|Home | About | Journals | Submit | Contact Us | Français|
Several probiotic strains of Bifidobacterium animalis subsp. lactis are widely supplemented into food products and dietary supplements due to their documented health benefits and ability to survive within the mammalian gastrointestinal tract and acidified dairy products. The strain specificity of these characteristics demands techniques with high discriminatory power to differentiate among strains. However, to date, molecular approaches, such as pulsed-field gel electrophoresis and randomly amplified polymorphic DNA-PCR, have been ineffective at achieving strain separation due to the monomorphic nature of this subspecies. Previously, sequencing and comparison of two B. animalis subsp. lactis genomes (DSMZ 10140 and Bl-04) confirmed this high level of sequence similarity, identifying only 47 single-nucleotide polymorphisms (SNPs) and four insertions and/or deletions (INDELs) between them. In this study, we hypothesized that a sequence-based typing method targeting these loci would permit greater discrimination between strains than previously attempted methods. Sequencing 50 of these loci in 24 strains of B. animalis subsp. lactis revealed that a combination of nine SNPs/INDELs could be used to differentiate strains into 14 distinct genotypic groups. In addition, the presence of a nonsynonymous SNP within the gene encoding a putative glucose uptake protein was found to correlate with the ability of certain strains to transport glucose and to grow rapidly in a medium containing glucose as the sole carbon source. The method reported here can be used in clinical, regulatory, and commercial applications requiring identification of B. animalis subsp. lactis at the strain level.
Probiotics are currently defined as live microorganisms which, when administered in adequate amounts, confer a health benefit on the host (12). Many of the organisms studied for their probiotic potential are members of lactic acid bacteria and the genus Bifidobacterium, which has resulted in their inclusion in a large variety of dietary supplements and food products. Relative to most bifidobacterial species of human origin, Bifidobacterium animalis subsp. lactis is less sensitive to stressful conditions (bile, acid, and oxygen) which might be encountered in the mammalian gastrointestinal tract or in fermented or acidified dairy products (7, 26, 28, 31, 37). B. animalis subsp. lactis is widely added to commercial products because it is better able to withstand the adverse conditions of starter culture and product manufacture and to maintain viability and stability during product shelf-life (30). Therefore, strains of B. animalis, specifically B. animalis subsp. lactis, have been found in the majority of probiotic-supplemented dairy products surveyed in North America (the United States and Canada) and Europe (Great Britain, France, Italy, and Germany) (6, 13-15, 21, 22, 28, 29, 32, 49).
When selecting a probiotic microorganism to add to supplements or foods, the strain must be identified at the genus, species, and strain levels (40). Proper characterization of a strain is important for safety and quality assurance, for identifying and differentiating putative probiotic strains, and for understanding the interactions among members of gut microbiota. In addition, proper characterization is important to maintain consumer confidence. Product labels often list invalid names of organisms or misidentify the species the product contains, leading to consumer confusion (6, 16, 20, 28, 29, 35, 38, 49). In the case of Bifidobacterium, most dairy products sold in the United States do not identify species, and many only refer to the invalid name “Bifid” or “Bifidus.” At the very least, added microorganisms should be accurately identified to the species level on product labels.
According to the FAO/WHO guidelines for probiotic use, specific health benefits observed in research using a specific strain cannot be extrapolated to other, closely related strains (12). Although most clinical studies of probiotic strains compare strains of different genera or different species, few studies have assessed the actual variability of expected health benefits within species or subspecies. However, it is reasonable to consider that health effects, like the phenotypic traits exhibited by strains within a species, are strain specific. Therefore, reliable techniques for the identification of probiotic organisms at the strain level are required.
Characterization to the strain level has several important potential applications. Understanding the complex interactions among microorganisms in the intestinal ecosystem requires methods of differentiating a strain of interest from other strains of the same species contained in the autochthonous microbiota. Strain differentiation techniques also aid in assessing survival of a probiotic organism through the gastrointestinal system, which is particularly important for clinical trials and regulatory purposes (17). The ability to uniquely identify a strain also lends credibility to statements made about the potential health benefits of consuming a particular product containing a strain with demonstrated probiotic effects and supports the licensing or intellectual property rights of the manufacturer.
