Bacterial strains, media and culture conditions
Bacterial strains used in this study are listed in Table 1, which includes six previously examined strains that had been isolated from human faeces and three strains isolated from cows . All strains were stored at −80°C in de Man-Rogosa-Sharpe (MRS) broth (Difco BD, Ireland), supplemented with 25% (v/v) glycerol as a cryoprotectant. Lactobacillus strains were grown anaerobically on MRS agar plates at 37°C for two days. Growth tests (stress resistance, aerotolerance, etc.) were initiated by growing Lactobacillus strains anaerobically in MRS broth at 37°C overnight . When required, MRS medium  was modified by the omission of dextrose and the addition of 0.5% (w/v) raffinose. MRS and Raffinose-MRS were used as plating media for the isolation of L. ruminis from porcine and equine faecal matter.
Carbohydrate-free MRS (cfMRS)  with added bromocresol purple was used as a basal screening medium to study the ability of the potential Lactobacillus ruminis strains to utilise various carbohydrates. These carbohydrates were used as a selective method to isolate L. ruminis based on its carbohydrate fermentation profile . The carbohydrate-free MRS was supplemented with 0.5% (v/v) of cellobiose, Raftilose P95 (Beneo-Orafti, Belgium), mannitol or ribose for screening the porcine faecal isolates while the additional carbohydrates glucose, lactose, raffinose, Raftiline HP (Beneo-Orafti, Belgium) and sucrose were used in the screening of the equine faecal isolates. Mannitol and ribose were used as negative controls, i.e. carbohydrates that L. ruminis is unable to metabolise. All carbohydrates were provided by Sigma Aldrich, Ireland unless stated otherwise.
To characterise the swarming phenotype of L. ruminis isolates, MRS medium was modified in several ways: (i) addition of increasing concentration of agar from 0.5% up to 3%; (ii) addition of increasing concentrations of Tween 80 from 0.2% up to 1%; (iii) minimal MRS containing 0.5% (w/v) of four different carbohydrates – glucose, lactose, cellobiose and Raftilose P95.
Animals and diets
Faecal samples were collected from four Large White x Landrace cross weanlings and sows. The animals were housed in the pig production unit of Teagasc Moorepark, Fermoy, Ireland. The weanlings were 10–12 weeks old. Their diets consisted of barley, wheat, maize, full-fat soya, soya hi-pro, fat, amino acids, vitamins and minerals.
Faecal samples were also collected from six mature racehorses, which were housed in a stable in Co. Limerick, Ireland. The horses were fed on diets containing forage and a high-starch concentrate . All samples were collected in accordance with current Irish legislation on animal handling.
Simulated gastric juice
To simulate the gastric environment, a sterile electrolyte solution  containing NaCl 6.2 gL−1, KCl 2.2 gL−1, CaCl2 0.22 gL−1 and NaHCO3 1.2 gL−1 was supplemented with lysozyme and pepsin (Sigma-Aldrich, Wicklow, Ireland) to final concentrations of 0.01% and 0.3% (w/v), respectively. The pH of the solution was reduced to pH 2.0 using 1 M HCl. Five millilitre volumes of each overnight culture were centrifuged at 4,000 × g for 10 min. The cell pellets were then re-suspended in the simulated gastric juice (SGJ) and incubated for 24 hours. Viable counts were determined by plate culture after 0-hr, 3-hr and 24-hr incubation.
Carbohydrate fermentation profiling
Newly isolated porcine and equine L. ruminis strains were tested for their ability to utilise twenty-eight carbohydrates and compared to previously determined carbohydrate utilisation profiles for 9 other L. ruminis strains . Each carbohydrate was filter-sterilised into cfMRS at a concentration of 0.5% (w/v). A Synergy 2 plate reader (BioTek Instruments Inc., Vermont, US) with Gen5 software was used to measure absorbance at the beginning (0 hr) and a second reading was taken after 48 hr. The carbohydrates tested include cellulose, dextran, esculin, lichenan, lyxose, Raftiline HP (Beneo-Orafti, Belgium), Raftiline ST (Beneo-Orafti, Belgium), ribose, sialic acid, sialyllactose, soluble starch, trehalose, melibiose, raffinose, GOS, GOS inulin, lactose, lactulose, beta-glucotriose B, cellobiose, Beneo P95 (Beneo-Orafti, Belgium), Raftilose P95 (Beneo-Orafti, Belgium), Raftilose Synergy 1 (Beneo-Orafti, Belgium), fructose, galactose, glucose, maltose, mannose, sucrose, the sources and preparation of which were as previously described .
