RESEARCH ARTICLE

Genome Analysis and Characterisationof the ExopolysaccharideProduced by Bifidobacteriumlongum subsp. longum 35624

TM

a11111

OPEN ACCESS

Citation: Altmann F, Kosma P, O’Callaghan A, Leahy S, Bottacini F, Molloy E, et al. (2016) Genome Analysis and Characterisation of the Exopolysaccharide Produced by Bifidobacterium longum subsp. longum 35624

TM

. PLoS ONE 11(9): e0162983. doi:10.1371/journal.pone.0162983 Editor: Mohamed N. Seleem, Purdue University, UNITED STATES Received: January 12, 2016 Accepted: August 20, 2016 Published: September 22, 2016 Copyright: © 2016 Altmann et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All chemical data is contained within the paper. The accession number for the 35624 genome sequence is CP013673 (NCBI - http://www.ncbi.nlm.nih.gov/genbank/). Funding: These studies were directly supported by a European Union Marie Curie training network, entitled “TEAM-EPIC”. In addition, the authors are supported by Swiss National Foundation grants (project numbers CRSII3_154488 and 310030_144219), Christine Ku¨hne Center for Allergy Research and Education, and by Science Foundation Ireland (SFI) (Grant No. SFI/12/RC/

Friedrich Altmann1, Paul Kosma1, Amy O’Callaghan2, Sinead Leahy2, Francesca Bottacini2, Evelyn Molloy2, Stephan Plattner3, Elisa Schiavi4,5, Marita Gleinser2, David Groeger5, Ray Grant5, Noelia Rodriguez Perez4, Selena Healy3, Elisabeth Svehla1, Markus Windwarder1, Andreas Hofinger1, Mary O’Connell Motherway2, Cezmi A. Akdis4, Jun Xu6, Jennifer Roper3, Douwe van Sinderen2, Liam O’Mahony4*

1 University of Natural Resources and Life Sciences, Vienna, Austria, 2 APC Microbiome Institute and School of Microbiology, University College Cork, Cork, Ireland, 3 Alimentary Health, Cork, Ireland, 4 Swiss Institute of Allergy and Asthma Research (SIAF), University of Zu¨rich, Davos, Switzerland, 5 Alimentary Health Pharma Davos, Davos, Switzerland, 6 Procter & Gamble, Cincinnati, United States of America

* [email protected]

TM

The Bifibobacterium longum subsp. longum 35624

strain (formerly named Bifidobacterium longum subsp. infantis) is a well described probiotic with clinical efficacy in Irritable Bowel Syndrome clinical trials and induces immunoregulatory effects in mice and in humans. This paper presents (a) the genome sequence of the organism allowing the assignment to its correct subspeciation longum; (b) a comparative genome assessment with other B. longum strains and (c) the molecular structure of the 35624 exopolysaccharide (EPS624). Comparative genome analysis of the 35624 strain with other B. longum strains determined that the sub-speciation of the strain is longum and revealed the presence of a 35624-specific gene cluster, predicted to encode the biosynthetic machinery for EPS624. Following isolation and acid treatment of the EPS, its chemical structure was determined using gas and liquid chromatography for sugar constituent and linkage analysis, electrospray and matrix assisted laser desorption ionization mass spectrometry for sequencing and NMR. The EPS consists of a branched hexasaccharide repeating unit containing two galactose and two glucose moieties, galacturonic acid and the unusual sugar 6deoxy-L-talose. These data demonstrate that the B. longum 35624 strain has specific genetic features, one of which leads to the generation of a characteristic exopolysaccharide.

2273). AO was supported by an enterprise partnership scheme of the Irish Research Council, while MOM is a recipient of an HRB postdoctoral fellowship (Grant No. PDTM/20011/9). The funder provided support in the form of salaries for authors [DG, E. Schiavi, RG – Alimentary Health Pharma Davos; SP, SH, JR – Alimentary Health; JX – Procter & Gamble], but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘author contributions’ section.

Competing Interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: David Groeger, Elisa Schiavi and Ray Grant are employees of Alimentary Health Pharma Davos AG, Jun Xu is an employee of Procter & Gamble, while Stephan Plattner, Selena Healy and Jennifer Roper are employees of Alimentary Health Ltd. Liam O’Mahony has received research funding from GSK and is a consultant to Alimentary Health Ltd. Cezmi Akdis has received research support from Novartis and Stallergenes and consulted for Actellion, Aventis and Allergopharma. Friedrich Altmann, Paul Kosma, Elisabeth Svehla, Markus Windwarder, Andreas Hofinger, Amy O’Callaghan, Sinead Leahy, Francesca Bottacini, Evelyn Molloy, Noelia Rodriguez Perez, Elisa Schiavi, Marita Gleinser, Mary O’Connell Motherway and Douwe van Sinderen have no competing interests. While competing interests are declared for some of the authors, the content of this article was neither influenced nor constrained by this fact and this does not alter our adherence to PLOS ONE policies on sharing data and materials.

