Preferential sugar utilization by bifidobacterial species

Bacterial strains and media

Eighty-one bifidobacterial strains were purchased from the Riken BioResource Research Center (JCM, Tsukuba, Japan), American Type Culture Collection (ATCC), and German Collection of Microorganisms and Cell Cultures (DSMZ) and used to inoculate MRS medium (Becton Dickinson and Company). After several passages, 40 strains that showed substantial growth in the medium were selected for this study, considering their taxonomic variety. Furthermore, five commercial strains of Bifidobacterium longum, Bifidobacterium bifidum, and Bifidobacterium breve were included. In total, 45 strains of bifidobacteria were used for sugar consumption analyses. Their scientific names, strain numbers, and isolation sources are listed in Table 1.

Bacterial culture

Preferential utilization of glucose or lactose was assessed as previously described^[20]. Modified MRS medium [1% w/v proteose peptone #3, 1% w/v beef extract, 0.5% w/v yeast extract (all from Becton Dickinson and Company Franklin Lakes, NJ, USA), 0.1% v/v Tween 20 (Nacalai Tesque Inc., Tokyo, Japan), 0.5% w/v K₂HPO₄ (Kokusan Chemical Co., Tokyo, Japan), and 0.5% w/v sodium acetate, 0.5% w/v di-ammonium hydrogen citrate, 0.02% w/v MgSO₄·7H₂O, and 0.005% MnSO₄·H₂O (all from Fujifilm Wako Pure Chemical Corporation, Tokyo, Japan), pH 6.5] was used for cultivation. The medium was supplemented with 0.8% (w/v) glucose (MRS-G), 0.8% lactose (MRS-L), or a mixture of 0.4% glucose and 0.4% lactose (MRS-GL) [Supplementary Table 1] prior to inoculation. We employed these sugar concentrations for determining the sugar preference of each strain because we found that these concentrations enable us to visually monitor the sugar consumption by high-performance thin-layer chromatography (HPTLC) analysis (see below).

Bacteria were pre-cultured in MRS medium at 37 °C for 24 h under anaerobic conditions, and then the cultures (400 µL per each) were centrifuged at 2,000 g for 3 min. The resulting pellets were washed twice with a saline solution (Otsuka Pharmaceutical, Tokyo, Japan), and 75 µL of each of the 400 µL suspensions was used to inoculate 2.5 mL of MRS-GL, MRS-G, or MRS-L medium supplemented with 0.05% (w/v) L-cysteine HCl (Kanto Chemical, Tokyo, Japan) in a 5 mL tube (Sarstedt, Nuembrecht, Germany). The culture broth was incubated at 37 °C for 72 h under anaerobic conditions (in N₂:H₂:CO₂, 80:10:10). Aliquots (0.2 mL per each) were taken and the optical densities at 600 nm (OD₆₀₀) were measured using a multi-microplate reader (SH-9000Lab, Hitachi, Tokyo, Japan). The cultures were centrifuged, and the resulting supernatants were used for high-performance thin-layer chromatography (HPTLC).

HPTLC analysis

Pre-coated silica gel-60 HPTLC aluminum plates (Merck, Darmstadt, Germany) were used to evaluate the sugar preferences of the bifidobacterial strains. The culture supernatants (0.5 µL per each) were loaded on the plate. The spotted plate was developed in a solvent system consisting of 1-butanol (Nacalai Tesque):acetic acid (Kanto Chemical):water (2:1:1 v/v). The plates were stained by spraying with diphenylamine-aniline-phosphoric acid reagent as previously described^[22] , and the sugars were visualized by baking the plates for 2 min.

Comparative genomic analysis

Genomic information of the type strains of the 40 Bifidobacterium species was retrieved from the NCBI public database [Supplementary Table 1]. Given its preference for lactose over glucose by repressing the glucose transporter glcP in the presence of lactose in the medium^[20], B. longum NCC2705 was also included in this in silico analysis as the strain has been shown to exhibit a preference for lactose over glucose by repressing the glucose transporter glcP in the presence of lactose in the medium^[20]. Amino acid sequence comparisons were performed using all-against-all bidirectional BLAST alignment (cut-off: E-value 0.0001, with at least 50% identity across at least 50% of either protein sequence). The sequences were then clustered into protein families by using the Markov clustering algorithm (MCL) implemented in the mclblastline pipeline v12-0678^[23]. An example of the scripts used for the analysis is available at https://github.com/frbot/Comparative-genomics/blob/master/comparative.sh. The upstream sequence of glcP was aligned using the ClustalW alignment tool^[24] with default parameters. The carbohydrate-active enzymes were annotated using the dbCAN3 server (https://bcb.unl.edu/dbCAN2/) and the HMMER: dbCAN tool (E-Value < 1e-15, coverage > 0.35). Intergroup differences were analyzed using Welch’s t-test, followed by Benjamini-Hochberg FDR calculations on R software (ver. 3.6.0) with the qvalue package (http://www.bioconductor.org/packages/release/bioc/html/qvalue.html).