The high degree of genome conservation observed between strains of B. animalis subsp. lactis in terms of size, organization, and sequence is indicative of a genomically monomorphic subspecies (2, 25; also HN019 GenBank project 28807). As an example, comparison of the complete genome sequences of two B. animalis subsp. lactis strains, DSMZ 10140 (the type strain) and Bl-04 (a commercial strain, also known as RB 4825) (2), identified 47 single-nucleotide polymorphisms (SNPs) in nonrepetitive elements, as well as 443 bp distributed among four INDEL sites: a 121-bp tRNA-encoding sequence, a 54-bp region within the long-chain fatty acid-coenzyme A ligase gene, a 214-bp region within the CRISPR (clustered regularly interspaced short palindromic repeats) locus, and a 54-bp intergenic sequence. Overall, this 99.975% genome identity explains the inability to differentiate these strains by techniques such as the sequencing of housekeeping genes, multilocus sequence typing, and pulsed-field gel electrophoresis (PFGE) (3, 9, 23, 39, 44-46, 50).
The strain specificity of reported health benefits of probiotics and the frequent use of B. animalis subsp. lactis as a probiotic in food products and supplements demands techniques with greater discriminatory power to identify and differentiate among strains within this highly homogeneous group. Unfortunately, strain level differentiation of B. animalis subsp. lactis presents several challenges. Although Ventura and Zink were able to differentiate strains of B. animalis subsp. lactis by sequencing the 16S-23S internal transcribed sequence (ITS) region (47), analysis of the four ITS operons between DSMZ 10140 and Bl-04 indicated complete identity (2). However, SNPs and INDELs do have potential for strain differentiation. According to Achtman, focusing on polymorphic SNPs is a desirable approach for the typing of monomorphic species (1). Therefore, the objective of the present study was to exploit the previously identified SNP and INDEL sites to develop a technique capable of differentiating among a collection of B. animalis subsp. lactis strains obtained from culture collections and commercial starter culture companies.
B. animalis subsp. lactis DSMZ 10140 and ATCC 27536 were obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ; The German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany) and the American Type Culture Collection (ATCC; Manassas, VA), respectively. Twenty additional strains were obtained directly from six commercial starter culture companies (Table (Table1)1) . By agreement with the suppliers, commercial strains are not identified by their trade names but are identified by a random four-digit number preceded by “RB” (e.g., RB 0171). Isolates of B. animalis subsp. lactis Bl-04 (RB 4825), BL-01 (RB 5251), Bi-07 (RB 5733), B420 (RB 7239), and HN019 were obtained from Danisco USA, Inc. Strains Bl-04 and RB 4825 represent chronologically distinct isolates of the same strain and were analyzed separately.
All strains were assayed for activity of fructose-6-phosphate phosphoketolase (F6PPK; EC 22.214.171.124) as described by Scardovi (42) and Biavati and Mattarelli (4). Genus and subspecies identification were confirmed by using PCR amplification of a region of 16S rRNA gene based on the method of Kaufmann et al. (24) and by PCR amplification of a B. animalis subsp. lactis-specific sequence in the 16S-ITS region as described by Ventura and Zink (47), respectively. Stock cultures were prepared in liver lactose (LL) (27), supplemented to 10% glycerol (vol/vol), and stored at −70°C.
Strains were grown anaerobically in tryptone-phytone-yeast extract broth (42) until turbid, and then the cells were pelleted by centrifugation. Lactic and acetic acid concentrations were determined in the supernatant by high-performance liquid chromatography (8). The activities of 19 enzymes were evaluated in strains grown on LL using API ZYM kits (bioMérieux, Inc., Durham, NC). A cell suspension was used to inoculate the test kits according to the manufacturer's instructions. The results were scored after 4.5 h of aerobic incubation at 37°C. Carbohydrate utilization was evaluated using API 50 CH kits (bioMérieux, Inc.). Commercial strains were prepared in LL or liver glucose (LG) medium. ATCC and DSMZ strains were prepared in LL, LG, or reinforced clostridial media. A cell suspension was used to inoculate API 50 CHL medium (bioMérieux, Inc.), which was then used to rehydrate the carbohydrate substrates in the test kits according to the manufacturer's instructions. Incubation was performed anaerobically at 37°C, and results were scored after 48 h. Glucose uptake assays were performed as previously described (8) by incubating mid-log-phase (OD600 ~0.5) cells with d-[U-14C]glucose.
