Arabinoxylan-based substrate preferences and predicted metabolic properties of Bifidobacterium longum subspecies as a basis to design differential media

interesting candidates to develop novel probiotic and synbiotic combinations with AX-based prebiotics. In this work we have deepen into the metabolic capabilities of bifidobacteria related to AX and AXOS metabolization through a combination of in silico an in vitro tools. Both approaches revealed that Bifidobacterium longum and, particularly, B. longum subsp. longum , appears as the better equipped to metabolize complex AX substrates, although other related subspecies such as B. longum subsp. infantis , also hold some machinery related to AXOS metabolization. This correlates to the growth profiles exhibited by representative strains of both subspecies in AX or AXOS enriched media. Based on these results, we formulated a differential carbohydrate free medium (CFM) supplemented with a combination of AX and AXOS that enabled to recover a wide diversity of Bifidobacterium species from complex fecal samples, while allowing easy discrimination of AX metabolising strains by the appearance of a precipitation halo. This new media represent an appealing alternative to isolate novel probiotic bifidobacteria, rapidly discriminating their capacity to metabolize structurally complex AX-derived prebiotics. This can be convenient to assist formulation of novel functional foods and supplements, including bifidobacterial species with capacity to metabolize AX-derived prebiotic ingredients.


Introduction
The increased consumer awareness and demand of foods with health beneficial properties, is fostering the research on novel probiotic strains, microorganisms which when administered in adequate amounts confer a health benefit (Hill et al., 2014), and prebiotics, non-digestible food ingredients that are selectively utilized by host microorganisms conferring a health benefit (Gibson et al., 2017).Prebiotics traditionally refer to non-digestible carbohydrates, and include a wide and rapidly growing diversity of oligo-and polysaccharides that are resistant to intestinal digestion and absorption (Ferreira-Lazarte et al., 2020).Among these, arabinoxylo-oligosaccharides (AXOS) derived from the hydrolysis of arabinoxylan (AX), which constitute the major hemicellulosic component in the cell wall of some cereal plants, have received increasing attention as emergent prebiotics (Mathew et al., 2018;Sabater et al., 2021).
The complex chemical structure of AX and AXOS is not degraded by human digestive enzymes, thus they reach the distal digestive system intact (Broekaert et al., 2011;Rivière et al., 2014;van den Abbeele et al., 2018).AX consist of a linear backbone of 1,500 to 15,000 β(1-4) Dxylopyranoside units, which can occur unsubstituted, monosubstituted with α-L-arabinofuranoside residues positioned on C-(O)-2 or C-(O)-3, or disubstituted with single α-L-arabinofuranoside units at C-(O)-2 and C-(O)-3 (Broekaert et al., 2011;Rivière et al., 2014).Xylose residues can be substituted to a lesser extent with glucuronic acid and its 4-O-methyl derivative, or with short oligomers like L-arabinose, D-xylose, D-galactose, D-glucose or uronic acids.Arabinose residues can be esterified with phenolic acids, such as ferulic acid and to a lesser extent p-coumaric acid or dehydrodiferulic acid (Broekaert et al., 2011;Rivière et al., 2014).The presence of these modifications mainly depends on the vegetable source and greatly influences the biological properties of AX and AXOS (Aachary & Prapulla, 2011;Sabater et al., 2021).On the other hand, AXOS or xylo-oligosaccharides (XOS) may be obtained by enzymatic action with β-endoxylanases produced by various microorganisms, and are characterized by their lower average degree of polymerization and their average degree of arabinose substitution.Therefore, AXOS are more soluble and readily available for fermentation by the gut microbiota than AX (Broekaert et al., 2011;Rivière et al., 2014;Suriano et al., 2017).
AX and AXOS are a source of energy for certain saccharolytic bacteria residing in the colon, including Bacteroides, Bifidobacterium, Clostridium, Lactobacillus and Eubacterium species.Accordingly, the intake of AXOS has been associated to an increase in several commensal and potentially beneficial species including members of the Lachnospiraceae and Oscillospiraceae families, and of the Bifidobacterium genus, that is the most strongly promoted genus following consumption of AXOS (Suriano et al., 2017).Bifidobacterium also includes some of the species most commonly used as probiotics in functional foods and nutraceuticals (Ivashkin et al., 2021;Russell et al., 2011;Sarkar and Mandal, 2016).Remarkably, some bifidobacteria can cross-feed with other gut commensals such as Eubacterium and Bacteroides species, as a mean to cooperatively degrade xylan what results in production of beneficial metabolites by third species (Fernandez-Julia et al., 2022;Liu et al., 2020;Rivière et al., 2014).Among bifidobacteria, Bifidobacterium longum is the species with a larger number of fully sequenced genomes available, the majority belonging to the subspecies B. longum subsp.longum and only a few representing B. longum subsp.infantis.Genomic data and in vitro growth tests have explored the diversity of the metabolic capabilities of these microorganisms with a focus on those traits involved in the colonization of the gastro-intestinal tract and their adaptation to different groups of human hosts based on their dietary patterns (Arboleya et al., 2018;Blanco et al., 2020;Odamaki et al., 2018;Díaz et al., 2021).Overall, B. longum subsp.infantis exhibits a better adaptation to the intestinal environment of breastfed infants as it is able to metabolise human milk oligosaccharides (HMO) (Sela et al., 2008).On the other hand, metabolization of the most complex carbohydrates that can be present in the adult diet, including AX and AXOS, require a complex arsenal of enzymes such as glycoside hydrolases (GH), which split the glycosidic bonds into poly and oligosaccharides, releasing shorter metabolisable products.Accordingly, the array of GH involved in metabolization of complex vegetable fibres is considerably larger in the genomes of B. longum subsp.longum (Blanco et al., 2020).Besides, available evidence suggests that, among bifidobacteria, the enzymatic activities involved in the metabolism of arabinans, AX and arabinogalactan may be restricted to some B. longum subsp.longum strains.This fact may explain the better adaptation of this microorganism to the intestinal tract of adults following a fiber-rich diet, as compared to other bifidobacterial species (Blanco et al., 2020;Komeno et al., 2019).On the other hand, low molecular weight AXOS can be fermented by a broader range of Bifidobacterium species, such as B. adolescentis, B. breve and B. longum (Pastell et al., 2009;Rivière et al., 2014).
In this context, the main objective of this work was to deepen into the differential metabolism of AX and AXOS by bifidobacteria and to take advantage of the variability in this metabolic trait to design a differential culture medium for the rapid discrimination of AX-metabolizing Bifidobacterium isolates, with a focus on B. longum subspecies as representatives of one of the most widely used probiotics.For this purpose, a combination of culture-dependent techniques and in silico methods, were used.