Fecal sample collection

Fecal samples of 10 infants at 2, 4, 6, and 8 weeks of age were collected from their diapers by their parents as soon as possible after defecation. The samples were stored at -20 °C at home until all samples from each subject were collected. The samples were then transported to the laboratory at less than -18 °C by logistics companies. Upon receipt, the samples were immediately stored at -80 °C until use. This study was approved by the institutional review board of the Japan Conference of Clinical Research (approval number: BONYU-01) and was conducted in accordance with the latest Declaration of Helsinki in 2013. Healthy mothers of the infants provided written informed consent for their participation in all procedures associated with the study.

Determination of lactose content in infant fecal samples

Lactose in the fecal samples was fluorescently labeled and quantified using high-performance liquid chromatography (HPLC). Specifically, the lyophilized fecal samples (3-5 mg) were rehydrated with 5% acetic acid (100 μL), fragmented in an ultrasonic bath for 10 min, and then mixed with a solution containing 120 µL of chloroform (Nacalai Tesque Inc.), 240 µL of methanol (Nacalai Tesque Inc.), and 20 µL of 10 mM melibiose (Fujifilm Wako Pure Chemical Corporation). Melibiose was used as the internal standard. After further mixing with 120 µL of chloroform and 120 µL of pure water (Milli-q), the supernatant obtained by centrifugation (21,500 g, 3 min, 4 ºC) was lyophilized after evaporation of methanol. The saccharides were then fluorescently labeled with 2-anthranilic acid and purified by solid extraction using Discovery DPA-6S columns (50 mg) (Supelco, Poole, UK) following a previously described method^[25]. HPLC was performed using a Thermo U3000 system (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a TSKgel Amide-80 HR column (4.6 mm × 250 mm, φ = 5 μm) (Tosoh, Tokyo, Japan) as previously described^[26]. The labeled sugars were detected using a Waters 2475 Fluorescence Detector (Waters Corp., Milford, MA, USA). Lactose concentrations were calculated using known concentrations of standard lactose and normalized to the internal standard melibiose.

Preferential utilization of glucose or lactose by each bifidobacterial strain

The ability of each Bifidobacterium strain to grow in the MRS-G and MRS-L media was evaluated [Figure 1]. Although most strains were able to utilize both glucose and lactose, the strains of B. asteroides (#2, the numbers are indicated in Table 1 and Figures 1 and 2), B. boum (#8), B. thermophilum (#11), B. pullorum subsp. gallinarum (#38), and B. pullorum subsp. saeculare (#39) showed limited growth on a lactose-containing medium (OD₆₀₀ < 0.1). In general, the strains of B. boum group and B. pollurum group showed poor lactose utilization. The preferential carbon utilization of each strain was then evaluated by incubating the strains in a medium containing both glucose and lactose and analyzing their sugar consumption behavior through HPTLC [Supplementary Figures 1 and 2]. The results definitively differentiated lactose- and glucose-preferring strains. Based on the phylogenomic tree of the genus Bifidobacterium reported by Lugli et al., we found that the B. asteroides, B. boum, B. bombi, and B. pullorum groups preferentially utilized glucose, whereas B. tissieri and B. adolescentis groups preferentially utilized lactose^[27]. The B. longum, B. bifidum, and B. pseudolongum groups contained glucose- and lactose-preferring strains. Evaluations were also conducted using commercialized strains belonging to three taxa: B. breve (#19 and 20), B. longum subsp. longum (#25), and B. longum subsp. infantis (#23), with no differences in sugar preference observed within the species or subspecies.

Figure 1. Saccharide utilization preferences and the ability of each Bifidobacterium strain to grow on each carbon source. Strains are sorted based on concatenation of 101 core protein sequences of the genus Bifidobacterium as previously reported. Numbering of each strain corresponds to Table 1.

