The majority of human gut microbes belong to Firmicutes and Bacteroidetes phyla, with Actinobacteria phylum being a minor but essential component of the gut microbiome [1
belonging to this phylum have been known to be essential and beneficial inhabitants of the human gut long before the era of molecular-genetic technologies. The clade plays important roles in vitamin production, protection against pathogens, regulation of immune system and lactose utilization. Bifidobacteria
confer functional benefits by cross-feeding other members of gut microbiota specializing on production of butyrate, an essential substance for colon epithelial cells with anti-inflammatory and anti-cancer properties [2
]. Low levels of bifidobacteria are associated with various adverse clinical conditions [3
]. Its abundance can be increased by consuming fermented dairy products (FDP) containing live microbes, probiotic supplements or by supporting bifidobacteria with prebiotics [6
]. Introduction of probiotic strains of bifidobacteria to human gut has been reported to improve clinical status in diseases like antibiotic-associated diarrhea [7
], necrotizing enterocolitis [9
], chronic pouchitis [10
]. Improvement has also been reported in allergic diseases including atopic eczema [11
] allergic rhinitis [13
] and allergic diarrhea [14
With the advent of 16S rRNA gene sequencing as a routine scientific method, it is now possible to investigate microbiota-mediated impact of probiotic strains of Bifidobacteria
and probiotic-fortified food products on human gut microbiome in more detail. Particularly, the approach allows efficient characterization of interactions between probiotic microbes and gut microbial species. Although recent culturomics efforts have succeeded in capturing the majority of gut microbial species [15
], in clinical studies the 16S rRNA gene sequencing is still superior to cultivation-based approaches as it allows to obtain information about the total community composition in an economic and high-throughput way. A recent survey demonstrated that intake of probiotic-fortified fermented milk products not only decreased levels of commonly associated gut pathobiont species but also directly improved the production of short chain fatty acids (SCFA)—possible biomarkers associated with healthy gut function [16
]. Interestingly microbiome changes in studies of patients with irritable bowel syndrome paralleled improvement in reported disease symptoms [16
]. Another probiotic dietary intervention study revealed that the persistence of probiotic strains in the gut of healthy subjects depends on the initial composition of the microbiota [18
]. Prediction of the individual response of gut microbiota to specific probiotics may allow development of personalized nutrition schemes intended to promote or maintain human wellbeing.
Here we examined the effect of a fortified fermented dairy product (FDP) intake on human gut microbiota composition using 16S rRNA sequencing of stool samples. In this controlled study, volunteers consumed FDP for 30 days; clinical data and stool samples were collected on the first and last days of the study.
16S rRNA gene sequencing yielded 34,663 ± 9641 reads per sample. Sequencing statistics and metadata are listed in the Table S2
. The fraction of the identified reads was 96.6 ± 2.6% confirming the high quality of the sequencing data and applicability of the selected classification algorithm (see Methods). In total 54 families, 126 genera and 519 species were detected in at least one sample (see Table S3
). Analysis of microbial community richness dynamics for each volunteer showed that there was no significant change in alpha-diversity after FDP intake (Shannon index 5.6 ± 0.6 and 5.6 ± 0.1, p
= 0.68, paired Welch’s test).
Individual microbial species cooperate with the other species within the gut community, making trophic chains and other connections, thus forming symbiotic subcommunities (cooperatives) [30
]. In order to determine the effect of FDP intake at this level, the cooperatives were identified based on correlation analysis of the abundance of microbial genera for all samples corresponding to Visits 0 and 3 (see Methods). As a result, five large potential cooperatives were identified along with a few smaller ones (see Figure 1
). Complete sets of genera in each cooperative are listed in Table S4
Comparison of overall microbial community structure before and after FDP consumption adjusted for the significantly associated factors showed that the gut microbiota composition of volunteers was significantly changed (PERMANOVA test, Bray-Curtis metric on the level of genera, p = 0.006, R2 = 0.51%), although the degree of the change was moderate—on average 1.4 times lower than the mean group variability level (Bray-Curtis measure: 0.41 ± 0.14 between the paired samples vs. 0.53 ± 0.11 between all possible samples, p = 0, Welch’s test).
