The microbiota composition in the human intestinal tract is determined by several factors, such as host genotype, health status, age, microbial interactions, and diet [3]. Based on numerous intervention studies, there is convincing evidence for the influence of the diet on the intestinal microbiota (e.g. [4], [5]). In contrast, although growing evidence indicates that host genetic background has a significant impact on the microbiota composition in the intestine, no specific genetic factors determining the intestinal microbiota composition have been established to date. Twin studies applying plate counts, PCR-DGGE fingerprinting or DNA microarrays have shown a higher similarity in the microbiota composition between monozygotic twins than between dizygotic twins, unrelated persons, marital couples and family members [6]–[7], thus clearly pointing to a strong effect of host genetics. In the study by Turnbaugh et al. [8], the pyrosequencing analysis also showed a higher level of similarity in the microbiota composition in twin pairs than between twins and their mothers or unrelated persons, although in their study the similarity of the microbiota between monozygotic twins did not differ from that of dizygotic twins.

The human intestinal tract is colonised with highly diverse and numerous microbiota which has an established role in maintaining the intestinal homeostasis. A particularly interesting group is bifidobacteria, which comprise the predominant intestinal microbiota in infants and are abundant also in the adult population comprising up to 6% of the normal intestinal microbiota [9]. An adult intestine is typically colonised with one to four bifidobacterial species [10], B. longum, B. adolescentis, B. bifidum and B. catenulatum being the most prevalent [11], [12]. Bifidobacteria have beneficial properties, such as immunomodulatory and pathogen inhibition effects (reviewed by [13]). They also are commonly incorporated in probiotic products.

The A, B and H blood group antigens are α1,2-linked fucose containing glycans present on glycoproteins and glycolipids of erythrocytes (red blood cells) in individuals representing A, B and H blood groups, respectively. The enzyme fucosyltransferase 1 encoded by the FUT1 gene is responsible for the synthesis of ABH antigens on erythrocytes. The ABH antigens are also expressed in mucus and other secretions, where their expression is generated by another enzyme, fucosyltransferase 2 (secretor type α1,2-fucosyltransferase) encoded by the FUT2 gene. In ABH secretor individuals (80% of Caucasians) fucosyltransferase 2 converts type 1 N-acetyllactosamine glycan chains to H antigen, which functions as a precursor for the A, B and Lewis b antigens. Non-secretor individuals do not express active fucosyltransferase 2 enzyme due to a non-sense mutation in the FUT2 gene and therefore they are not able to express the ABH antigens in their mucus and other secretions. The FUT2 gene together with the FUT3 gene encoding fucosyltransferase 3 (Lewis type α1,3/4-fucosyltransferase), is also required for the synthesis of Lewis b histo-blood group (Le a⁻b⁺) antigens in secretions. In non-secretor individuals, the FUT3 gives rise to the Lewis a histo-blood group (Le a⁺b⁻) antigen due to non-functional fucosyltransferase 2. Lewis negative (Le a⁻b⁻) individuals have a mutation in FUT3 gene leading to Lewis null phenotype, irrespective of the secretor status or FUT2 gene. These mucosal ABH and Lewis histo-blood group antigens are known to serve as an energy source [14] and adhesion receptors for many microbes [15], and thus could play a role in shaping the microbiota composition of the host.

The secretor status determined by the FUT2 gene was strongly associated with the bifidobacterial diversity and composition. The non-secretor individuals only had about half of the bifidobacterial diversity and richness present in the secretor individuals based on the PCR-DGGE analysis. In addition, the non-secretor individuals had significantly reduced bifidobacterial abundance in comparison with the secretor individuals as measured by qPCR. Moreover, the non-secretor individuals lacked, or were rarely colonised by, several bifidobacterial DGGE genotypes, which were related to species B. adolescentis, P. catenulatum/pseudocatenulatum and B. bifidum, and were common in the secretor individuals. We applied the PCR-DGGE to compare the bacterial diversity and community structure between the secretor and non-secretor individuals. The PCR-DGGE method is known to detect only the predominant part of the bacteria present in a complex sample. Bifidobacterial population is usually composed of limited number (0–4) of species [10] and thus, could be captured by the PCR-DGGE analysis with bifidobacterial specific primers. We also showed that bifidobacterial-DGGE profiles were highly reproducible. Moreover, we isolated bifidobacterial strains from the non-secretor and secretor individuals and analysed their 16S rRNA gene fragments in a DGGE gel along with faecal samples. The isolated strains corresponding to the most common (present in >10% samples) band positions in bifidobacterial DGGE gels were found (data not shown), reducing the likelihood that the detected DGGE bands originate from PCR artefacts sometimes occurring in the PCR-DGGE. In contrast to the bifidobacteria-specific DGGE, it is likely that methodological limitations hinder the interpretation of the PCR-DGGE targeted at dominant bacteria (universal PCR-DGGE). It is probable that several secretor-status associated genotypes, which are present at low levels, are missed in the PCR-DGGE using universal primers and their associations with the secretor status could, thus, not be detected. Nevertheless, our finding on the differences in the dominant microbiota between the non-secretor and secretor individuals suggests that the association between microbiota and secretor status is not limited to bifidobacteria only. It remains to be seen which other bacterial groups and species are associated with the secretor status.