Strains and growth conditions
Four strains of B. longum subsp.longum (NCIMB8809, and three IPLA collection strains recovered from colonic biopsia or contents from healthy adult volunteers, 1CCM4, 2BCM3 and 7BCM1) and four strains of B. longum subsp.infantis (DSM20090, DSM20218, CECT4551 and LMG18902) were used.These bifidobacterial strains were identified through Internal Transcribed Spacer (ITS)-bifidobacterial profiling by following a methodology previously described (Milani et al., 2014).Strains were routinely grown in MRS (Biokar Diagnostics, Beauvois, France) supplemented with 0.25% L-cysteine (Sigma-Chemical Co., St. Louis, MO, USA) at 37 • C in an anaerobic chamber MG500 (Don Whitley Scientific, Yorkshire, UK) under an 80% N 2 , 10% CO 2 and 10% H 2 atmosphere.Overnight cultures (18 h) were used to prepare the bacterial inoculum for all experiments, which were obtained after collecting cells by centrifugation and suspending them in the same volume of the medium without added carbon source as described below.

Growth curves in liquid medium with different carbohydrates
Growth curves were performed in a carbohydrate-free medium (CFM) which contained 10 g/L proteose peptone N • 3 (Difco, BD, Biosciences, San Diego, CA), 10 g/L beef extract (Difco), 5 g/L yeast extract (BD, Bacto, NJ, USA), 1 g/L polysorbate 80 (Sigma-Aldrich, Merck, St. Louis, MO, United States), 2 g/L ammonium citrate, 5 g/L sodium acetate (Scharlau, Barcelona, Spain), 0.05 g/L manganese sulphate (Sigma-Aldrich, Merck), 2 g/L dipotassium phosphate (Merck), 0.1 g/L magnesium sulphate (Sigma-Aldrich, Merck) and 2.5 g/L L-cysteine-HCl (Sigma-Aldrich, Merck).A final 1% (w/v) concentration of one of the following different sugars was incorporated to the medium prior to bacterial inoculation: D(+)-xylose (Sigma-Aldrich, Merck, St. Louis, MO, United States), xylan (Megazyme, Wiclow, Ireland), arabinan (Megazyme), L-arabinose (BD, Bacto, NJ, USA), AX (Megazime, Wiclow, Ireland), AXOS enriched powder (kindly provided by Cargill, Minnetonka, Minnesota, U.S. USA) or glucose (Sigma-Aldrich, Merck).Freshly prepared media, supplemented with one of the carbon sources previously listed, were distributed in round-bottom 96-microwell plates.Plates were inoculated with freshly grown overnight cultures of the different bifidobacterial strains, prepared as previously described, at 1% (v/v), corresponding to a final inoculum of 8.11 ± 0.29 and 7.75 ± 0.62 log CFU/mL for B. longum subsp.longum and B. longum subsp.infantis strains, respectively.Growth curves for each strain/carbohydrate combination were conducted on independent biological triplicates.Fermentations were carried out at 37 • C in the anaerobic chamber MG500 under the same atmosphere previously described.Using an EPOCH instrument (Agilent-Biotek, Santa Clara, U.S.A.) installed within the anaerobic chamber, the absorbance was measured at 600 nm every hour during 24 h, in order to monitor the growth of the bifidobacterial strains.Growth rates were estimated by fitting the growth data to the equation Nt = N0 × e µt , in which Nt and N0 are the cell densities at time t and time zero in the exponential growth phase, respectively.
Determination of viable cells at early stationary growth phase was performed on independent biological triplicate growth curves performed on 5 ml of CFM supplemented with 1% lactose, AX or AXOS and incubated at 37 • C under the above indicated conditions.