Reconstituted skimmed milk (RSM) was prepared as a 10% (w/v) solution and autoclaved at 121°C for 10 minutes. Strains were inoculated into the RSM at 1% (v/v) and incubated for 72 hours at 37°C. Following the incubation period the pH of each culture was recorded and any change was adjusted by the change value of the negative control to identify the net pH change.
For aerobic growth, strains were inoculated as a 1% (v/v) inoculum in 5 ml of MRS and grown overnight aerobically at 37°C. Optical density (OD) readings were recorded at 0 hr and 24 hr.
The API-ZYM kit (bio-Merieux, France) was used to characterise the enzyme activity in whole bacterial cells. The tests were carried out in duplicate following the manufacturer’s instructions. Beta galactosidase activity was assayed in duplicate using OPNG disks (Sigma Aldrich, Co. Wicklow, Ireland) as per the manufacturer’s instructions.
Bile salt resistance, low pH tolerance and EPS production
To assess the effect of increasing concentrations of porcine bile salts (Sigma Aldrich, Wicklow) and lowered pH on L. ruminis viability, modifications were made to MRS. For the bile salt assay MRS was supplemented with 0.25-5% (w/v) porcine bile salts. In the acid tolerance assay, the pH was reduced using acetic acid from pH 5.5 to 3.0 in step-wise pH 0.5 increments.
Exopolysaccharide production was analysed using modified MRS supplemented with 70% (v/v) of filter sterilised glucose, sucrose and lactose.
Rifampicin and chloramphenicol were chosen as exemplars of broad-spectrum antibiotics. Each antibiotic was tested using sterile disks (Sigma Aldrich, Wicklow, Ireland) on MRS agar plates supplemented with each test strain. The disks were saturated with rifampicin (0.1-1 μg/ml) and chloramphenicol (1-4 μg/ml). The test plates containing the disks were then grown at 37°C for 48 hr. A strain was considered resistant if no zone of clearing was present surrounding the antibiotic disk.
DNA extraction, 16S rRNA gene amplification and sequence analysis
DNA was extracted from bacterial isolates using the Sigma Genelute Bacterial genomic DNA kit (Arklow, Wicklow, Ireland). The primers used in this study are listed in Additional file 1. Universal primers 27 F and 1492R  were used to amplify the 16S rRNA gene in a 50 μl reaction mixture consisting of 45 μl Platinum High Fidelity Supermix (Invitrogen, USA), each primer at 25 μM, 20 ng of template DNA and water to make the reaction up to 50 μl. Amplification conditions for the PCR included an initial denaturation step of 94°C for 2 min, followed by 35 cycles of 94°C for 20 s, 52°C for 30 s, and 68°C for 2 min and a final extension step of 68°C for 10 min. PCR products were checked for size and purity on a 1% (w/v) agarose gel using gel electrophoresis. PCR products were purified with the QIAquick PCR purification kit (Qiagen, USA). DNA sequencing of the amplified 16S rRNA gene region was carried out by Beckmann Coulter Genomics (Takely, UK). Sequence alignments were performed using the ClustalW application in BioEdit . MEGA (version 5)  was used to construct trees by using the neighbour-joining algorithm and the Kimura two-parameter substitution model. Branch support was measured by 1,000 replicate bootstrap tests for each analysis.
Assessment of flagellum production and motility
L. ruminis cells were stained with a crystal violet-based flagellar stain (BD Diagnostics). The procedure was carried out as outlined by the manufacturer. Stained cells were then examined by light microscopy under oil immersion using 1,000X magnification, and images were captured using an Olympus DP50 camera attached to the microscope.
