Fucosyllactose and L-fucose utilization of infant Bifidobacterium longum and Bifidobacterium kashiwanohense

Background Human milk oligosaccharides (HMOs) are one of the major glycan source of the infant gut microbiota. The two species that predominate the infant bifidobacteria community, Bifidobacterium longum subsp. infantis and Bifidobacterium bifidum, possess an arsenal of enzymes including α-fucosidases, sialidases, and β-galactosidases to metabolise HMOs. Recently bifidobacteria were obtained from the stool of six month old Kenyan infants including species such as Bifidobacterium kashiwanohense, and Bifidobacterium pseudolongum that are not frequently isolated from infant stool. The aim of this study was to characterize HMOs utilization by these isolates. Strains were grown in presence of 2′-fucosyllactose (2′-FL), 3′-fucosyllactose (3′-FL), 3′-sialyl-lactose (3′-SL), 6′-sialyl-lactose (6′-SL), and Lacto-N-neotetraose (LNnT). We further investigated metabolites formed during L-fucose and fucosyllactose utilization, and aimed to identify genes and pathways involved through genome comparison. Results Bifidobacterium longum subsp. infantis isolates, Bifidobacterium longum subsp. suis BSM11-5 and B. kashiwanohense strains grew in the presence of 2′-FL and 3′- FL. All B. longum isolates utilized the L-fucose moiety, while B. kashiwanohense accumulated L-fucose in the supernatant. 1,2-propanediol (1,2-PD) was the major metabolite from L-fucose fermentation, and was formed in equimolar amounts by B. longum isolates. Alpha-fucosidases were detected in all strains that degraded fucosyllactose. B. longum subsp. infantis TPY11-2 harboured four α-fucosidases with 95–99 % similarity to the type strain. B. kashiwanohense DSM 21854 and PV20-2 possessed three and one α-fucosidase, respectively. The two α-fucosidases of B. longum subsp. suis were 78–80 % similar to B. longum subsp. infantis and were highly similar to B. kashiwanohense α-fucosidases (95–99 %). The genomes of B. longum strains that were capable of utilizing L-fucose harboured two gene regions that encoded enzymes predicted to metabolize L-fucose to L-lactaldehyde, the precursor of 1,2-PD, via non-phosphorylated intermediates. Conclusion Here we observed that the ability to utilize fucosyllactose is a trait of various bifidobacteria species. For the first time, strains of B. longum subsp. infantis and an isolate of B. longum subsp. suis were shown to use L-fucose to form 1,2-PD. As 1,2-PD is a precursor for intestinal propionate formation, bifidobacterial L-fucose utilization may impact intestinal short chain fatty acid balance. A L-fucose utilization pathway for bifidobacteria is suggested. Electronic supplementary material The online version of this article (doi:10.1186/s12866-016-0867-4) contains supplementary material, which is available to authorized users.