Analysis of carbohydrates consumption in liquid cultures
In order to determine the pattern of carbohydrates consumption in the presence of complex AX and/or AXOS mixtures, the eight model strains were inoculated at 1% in CFM, CFM with 1% AX (w/v), CFM with 1% AXOS (w/v) and CFM with 1% (w/v) AX and 1% AXOS (w/v).These cultures were grown in the anaerobic chamber at 37 • C, and samples were collected at 0, 6, 10, 24 and 48 h.Then, samples were centrifuged at 4000 rpm and the supernatants were purified with Carrez salts (Moreno et al., 1999) and analyzed by GD-FID following the methodology previously described by Sabater, Corzo, Olano & Montilla (2018), with an oven initial temperature of 120 • C, increased at a rate of 3 • C/ min to 380 • C and held for 10 min.
Independent biological duplicates were performed for each culture.The characterization of AX and AXOS was based on the determination of the low molecular weight carbohydrates (LMWC, from mono-to heptasaccharides) content, and the monomeric composition after hydrolysis with trifluoroacetic acid by GC-FID (Sabater, Corzo, Olano, & Montilla, 2018).

Growth in solid medium
Among the sugars used in the growth curves, AX and AXOS were selected to formulate a solid differential medium to aid the rapid discrimination of AX-metabolizing capabilities in the two B. longum strains subspecies.Different concentrations of these substrates (0.5%, 1% and 2% w/v) were incorporated to CFM agar plates, and the capability of the 8 Bifidobacterium strains to grow on each media was evaluated.For this purpose, actively growing overnight cultures in MRSc of each strain were washed in CFM without added carbon source, appropriately diluted, and spread on CFM agar plates supplemented with different concentrations of AX or AXOS.Seeded plates were grown in an anaerobic chamber at 37 • C for 48 h prior to evaluation of colony and precipitation halo formation.

Growth recovery and discrimination of Bifidobacterium longum subsp. longum strains from human fecal microbiota samples using CFM-AX-AXOS
Since the different subspecies under study, B. longum subsp.longum and B. longum subsp.infantis, originated colonies with different visual appearance, a novel CFM medium combining 1% (w/v) of both AX and AXOS (CFM-AX-AXOS) was evaluated as a differential media to support growth of both subspecies, while facilitating the discrimination of colonies belonging to the two subspecies by visual inspection.In order to verify, as a proof-of-principle, whether the formulated CFM-AX-AXOS media could aid to identify and discriminate AX-metabolizing B. longum subsp.longum strains and non AX-metabolizing B. longum subsp.infantis strains when present in a complex fecal microbiota sample, representative strains of both subspecies were spiked in a human fecal sample.For this purpose, a preserved fecal sample previously analyzed by means of 16S rRNA sequencing and which had been demonstrated not to harbor any bifidobacterial assigned reads (sample HD11 from Hevia et al., 2016), was used.This sample was supplemented with 10 8 cells of either B. longum subsp.infantis (CECT4551), B. longum subsp.longum (NCIMB8809), or with a combination of both strains.The artificially Bifidobacterium enriched fecal samples were vortexhomogenized in the anaerobic cabinet and serial dilutions were spread on CFM-AX-AXOS plates, which were further incubated anaerobically for at least 48 h.
The applicability of CFM-AX-AXOS media to recover and discriminate B. longum subsp.infantis and B. longum subsp.longum colonies from fecal samples was further validated on freshly collected fecal samples.For this purpose, fecal samples from 8 healthy lactating infants (with ages ranging from 3 months to 2 years), which had not taken any antibiotics in the previous 6 months, were collected.Participants were recruited under the project PID2019-104546RB-100, and the study was approved by the Regional Ethics Committee for Clinical Research of Principality of Asturias (Ref. 343/19).These samples were homogenized in phosphate buffered saline (PBS) and serial dilutions were spread on CFM-AX-AXOS supplemented with lithium mupirocin solution (MUP, Merck, Germany: 50 mg/L) (CFM-AX-AXOSm) to increase its selectivity towards bifidobacteria (Margolles and Ruiz, 2021).Plates were incubated in an anaerobic chamber under the aforementioned conditions.
In order to confirm the species and subspecies assignation of the colonies recovered from CFM-AX-AXOS media, a random selection of colonies (at least 10 from each fecal sample) were recovered and identified through partial 16S rRNA sequencing.For this purpose, colonies were subjected to small-scale total DNA extraction with chloroform (Ruiz-Barba et al., 2005) and 16S rRNA region was amplified and sequenced by Sanger technology as previously described by using primers 27F and 1492R (Frank et al., 2008).Further confirmation of the assignation to B. longum subsp.longum and B. longum subsp.infantis subspecies, was performed through PCR detection of the gene encoding the urease (subunit C), as a genetic marker commonly associated to B. longum subsp.infantis (Schimmel et al., 2021).For this purpose the oligonucleotides ureC-Fwd (5 ′ -GCCGGATTGAAGATCCACG-3 ′ ) and ureC-Rev (5 ′ -CCGCGATGGTGTCATTG-3 ′ ), designed to amplify an internal fragment of the urease subunit C gene in B. longum subsp.infantis, were used.

Genome sequencing of Bifidobacterium longum strains
Commercial kit DNeasy Blood & Tissue from QIAGEN was used for DNA extraction (Qiagen GmbH, Germany) of B. longum subsp.longum strains (1CCM4, 2BCM3 and 7BCM1) and B. longum subsp.infantis strains (DSM20090, DSM20218 and LMG18902).DNA extraction protocol was followed according to the manufacturerś instructions with the following modifications.A prior enzymatic lysis step was conducted by suspending the cells in a lysis buffer composed of 20 mM Tris-HCl (GE Healthcare, Illinois, USA) (pH 8), 2 mM sodium EDTA (AlfaAesar, Massachusetts, USA), 1.2% triton X-100 (Sigma-Aldrich, Merck), 20 mg/ mL lysozyme (Sigma-Aldrich, Merck) and 25 U/mL mutanolysin (Sigma-Aldrich, Merck), and incubating the suspensions for 30 min at 37 • C. Afterwards, lysed suspension was used for DNA purification by following the manufacturer's instructions.