The gut microbiome contributes to host health by multiple mechanisms, including digestion, competitive exclusion of pathogens, degradation of mucins, enhancement of epithelial cell differentiation and promotion of mucosa-associatedlymphoid tissue proliferation [1, 2]. Furthermore, accumulating evidencesuggests that the composition and metabolic activity of the gut microbiome has profound effects on proinflammatory activity and the induction of immune tolerance within mucosal tissue [3–6]. Certainmicrobes induce regulatory responses, while others induce effector responses, resulting in the case of healthy individuals in a balanced homeostatic immunologicalstate, which protects against infection and controls aberrant, tissue-damaging inflammatory responses [7].

The Bifidobacterium longum subsp. longum 35624

strain (formerly named B. longum

TM

subsp. infantis) is known to induce tolerogenic responses within the gut and reduces disease symptoms in irritable bowel syndrome patients [8–10]. Induction of T regulatory cells by B. longum 35624 strain (35624) in mice is associated with protection against colitis, arthritis,

allergic responses and pathogen-inducedinflammation [11–15]. Administration of this bacterium to humans increases Foxp3+ lymphocytes in peripheral blood, enhances IL-10 secretion ex vivo, and reduces the level of circulating proinflammatory biomarkers in a wide range of

patient groups and healthy volunteers [16, 17]. A number of host mechanisms have been described,which contribute to the anti-inflammatory activity of this microbe, including TLR-2 and DC-SIGN recognition,and retinoic acid release by dendritic cells [16, 18].

Bifidobacterialstrain-specificstructures that interact with the host are recently being better describedand bifidobacterial-associatedexopolysaccharides (EPS) are of particularinterest in the field. The genetic content and structure of bifidobacterialEPS clusters has already been describedas being highly variable [19]. A number of in vitro studies have demonstrated how EPS from members of several bifidobacterialspecies elicit different imunologicalresponses [20]. For example, an in vivo study of EPS produced by Bifidobacterium breve UCC2003 showed how it promotes persistence in the gut following colonization through evasion of the adaptive host immune response [21].

In this study we wished to identify and characterize 35624 strain-specificfeatures, such as the 35624 EPS. We sequencedthe genome of 35624 and following core genome-basedcomparisons to related bifidobacterialstrains, 35624 was shown to clearly belong to the B. longum subsp. longum phylogenetic group. Furthermore, a 35624-specificgene cluster, designated eps624, predicted to be responsible for the production of a exopolysaccharide (EPS624), was identified.Following detailedchemical analysis, we identifiedthat EPS624 contained a branched hexasaccharide repeating unit with two galactoses,two glucoses,galacturonic acid and the unusual sugar 6-deoxytalose.

The complete genome sequence of the 35624 strain was determinedto be 2.26 Mb in size with a G+C % content of 59.34% (for salient features of this genome see Table 1; S1 Fig), consistent with values reported for other bifidobacterialgenomes [22]. A total of 1,735 open reading frames (ORFs) were identified,and of these a functioncould be assigned to 1,370 based on similarity searches to public data bases. In order to accurately evaluate 35624’s phylogeny and (sub)speciesassignment, a B. longum phylogenetic supertreewas constructedbased on the deducedprotein sequences of the B. longum core-genome [23]. From the resulting supertree,

1. Table 1. Bifidobacterium longum general genome features.

1. doi:10.1371/journal.pone.0162983.t001

two major clades were identified(Fig 1), of which the largest encompasses twenty six strains of the B. longum subsp. longum phylogenetic group, including the type strain of this subspecies

1. [23]. 35624 had, similar to strains B. longum 157F and B. longum CCUG52486,originally been classifiedas a subspeciesinfantis member. However, classification as based on the phylogenetic supertreeapproach clearly shows that these three strains are positioned within the B. longum subsp. longum phylogenetic group, and therefore represent bona fide members of the subspecies longum, their original miss-assignment being due to the sole use of (partial) 16S rRNAbased taxonomic classification [23].
2. [24]. Interestingly, 35624 is not predicted to encode so-calledsortase-dependentpili, which have previously been shown to play a role in host-microbe interactions [25]. Comparative analysis was performedbetweenthe 35624 genome and other B. longum genome sequences in the hope to locate genetic elements specifyingother extracellular structures that may be involved in such interactions and that would exhibit 35624-specificattributes. The main genomic difference between35624 and other B. longum genomes is the presence of a 26.2 Kb gene cluster,

designated here as eps624 (Fig 2), predicted to encode the biosynthetic machinery for EPS biosynthesis. Comparative analysis of complete and publicly available B. longum subsp. longum genomes shows that a predicted EPS-specifyinggene cluster is present in the same location in other B. longum subs. longum genomes, but that the genetic composition of these gene clusters is very diverse at intraspecies level and that the eps624 gene cluster only displays partial similarity to three other EPS-specifyinggene clusters (Fig 2). This analysis furthermorereveals that some of these EPS gene clusters lack certain critical functions required for EPS synthesis (Fig 2). For example, the EPS cluster in B. longum subsp. longum JCM1217 appears to lack a flippase-encodinggene which is responsible for the export of EPS precursors across the cell membrane. These findings are in agreement with a recent study [23].