Background
Bifidobacteria are universally distributed in organisms that raise offspring by parental care including mammals, birds and social insects. Bifidobacteria are highly specialized organisms in using non-digestible oligosaccharides and a major part of their genomes is devoted to the utilization of carbon sources [1][2][3][4][5]. The proportion of genes related to carbohydrate transport and metabolism is higher in bifidobacteria than in Bacteroides, which are also characterized by their ability to utilize a variety of polysaccharides [6]. Host-specific adaption in regard to carbohydrate degradation has been suggested [7,8]. Adult species, such as Bifidobacterium adolescentis and Bifidobacterium longum subsp. longum, are well equipped to degrade plant derived polysaccharides [4,9]. Infant species, such as Bifidobacterium longum subsp. infantis and Bifidobacterium bifidum, are adapted to utilize human milk oligosaccharides (HMOs), one of the major glycan sources of breast milk [3,[10][11][12]. Primary components of HMOs are D-glucose, D-galactose, L-fucose, N-acetylglucosamine, and sialic acid. Lactose constitutes the reducing end, its galactose moiety can be fucosylated or sialylated to form 2′-or 3′-fucosyllactose (2′-FL or 3′-FL) or 3′-and 6′-sialyl-lactose (3′-SL or 6′-SL). Lactose can also be elongated with units of N-acetyllactosamine (Gal-β1-4GlcNAc) with its simplest form being Lacto-N-neotetraose (LNnT) [13].
Bifidobacteria metabolize hexoses via the 'bifid shunt' with fructose-6-phosphoketolase being the key enzyme to theoretically yield 1.5 mol acetate, 1 mol lactate and 2.5 ATP from 1 mol glucose [17]. The ratios of lactate and acetate formed may vary with carbohydrate source and species, depending on whether the intermediate pyruvate is cleaved to acetyl phosphate and formate, or reduced to lactate [18]. Pentoses, such as xylose, are also fermented to lactate, acetate and possible formate [19]. There is little information available about bifidobacterial metabolism of desoxyhexoses, and rhamnose was not used by various species tested [6].
In a previous study, several Bifidobacterium strains were isolated from Kenyan infant stool, that were identified as B. longum, B. bifidum, B. breve, Bifidobacterium kashiwanohense, and Bifidobacterium pseudolongum [20]. B. kashiwanohense has only been isolated from a healthy Japanese infant [21] and from Kenyan anaemic infants [20]. B. pseudolongum has been frequently recovered from animal feces [22]. B. kashiwanohense, and B. pseudolongum are species not commonly associated with the infant bifidobacteria community, however, the presence of additional species in infant feces from developing countries, might reflect variations in diet and sanitary status. Little is known about the ability of these species to utilize HMOs.
Therefore it was the aim of this study to investigate HMO degradation by these newly obtained isolates. As we observed that beside B. longum subsp. infantis and B. bifidum, an isolate of B. longum subsp. suis, and strains of B. kashiwanohense were able to metabolise fucosyllactose, we further investigated metabolite formation during growth on L-fucose and fucosyllactose. Furthermore we analyzed genomes of the studied strains to elucidate possible genes and pathways involved in fucosyllactose degradation and L-fucose utilization through genome comparison.

Bacterial strains
Twenty-nine bifidobacterial strains were included in the initial HMO utilization screening (Table 1). Nineteen strains originated from stool samples of Kenyan infants [20] and ten reference strains were obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ, Braunschweig, Germany, Table 1). Kenyan isolates that had been previously typed to species level were additionally characterized on subspecies level using the (partial) 16S rRNA gene as marker. Briefly, DNA was extracted from overnight cultures using the PrepMan® Ultra protocol for pure culture (Thermo Fisher Scientific, Reinach, Switzerland). PCR amplification of partial 16S rRNA genes was performed using universal primers 518 F (5′-CCAGCAGC CGCGGTAATACG-3′) and 1391R (5′-GACGGGCGG TGTGTRCA-3′). PCR reaction mixtures (25 μL) contained 12.5 μL of 2× PCR MasterMix (Thermo Fisher Scientific), 0.2 μM of primers (Microsynth AG, Balgach, Switzerland) and 1 μl of template DNA. The cycling programme consisted of an initial denaturation of 5 min at 95°C, followed by 32 cycles of denaturation for 30 s at 95°C, annealing for 30 s at 52°C, and extension for 1 min at 72°C. Amplicons were sequenced by GATC Biotech (Konstanz, Germany). For generation of phylogenetic trees, 16S rRNA gene sequences were aligned and cut using CLUSTALW implemented in BioEdit Version 7. Phylogenetic analysis of partial 16S rRNA gene sequences (772 bp) was performed using Maximum Likelihood Analysis implemented in MEGA6 [23], applying the Jones-Taylor-Thornton substitution model and default settings. Bootstrap support was calculated for 500 replicates, strains of Lactobacillus were applied as outgroup. Sequences are listed in the Additional file 1.