16S rRNA and bifidobacterial ITS profiling of infant fecal samples
DNA from the infant fecal samples collected for Bifidobacterium strains isolation, was extracted by using the DNeasy PowerSoil Pro Kit (Qiagen GmbH, Germany) according to manufacturer's instructions.Then, 16S rRNA V3 regions were amplified using the primer's pair Probio_Uni and /Probio_Rev, as previously described (Milani et al., 2013) and samples were submitted to 2 × 250 bp paired-end sequencing in an Illumina MiSeq System instrument (Illumina) at GenProbio S.R.L. (Parma, Italy).Sequence reads were quality filtered by the Illumina software and then trimmed, and filtered sequences were processed with a custom script based on the QIIME software suite (Caporaso et al., 2010).On the other hand, the bifidobacterial community composition of the fecal samples was analyzed through targeted amplicon sequencing of ITS regions, which were sequenced and analyzed according to previously described procedures (Milani et al., 2014).Raw sequences generated are deposited in the SRA of the NCBI (https://www.ncbi.nlm.nih.gov/sra, accessed on 9 March 2023) under BioProject code PRJNA851776.

Bioinformatic analysis
To gain a better understanding of metabolic capabilities of Bifidobacterium species associated to AX and AXOS consumption, several bioinformatic analyses were carried out.In this sense, a comprehensive dataset (n = 707) comprising reference genomes and genome assemblies from Bifidobacterium species that have the Qualified Presumption of Safety (QPS) status according to the European Safety Authority (EFSA) was retrieved from the European Nucleotide Archive (ENA) (Supplementary material Table S1) These reference sequences from a total 12 representative Bifidobacterium species were mapped against the Carbohydrate-Active en-ZYmes Database (CAZy, https://www.cazy.org/lastaccessed: February 11, 2022) using "run_dbcan" software developed by Zhang et al. (2018).This computational pipeline integrates HMMER software for biosequence analysis using profile hidden Markov models, allowing functional domain annotation of bacterial glycosidases.To ensure the quality of the data generated, only glycosidase domains showing coverage values higher than 0.95 were chosen.Bifidobacterium species were grouped according to the glycosidase domains found in their genomes through complete linkage hierarchical clustering.In this method, all pairwise dissimilarities between the elements in each cluster (i.e., glycosidase profiles of Bifidobacterium species) are calculated using the basic function "hclust" from R v.3.6.2 programming environment.Then, the largest dissimilarity value is chosen as the distance between clusters.Therefore, this method produces more compact clusters.Then, heatmaps illustrating the presence and absence of functional domains were generated.
Once glycosidase profiles of bifidobacteria were elucidated at species level, a second bioinformatic analysis focusing on the genome sequences of the model B. longum subsp.longum strains (1CCM4, 2BCM3, 7BCM1 and NCIMB8809) and B. longum subsp.infantis strains (CECT4551, DSM20090, DSM20218 and LMG18902) utilized in the in vitro growth experiments previously described was performed.These sequences were analysed following the computational pipeline above described.

In silico study of glycosidase profiles of bifidobacteria related to carbon sources containing arabinose and xylose
As a first approximation, an in silico analysis was conducted to delineate the variation in the genetic attributes that could sustain the metabolism of AX, AXOS, XOS and glucan-oligosaccharides, such as cello-oligosaccharides (COS), commonly found in AX rich biomasses, across genomes representative of various species and subspecies of the genus Bifidobacterium.For this purpose, functional domain prediction and annotation of glycosidase activities capable of degrading the aforementioned carbohydrate structures was conducted in genomes from Bifidobacterium species that have the QPS status (n = 707), as well as other bifidobacterial species commonly present in the human gut (n = 320) (Fig. 1).
According to the results obtained, most bifidobacterial species that have the QPS status exhibit similarities in their carbohydrate metabolism capabilities; yet they show notable differences in the presence of specific glycosidase domains.Specific domains involved in the degradation of AX, AXOS, XOS, β-gluco-oligosaccharides, such as COS, and some of its constituent residues such as xylose or arabinose, were identified across the whole dataset, although their prevalence significantly differed across species.For instance, glycosidase domains acting on arabinose residues were less prevalent in B. animalis subsp.animalis and B. gallicum genomes; while those acting on xylose residues were less prevalent in the genomes from B. longum subsp.infantis, B. bifidum, B. adolescentis and B. pseudocatenulatum (Supplementary material Table S2).On the other hand, among the species recognized with the QPS status, B. longum subsp.infantis, B. breve and B. bifidum genomes displayed the lowest prevalence of functional domains involved in AXOS and XOS hydrolysis, suggesting a poor capacity to metabolise those substrates.Interestingly, a higher number of genomes from Bifidobacterium longum subsp.longum showed glycosidases of interest as compared to B. longum subsp.infantis (Supplementary material Table S2).These activities involve functional domains acting on: i) AXOS, COS and XOS (89.2 and 66.1% genomes from Bifidobacterium longum subsp.longum and B. longum subsp.infantis, respectively), ii) AXOS and XOS (69.4 and 13.6%, respectively), iii) XOS (69.8 and 20.0%, respectively).Therefore, notable differences in the metabolism of arabinose-and xylose-containing polysaccharides of Bifidobacterium longum subsp.longum and B. longum subsp.infantis have been determined.These metabolic activities could be of great interest to discriminate these two subspecies and to design target-specific prebiotics for B. longum subsp.longum.