Paired comparison of the abundance of individual microbial taxa before and after the intervention using metagenomeSeq [31
] (see Supplementary Materials
) revealed 39 taxa as significantly increased and 24, decreased (see Table S5
). At a family level, Coriobacteriaceae, Bifidobacteriaceae, Staphylococcaceae, and Erysipelotrichaceae, increased their fraction. Moreover, the list of significantly increased genera and species included Bifidobacterium
, B. adolescentis
, B. animalis
, B. bifidum
, B. longum
), and other taxa. However, only two genera and one family decrease (Lachnoclostridium/unclassified
genera and Acidaminococcaceae family) were associated with the FDP consumption. Interestingly, when a similar analysis was performed at the level of microbial cooperatives, no significant changes were detected for any of the cooperatives.
The impact of the FDP intake on microbiota of the volunteers was also assessed at the level of functions—via the analysis of changes in relative abundance of metabolic pathways (see Methods). In total two pathways were significantly increased and 24, decreased (see Table S6
). The pathways with the most profound changes (for which the highest fraction of genes were affected) included an increased “Phosphotransferase system (PTS)” pathway—the transport systems specific for the Firmicutes phylum—and also the decreased pathways “Bacterial chemotaxis” and “Flagellar assembly”—reflecting the effects of the decreased Gram-negative:Gram-positive microbes ratio after the FDP consumption. Moreover, there was an increase in pathways associated with starch and simple sugars transport and amino acids synthesis. At the module level (see Table S7
), FDP intake was associated with increased lactose transport system genes (PTS system, lactose-specific II component)—in agreement with the observed fraction of lactose-fermenting bacteria at the taxonomic level. Among the decreased modules, there is a module related to the synthesis of lipopolysaccharides (Lipopolysaccharide biosynthesis, KDO2-lipid A)—immunogenic components of Gram-negative bacteria cell walls.
Response of gut microbiota community structure to intake of probiotics can vary across individuals [18
]. In order to explore variation of response to FDP intake on subject level for our cohort, firstly we assessed the changes in relative abundance of the major target group of microbes that was expected to react—lactose-fermenting taxa (LFT): the list included Bifidobacterium, Lactobacillus
and unclassified Enterobacteriaceae
(see Methods). Cluster analysis of the subjects’ microbiomes before the FDP course based on the extent of changes in total LFT abundance showed that the group of volunteers formed two clusters (ASW = 0.2)—cluster #1 (n
= 75 subjects) and cluster #2 (n
= 58). Therefore, subjects can be divided into two groups in which the pool of LFT demonstrated two different types of response to FDP consumption. For subjects from the cluster #1, the microbiota manifested a significantly weaker increase of the levels of LFT in comparison with the cluster #2 (change −0.10 ± 1.20% vs. 0.51 ± 1.26, respectively, p
= 0, Welch’s test).