Distribution of genes involved in lactose and glucose uptake among bifidobacterium species

The distribution of the lactose symporter gene lacS^[28] was examined to explore the genetic basis of preferential sugar utilization by bifidobacteria. All of the strains that showed a preference for lactose over glucose had a lacS homolog with amino acid identities higher than 50%, except for B. choerinum ATCC 27686 (#43) [Supplementary Table 2]. The exceptional strain B. choerinum ATCC 27686 (#43), whose genomic sequence was complete, showed the lowest growth rate in the medium supplemented with glucose. A lacS homolog search also revealed that not all strains with a lacS homolog preferentially utilized lactose. For example, B. reuteri JCM 17295 (#27) and B. pseudolongum ssp. pseudolongum JCM 1205 (#45) preferred glucose over lactose. Next, we investigated the homolog distribution of the glcP, ptsG, and licT gene cluster (BL1631, BL1632, and BL1633 in B. longum NCC2705), which are reportedly involved in the preferential utilization of lactose in NCC2705^[20]. Of the 40 isolates, 12 strains had all three homologs, and six had three genes within the same cluster. However, we found no correlation between the presence of these genes and the preferential utilization of lactose [Supplementary Table 2]. DNA sequence alignment was conducted for the regulatory regions of glcP, which comprise a potential rho-independent terminator and a ribonucleic anti-terminator (RAT) element, to identify differences that could explain the variability in the preferential utilization of sugars. However, no distinguishing features were detected [Supplementary Figure 3]. Gene-trait matching based on MCL clustering did not reveal any corresponding gene family [Supplementary Table 3]. We also examined the distribution of glycosyl hydrolases using dbCAN; however, no clear relationship was detected between the preferential sugar utilization phenotype and the presence of these enzymes [Supplementary Table 4].

Link between sugar preference by bifidobacterial strains and their isolation source

The relationship between the origin of the bifidobacterial isolates and their sugar preferences was investigated by sorting them based on their source [Figure 2, Supplementary Figures 1 and 2]. All human-derived bifidobacterial strains belonging to seven species/subspecies demonstrated a preference for lactose, whereas all strains isolated from bees and sewage exhibited a preference for glucose. The other bifidobacterial strains isolated from different sources showed varied preferences for glucose and lactose; however, four of the five strains isolated from marmosets preferred lactose. In addition, B. animalis ssp. lactis (#41) and B. mongoliense (#12) isolated from yogurt and horse milk wine, respectively, preferentially utilized lactose.

Figure 2. Preference of glucose and lactose utilization and the ability of each Bifidobacterium strain to grow on each carbon source. Strains are sorted based on their isolation source.

Relationship between lactose content in host milk and lactose-preferring bifidobacterial strains

Considering that B. animalis ssp. lactis and B. mongoliense are commonly found in lactose-rich environments, we hypothesized that the lactose content in the milk of their host animals is associated with their preference for lactose utilization. A literature search was conducted to obtain data on the lactose content in mammalian milk [Table 2]. The lactose concentration in the breast milk of humans and marmosets was relatively higher (≥ 75 g/L) than that in the milk of other animals (< 75 g/L), and most strains isolated from humans and marmosets showed a relatively strong preference for lactose [Figure 2 and Supplementary Figure 4].

Lactose contents in the fecal samples of breastfed infants

Finally, to infer the lactose availability in the gut microbial ecosystem of human infants, lactose concentrations were determined in fecal samples obtained from 10 breastfed infants at 2, 4, 6, and 8 weeks of age. Lactose was detected in 37 out of 40 samples, with median concentrations of 0.47, 0.42, 0.29, and 0.22 mg/g feces at the respective time points [Figure 3].

Figure 3. Lactose concentration in fecal samples collected from 10 breastfed infants at 2, 4, 6, and 8 weeks of age. (A) Boxplots of lactose concentrations. Medians with interquartile range are shown. The whiskers represent quartile range; (B) Changes in fecal lactose concentrations in individuals over 6 weeks.

Upper heatmap displays the optical density (OD_600nm) values of Bifidobacterium cultures in MRS medium supplemented with 0.8% w/v glucose or lactose. Lower heatmap shows the OD_600nm values of Bifidobacterium cultures in MRS medium supplemented with a mixture of 0.4% w/v glucose and 0.4% w/v lactose. Colored bars indicate the phylogenetic group to which each strain belongs. Black bars indicate the strains that were not classified into any group. Pie charts represent the ratios of glucose- (blue) and lactose-preferred (orange) strains within the respective taxonomic groups.

This figure is a reorganization of Figure 1 based on the isolation source rather than the phylogenetic group. The light gray bar indicates strains that were not classified into any group. The number of each strain is listed in Table 1.