DNA was extracted for PCR as described by Vincent et al. (48). Briefly, overnight cultures of bifidobacteria in LL broth were harvested by centrifugation, washed, and lysed. DNA was extracted with chloroform-isoamyl alcohol (24:1) three times and precipitated with cold isopropanol. DNA concentration was estimated by measuring A260 with a DU-650 spectrophotometer (Beckman Coulter, Inc., Somerset, NJ).
A rapid pulsed-field gel electrophoresis (PFGE) method (9) was used to further characterize strains of B. animalis subsp. lactis, comparing restriction patterns of chromosomal DNA digested with XbaI or SpeI. Strain comparison by randomly amplified polymorphic DNA-PCR analyses (RAPD-PCR) was performed using seven different primers. Primers 103 (5′-GTGACGCCGC-3′), 127 (5′-ATCTGGCAGC-3′), and 173 (5′-CAGGCGGCGT-3′) are from Sakata et al. (41). Primers AB-1 (5′-GGTGCGGGAA-3′) and AB-5 (5′-AACGCGCAAC-3′) are from the Amersham Biosciences Ready-to-Go RAPD analysis beads technical insert. Primers OPV-07 (5′-GAAGCCAGCC-3′) and OPR-13 (5′-GGACGACAAG-3′) are from Mayer et al. (33).
Primers were designed in the regions flanking the SNP or INDEL previously identified in the alignment of the DSMZ 10140 and Bl-04 genome sequences (2). Sequences surrounding the SNP or INDEL were entered into Primer3 (http://frodo.wi.mit.edu/), and primers were selected. Primer sets were evaluated by using nucleotide BLAST against the DSMZ 10140 genome to ensure that each primer only annealed at one position. All primers were designed with an annealing temperature of 60°C.
Amplification mixtures (50 μl) consisted of 10 μl of 5× colorless GoTaq reaction buffer (Promega, Madison, WI) with a final concentration of 1.5 mM MgCl2, 300 μM concentrations of each deoxynucleotide triphosphate (Promega), 0.5 μM concentrations of each forward and reverse primer (see Table S1 in the supplemental material; Integrated DNA Technologies, Coralville, IA), 100 ng of template DNA, and GoTaq DNA polymerase (1.5 U). Amplifications were performed with 1 cycle of 95°C for 5 min; followed by 35 cycles of 95°C for 1 min, 58°C for 45 s, and 72°C for 1 min; and then a final cycle of 72°C for 7 min. Amplicons of the reaction mixtures were separated on a 1.0% agarose gel using 0.5× Tris-borate-EDTA buffer. A 100-bp DNA ladder (Promega) was included as a molecular weight marker. Electrophoresis was performed by using a submerged horizontal gel electrophoresis system at 110 V for 90 min. After staining in a solution of ethidium bromide for at least 1 h, bands were visualized on a UV transilluminator, and images were captured by using an AlphaImager 3300 gel documentation system (Alpha Innotech, San Leandro, CA). Amplicons of the appropriate size were extracted from the agarose gel and purified by using the QIAquick gel extraction kit (Qiagen, Valencia, CA). Purified products were sequenced at the Genomic Core Facility at The Pennsylvania State University using 3′ BigDye-labeled dideoxynucleotide triphosphates (v 3.1 dye terminators; Applied Biosystems, Foster City, CA) and run on an ABI 3730XL DNA analyzer using the ABI data collection program (v2.0). Sequences were analyzed in SeqMan version 6.0 (DNASTAR, Madison, WI) and aligned by using CLUSTAL W. For quality assurance purposes, each time PCR and sequencing were performed, DSMZ 10140 and Bl-04 were included as controls.