Utilization of selected sugars and metabolite formation
Isolates derived from −80°C stock cultures were streaked on supplemented Wilkins-Chalgren agar and were incubated anaerobically at 37°C for two days. Single colonies of each isolate were subsequently incubated twice in supplemented Wilkins-Chalgren broth (10 ml, 1:10) at 37°C for 20 h. To obtain working cultures, the supernatant was removed from overnight cultures, cells were washed, and re-suspended in same volume of 50 mM phosphate buffer, pH 6.5 (PB).
To investigate growth on fucosyllactose and L-fucose of selected strains, cell suspensions (50 μl) were added to 950 μl API medium supplied with 30 mM L-fucose, 2′-FL, or 3′-FL (28.0 and 27.0 mM, respectively). Strains were grown in independent triplicates under anaerobic condition at 37°C for 48 h.

Genome assembly and annotation
Genomes were assembled using Abyss v.1.9.0 for pairedend libraries implemented in Bio-Linux 8. The partial genomes were functionally annotated with RAST using default settings [24]. RAST annotations of genes of interest were verified using the BLAST tool implemented in RAST. Average nucleotide identity (ANI) was calculated using the online tool supplied by Rodriguez-R and Konstantinidis [25]. Carbohydrate-active enzymes were selectively confirmed based on similarity to the carbohydrate active enzyme (CAZy) database entries, and Pfam alignments implemented at the CAZymes Analysis Toolkit (CAT) [26]. Additionally, dbCAN was used for identification of carbohydrate active proteins which is based on a search for signature domains of every CAZyme family [27].

Utilization of HMOs
We investigated growth of 19 bifidobacterial isolates of Kenyan infants and 10 culture collection strains (Table 1) in the presence of individual HMOs: 2′-FL, 3′-FL, 3′-SL, 6′-SL, and LNnT, and combined HMOs in API medium. All isolates were able to grow in the presence of glucose or lactose confirming the suitability of the assay ( Table 2). Growth correlated with the degradation of the supplied HMOs as determined with HPAEC-PAD for selected strain-HMO combinations ( Table 2). B. longum subsp. infantis utilized of 2′-FL, 3′-FL, 3′-SL and LNnT and degraded all HMOs when supplied together. B. bifidum grew in the presence of 2′-FL, 3′-FL and LNnT and also utilized 3′-SL and 6′-SL in HMO mixtures confirming adaptation of both species to HMO utilization, as reported before [3,12]. Strains of B. bifidum liberated L-fucose and a second degradation product ( Fig. 1, peak y) in the supernatant when grown in the presence of fucosyllactose while L-fucose accumulation or the release of any other degradation intermediate was not observed for B. longum subsp. infantis strains [28] (Fig. 1).
All B. breve isolates were able to utilize LNnT as shown previously [15]. B. breve DSM 20213 also degraded 2′-FL and 3′-FL when grown with HMO mixtures.
Also, B. kashiwanohense DSM 21854 and the Kenyan isolates grew in the presence of 2′-FL and 3′-FL, thereby accumulating L-fucose and releasing compound y (Fig. 1). The amount of L-fucose released by B. kashiwanohense isolates was only about 12 % compared to the complete release of B. bifidum.
Strains of B. pseudolongum did not metabolize with any of the HMOs tested.
The ability to use fucosyllactose was thus identified as being a trait of several bifidobacteria species. B. longum subsp. suis and B. kashiwanohense have not considered infant bifidobacteria species, yet, the ability to utilize fucosyllactose points at adaptation to the infant gut.