Notably, the eps624 cluster encodesa number of key enzymes that are predicted to be required for EPS production by means of the so-calledWzx/Wzy-dependent pathway, which typically employs a priming glycosyltransferase (pGT), one or more glycosyl transferases (GHs), a flippase, and a polymerase to produce an extracellular heteropolysaccharide [19, 26]. The first gene (corresponding to locus tag BL_0342 and designated here as pgt624) of the eps624 gene cluster is predicted to encode the pGT, which adds the first monosaccharide to a cytoplasmic, membrane-bound carrier moleculeundecaprenyl as part of the oligosaccharide

doi:10.1371/journal.pone.0162983.g001

subunit biosynthesis [27]. The eps624 cluster encodes five additional GTs (corresponding to locus tags BL_0345, BL_0346, BL_0349 and BL_0352; Fig 2), which are predicted to each add one monosaccharide to the carrier moleculeso as to complete the oligosaccharidesubunit, prior to its export to the external side of the membrane by a flippase (predicted to be encoded by a gene corresponding to locus tag BL_0355) and its subsequent use by a polymerase (putatively specifiedby locus tag BL_0353) to produce the EPS polymer. Interestingly, two adjacent genes of the eps624 cluster, corresponding to B624_0347 and B624_0348, are predicted to encode a UDP-glucuronate 5'-epimerase and a UDP-glucose 6-dehydrogenase, suggesting that one of the incorporated monosaccharides of the EPS is an epimer of glucuronic acid, e.g. galacturonic acid or mannuronic acid. Three genes located within the eps624 cluster, corresponding to locus tags B624_0360 through to B624_0362 (Fig 2), encode enzymes known to be

1. Fig 2. EPS gene cluster. Illustration of the EPS cluster located in the B. longum 35624 genome and comparison to similar clusters located in B. longum 105-A, B. longum subsp. longum JCM1217 and B. longum subsp. longum NCC2705. Each gene is colour-coded according to function which is indicated in the legend located at the end of the page. Percentages represent the percent of sequence similarity at the protein level with corresponding genes in the B. longum 35624 genome. The locus tags of the first and last genes located in the EPS clusters of B. longum 105-A, B. longum subsp. longum JCM1217 and B. longum subsp. longum NCC2705 are also indicated in the illustration. doi:10.1371/journal.pone.0162983.g002

involved in the biosynthesis of dTDP-L-rhamnose [19, 28, 29], while the deducedprotein products of B624_0350 and B624_0357 are predicted to encode NAD-dependent reductase/ epimerase enzymes. Such enzymes have been shown to be involved in the rerouting of the dTDP-L-rhamnose biosynthesis pathway towards the production of dTDP-D-fucose or dTDP6-deoxy-L-talose[30–32]. The eps624 cluster also encodes two predicted acetyl transferases, similar to enzymes that have previously beenshown to perform O-acetylation reactions on particular sugar components (such as 6-deoxy-L-talose)in polysaccharides [33]. Furthermore, the genes with locus tags B624_0344 and B624_0366 represent putative tyrosine kinase and phosphotyrosine protein phosphatase activities,respectively, which have been associated with

controlling EPS export, polymerisation and production [34, 35]. Therefore, based on the gene content of the eps624 cluster, we predict that the EPS produced by the 35624 strain is composed of a repeating subunit that consists of six monosaccharides,of which one is an epimer of glucuronic acid, one or two others are either D-fucoseor 6-deoxy-talose,and some of which may be O-acetylated. We show below that this prediction is quite correct.

We performedelectron microscopy analysis, which indeed revealed the presence of a thick EPS layer on the cell surface of the 35624 strain (Fig 3A). Following harvesting of 35624 cells, which were grown on agar plates to minimize carryover of media components, an EPS solution was generated by agitating the cells in PBS. The harvested EPS solution was mixed with ethanol and the EPS aggregated at the center of the surface of the ethanol solution, which facilitated harvesting of the EPS without the need for centrifugation.The EPS aggregations were taken with a spatula, resuspended in water and dialysed against water to remove contaminants and residual ethanol. Further purification using reverse phase columns resulted in a highly purified polysaccharide, with no detectable proteins or lipids remaining. The precipitated and purified EPS is illustrated in Fig 3B.

Comparison of the purified35624 EPS with dextran standards by high performance size exclusion chromatography (SEC) indicated an average mass in excess of 1 MDa (Mw) (Fig 4A). Monosaccharide analysis of SEC fractions gave the same composition for all fractions of the broad peak. Anion exchange of the EPS revealed it as being negatively charged. 1H NMR data recorded at 90°C in D2O showed severe line broadening (data not shown) but nevertheless indicated the presence of methyl groups corresponding to acetyl and 6-deoxy protons in an approximate 1:1 ratio.

1. Fig 3. B. longum 35624 electron microscopy. (A) A layer of extracellular polysaccharide is clearly visible by electronic microscopy of the 35624 strain. (B) The isolated and purified EPS is illustrated.

doi:10.1371/journal.pone.0162983.g004

chromatography-mass spectrometry(GC-MS) of partially methylated alditol acetates identified it as 6-deoxy-hexose (data not shown). With NMR data pointing at 6-deoxy-talose (6dTal), this rare sugar was synthetized from L-fucose by epimerization at C-2 and deoxy-hexose indeed co-eluted with 6dTal (Fig 4B). Adding the peak area of the aldobiuronic acid to that of Gal and GalA, the molar ratio of the constituent sugars Glc, Gal, GalA and a 6dTal approximated 2: 2: 1: 1 respectively.