All allelic sequence data resulting from the resequencing of the 47 SNPs and three INDELs in 24 strains were uploaded in JMP Genomics (SAS Institute, Cary, NC). Sequences from a strain that matched the sequence from DSMZ 10140 or Bl-04 were assigned values of 1 and 3, respectively. A unique sequence from a strain that did not match either DSMZ 10140 or Bl-04 was assigned a 2. A maximum of three sequence types were observed for any locus investigated. Hierarchical two-way clustering was carried out using the fast Ward algorithm to analyze both strains and allelic profiles simultaneously and visualize polymorphism of the 24 strains across the 50 genetic loci. These analyses allow identification of the number of sequence types, clustering of strains in similar genotypes, and provide an overview of variability across genetic loci.
The discriminatory power of the method (19) with 95% confidence intervals was calculated by using EpiCompare version 1.0 (Ridom GmbH, Wurzburg, Germany).
All strains in the study were confirmed as Bifidobacterium by the presence of F6PPK activity and by PCR with genus-specific primers. Metabolism of glucose by bifidobacteria involves the action of F6PPK, producing lactic and acetic acids as the final end products in a theoretical molar ratio of 1.5:1 (43). When grown in tryptone-phytone-yeast extract broth, the strains in the present study exhibited production of both lactic and acetic acids (data not shown). All strains were further confirmed as B. animalis subsp. lactis by PCR using subspecies-specific primers targeting a region in the 16S-ITS (data not shown).
In the present study, all of the B. animalis subsp. lactis strains examined exhibited C4 esterase, leucine arylamidase, acid phosphatase, naphthol-AS-BI-phosphohydrolase, β-galactosidase, α-glucosidase, and β-glucosidase activities (data not shown). This is consistent with previously published work on enzymatic activities of bifidobacteria isolated from commercial products (15).
With respect to carbohydrate utilization, when harvested from LL, all commercial strains of B. animalis subsp. lactis fermented d-ribose, esculin, d-maltose, d-lactose, d-melibiose, sucrose, and d-raffinose, as did DSMZ 10140 and ATCC 27536. Ten of the commercial strains, along with the two culture collection strains, also fermented glucose, while the other ten commercial strains did not (Table (Table1).1). All of the strains evaluated differed from the description of B. animalis subsp. lactis in the literature (34) because they were unable to ferment l-arabinose, d-xylose, amygdalin, gentibiose, and potassium 5-ketogluconate. According to Gueimonde, et al. (15), carbohydrate fermentation patterns for B. animalis subsp. lactis strains isolated from commercial products were all positive only for glucose, esculin, and raffinose and most strains could be differentiated by utilization of at least one carbohydrate. However, this was not observed with the strains examined in the present study. The results of carbohydrate assays can vary depending on the medium used to prepare the inoculum, which was also observed in this work. For example, DSMZ 10140 fermented amygdalin when cells were harvested from RC medium, but not from LL or LG media and ATCC 27536 did not ferment lactose when harvested from RC medium, but did when harvested from LL medium. It is possible differences observed among the commercial B. animalis subsp. lactis strains evaluated by Gueimonde et al. and the strains evaluated in the current study are related to preparation on glucose-based (MRS) and lactose-based (LL) media, respectively.
Strains of B. animalis subsp. lactis were grown in LL and then evaluated for glucose uptake. A correlation between glucose fermentation pattern with the API CH 50 kit and glucose uptake was observed (Table (Table1).1). Strains that yielded a positive result for glucose fermentation also exhibited glucose uptake greater than 4.0 nmol/min/mg of cell protein, while strains with a negative result for glucose fermentation exhibited glucose uptake less than 1.5 nmol/min/mg of cell protein. The results of the carbohydrate fermentation pattern and glucose uptake represented the main phenotypic difference among the strains. Based on these characteristics alone, the strains could be differentiated into two groups (Table (Table11).
For differentiation of bacterial strains, PFGE has been considered the “gold standard” (5, 10, 12, 36). Therefore, PFGE was used in an attempt to differentiate among commercial strains of B. animalis subsp. lactis. PFGE patterns (Fig. (Fig.1)1) with XbaI were identical for all 22 strains examined, whereas ATCC 27536 exhibited a two-band difference with SpeI (Fig. (Fig.1,1, lane 2), likely from the loss of a restriction site.