Next, we compared the clusters by the change in total taxonomic composition (not just LFT) after FDP consumption. The total taxonomic composition for members of cluster #1 did not change significantly (PERMANOVA test, p
= 0.0813, R2
= 0.61%), while for cluster #2 the change was significant (p
= 0.0004, R2
= 2.07%). Further, changes in composition were compared between clusters on a more detailed level, for individual taxa and cooperatives; results suggest that the two clusters are different by the type of response of not just LFT but also the other microbial taxa (see Figure 2
, Table S8
Cluster #1 showed fewer changes (n
= 9 taxa/cooperatives), the most pronounced of which was a decrease of Lactococcus
genus abundance (including L. plantarum/raffinolactis
). For cluster #2, the number of affected taxa/cooperatives was higher (n
= 108). Therefore, cluster #2 can be considered responders in comparison with the other samples (i.e., cluster #1) (Figure 2
After the responders were determined as members of cluster #2, we sought to identify distinctive features of the responders’ microbiota that might be predictive in the general population. For this purpose, baseline microbiota composition (before intervention) was compared between responders and non-responders using the MaAsLin method (see Methods), results are listed in Table 1
These results show that the microbiota of responders contains a lower fraction of lactose-fermenting taxa, while the fraction of taxa from cooperative #2, particularly, the members of Bacteroidaceae, is increased (Noteworthy, this effect cannot be explained solely by the compositionality of the microbiome data because the microbiota of the volunteers contains many other species besides LFT and members of cooperative #2). However, none of the physiological factors significantly differed between the two clusters at baseline.
Semiquantitative microbiota composition profiles obtained using metagenomic analysis of stool samples give the most complete picture of gut microbial community structure independently of whether the species are cultivable. These profiles were used to estimate the change in gut microbiota composition after FDP consumption.
The lists of taxa differentially abundant before and after FDP consumption show significant overlap between multiple methods of statistical analysis, thus confirming the validity of the findings. Among the decreased taxa, there are various species from the Firmicutes and Bacteroidetes phyla, usually comprising up to 90% of total bacterial abundance in gut microbiota of healthy people [32
]. At the same time, there was a pronounced increase in the abundance of the third most dominant phylum, Actinobacteria, including Bifidobacterium
. Many members of this genus are probiotic microorganisms, and their role in anti-inflammatory activity, protection from pathogenic microorganisms and vitamin production has been noted by others [33
]. Interestingly, recent study has shown that this genus commonly associated with infant microbiota can also dominate microbiome of adult individuals from some world populations [34
]. The Bifidobacterium
is among the genera that increased their abundance significantly after FDP consumption. Noteworthy, in addition to the increase of B. animalis
fraction, there was also a significant increase in the abundance of another bifidobacteria—B. bifidum, B. adolescentis, B. animalis, B. longum
—that were not included in the FDP starter culture composition. This indicates that the FDP consumption not just leads to increased presence of B. animalis
due to its direct introduction but also potentially affects the ecology of the gut microbiome by supporting its resident bifidobacterial species. The Streptococcus
genus (including Streptococcus thermophilus
, a component of the starter culture) was also increased in abundance. Further studies including a control group consuming a placebo product with identical formulation but lacking the probiotic are required to dissect the effects of the probiotic from the effects of the fermented product itself. Additionally, microbiome analysis of stool samples collected several weeks after the end of FDP consumption will provide clues to assess the persistence of the observed shifts in species populations.
Interestingly, there was an increase of other Actinobacteria
, including several species of the Coriobacteriaceae
family. This effect could be attributed to the increased levels of lactose in the diet provided by regular intake of the test product. These taxa have a specific ability to metabolize lactose to lactate; meanwhile, a part of the lactose originally derived from dairy products remains intact during preparation of FDP, is incompletely digested in small intestine and reaches the large intestine. Actinobacteria
possess specific metabolic function that contribute to general human health by participation in metabolism of food components that increase antioxidant capacity. Specifically, Adlercreutzia equolifaciens
and Slackia isoflavoniconvertens
are active participants in isoflavone metabolism. The main dietary source of these substances are legumes, mainly soybeans that contain the isoflavones genistein and daidzein. These substances themselves are phytoestrogens, and a number of studies indicated an association between their consumption and improved reproductive functions, as well as with a reduced risk of breast cancer in women, and antioxidant properties [36
are capable of metabolizing daidzein into equol [39
]. Equol is an isoflavandiol manifesting phytoestrogenic activity with a potentially positive effect on human health, including hormonal and cardiovascular functions [40
] and anticancer activity [41
The ability to metabolize isoflavones into equol is a quite specific microbial feature: it is estimated that only about a third of the world population harbor such microbial species in their gut. Thus, a hypothetical therapeutic diet with a high content of soybean products might not be effective for a large part of the population. Based on these facts, we can conclude that the observed increase of Adlercreutzia and Slackia genera after FDP consumption may improve the capacity of human microbiota for responding to a diet rich in isoflavones, including soy-based products. These observations suggest an opportunity to design food products and/or diets containing not only dairy components enriched with bifidobacteria and lactobacilli, but also soy.