AX and AXOS metabolism in a selection of B. longum strains: In silico and in vitro analyses
A complementary in silico analysis on the genomes of a selection of eight B. longum model strains available at the IPLA collection was performed so as to deepen in the genetic signatures that may determine a differential AX/AXOS metabolization by both subspecies.The glycosidase-encoding genes from these strains, enable to cluster them according to their subspecies classification (B.longum subsp.longum and B. longum subsp.infantis) suggesting characteristic patterns in the carbohydrate metabolism of both subspecies (Fig. 2).Whereas glycosidase domains capable of degrading arabinose monomers from AXOS as well as unspecific domains acting on AXOS, COS and XOS were present in the genome sequences of all strains, functional domains associated to AXOS and XOS metabolism, such as arabinofuranosidases and xylanases, were mainly found in B. longum subsp.longum strains (72.7-100.0% of total domains acting on AXOS and XOS were present in the genomes from B. longum subsp.longum strains; Fig. 2 and Supplementary material Table S4).Besides, a wider variety of these domains was found in genomes from strains 1CCM4 and NCIMB8809 (100.0 and 90.9% of total domains in 1CCM4 and NCIMB8809 genomes, respectively; Supplementary material Table S4) than in the strains 7BCM1 and 2BCM3 (72.7 and 81.8% of total domains in 7BCM1 and 2BCM3 genomes, respectively; Supplementary material Table S4) (Fig. 2).The characteristic enzyme patterns elucidated for B. longum subsp.longum at strain level (Fig. 2) are consistent with those obtained at species level (Fig. 1), and they suggest that B. longum subsp.longum might be better equipped to metabolise more structurally complex AX chains than B. longum subsp.infantis, in agreement with previous bioinformatic studies (Blanco et al., 2020).
In order to establish whether the differential metabolic traits in silico predicted correlate with actual AX/AXOS metabolization patterns, the ability of the eight B. longum subsp.longum and B. longum subsp.infantis strains to grow in the presence of a range of arabinose and xylose containing oligo-and polysaccharides was elucidated in vitro.Results demonstrated that most B. longum subsp.longum tested strains were able to grow on the majority of the substrates under study (xylose, arabinose, arabinan, AX and AXOS), initiating their exponential growth phase at around 6 h and reaching OD 600 values above 0.5 within 24 h (Fig. 3).Remarkably, all the strains tested grew in media supplemented with AXOS (exhibiting growth rates of 0.35 ± 0.03 and 0.31 ± 0.02 in B. longum subsp.longum and B. longum subsp.infantis strains, respectively); although only B. longum subsp.longum strains grew in AXsupplemented media, reflected by growth rates of 0.42 ± 0.06 and 0.02 ± 0.01 in B. longum subsp.longum and B. longum subsp.infantis strains, respectively.The differential ability of representative B. longum subsp.longum and B. longum subsp.infantis strains to grow on AX was confirmed through colony counting from independent cultures grown for 14 h in the presence of AX (Supplementary Fig. S1).Of note, no significant differences in the growth rates were observed in the carbon sources used as positive control (lactose: 0.51 ± 0.12 and 0.43 ± 0.09 in B. longum subsp.longum and B. longum subsp.infantis strains, respectively; maltose: 0.46 ± 0.07 and 0.39 ± 0.14 in B. longum subsp.longum and B. longum subsp.infantis strains, respectively), neither in the final viable counts obtained following growth in lactose (9.78 ± 0.07 and 9.49 ± 0.03 log CFU/mL in B. longum subsp.longum and B. longum subsp.infantis strains, respectively) (Supplementary Fig. S1).These results suggest that, in agreement with previous studies and with the in silico analysis previously described, the enzymatic machinery in B. longum subsp.longum strains enable the cells to metabolize more complex polysaccharides, such as AX, unlike other bifidobacterial species and subspecies (Komeno et al., 2019).

Carbohydrate analysis and metabolization by a collection of B. longum strains
Characterisation of AX and AXOS.
In an attempt to identify the specific AX and AXOS fractions that were metabolized differentially by the B. longum subsp.longum and B. longum susbp.infantis strains investigated GC-FID analysis was performed on both AX and AXOS powders, to determine LMWC and monomeric composition prior acid hydrolysis (Supplementary material Table S3), in order to have a more detailed information on their composition.The analysis of the AXOS powder allowed the detection and quantification of AXOS chains with a degree of polymerisation (DP) ranging from 2 to 7, which comprised 76% of the total carbohydrates present in the mixture, based on monomeric composition (Supplementary material Table S3).In addition, β-gluco-oligosaccharide fractions (mainly, COS) exhibiting polymeric series (DP from 2 to 5) were also found at lower concentrations than AXOS, comprising 22% of total carbohydrates, based on monomer composition.Concerning the AX powder, no oligosaccharide fraction could be determined in the GC-FID analysis.Monomeric composition analysis evidenced that AX was an arabinose-enriched substrate (arabinose:xylose ratio 1:2) while glucose was present in negligible amounts (arabinose + xylose:glucose ratio 59:1).On the contrary, AXOS powder contained a much more reduced content in arabinose (arabinose:xylose ratio 1:14), as compared to AX, Bifidobacterium species that may be present in human gut microbiota.The percentage (%) of reference genomes and genome assemblies from each species containing each functional domain is shown.Specifically, functional domains involving glycosidases capable of degrading arabinoxylan and arabinoxylo-oligosaccharides (AXOS), cello-oligosaccharides (COS) and xylo-oligosaccharides (XOS) are illustrated.Glycosidase functional domains showing coverage values higher than 0.95 were annotated.Codes corresponding to the CAZy family of each enzyme have been assigned.More information about these families can be found in the Carbohydrate-Active enZYmes Database (CAZy).Similarities between Bifidobacterium species are also illustrated in a dendrogram and expressed as distances between their characteristic glycosidase profiles calculated by the complete linkage method (vertical axis).Some of these species have the Qualified Presumption of Safety (QPS) status.also having considerable levels of glucose residues derived from COS (arabinose + xylose:glucose ratio 3.5:1).