Homogeneity of PFGE patterns among strains of B. animalis subsp. lactis has been reported previously in the literature (7, 11, 14). However, different PFGE patterns were observed for two commercial strains of B. animalis subsp. lactis, HN019 and Bifidobacterium sp. strain 420, by Mayer et al. (33), although the authors acknowledged these results contradicted previous reports. One additional study reported differentiation among strains of B. animalis subsp. lactis isolated from commercial products, with four PFGE pattern types (28). The results in the current study indicate high genetic homogeneity among all commercial strains of B. animalis subsp. lactis and suggest that the observation of different patterns resulted from either instability in the B. animalis subsp. lactis genome, a more diverse strain collection, or analytical artifacts.
In an additional attempt to differentiate the strains, RAPD-PCR was performed using seven different primers. A representative gel obtained with primer 103, and the B. animalis subsp. lactis strains is shown in Fig. Fig.2.2. Inter- and intraspecies differences were observed with RAPD-PCR primers among other species of bifidobacteria (data not shown); however, no differences were observed among the strains of B. animalis subsp. lactis with any of the seven RAPD-PCR primers evaluated. Three of the RAPD primers selected for evaluation (primers 103, 127, and 173) were included because they had been used to differentiate among and within strains of B. longum, B. infantis, and B. suis (41). Two of the RAPD-PCR primers (OPV-07 and OPR-13) were selected for evaluation because they have been shown to differentiate B. animalis subsp. lactis LAFTI B94 from other commercial strains of B. animalis subsp. lactis (33). B. animalis subsp. lactis LAFTI B94 was obtained directly from the supplier and included in the present study among the RB strains; however, OPV-07 and OPR-13 could not differentiate B94 from other strains in the group, suggesting alternative primers are needed to differentiate these strains. It is important to note that RAPD-PCR is very sensitive and slight differences in reagents or DNA preparation may explain the results between the present study and the one by Mayer et al.
Differentiating among a genomically monomorphic group, such as this collection of B. animalis subsp. lactis strains, was not possible with a variety of conventional phenotypic and nucleic acid-based techniques. Although glucose uptake allowed separation of the strains into two major groups, PFGE and RAPD-PCR were ineffective. With genome sequences for DSMZ 10140 and Bl-04 recently available (2), in silico analysis of restriction digests of DSMZ 10140 and Bl-04 with XbaI and SpeI confirmed the banding patterns observed by PFGE (data not shown). These strains did not differ in terms of the number of predicted fragments with XbaI and SpeI (29 and 24, respectively), and comparison of the fragments indicated no loss or gain of restriction sites. Although the four INDELs between the two strains did result in a few fragments with different sizes, the differences were relatively too small (≤0.2 kb) to be discerned on the PFGE gels under the electrophoresis conditions used.
Comparative analysis of the DSMZ 10140 and Bl-04 genomes revealed little diversity—47 SNPs and four INDELs—between these two strains (2). Forty-seven SNPs and three INDELs (INDELs 1 to 3) were examined across all 24 of the strains in our collection, serving as a basis for the development of a strain-specific typing method. Sequencing reactions for INDEL 4 failed to consistently provide unambiguous data for certain strains; therefore, this locus was omitted from the analysis.
Fourteen distinct genetic clusters were identified across the 24 strains based on hierarchical clustering (Fig. (Fig.3).3). Of these 14 strain clusters, 10 are comprised of a single strain, while four contain two, three, four, or five strains that cannot be differentiated from the other strains within the group. Multistrain groups may represent different isolates of the same strain, as is known for the chronologically distinct isolates RB 4825 and Bl-04, or may differ at loci that were not examined. Perhaps not surprisingly, widely used commercial strains, such as BB-12, Bl-04, and Bi-07 (RB 5733), fall within multistrain clusters (see Fig. Fig.3,3, strain clusters 3, 9, and 12, respectively). Other loci would need to be analyzed in order to determine whether these strains are genetically different or represent different isolates of the same strain.
