, Subdoligranulum, Tyzzerella and Adlercreutzia than other microbiota groups (Supplementary Figure S2). The remaining three microbiotas had individual compositions different from Pr and Ba-types. Characteristics of the three individual patterns were as follows: Ch-dominated microbiota had higher abundances of Anaerostipes, Eubacterium siraeum, Intestinimonas, Cloacibacillus and methane-producing microorganisms – Methanosphaera and Methanobrevibacter, Fu-dominated microbiota had higher levels of Bifidobacterium, Anaeroplasma and Turicibacter, Ph-dominated microbiota comprised increased levels of Bilophila, Senegalimassilia and Clostridium sensu stricto 1.

Overall, the composition of individual microbiota remained quite stable throughout the entire 21-week study, regardless of the increased fiber consumption (Fig. 3). However, we observed individual fluctuations in the composition of the fecal microbiota. The most abundant genera were present in all studied gut microbiotas in amounts of at least 70% of total bacteria. These genera were relatively stable during the base period and stayed in the range of normal fluctuations until the end of the study (last wash-out). The most stable genera were Blautia, Coprococcus and Faecalibacterium with average relative standard deviation (rel-Stdev) below 20%, while Prevotella and Christensenellaceae exhibited much higher fluctuations (rel-Stdev 100%) (Supplementary Figure S3). Comparing the bacterial abundances at baseline to that of the last wash-out period, about half of the studied gut microbiota types (Ba1, Ph1, Pr3, Pr4, Pr5) were relatively resilient as more than 90% bacteria remained to baseline level (absolute z-scores above 3). The consortium Pr2 was the least stable where 60% of genera were out of the normal range after the last wash-out.

Fig. 3
figure 3

Dynamics of relative bacterial abundances of the top 50 taxa of each participant over the whole study period. Each stacked bar illustrates the pattern of a single sample at time-points indicated as P1-P31 in Fig. 1.

Full size image

Modulation of gut microbiota by dietary fibers

In this study, the four-phase diet intervention was carried out (Fig. 1). Typically, the changes in the Pr-type microbiota were more pronounced compared to those of the Ba-type microbiota (Figs. 3, 4). Similarly, notable shifts in microbial diversity (Shannon index) were observed in individuals with Pr-type microbiota. Comparable shifts in microbial diversity were also observed in the participant with Ph-type microbiota. These observations highlight a significant effect of dietary fibers on the microbial composition dependent on the individual gut microbiota.

Fig. 4
figure 4

Shannon diversity of fecal microbiota after intervention and washing-out periods. Grey dots represent values after the base and washing out (WO) periods. Colored dots illustrate the Shannon diversity values after intervention periods: orange—beta-glucan, red—fiber mixture “Red”, green—fiber mixture “Green”, blue—rye bran).

Full size image

Longitudinal sampling over five months including two one-week sequential everyday sampling periods enabled the analysis of the dynamics of specific gut bacteria in response to four dietary fiber intervention periods. To analyze the rate of microbiota reaction on fiber intake, the effect of oat beta-glucan, the most studied dietary fiber, was followed in more detail. The abundance of Roseburia begun to rise three to seven days after the start of beta-glucan intake intervention only in three cases (Ba1, Ba2, Ph1) out of twelve (Fig. 5). Although Roseburia was an abundant genus in gut microbiota of all participants, it responded to beta-glucan only in some cases. On the contrary, the proportion of another abundant taxon Prevotella fluctuated significantly independent of the dietary fiber intervention. The participant Ph1 was an exception whose fecal Prevotella increased steadily over the first week of the beta-glucan period becoming a dominant genus (20%) by the end of the observational study. This is a single case of the microbiota type changing from Ph to Pr-type during the study. The Ba-type fecal microbiota remained relatively stable during the entire beta-glucan intervention period. However, butyric acid-producing bacteria Faecalibacterium or Agathobacter abundance fluctuations were seen in Ba-type microbiota (Supplementary Figure S3). For some participants (Ba3, Pr6), the abundance of Collinsella was decreased while the Bifidobacterium decreased for Ba3 and Fu1 in response to beta-glucan consumption. Thus, the response of increased beta-glucan consumption to the abundance of specific bacteria clearly depended on the individual microbiota composition but not on the particular taxon. Moreover, the alteration of gut microbiota was not always persistent in response to the continuous daily consumption of beta-glucan, and the effect sometimes disappeared after the finishing of dietary fiber intake.