family of bacteria were also significantly increased after FDP consumption (including species related to Eubacterium dolichum
and Catenibacterium mitsuokai
). While data on clinical associations for this bacterial family are ambiguous, there are a number of studies linking their increased abundance to inflammatory bowel diseases, as well as obesity, while other studies suggest opposite associations [42
The groups of bacterial taxa detected during correlation analysis represent potential symbiotic cooperatives of species (Figure 1
). Observed cooperatives vary by phylogenetic composition, and many features of their content are consistent with published data [44
Half of cooperative #1 is formed by Clostridiales bacteria (Eubacterium, Anaerostipes, Blautia, Dorea
) known as prominent producers of butyrate determining their anti-inflammatory activity and association with the healthy gut [46
]. Studies of the effect of diet on microbiota composition show that levels of these taxa are increased when the diet is rich in fiber, commonly found in vegetables, cereals and other products [47
The dominance of cooperative #2 is formed by members of Bacteroidaceae
family including; Bacteroides
and others. Increased prevalence of these groups have been associated with the “Western diet”, a diet rich in animal fats, meat and sugars, as well as deficient in non-digestible dietary fiber [48
]. Additionally, two members of a related family Porhyromonadaceae
) are present.
Cooperative #3 includes butyrate-producing clostridia associated with decreased risk of inflammatory bowel diseases as well as a number of distantly related microbes with symbiotic links. Among them are the microbes suggested to be significant in the regulation of weight and dietary behavior. The Christensenella
genus (a representative species is C. minuta
) is the most inheritable gut microbe; it is also associated with normal weight and prevents obesity in mice models [44
]. The M. smithii
archaeon facilitates a more efficient fermentation of dietary fiber and may contribute the regulation of weight and dietary behavior. Methanobrevibacter
were previously observed to be inherited together; both are associated with normal BMI [44
], although the mechanisms underlying this co-occurrence have not been identified yet.
Generally, none of the microbial cooperatives significantly changed their abundance as the result of FDP consumption. This difference from the results obtained during species-level analysis may be due to the observation that many of the differentially abundant species are not included in any of the cooperatives (only large cooperatives were examined). Secondly, it may be related to the fact that FDP consumption represents a relatively small change in total daily dietary intake of volunteers—in comparison with changing one’s diet to follow certain recommendations, e.g., aiming to increase total dietary fiber consumption by including more fruit, vegetables, and whole grains into diet [50
However, there were significant changes for the cooperatives when subjects were stratified into responders and non-responders (Figure 2
). Namely, responders had increased baseline levels of cooperative #2 enriched in Bacteroidaceae
and decreased—of cooperative #1 enriched with Clostridiales. The taxa included in cooperative #2 are reported to be associated with long-term “Western diet” [48
]. While no significant associations between responder/non-responder clusters and physiological factors were identified, one can speculate that FDP consumption may have a more profound positive impact on the microbiota of individuals on “Western diet”.
In addition to the evaluation of impact of FDP consumption on species-level composition, microbiota analysis allowed evaluation of impact on microbiota functions by investigating selected changes in metabolic potential. Significant functional changes were observed reflecting increased capacity of the community to metabolize lactose, other simple sugars, starch, as well as to synthesize amino acids. This was accompanied by a decrease in synthesis of immunogenic molecules (lipopolysaccharides) that can be interpreted as a reduction of the proinflammatory potential of the microbiota.