The products of mild acid treatment were analysed by porous graphitic-liquid chromatography-electrospray ionization-mass spectrometry(PGC-LC-ESI-MS).The occurenceof several

isobaric fragments indicated lack of a preferred cleavage site. The smallest of the major fragments was the tetrasaccharide os211 consisting of 2 Hex, 1 GalA and 1 dHex residue. ESI-MS/ MS of the BD4-reduced oligosaccharideidentifiedits reducing end as hexose. Os211 was present as a single isomer, which allowed the preparative isolation of os211 by hydrophilic interaction HPLC for NMR analysis. The os311 and os411 fragments, when analyzed by PGC-LCESI-MS/MS turned out to contain at least 3 isomers with hexose and 1 with deoxyhexose at the reducing end (Fig 5A). Reducing end dHex on os411 and larger fragments implies this sugar to be part of the main chain rather than a side arm. These isomers were isolated by preparative PGC for MS/MS in permethylated form.

Permethylation of the intact polysaccharide failed due its insolubility in DMSO. Results were obtained upon gentle hydrolysis with 50 mM TFA (4 h 80°C). Peaks for a 4-substituted 6-deoxy-hexopyranose (or a 5-substituted furanose),terminal Glc, 4-substituted Gal and 2,4-disubstituted Gal were obtained in a ratio of approximately 0.8: 1: 0.6: 1: 1. A uronic acid cannot be seen by this approach. Taking into account the incomplete hydrolysis of the uronic acid´s glycosidic linkage as its acid function is regained during the hydrolysis step and the higher volatility of dHex, this insinuates that each linkage variants occurs just once in the repeating unit. The low yield of 4-substituted Gal pointed at its substitution by uronic acid, which results in a particularilyacid stable linkage.

Permethylation linkage analysis of fragment os211 revealed terminal Glc, 4-substituted 6-deoxy-hexose and a 4-substituted Gal at the reducing end. The occurrenceof two singly substituted and one terminal residue necessitates a linear topologyof os211. As the occurrence of aldobiuronic acid implies a HexA-Hex sequence,the only topologyof os211 consistent with the hitherto findings is Glc-d6Tal-GalA-Gal.

The chromatographically separated, reduced and perdeuteromethylated EPS fragments were analyzed by MS/MS on a MALDI-TOF/TOF instrument (Fig 5B). The single os211 fragment gave one y and one b fragment, that underpinnedthe Glc-dHex-GalA-Gal topology. In the spectraof os311a and os411a, y-fragments with two and three hexoses and with three hexoses plus uronic acid were found. Necessarily, these y-fragments must have harbored the branching Gal residue bearing the unsubstituted Glc residue. The topologyGalA-(Glc-)GalGal, however, could not lead to a fragment of m/z 503.3, with only one fragmentation char. Rather, the MS/MS spectrafor os311a and os411a could be reconciledwith the fragment structure GalA-Gal-(Glc-)Gal. Os311a did not show a b-fragment from the non-reducing terminus but os411a presented the disaccharide Hex-dHex (= Glc-dTal). Thus, together with the insights from linkage analysis of os211, os311a could be assigned the sequence d6Tal-GalA-Gal-(Glc-) Gal and os411a Glc-d6Tal-GalA-Gal-(Glc-)Gal, which is the hexasaccharide proposed to constitute the repeating unit of 35624 EPS and which would also be fully consistent with the genetic content of the eps624 cluster (see above).

The 600 MHz 1H-NMR spectraof very mild (as for permethylation) acid-treated exopolysaccharide sample were recorded in D2O at 297 K and 338 K, respectively. Since the latter condition led to better resolved signals, the ensuing 13C NMR as well as COSY, TOCSY, HSQC-TOCSY, HMBC and ROESY spectra were recorded at 338 K throughout. The proton spectrum(Fig 6A) revealed inter alia six signals of equal intensity attributable to anomeric protons, which gave HSQC-correlationsto connectedcarbons in the range of 97–104 ppm. The absence of anomeric carbon signals shifted to lower-field than 104 ppm and of any signals of non-anomeric carbons at a field lower than 82 ppm confirmedthat furanose forms were not present [36]. Five of the anomeric signals (B1-F1) had small homonuclear nuclear coupling constants, while residue A displayed a larger coupling constant J1,2 (~ 8 Hz). The assignment of the α-anomeric configurationfor pyranose residues B-F was confirmedby the values of the heteronuclear coupling constant JC-1,H-1 (observedin an HMBC experiment) that were found in a range of 170–176 Hz. The coupling constant JC-1,H-1 for residue A was consistent with the β-anomeric configuration(167.9 Hz). In the high-fieldsegment of the 1H-NMR spectruma methyl group signal being characteristic of a 6-deoxy-pyranose was found at 1.22 ppm.