Fig. 5
figure 5

Dynamics of relative bacterial abundances of the five taxa significantly changed during beta-glucan intervention. For each participant all samples over the whole study period are shown. Each stacked bar illustrates the pattern of a single sample at time-points indicated as P1-P31 in Fig. 1.

Full size image

To analyze the effects of dietary fibers on specific gut bacteria compared to the baseline period, a Wilcoxon test was performed (Fig. 6). We observed significant decreases in the appearance of Dorea, Marvinbryantia, and Intestininmonas taxa and increased level of Adlerecreutzia, and Eubacterium ventriosum group abundances in response to beta-glucan intake. “Red” dietary fibers had an impact on Roseburia, and Romboutsia levels, while “Green” mixture changed Subdoligranulum, Parabacteroides, Prevotella-9, and Streptococcus abundances. The distributions of Paraprevotella and Adlercreutzia were specifically modified by rye bran consumption.

Fig. 6
figure 6

Boxplot showing the distribution of relative bacterial abundances after each intervention period. Only significant changes after a specific dietary fiber period compared to the baseline period are shown. Color code: orange—beta-glucan, Red—fiber mixture Elsavie Red, Green—fiber mixture Elsavie Green, Blue—rye bran. Dots represent the outliers.

Full size image

Aside from dietary fiber interventions, changes in gut microbiota can be influenced by eating habits. For instance, in March during the beta-glucan period, there was a noticeable decrease in the consumption of meat and eggs, coupled with an increased fish intake (Supplementary Figure S4). During the same period, more fruits and less vegetables were consumed. These significant changes might mask the effect of increased beta-glucan consumption on the gut microbiota. During the rye bran intervention period (end of May and beginning of June), an increased intake of strawberries and cherries and a decreased intake of apples was recorded. As a result, the rise in fecal acetate concentration (Fig. 7) could potentially obscure the impact of rye fiber.

Fig. 7
figure 7

The effect of dietary fibers on fecal pH (black line) and concentrations of organic acids (μmol/g-feces, upper panel) and correlation between pH and concentrations of organic acids (μmol/g-feces, lower panel). Each bar or line represents the values after the end of the corresponding study period. Colors represent specific organic acids.

Full size image

Change of fecal pH and metabolite patterns in response to dietary fiber interventions

The fecal pH and the amounts of organic acids varied notably between the study subjects. The average pH of all fecal samples deviated between 7.3 and 8, with the min/max individual values from 6.1 to 8.3 (Fig. 7). The fecal pH of some participants (Ba3, Fu1, Ph1, Pr1) fluctuated in a wide range, while that of the others (Ba2, Pr5, Ch1) was relatively stable throughout the study. pH from the end of base and wash-out periods of the same person also varied, possibly because of microbiome instability, too short stabilization period or shifts in the diet.

In most cases, the decrease in fecal pH is correlated with an increase in the amounts of acetate, butyrate and propionate (Fig. 7). Acetate was the most influential acid regarding the decrease of fecal pH, although no significant effect was found between pH and consumption of certain fibers. For Fu1, enhanced acid production was observed during the beta-glucan period, while the rye bran period enhanced the acid production of Pr1 and Pr2. The highest amounts of fecal acetate were seen in the samples of Fu1, followed by those of Pr1 and Ph2. Depending on the fiber and subject, dietary fiber intervention resulted in a pH rise or drop. The fecal pH of participant Pr1 responded to all dietary interventions with a decrease of over 0.5 pH units. In half of the cases, fecal pH decreased during the beta-glucan intervention period (6 out of 12). However, in general, the drop in fecal pH was in accordance with the increase in total acids (Fig. 7). Only in three cases, we did not observe an association between pH change in a range of one unit and the total amount of organic acids (Fig. 7). The fiber interventions had the most impact on fecal pH for four participants Pr1, Pr4, Pr5 and Ba1, who also had the highest amounts of total organic acids, while in Ph1 and Fu1, the fecal pH and acid concentrations deviated largely (Fig. 7).

The highest variation in succinate concentration was observed in the fecal samples of Fu1, where the increase of succinate production by Fusicatenibacterium was reflected by an increased amount of succinate in this subject. A high abundance of two succinate-consuming was observed in Fu1 fecal microbiota (Phascolarctobacterium – 1.0–2.1% and Dialister 1.0–2.9%).