AXOS and AX metabolism by Bifidobacterium longum
GC-FID analysis was also conducted to determine AXOS consumption on culture supernatants of the eight model strains following growth in the presence of either AXOS or a combination of AX and AXOS (Fig. 4).On the one hand, all B. longum subsp.longum strains showed a higher consumption of AXOS along the first 24 h of growth, as compared to the B. longum subsp.infantis strains, both in AXOS and AXOS + AX supplemented media.With regard to B. longum subsp.infantis, their AXOS carbohydrate consumption rates were generally lower than those exhibited by B. longum subsp.longum cultures.On the other hand, all B. longum subsp.longum strains showed a similar behavior after 48 h in AXOS-supplemented medium.However, most B. longum subsp.longum strains showed lower AXOS consumption when grown in the presence of AXOS + AX, than those observed in media supplemented only with AXOS.This fact may be attributed to its plausible utilization of AX, supported by a larger number of glycosidase domains involved in AX metabolization, and that could lead to the preferential consumption of enriched-arabinose polysaccharide fractions from AX instead of smaller AXOS molecules.This also suggests that monomeric composition could play a key role in AX and AXOS fermentability.It is also worth noting that AX metabolization by B. longum subsp.longum strains may release AXOS, leading to AXOS augmentation, as observed for strains 1CCM4 and 7BCM1 at 48 h (Fig. 4D), thus masking the actual AXOS consumption in cultures.On the contrary, AX structures might not be easily metabolized by most B. longum subsp.infantis strains, which predominantly metabolize low-DP AXOS rich in xylose.Taking into account that AX showed higher arabinose contents than AXOS (Supplementary material Table S3), its preferential metabolization by B. longum subsp.longum strains might be linked to a most prevalent representation of activities associated to arabinose metabolism in this subspecies as compared to B. longum subsp.infantis strains, as previously suggested by other works (Blanco et al., 2020) and by the in silico analysis previously described.

Formulation of solid medium to discriminate AX and AXOS metabolization capacity in bifidobacteria
In view of the differential ability of B. longum subsp.longum and B. longum subsp.infantis strains to grow and metabolise AX and AXOS, as supported by growth profiles, sugar consumption and in silico genome analyses, we subsequently aimed at formulating a medium to aid rapidly discriminating bifidobacterial isolates based on its AX metabolism.For this purpose, the growth of selected strains of both subspecies was evaluated on CFM solid media supplemented with AXOS and/or AX as the only carbon sources.This analysis confirmed that, while B. longum subsp.longum was capable to grow and form colonies in all the media tested, B. longum subsp.infantis did only grow with AXOS, not being able to form any colonies on AX (Supplementary material Table S5 and Supplementary Fig. S2).Besides, it was very characteristic the appearance of a halo of precipitation surrounding the colonies of B. longum  and LMG18902).The percentage (%) of genomes from each species containing each functional domain is shown.Specifically, functional domains involving glycosidases capable of degrading arabinoxylan and arabinoxylo-oligosaccharides (AXOS), cello-oligosaccharides (COS) and xylo-oligosaccharides (XOS) are illustrated.Glycosidase functional domains showing coverage values higher than 0.95 were annotated.Codes corresponding to the CAZy family of each enzyme have been assigned.More information about these families can be found in the Carbohydrate-Active enZYmes Database (CAZy).Similarities between Bifidobacterium longum strains are also illustrated in a dendogram and expressed as distances between their characteristic glycosidase profiles calculated by the complete linkage method (vertical axis).These strains have the Qualified Presumption of Safety (QPS) status.
subsp.longum strains when grown in CFM-AX (Supplementary Fig. S2), which was not observed for any of the strains in CFM-AXOS.In view of these differences, we subsequently aimed at formulating a CFM-based medium, which could support the growth of both subspecies simultaneously, while enabling to discriminate AX metabolizing colonies by macroscopic inspection.In this regards, a CFM media supplemented with a mixture of AX and AXOS, CFM-AX-AXOS, was demonstrated effective to recover and discriminate by visual inspection AXmetabolizing B. longum strains from complex samples (fecal samples artificially enriched in one or other subpsecies), as B. longum subsp.longum form larger colonies surrounded by a precipitation halo, whereas B. longum subsp.infantis form much smaller colonies with no precipitation halo (Fig. 5).The appearance of this precipitation may be attributed to an incomplete metabolization of AX by B. longum subsp.longum, and may correspond to non-degraded and/or partially degraded AX chains, and allows discriminating strains co-existing in the same sample and harboring differential AX metabolization capabilities.The subspecies identity of a random selection of colonies obtained from the sample containing both subspecies was confirmed through colony recovery and partial 16S rDNA sequencing (data not shown).
CFM-AX-AXOS media performance to isolate and discriminate bifidobacterial isolates, based on their AX-metabolizing capacity, was also evaluated on fecal samples freshly collected from lactating infants (n = 8).Macroscopic visualization of the colonies and isolates identity confirmation through partial 16S rRNA sequencing, verified that a number of bifidobacterial species different from B. longum can grow in CFM-AX-AXOSm.These include B. adolescentis, B. bifidum,B. breve,B. dentium,B. pseudocatenulatum and B. suillum (Fig. 6).Remarkably, 100% of the colonies displaying the characteristic precipitation halo were further identified as B. longum subsp.longum, based on 16S rRNA sequencing, while all of the non-halo formers were presumptively identified as belonging to B. longum subsp.infantis, based on 16S rRNA sequencing and on the detection of the urease gene, or to bifidobacterial species other than B. longum (Fig. 6 and Supplementary material Table S6).This corroborates CFM-AX-AXOS is a valuable media to rapidly detect AX-metabolising bifidobacterial species including B. longum subsp.longum, as well as the key role of this subspecies as primary degrader of complex AX.Besides, the pattern of bifidobacterial species recovered from the samples is highly consistent with the species detected through bifidobacterial ITS-sequencing in the corresponding samples (Supplementary Fig. S3).Thus these results confirm that CFM-AX-AXOS media enable to recover a wide range of bifidobacterial  species from complex samples, yet they enable an easy and rapid discrimination with the naked eye of AX-metabolizing B. longum subsp.longum isolates.