In the low-field sectionof the spectrum,two additional, non-anomeric and spin-coupled proton signals were seen (C4, C5). The signal observedat 4.62 ppm (C5) revealed an HMBC connectivityto a carbon signal at 175.3 ppm, thus indicating the presence of a pyranosyluronic acid (C).

doi:10.1371/journal.pone.0162983.g006

The presence of a hexasaccharide repeat unit was confirmedby analysis of the 150 MHz 13C NMR spectrumwhich contained four signals of anomeric carbons in the region of 100– 104 ppm (A-D), and a slightly high-fieldshifted signal of double intensity at 97.5 ppm (E,F) (Fig 6B). Pyranose ring carbon signals were present in the range between65.9 and 81.5 ppm, while methylene carbons (identifiedvia an APT experiment) were shown as two broadened singlets of double intensity at 61.6 and 60.9 ppm, respectively. Additionally, a carbonyl signal originating from the uronic acid moiety was observedat 175.3 ppm, as well as the carbon signal of the methyl group of the 6-deoxy sugar at 16.4 ppm.

Proton and carbon signals were then assigned using COSY, TOCSY, HSQC, HSQC-TOCSY (Fig 7A and 7B), HMBC and ROESY experiments (Table 2). Glycosylation shifts were seen for units A-E, whereas residue F occurredas an unsubstituted sugar. Glycosylation sites were identifiedat position 4 of residues A-E. Residue E was found to be additionally substituted at carbon 2, based on an HMBC correlation from H-1 of F to C-2 of E. Furthermore, both anomeric protons of residues E and F gave inter-residue ROEs in support of a close spatial proximity. The observedsubstitution pattern was thus in full agreement with the results of the permethylation analysis.

The absolute configuration of constituent sugars was derived from the analysis of the trimethylsilylated L-cysteine methyl ester [37] and the results of the monosaccharidecomposition analysis, this led to the assignment of a D-gluco configuration for residues A and F and a Dgalacto configurationfor residues C, D and E, respectively.

Key assignments were then obtained and confirmedfrom the tetrasaccharidefragment os211 generated by acid treatment (0.25 M TFA, for 3 h at 80°C) of the EPS followed by borohydride reduction and chromatographic purification.While data of the reduced hexitol component could not be extracted due to significant signal overlap, the proton signals of three pyranose units A, B, C could be fully assigned based on COSY and TOCSY experiments.This

allowed to identify the configuration of the constituent sugars based on homonuclear JH,H coupling constants as well as 1H chemical shift information (Table 2). Residue A was assigned as a β-glucopyranosyl residue as seen from the J2,3 and J3,4 coupling constants being consistent with

a trans-diaxial arrangement of the respective protons. Signals arising from residues B and C were close to those observedfor the related units in the polysaccharide chain, whereas H-3 and H-4 signals of unit A were shifted to higher field, indicating that hydrolysis had occurredat the residue preceding unit A. The small values of the coupling constants J1,2, J3,4 and J4,5 in conjunction with the large values seen for J2,3, identifiedresidue C as α-galacturonic acid, whereas residue B corresponds to an α-anomeric 6-deoxy-sugar. The coupling constants for the lowfield shifted H-5 and H-3 protons of residue B revealedsmall values for J4,5, J3,4 and J2,3, respectively, which are compatible with the configuration of a 6-deoxy-taloseor a 6-deoxy-gulose.

The absolute configuration of the 6-deoxy-sugar was eventually based on the molar rotation

values calculated for the polysaccharide. Published optical rotation values of methyl glycosides were used for the calculation [38]. For a hexasaccharide repeat unit containing an α- and β-Dglucopyranosyl unit, two α-D-galactopyranosylunits, one α-D-galactopyranosyluronic residue

and a 6-deoxy-α-talopyranose unit, the calculated molar rotation [MD] gives 1221.5 for 6-deoxy-α-D-taloseas constituent sugar and 844.3 for the presence of 6-deoxy-α-L-talose, which would correspond to optical rotation values of +126 and +87, respectively. The mea-

sured optical rotation value [α]D20 +84.7 (c 0.96, H2O) was in full agreement with the presence of a 6d-L-talopyranose. This assignment was furthersupported by characteristic chemical shift differences observedfor anomeric carbons when engaged in linkages betweenL,D- and D,Dconfiguredsugars [39, 40]. The presence of rhamnose, fucoseand 6-deoxy-quinovose was excluded based on the results of the HPLC monosaccharideanalysis, while the 13C NMR chemical shifts observedfor residue B were not compatible with the presence of a 6-deoxy-gulose, 6-deoxy-altrose and 6-deoxy-allose[39]. The NMR data of the 6-deoxy-pyranose showed a diagnostichigh-fieldshifted 13C NMR signal as seen for C-3 of talose1 and were also in good agreement with published NMR features of a disaccharide fragment β-D-Glcp-(1!4)-α-L6dTalp occurringin Burkholderia caribensis strain MWAP71 [41]. The identity of the deoxysugar was eventually proven using a synthetic sample of L-6dTal.

HMBC data (two experiments with values of JCH-coupling constants of 5 and 11 Hz, respectively, were performed)confirmedthe assignment of spin systems with characteristic intraresidue correlations betweenthe anomeric protons of residues C and F to carbon 3, and anomeric

doi:10.1371/journal.pone.0162983.g007

1. Table 2. 1H and 13C NMR chemical shifts (δ, ppm) of the exopolysaccharide (recorded at 338 K) and the tetrasaccharide os211 (recorded at 300 K) from B. longum 35624.