The composition of fecal microbiota was related to fecal pH

The availability of dietary fiber may provide an advantage to certain bacterial species able to degrade and metabolize this fiber. However, the metabolites may enhance the growth of other taxa, known as cross-feeding mechanisms. Although we cannot discriminate the cross-feeding mechanisms using correlation analysis, we can estimate the pairwise associations independently from dietary fibers and fermentation products or pH by using multiple data points from the same participant. Among the prevalent taxa, the abundance of Prevotella and Roseburia decreased as the fecal pH increased in most of the participants (Fig. 8). On the contrary, a positive association between pH and the abundance of Clostridium family XIII was observed (p < 0.05 for Pearson correlation for four participants). Relationships between the bacterial abundances and the concentrations of specific organic acids were subject-specific (Supplementary Table S2).

Fig. 8
figure 8

Relationships between fecal pH and organic acids or bacteria throughout the study. Bacteria and pH pairs were selected based on Pearson correlations (p < 0.05 for at least four participants, excl. pairs of Prevotella where p < 0.1 were selected as Prevotella was present in eight participants). The color indicates the participant, and each dot represents a sample taken at the end of the corresponding study period.

Full size image

Based on the subject-level analysis, we observed similar co-associated patterns of the same bacterial pairs (Supplementary Figure S4). This indicates mutual relationships between certain bacteria unrelated to the available substrates. Alistipes and Prevotella had the highest number of associations with other abundant genera or several prevalent taxa (Supplementary Figure S4, Supplementary Table S2). The abundance of Alistipes, the key genus in Fu-dominated microbiota, was positively associated with Bacteroides, Barnesiella and Paraprevotella. On the other side, Prevotella, which abundance significantly increased in certain cases (e.g. from 10 to 44% in Ba2), only showed negative associations with other bacteria. Moreover, Prevotella had negative associations with Christensenella, Clostridium Family XIII, Enterorhabdus and Senegalimassilia while Christensenella positively associated with those of Methanobrevibacter and Enterorhabdus. Among the Ph-dominant microbiota, Ruminococcaceae UCG-002, Phascolarctobacterium or Butyricimonas were the most co-associated bacteria.

Discussion

To the best of our knowledge, this study is the first to follow changes in the individual gut microbiota of healthy subjects by gathering up to 29 stool samples from each participant over five months. The results of this longitudinal interventional study on the consumption of four various single or blended dietary fibers stressed the high interindividual variability among the participants. Massive, collected data on the individual gut microbiome before the beginning of fiber consumption, during the intervention, and in wash-out periods showed that despite the personal unique changes in bacteria composition connected with increased fiber intake, they are temporary, and gut microbiota stabilizes to its basal level after the intervention ends.

There are only a few studies on the stability of the gut microbiota over long periods. By analysis of sequential fecal samples, Johnson et al.4 showed that species of the genera Bacteroides and Alistipes correlated with a stable gut microbiota while those of Prevotella and Dorea were characteristic of an unstable microbiota. The stability of the Ba-type microbiota and fluctuations of the Pr-type microbiota have been also shown in a large dataset analysis7. A study by Kovatcheva-Datchary et al.24 indicated that increased barley consumption led to higher abundance of fecal Prevotella and improved blood glucose tolerance in Pr-type subjects. Similarly, the diet supplemented with wheat arabinoxylo-oligosaccharides primarily modulated the Prevotella-positive fecal microbiota in elderly people25.

Besides commonly accepted Prevotella and Bacteroides enterotypes, we observed various more diverse populations of individual gut microbiotas. One of them was dominated by Fusicatenibacter. Fusicatenibacter can degrade several saccharides and produce acetate, lactate, formate and succinate26. It was shown that Fusicatenibacter can play an important role in cross-feeding processes by producing these intermediate metabolites. It may support the growth of propionate or butyrate-producing asaccharolytic bacteria such as Phascolarctobacterium27 and Dialister28 thereby enhancing the formation of propionate and butyrate. Fusicatenibacter has been associated with healthy phenotype, indicating better glucose tolerance29. Fusicatenibacter is found in lower abundance in the feces of people with Parkinson’s disease30, as well as in constipated31 and ulcerative colitis patients32. Another genus that can benefit from growing with Fusicatenibacter is Alistipes. Alistipes is a succinate-producing bacterium, hence, could also provide the substrate for propionate and butyrate-producing taxa. The dominance of Christensenellaceae and Phascolactobacterium and the incapability to form a definite cluster with accepted gut enterotypes in two other participants indicates the more complex nature and variability of individual gut microbiota and raises the question about the definition of healthy gut.