L-fucose metabolism of bifidobacteria
Similar to B. bifidum, B. kashiwanohense excreted Lfucose into the supernatant [28]. L-fucose accumulation was not observed when B. longum subsp. infantis isolates and B. longum subsp. suis BSM 11-5 were grown in the presence of fucosyllactose.
infantis DSM 20088 produced a lactate:acetate ratio of 2:3 as expected of the metabolism of hexoses through the bifid shunt [17] in addition, this strain produced 1,2-PD (Table 4).
In contrast, the ratio of lactate:acetate of B. longum subsp. suis BSM 11-5 grown with 2′-FL and 3′-FL was approx. 1:1 and 1:3 respectively. B. longum subsp. suis BSM 11-5 synthesized 1,2-PD mainly from 3′-FL, and accumulated 5 mM L-fucose when grown in the presence of 2′-FL. L-fucose might have been accumulated during growth in the presence of 28 mM 2′-FL as glucose and galactose became also available after fucosyllactose degradation.
B. kashiwanohense DSM 21854 grew in the presence of 2′-FL and 3′FL and accumulated approximately 10 mM L-fucose but did not produce any 1,2-PD ( Table 4). The ratio of lactate:acetate was approx. 1:2.
B. longum subsp. infantis degrades HMOs internally [3,34]. The gap in substrate consumption, L-fucose release and/or 1,2-PD formation observed for B. longum Table 2 Degradation of HMOs by selected strains Growth is indicated by grey shading. Degradation of HMOs of selected samples was investigated by HPAEC-PAD Plus (+) indicates degradation of HMO tested, minus (−) no degradation HMOs that were used during growth in the presence of HMO combinations (2′-FL, 3′-FL, 3′-SL, 6′-SL, LNnT) are indicated in the respective column subsp. infantis and B. longum subsp. suis might be due to the intracellular which were not released in the supernatant. In contrast, fucose and an additional compound were detected in supernatants of B. bifidum which harvests fucosyllactose extracellularly [28,34].
We here identified 1,2-PD as a metabolite of bifidobacteria fucosyllactose respective L-fucose degradation. L-fucose derived 1,2-PD can be further metabolized to propionate and propanol by other gut microbes such as Eubacterium hallii [35,36]. It was estimated that in adults approximately 30 % of propionate might derive from 1,2-PD, but no data exists for infants [36,37]. Nevertheless, the bifidobacterial formation of lactate and 1,2-PD as precursors of short chain fatty acids butyrate and propionate, respectively, contributes to the trophic interactions of the infant gut microbiota [38].
B. longum subsp. infantis DSM 20088 lacks the genes encoding proteins to use fucose via phosphorylation [3]. FucI, FucK and FucA were also not detected in the other genomes analysed here. To investigate whether bifidobacteria might utilize L-fucose similar to X. campestris, we searched for the corresponding proteins of X. campestris in bifidobacteria genomes using BlastP, and also collected enzymes related to fucose metabolism that were annotated by RAST.
The majority of genes was located on two genomic regions (Fig. 3). In contrast, in X. campestris all responsible genes were located on an operon XCC4065-XCC4070 [44]. Region 1 encompassed a L-fucose mutarotase, a L-2-keto-3-deoxy-fuconate hydrolase, and a L-fuconate dehydratase (Fig. 3). The gene cluster of region 1 also contained genes encoding the αfucosidases BLON_2335 and BLON_2336 and is part of   Table 7 Identification of B. longum and B. kashiwanohense genes related to L-fucose metabolism L-fucose related genes were identified by blastP search of homologous proteins of X. campestris, and by annotation by RAST using default settings. Shown are gene ID and in brackets bit scores and e-values of the obtained hits. Genes encoding these enzymes were predominantly located on two genomic regions shaded in light grey (region 1) and dark grey (region 2) + A recent transcriptomic study investigated gene expression of B. longum subsp. infantis DSM 20088 in the presence of 2′-FL and 3′-FL, α-fucosidases that were overexpressed are indicated [34] the B. longum subsp. infantis HMO utilization operon H1 [3,11]. A possible L-fuconolactone hydrolase, and paralogs of fuconate dehydratase and L-fucose mutarotase were located elsewhere on the genome in close proximity to a putative fucose permease (Fig. 3). Expression of the fuconate dehydratase of region 1, and of the putative L-2-keto-3-deoxy-fuconate hydrolase was recently reported to be upregulated when B. longum subsp. infantis DSM 20088 was grown in the presence of fucosyllactose providing strong support to the proposition that L-fucose is metabolized via this pathway [34].
B. kashiwanohense DSM 21854 and PV20-2 only possessed a gene segment similar to region 1 with a truncated L-fucose-mutarotase, and lacked region 2 which encompassed the fucose permease. This might be the reason why strains of B. kashiwanohense were not able to utilize L-fucose.
Taken together these results suggest that B. longum subsp. infantis TPY12-1, DSM 20088 and B. longum subsp. suis BSM11-5 metabolize fucose via a pathway with non-phosphorylated intermediates as previously described for Campylobacter sp. and X. campestris.