1. doi:10.1371/journal.pone.0162983.t002

protons of residues A, B and E to carbon 5, respectively. In addition the following inter-residue connections were observed:H-1 of A and C-4 of B, H-1 of B and C-4 of C, H-4 of E and C-1 of D, H-1 of F and C-2 of E.

Based on the combined evidenceof sugar analysis, methylation data and NMR-data, the 35624 EPS structure is illustrated in Fig 8.

Bifidobacteriacomprise a significant proportion of the gut microbiota, in particularin infants, and many strains are used as probiotics. Strain-specificbeneficialeffects are well established and the genetic or structuraldifferences that associate with such bifidobacterialstrains are beginningto emerge. In this report, we describea 35624-specificgene cluster, responsible for the production of a cell surface-associatedexopolysaccharide (EPS). The 35624 EPS consists of a branched hexasaccharide repeating unit with two galactoses,two glucoses,galacturonic acid and the infrequent sugar 6-deoxytalose.

Bifidobacterialcell surface-associatedpolysaccharides have previously been proposed to (i) mediate some of their health benefits,(ii) to aid in their tolerance to the harsh conditions within the gut, and (iii) to influence composition of the gut microbiome by being used as a growth substrate by other microbes [21, 27, 42, 43]. In general, bacterialEPS consists of repeating mono- or oligosaccharidesubunits connectedby varying glycosidic linkages, which are structurallydiverse. As previously reported, and as indicated in our genome comparisons, the identifiedgenes associated with the EPS biosynthetic machineryare highly diverse among the analyzed bifidobacteria,likely contributing to strain-specifictraits due to the expected

doi:10.1371/journal.pone.0162983.g008

structuraland therefore functionaldiversity of such EPS molecules[19]. Of note, pathogenassociated EPSs have long been known to be critical in host–microbeinteractions, where they facilitate adherence and colonization within the human host, with additional immunomodulatory effects [19, 44, 45]. Regarding the role of EPS in counteracting pathogens, a role of scavenger and trapping system to recruit and present pathogens to host cell receptors has been proposed [46]. With respect to gastrointestinal infections,the EPS produced by B. longum BCRC 1464 has been shown to possess antimicrobial activity against pathogens and immunemodulating activity, which causes release of the anti-inflammatory cytokine IL-10 [47]. Furthermore, EPS of B. animalis subsp. lactis has been shown to act as an adherent surface trapping pathogens on the mucus layer [46] and exopolysaccharide capsule production by a B. breve strain was also shown to be linked to the evasion of adaptive B-cell responses [21].

Previously, the 35624 strain has been describedas B. longum subsp. infantis according to the most up to date information available at that time for classifying bifidobacteriasubspecies. However, following the full sequencing of the bacterialgenome and the use of its core genome for phylogenetic taxon assignment, it is now apparent that this bacteriumshould be reclassified

as B. longum subsp. longum. Clearly, the reclassification of the strain does not affect the strainspecificbeneficialeffects previously describedfor 35624. However, very few probiotic strains that are commercially available have been characterizedto the same extent in terms of mechanism and genome identification as the 35624 strain and it is likely that without such in-depth analysis many probiotic products have been incorrectly classifiedand labeled. Lack of clarity on the key mechanism of action associated with other probiotic products is compounded by the fact that many contain multiple strains with unknown interactions.

Polysaccharides from B. bifidum, B. breve, B. infantis and B. longum strains typically contain glucose and galactose,with rhamnose occasionallybeing described[48, 49]. The presence of 6-deoxy-L-talosein bifidobacterialpolysaccharides has only been previously observedonce in a B. adolescentis strain [50]. However, the repeating unit describedherein for the 35624 strain has never been previously describedin bifidobacteria.The occurrenceof 6-deoxy-L-talosehas been previously describedin unrelated microbial species, which include Agrobacterium rubi, Rhizobium loti, Treponema pectinovorum, Burkholderia pseudomallei, Mycobacterium intracellulare, Mycobacterium smegmatis, Aeromonas hydrophila, Pseudoalteromonas flavipulchra, Hafnia alvei, Actinobacillus actinomycetemcomitans, and Proteus strains [51–62]. Many of these organisms are Gram-negative bacteria and the presence of 6-deoxy-talosein their LPS and in the EPS from the Gram-positive B. longum 35624 suggests that specificmonosaccharides can be shared in the surface associated polysaccharides of widely divergent bacteria.In most cases, 6-deoxy-L-talosehas been reported to be O-acetylated at position 2 and 4, respectively.

In conclusion, we have identifieda specificgenetic locus associated with the biosynthesis of a EPS in the human commensal B. longum 35624 strain. This EPS possesses a chemical structure that has not been previously describedfor other bifidobacterialstrains. Future studies will assess if the 35624 EPS has any immunoregulatory effects.