All dietary fibers consumed in this study resulted in significant changes in gut microbiota and its fermentation pattern. Nevertheless, we could not find common trends partly because of the limited number of participants or due to high individual variability between the participants. Also, the baseline fluctuations of several bacterial taxa were several folds overshadowed the up- or downshifts of bacterial abundances after the dietary fiber intervention. It emphasizes the necessity of collecting a greater number of samples per participant to draw more robust conclusions. Variability in eating patterns and the amount of dietary fiber could also have influenced the results of the study. Another limitation of this study is the predominance of middle-aged females. Our participant group is not unique in this regard as women are more often interested in a healthy lifestyle and capable of following the design of long intervention till the end. However, despite all these biases, we noticed significant shifts in individual microbiome behavior in response to a definite fiber intake. Thus, despite a relatively unchanged bacterial composition in Ba-type microbiota, the concentrations of their metabolites were altered. A similar prevalence of metabolic shifts over changes in the gut Ba-type microbiota has been observed by others33. The accompanying bacteria can play a crucial role in metabolic alterations. For instance, the Subdoligranulum is one of the genera whose species can produce various compounds such as predominant lactic and butyric acids, however, depending on the conditions, acetate and succinate can be formed34. In our study, Subdoligranulum was abundant (more than 5%) in the fecal samples of participant Ba3, with the most remarkable metabolic shifts including reduced succinate and increased butyrate production.

One of the variables that most importantly modulates the gut community is pH. We noticed lower fecal pH values accompanying higher amounts of fermentation products. High concentrations of bacterial metabolites can inhibit or promote the growth of sensitive. However, in some cases, pH change was detected without any changes in organic acid concentrations. It can be explained by bicarbonate secretion and absorption or processes driven by bile salt metabolism, which are tightly linked to fat content in the diet.

Responsiveness of the Pr-type microbiota to the dietary changes was more pronounced compared to that of the Ba-type microbiota, which can be explained by the better adaptation capacity of the latter to environmental changes, including pH shifts. According to in vitro data by Duncan et al.35, Bacteroides species are well adapted to the gut environment, even though their abundance might be three times lower at pH 5.5 compared to that at pH 6.7 (27% and 86%, respectively). On the contrary, butyrate-producing Eubacterium rectale (currently Agathobacter) became dominant at pH 5.5 (50%) but was below the detection level at pH 6.735. Our data also showed high abundancies of Bacteroides at pH range between 6 and 8 while butyric acid-producing Faecalibacterium and Coprococcus were rapidly washed out of the chemostat culture (dilution rate 0.05 1/h) above pH 7.536. These findings from in vitro studies have also been confirmed in nutrition surveys. Results from ten food intervention trials confirmed a negative correlation between relative abundances of butyrate-producing species Roseburia and Eubacterium rectale and fecal pH (LaBouyer et al.37), the phenomenon we also observed. For example, Prevotella-dominated gut microbiomes could respond to diets rich in pectin and beta-glucan by increasing the production of organic acids, thereby optimizing colonic pH. On the other hand, different dietary fibers may promote these beneficial changes in Bacteroides-dominated gut microbiota. This is an important step towards personalized nutrition and the mechanisms behind this phenomenon are worth further investigation.

In conclusion, it can be stated that the study subjects and the Estonian population predominantly exhibit two main microbiota types, dominated by either Prevotella or Bacteroides. However, it is important to note that the microbiota is not limited to only two enterotypes. Complex composition and individual variability characterize the individual nature of a healthy gut microbiome. Thus, specific community patterns dominant with Phascolarctobacterium, Fusicatenibacter, or Christensenellaceae were also described. Along with the major bacteria, several associated taxa were identified, and their concomitant dynamic changes during dietary fiber interventions were shown. The strategy to cluster gut microbiomes before nutrition guidance is a powerful tool to improve health outcomes, however, the biological meaning of the dynamic changes in the complex gut consortia needs to be elucidated in further studies. The results of this long-term study demonstrate that dietary choices have a specific and immediate impact on the gut microbiome, while the overall gut microbiota remains relatively stable. In investigating the effects of the intervention, it is crucial to consider the specific timepoints at which samples are used for analysis. Many studies typically gather data only at the endpoint of a designated period. However, in this study, daily microbiota samples revealed significant fluctuations. Consequently, some effects observed in endpoint analyses may merely reflect random variability rather than true intervention impacts.