Discussion
Bifidobacterium longum is commonly present in the human gut and several strains have interest as probiotics, with different subspecies exhibiting adaptation to carbohydrates characteristic in either the infant or adult diet (Russell et al., 2011).Among the latest, AX is a major constituent of dietary fiber from cereal grains that is mainly metabolised in the large intestine by bifidobacteria, what generally involves an initial arabinose cleavage, with or without subsequent xylose backbone utilisation (Liu et al., 2020;Lynch et al., 2021;Rivière et al., 2014;Saito et al., 2020).Besides, AX and particularly AXOS, are considered emergent prebiotics that stimulate the growth of bifidobacteria and the production of butyrate in the human colon (Rivière et al., 2015;Sabater et al., 2021).Their capacity to impact the gut microbiota has been confirmed through several clinical trials that revealed that B. longum was significantly more abundant in the microbiota of healthy, elderly and overweight volunteers following AXOS supplementation (Chung et al., 2020;Benítez-Páez et al., 2019;Cloetens et al., 2010).Therefore, AXOS  may be considered a viable prebiotic option.However, only particular bifidobacterial strains belonging to a few species, including B. longum, exhibit the required capabilities to achieve its metabolization (Song et al., 2020).For these reasons, understanding AX and AXOS metabolization by beneficial commensals and probiotics species, such as those belonging to Bifidobacterium, is of great interest to develop novel functional foods including AX-metabolising probiotic species, AX-based prebiotics or synbiotics.
In silico analysis on genome sequences from Bifidobacterium species commonly found in the human gut, confirm a broader metabolic capacity to metabolise AX, AXOS and XOS of the species B. longum and, specifically, of B. longum subsp.longum as compared to B. longum subsp.infantis.These results agree with results from in vitro tests herein performed with a selection of strains and corroborate a better ability of the subspecies longum to metabolise the more structurally complex AX.Some studies have suggested that AX can be primarily metabolized by B. longum, which in this way may establish symbiotic relationships with other bifidobacteria like B. adolescentis that preferentially metabolise AXOS (Pastell et al., 2009).Indeed, some B. longum strains appear to prefer AX or AXOS with relatively high molecular weight and could grow well with high molecular weight AX as the sole carbon source (Song et al., 2020).This pattern has also been observed in our work, since only B. longum subsp.longum strains tested were capable to grow on AX, whereas B. longum subsp.infantis were not able to metabolise and grow on the high molecular weight AX (Fig. 3 and Supplementary Table S5).Interestingly, bioinformatic analysis performed in this work revealed that, across the analysed species, B. longum subsp.longum showed the highest number of glycosidases that may hydrolyse xylose backbones, although domains comprising arabinofuranosidases and xylosidases acting on AXOS have also been found in B. pseudocatenulatum genomes in agreement with other authors (Saito et al., 2020).
Differential glycosidase profiles may be useful to provide a theoretical background for metabolic cooperation between bifidobacteria; as Fig. 6.Representative CFM-AX-AXOSm plates (carbohydrate-free medium (CFM) supplemented with 1% arabinoxylan (AX), 1% arabinoxylooligosaccharides (AXOS) and mupirocin) spread with an infant fecal sample, showing appearance of colonies surrounded by a precipitation halo as well as colonies lacking the precipitation halo (A).Distribution of Bifidobacterium species among the isolates identified (n = 88) (including halo formers and non-halo formers) from infant fecal samples in CFM-AS-AXOSm.Species were identified through 16S rRNA sequencing and PCR detection of urease gene was used for identification of B. longum subsp.infantis (B).Distribution of Bifidobacterium species exhibiting a halo of precipitation surrounding the corresponding colonies, as determined through 16S rRNA sequencing (C).
well as to assist the rational design of novel probiotics, prebiotics and synbiotic combinations.Interestingly, some B. longum subsp.longum strains have been suggested to establish symbiotic relationships with intestinal commensals other than bifidobacteria, such as Eubacterium rectale or Bacteroides species, leading to beneficial short-chain fatty acids production and overall modification of the fecal microbiota profiles (Fernandez-Julia, Commane, van Sinderen, & Munoz-Munoz, 2022;Liu et al., 2020;Moens, Weckx, & De Vuyst, 2016;Rivière, Gagnon, Weckx, Roy, & De Vuyst, 2015;Rivière, Selak, Geirnaert, Van den Abbeele, & De Vuyst, 2018).Hence, bifidobacterial strains with capacity to primary degrade complex AX-based prebiotics have great interest due to its potential to beneficially impact the gut microbiota ecosystem upon AX consumption, promoting the metabolism of key commensal species other than bifidobacteria.Results from this work show that B. longum subsp.longum strains showing several arabinoxylan-degrading domains are able to efficiently metabolise AX chains from media containing AX as the only carbon source; whilst B. longum subsp.infantis strains exhibiting a much-reduced distribution of xylan-degrading domains on their genomes, were not able to grow on AX-containing media but on AXOScontaining media.This might be attributed to a preference of B. longum subsp.infantis for low molecular weight AXOS, or for oligoand polysaccharides with low arabinose contents.
Based on these observations, an AX and AXOS containing media, CFM-AX-AXOSm, was demonstrated effective to facilitate isolation and discrimination of potential probiotic bifidobacteria with AX metabolizing capacity.CFM-AX-AXOSm enabled selective isolation of bifidobacteria, and rapid discrimination of AX metabolizing bifidobacterial isolates as it led to a characteristic precipitation formation around the colonies when grown in solid AX-containing media.Its usefulness to efficiently isolate bifidobacteria and to discriminate AX utilizing bifidobacterial isolates was further validated on infant fecal microbiota samples.Precise identification of the specific genes/mechanisms responsible for the differential pattern of AX metabolization among B. longum isolates and the differential macroscopic colony aspect on AXcontaining media would require comparative genome analyses and gene-trait matching analysis on a larger number of bifidobacterial isolates.
In conclusion, in this work, we have deepened into AX and AXOS metabolism in bifidobacteria and, particularly, in the most prevalent B. longum subspecies, B. longum subsp.longum and B. longum subsp.infantis, through a combination of in silico and in vitro analyses.We have also taken advantage on the macroscopic appearance of AXmetabolising bifidobacteria on AX-containing media to formulate a media that enables to recover a wide range of bifidobacterial species from complex samples, while facilitating rapid macroscopic discrimination of AX-metabolising isolates.The AX and AXOS-supplemented culture medium herein described could represent an attractive alternative to isolate novel probiotics and to rapidly discriminate their capacity to metabolize structurally complex AX-derived prebiotics.This can be convenient to assist formulation of novel functional foods, including bifidobacterial species with capacity to metabolize AX-derived prebiotic ingredients.