Bifidobacteriawere routinely cultured in either de Man Rogosa and Sharpe medium (MRS; Oxoid Ltd., Basingstoke, Hampshire, United Kingdom) supplemented with 0.05% cysteineHCl or reinforced clostridialmedium (RCM; Oxoid Ltd.). Bifidobacterialcultures were incubated at 37°C under anaerobic conditions in a Don Whitley anaerobic Chamber. Modified Man Rogosa and Sharpe agar plates, containing 3% glucose,were used to grow 35624 for EPS extraction and purification.

Chromosomal DNA from bifidobacteriawas isolated as previously described[63]. The genome sequence of 35624 was sequencedusing a Roche 454 FLX Titanium instrument by the commercial sequencing serviceproviders Agencourt Bioscience(Beverly, MA) and Eurofins MWG Operon (Germany) and then assembled, after which remaining gaps were closed using Sanger Sequencingof PCR products. Sequencereads were initially assembled using Phred [64, 65], Phrap (P. Green, University of Washington; http://www.phrap.org/), RepeatMasker (AFA. Smit, R. Hubley, & P.Green; www.repeatmasker.org/) and the Staden software package [66]. Final assembly of the 35624 genome was verifiedusing Newbler v 2.3 (http://454.com/ products/analysis-software/index.asp). The accession number for the 35624 genome sequence is CP013673 (NCBI—http://www.ncbi.nlm.nih.gov/).

Prediction of putative open reading frames (ORFs) was performedusing PRODIGAL prediction software (http://prodigal.ornl.gov/)and supported by BLASTX [67] alignments. Results of Prodigal/BLASTXwere combined manually and a preliminary identification of ORFs was performed on the basis of BLASTP analysis against a non-redundant protein database provided by the National Centre for Biotechnology(http://www.ncbi.nlm.nih.gov/). Using the ORF finding outputs and associated BLASTP results, Artemis [68] was employed for visualisation and manual editing in order to verify, and if necessary, redefine the start of everypredicted coding region, or to remove or add coding regions. The assignment of protein function to predicted coding regions was performedmanually. In addition, the individual members of the revised gene/protein data set were searched against the protein family (Pfam) [69] and COG [70] databases. RibosomalRNA (rRNA) and transfer RNA (tRNA) genes were detected using RNAMMER (http://www.cbs.dtu.dk/services/RNAmmer/) and tRNA-scanSE (http://lowelab.ucsc. edu/tRNAscan-SE/), respectively.

The computation of a phylogenetic supertreewas performedbased on the alignment of a set of orthologous proteins definedby the pan-genome computation. Each protein family was alignedusing CLUSTAL_W v1.83 [71]. Phylogenetic trees were computed using the maximum-likelihoodin PhyML v3.0 [72] and concatenated; the resulting consensus tree was computed using the Consense module from the Phylip package v3.69 using the majority rule method (http://evolution.genetics.washington.edu/phylip.html).

The 35624 strain was cultured on MRS agar plates at 37°C under anaerobic conditions for 60 h. Cells were harvested using cell scrapers and resuspended for 2 h with constant rotary agitation in PBS containing RibonucleaseA (1 μg/ml) and DeoxyribonucleaseI (5 μg/ml Sigma, St. Louis, MO, USA). Whole cells were removed by centrifugationfor 30 min at 4°C and 20,000 x g (SorvalRC6 plus) and the supernatant was filtered through 0.45 μm syringe filters. The EPS-containing solution was poured into 3 volumes of chilled ethanol, gently stirred for 30 seconds and incubated at 4°C for 2 h. The precipitated EPS located at the surface of the ethanol was removed with a spatula and resuspended in H20 under constant, yet gentle shaking for 2 h. The solution was dialysed against H20 in a dialysis tubing (14 kDa, Sigma) at 4°C and consequently applied 2 times on SPE C18 columns (Bakerbond)as indicated by the manufacturer using a HyperSep-96

vacuum manifold (Thermo Scientific).The flow-throughfraction was collectedand filtered through 0.45μm syringe filters. The solution was aseptically filledin glass vials and freeze-dried(Virtis Genesis). Dried EPS was stored at -80°C until furtheranalysis.

TM

Size exclusion chromatography was performedwith a PL aquagel-OH MIXED-H 8 μm column (300 x 7.5 mm, Agilent, Waldbronn, Germany) at flow rate of 0.5 mL/min at room temperature with 25 mM ammonium acetate (pH 8.5) and a refractive index detector. Dextran standards (Sigma) with nominal masses (Mw) of 25, 150 and 1185 kDa were used for comparison.

Anion exchange chromatography was performedwith EconoPac-Q cartridge(Bio-Rad, Vienna, Austria). The EPS was applied in 50 mm ammonium acetate and eluted with a NaCl gradient. Fractions were analyzed for carbohydrate by the orcinol-sulfuricacid method.