Ten distinct clusters of genetic loci were identified across the 50 alleles evaluated based on hierarchical clustering (Fig. (Fig.3).3). Of these 10 clusters, 7 are comprised of a single allelic distribution across the strains. In contrast, three clusters (genetic locus clusters 6, 4, and 7 comprised of 2, 6, and 35 genetic loci, respectively) exhibit conserved allelic distributions across the strains within an individual cluster. This analysis allows selection of the minimum number and identity of genetic loci that must be evaluated for maximum genetic differentiation of this collection of strains. Accordingly, a minimum of nine genetic loci used in combination (Fig. (Fig.3,3, one from each of the genetic loci clusters 1 to 6 and 8 to 10) will differentiate the 24 strains in this collection into 14 strain clusters. Thirty-five of the alleles evaluated are DSMZ 10140-specific (genetic loci cluster 7 on Fig. Fig.3)3) and one allele, Balat_0141, is specific for RB 0171.
In order to assess the ability of a typing scheme to separate strains, discriminatory power is often calculated (19). The discriminatory power of the allelic profiling method described was calculated to be 0.92 with a 95% confidence interval (CI) of 0.852 to 0.988, indicating that there is a 92% chance that any two randomly selected strains will be placed into two different groups. For comparative purposes, the discriminatory power calculated for the same strain set using the glucose phenotype was a discriminatory power of 0.52 (95% CI, 0.481 to 0.558).
A nonsynonymous SNP was identified in a putative glucose uptake gene (glcU; DSMZ 10140 position 1260073). Upon sequencing, it was determined each of the strains that exhibited slow growth on glucose and low glucose uptake possess one genotype, whereas the strains that exhibited normal growth on glucose and greater glucose uptake possess another genotype (Table (Table1).1). This SNP has the potential to explain the differences observed in glucose utilization and transport among these strains since the genotype is correlated to phenotype for the entire strain set. In addition, during analysis of the strain collection, a second SNP identified in the glcU gene (DSMZ position 1260380) was able to differentiate RB 1791 and RB 7239 from all other strains. Based on genotypic analysis at these two SNP sites, it is possible to break the collection into three distinct groups (Table (Table1),1), whereas it was only possible to generate two groups based on the phenotype of glucose uptake.
A collection of 24 strains of B. animalis subsp. lactis, including reference and commercial strains, could not be differentiated by PFGE or RAPD-PCR, which are considered to be discriminatory typing methods. The genetic dynamic range (1) of this collection is defined by the distance between Bl-04 and DSMZ 10140, the two reference genomes for this subspecies. Visual analysis of the allelic distribution of the 24 strains across the 50 genetic loci clearly indicates that DSMZ (the type strain) is the most unique strain with 35 distinct alleles (genetic locus cluster 7 on Fig. Fig.3)3) and that all other strains are more similar to Bl-04. However, in addition to DSMZ 10140, it is also clear that two distinct families of strains appear to exist (Fig. (Fig.33).
Overall, the SNP/INDEL analysis revealed polymorphism in the culture collection and demonstrates that a combination of only nine genetic loci needs to be analyzed to differentiate the 24 strains of this collection into 14 clusters. These nine genetic loci provide maximum genetic discrimination among the collection, which includes widely used commercial and reference strains. Although this collection of 24 isolates of B. animalis subsp lactis exhibited a high degree of relatedness, one phenotypic difference, related to glucose uptake, was observed and correlated with the glcU genotype.
A functional overview of the 50 variable genetic loci (Table (Table2)2) indicates that transporters and CRISPR elements are highly represented in the differential genetic content. This might indicate selective evolutionary pressure on hypervariable loci (CRISPR) and genes involved in the adaptation to the environment (transcriptional regulators, carbohydrate uptake, and metabolism). This is consistent with previous reports indicating the polymorphic nature of CRISPR loci in lactic acid bacteria, notably in Bifidobacterium (18).
This typing method can be used in clinical, regulatory, and commercial applications requiring identification of B. animalis subsp. lactis at the strain level. Also, it can be exploited in analyses and surveys of environmental samples.
We thank Allen Phillips for assistance with the glucose uptake assays.
This research was funded in part by a seed grant from the Penn State University College of Agricultural Sciences. R.B. is supported by Danisco USA, Inc.
Published ahead of print on 2 October 2009.
†Supplemental material for this article may be found at http://aem.asm.org/.