The altered microbiota composition in the non-secretor individuals shown in this study may be an important factor contributing to the non-secretor disease susceptibility. Polymorphism of the FUT2 gene, determining the secretor status, has been suggested to modulate innate immune responses and even have an evolutionary role in humans' survival during different pathogen outbreaks [16],[17]. The non-secretor phenotype has been genetically associated with increased risk for Crohn's disease [18], [19], and necrotizing enterocolitis [20]. Secretor status is also associated with the susceptibility to several infectious diseases. Non-secretors have an increased risk for urinary tract infections [21], [22] and vaginal candidiasis [23], [24], but a reduced risk for diarrhoea caused by certain genotypes of Norovirus [25]. Interestingly, many of these secretor status associated diseases, such as Crohn's disease [26], [27], urinary tract infection [28], and NEC [29] have also been connected to changes in the intestinal microbiota composition or activity. Bifidobacteria have been shown to have health promoting effects on humans [13]. Bifidobacteria or bifidobacteria-containing strain mixtures have shown promising results e.g. in the alleviation of the symptoms of irritable bowel syndrome (IBS) [30], inflammatory bowel disease [31], and diarrhoea [32], although the mechanism of action is largely unknown. Reduced bifidobacterial abundance has been connected to intestinal disorders such as irritable bowel syndrome [33] and inflammatory bowel disease [26]. Taken together, the properties of bifidobacteria and the results of this study suggest that the secretor status, by effecting bifidobacterial diversity, may also play a role in susceptibility to the diseases associated with the decreased bifidobacterial abundance in the intestine.

Metagenomics studies indicate that a considerable number of intestinal microbiota genes are involved in carbohydrate metabolism. Kurokawa et al. [34] reported that the carbohydrate metabolism genes of microbiota are enriched in the intestine (24% of genes in adults and 34% in children) in comparison with the microbiota originating from other environments, such as soil and sea. Both plant polysaccharides and host derived glycans are important energy sources for intestinal bacteria. Fucosylated histo-blood group antigens, such as the ABH and Lewis b histo-blood group antigens, are terminal epitopes of glycan chains in glycoproteins and glycolipids mediating the interaction between host and both commensal and pathogenic intestinal bacteria [35], [36]. Non-secretor individuals have null-allele of FUT2 gene and do not express such α1,2-fucose containing glycan structures in their intestinal mucosa. Bacteria that can interact with these epitopes and compete for adhesion sites or to use them as energy sources have a better colonisation ability in secretor individuals than in non-secretor individuals (e.g. bifidobacteria in this study). Intestinal bifidobacteria, whose abundance and diversity were higher in the intestine of secretor individuals than non-secretor individuals in this study, are adapted to utilise glycans present in mucins and human milk [37]. It is known that some microorganisms secrete glycosidases capable of degrading histo-blood group antigens [14]. Comparatively small populations of human faecal bacteria produce α-glycosidases capable of degrading terminal ABH and/or Lewis groups in glycans [14]. Among them are bifidobacteria with 1,2-α-fucosidase to hydrolyse α-1,2-fucosyl linkages present in various glycans, such as the above-mentioned histo-blood group antigens [38]. Recently, Turroni et al. [39] showed, using genomic, proteomic and transcriptomic analysis of B. bifidum PRL2010, the existence of enzymes allowing further degradation of many core glycan chains and they concluded that the property is important for intestinal colonisation of B. bifidum. Such degradation of glycan cores may require the initial removal of terminal α-fucose, enabling subsequent processing of glycan chain by β-galactosidase and/or β-N-acetylhexosaminidase and endo-α-N-acetylgalactosaminidase, which catalyses the release of GalNAc from serine/threonine residues of various mucin-type glycoproteins, all of these enzymes being encoded in B. bifidum PRL2010 genome [39]. In addition, bifidobacteria but typically not the other common commensals, have lacto-N-biosidase degrading type 1 glycan chains, which are precursors of fucosylated histo-blood group antigens in the human intestine [40]. Therefore, it can be concluded that bifidobacteria have very specific strategies for the utilization of host glycans.

Table S1