EPS was hydrolyzed with 4 M trifluoro acetic acid at 100°C for 4 h. The monosaccharides were derivatizedwith anthranilic acid and analyzed by reversed phase HPLC using a 5 μm Kinetex C18 core-shell column (Phenomenex, Torrens, CA) and an acetonitrile gradient in either 0.2% 1-butylamin, 0.5% phosphoric acid, 1% tetrahydrofuran [73] or in 0.3% formic acid buffered to pH 3.0 with ammonia in order to allow subsequent mass spectrometricverification of peaks [74]. High-performanceliquid chromatography of monosaccharides as 1-phenyl-3-methyl-5-pyrazolone (PMP) derivatives was performedon a Hypersil ODS column as described[75]. Gas chromatographic analysis of alditol acetates was performedwith a 60 m OPTIMA1 1 MS Accent column with 0.25 mm inner diameter and 0.25 μm film thickness (Macherey-Nagel, Düren, Germany) with mass spectrometricdetectionon a GC System 7820A with coupled to a MSD5975 (both Agilent, Waldbronn, Germany) [76].

For determination of the D/L configuration of Gal and Glc monosaccharides were reacted with L-cysteine methyl ester [37]. The dried sample was then derivatizedwith N-methyl-N-trimethylsilyltrifluoroacetamidecontaining 1% trimethylchlorosilane. The sugars were immediately analyzed on an Optima 1MS Accent analytical column (Macherey-Nagel, Germany, 60 m×0.25 mm i.d., 0.25 μm film thickness, 100% dimethylpolysiloxane stationary phase) and detectedwith a 7200 GC-QTOFMS system (Agilent, Waldbronn, Germany) [77]. Notably, compounds were chemically ionized using methane to give molecularions such that peaks arising from hexoses, deoxyhexoses and uronic acids could be discriminated.The dominating ion for hexoses was the penta-trimethylsilylderivative associated with an m/z of 658.2931.

Linkage analysis by gas chromatography with electron impact-mass spectrometry(GLC-MS) was conducted after hydrolysis of EPS under the mildest conditions that provided solubility in dimethyl sulfoxide, i.e. 50 mM trifluoroaceticacid at 80°C for 4 h. Permethylation was achieved with sodium hydroxide / methyl iodide followed by hydrolysis, reduction with sodium borodeuteride and acetylation. The resulting partially methylated alditol acetates were separated on a 60 m OPTIMA1 1 MS Accent column with 0.25 mm inner diameter and 0.25 μm film thickness (Macherey-Nagel, Düren, Germany). Retention time standards were obtained by derivatizing glucose or galactose with sub-stoichiometric amounts of methyl iodide, which worked well for tetra- and tri-methyl ethers.

EPS was hydrolyzed with 0.25 M TFA at 80° for various times with 4 h giving the best yield of fragments of suitable size. The samples were purifiedfor analysis with porous graphitic carbon (PGC) cartridges(25 mg, Hypersep Hypercarb, Thermo Scientific,Waltham, MA) equilibritated with 0.3% formic acid buffered to pH 3.0 with ammonia and eluted with 50% acetonitrile in this buffer. The eluate was dried, redissolved and separated by PGC HPLC on a 100 x 0.32 mm column (Hypercarb, Thermo Scientific)with a 30 min gradient from 5 to 20% (apart from sample application at 1% and column cleaning up to 50%) acetonitrile in the above formic acid/formate buffer [78]. Detectionof compounds was accomplished by electrospray ionization-mass spectrometry(ESI-MS/MS) on a Bruker Maxis G4 Q-TOF mass spectrometeroperated in the data dependent acquisition mode. Selectedoligosaccharidesobtained by preparative HPLC on a 100 x 3 mm hypercarb column (Thermo Scientific)applying the conditions used with the analytical column. Fractions containing relevant oligosaccharideswere pooled,dried and permethylated as describedabove, however, with perdeutero-methyl iodide to allow discrimination betweenmethyl groups and uronic acids. The permethylated oligosaccharides were analyzed as [M+Na]+ ions by matrix assisted-time of flight-mass spectrometryusing

2,5-dihydroxybenzoicacid as matrix and an Autoflex MALDI-TOF/TOF-MS (Bruker, Bremen, Germany) in the positive reflectronor in LIFT mode.

Preparative isolation of fractionos211 was done by hydrophilic interaction HPLC on a TSKgel Amide-80 column (Tosoh Bioscience,Griesheim, Germany) as recommendedby the supplier.

NMR spectraof the polysaccharide and tetrasaccharidesample were obtained for solutions in 99.9% D2O (0.6 mL) at 338 K on a Bruker Avance III 600 instrument (600.2 MHz for 1H, 150.9 MHz for 13C) equipped with a BBFO broad-band inverse probe head and z-gradients using standard Bruker NMR software TopSpin 3.0. 1H spectrawere referenced using DSS as standard (δ = 0); 13C spectrawere referenced using 1,4-dioxane as external standard (δ = 67.40). In general, sweep widths of 5000–6000 Hz for 1H and 32000–36000 Hz for 13C were used. 1H,

1. 1H-COSY experiment were measured using the pulse program cosygpqf. 1H, 13C-HSQC spectra were obtained using the pulse program hsqcedetgpwith 1024 x 64 k data points and 600
2. 2-ulose proceededsmoothly and afforded the talo-product.Deprotection via hydrolysis of the thioglycosideand cleavage of the 3,4-O-acetonide afforded the efficient production of 6-deoxyL-talose [74].

is a trademark of Alimentary Health Ltd. We would like to thank Katharina Wozny for help with monosaccharide analysis.

TM