Fig. 1 .
Fig. 1.Heatmap showing the presence of different glycosidases (indicated as grey and black cells) in the reference genomes and genome assemblies (n = 1027) of 14Bifidobacterium species that may be present in human gut microbiota.The percentage (%) of reference genomes and genome assemblies from each species containing each functional domain is shown.Specifically, functional domains involving glycosidases capable of degrading arabinoxylan and arabinoxylo-oligosaccharides (AXOS), cello-oligosaccharides (COS) and xylo-oligosaccharides (XOS) are illustrated.Glycosidase functional domains showing coverage values higher than 0.95 were annotated.Codes corresponding to the CAZy family of each enzyme have been assigned.More information about these families can be found in the Carbohydrate-Active enZYmes Database (CAZy).Similarities between Bifidobacterium species are also illustrated in a dendrogram and expressed as distances between their characteristic glycosidase profiles calculated by the complete linkage method (vertical axis).Some of these species have the Qualified Presumption of Safety (QPS) status.

Fig. 2 .
Fig. 2. Heatmap showing the presence of different glycosidases (indicated as black cells) in the genome sequences of 4 Bifidobacterium longum subsp.longum strains (1CCM4, 2BCM3, 7BCM1 and NCIMB8809) and 4 B. longum subsp.infantis strains (CECT4551, DSM20090, DSM20218and LMG18902).The percentage (%) of genomes from each species containing each functional domain is shown.Specifically, functional domains involving glycosidases capable of degrading arabinoxylan and arabinoxylo-oligosaccharides (AXOS), cello-oligosaccharides (COS) and xylo-oligosaccharides (XOS) are illustrated.Glycosidase functional domains showing coverage values higher than 0.95 were annotated.Codes corresponding to the CAZy family of each enzyme have been assigned.More information about these families can be found in the Carbohydrate-Active enZYmes Database (CAZy).Similarities between Bifidobacterium longum strains are also illustrated in a dendogram and expressed as distances between their characteristic glycosidase profiles calculated by the complete linkage method (vertical axis).These strains have the Qualified Presumption of Safety (QPS) status.
. Reordering of the final contigs compared with National Center for Biotechnology Information (NCBI) template genomes of B. longum subsp.longum and B. longum subsp.infantis were performed with Mauve v2.3.1.Average coverage ranged from 71 to 133 while the number of contigs ranged from 19 to 46.Raw sequences are deposited in Sequence Read Archive (SRA) of the NCBI (https://www.ncbi.nlm.nih.gov/sra,accessed on 9 March 2023) under BioProject code PRJNA851776, Bio-