Multi-locus sequence typing
The nucleotide sequences of the following genes were used for MLST analysis: ftsQ, nrdB, parB, pheS, pstB and rpoA. Primer pairs for each locus were designed using BioEdit . An approximately 800 bp internal fragment of each gene was amplified, which allowed subsequent sequencing of an internal 600–760 bp fragment within each amplicon, using the primers specified in Additional file 1. Individual PCR products were sequenced (Beckman Coulter genomics, Takely, UK) and trimmed using Bioedit. Different allelic sequences, with at least one nucleotide difference per allele, were assigned arbitrary numbers. A combination of six alleles defined the allelic profile of each strain, and a unique allelic profile was designated with a sequence type (ST). Split decomposition analysis of the allelic profile data and individual alleles was performed using SplitsTree 4.8 . Concatenated sequences (4,103 bp) of the loci (ordered as ftsQ, nrdB, parB, pheS, pstB and rpoA) were generated using the Sequence type Analysis and Re-combinatorial Tests (START2) software . One thousand replicate neighbour-joining bootstrap trees were constructed using the Kimura 2-parameter method  in MEGA version 5  to determine phylogeny. The relatedness of the isolates was assessed using START2; related STs were clustered in groups or lineages using BURST analysis. START2 was also used to determine the ratio of non-synonymous to synonymous polymorphisms (dN/dS ratio) for each locus . Statistical comparisons were carried out using the maximum chi-square analysis application in the START2 package.
Genome sequencing and comparative genomics
Sequencing was carried out for strains, S23 (human isolate), DPC 6832 (equine isolate), DPC 6830 (porcine isolate) and ATCC 27780 (bovine isolate) using the Illumina HiSeq 2000 reversible dye terminator system with read lengths of 101 bp. The functional assignment of predicted genes was performed using Metagene  to predict open reading frames (ORFs) and BLASTP was used forannotation against the NCBI non-redundant protein database . Whole genome comparisons were made using the Artemis Comparison Tool (ACT) . BLAST ring image generator (BRIG)  was used to create an image of whole genome comparisons.
QuartetS, which uses a reciprocal-best BLAST approach followed by 2-stage clustering, was used to predict orthologs. A core genome of 907 genes was identified. ClustalW  was used to create PHYLIP files from the amino acids of each core gene, which were imported into Clann (version 3.2.3)  to create a bootstrapped supertree.
RNA isolation and RNA-seq
L. ruminis ATCC 27782 and DPC 6832 were cultured anaerobically at 37°C for 18 hours in 5 ml aliquots of MRS media (swimming cells) and also on MRS agar plates containing 0.5% (w/v) agar (swarming cells) and 2% (w/v) agar (stationary cells) for 48 hours. The broth cultures were centrifuged at 4°C to harvest the cells that were immediately resuspended in 10 ml of RNAprotect Bacteria Reagent (Qiagen, Germany). To each agar plate 10 ml of RNAprotect Bacteria Reagent was added and the cells gently harvested using sterile spreaders and removed from the plate using a wide-bore pipette tip into a fresh 50 mL Falcon tube. Each tube was centrifuged at 4,000 × g for 15 min at 4°C. Total RNA was isolated according to the protocol for Gram-positive bacteria outlined by the Roche High Pure Isolation kit (Roche, Indiana, USA), but with minor modifications. The lysozyme step was shortened and the concentration was increased to 100 mg/ml. Additionally, this step was also merged with a bead-beating step to ensure complete cell lysis; the cells were incubated for 60 min at 37°C while shaking at 1400 rpm in a 2 ml stock tube containing 0.1 mm zirconia beads in an Eppendorf Thermomixer. Contaminating DNA was removed with the Turbo DNA-free kit (Invitrogen, Dun Laoghaire, Ireland). The total RNA was ribo-depleted using the Gram-Positive Bacteria Ribo-Zero™ Magnetic Kit (Cambio Ltd., Cambridge, UK) and cleaned using the RNA Clean & Concentrator™-5 (Cambridge Biosciences, Cambridge, UK). The mRNA fragmentation, random-primed cDNA synthesis, adapter ligation, adapter-specific PCR amplification and pooling of the six tagged libraries into one pool were carried out by GATC Biotech (Konstanz, Germany).
Each sample was run on an Illumina HiSeq sequencer (GATC Biotech, Konstanz, Germany) to generate 101 bp length reads using paired-end sequencing. FastaQC was used to identify the quality of the RNA-seq reads from each treatment (www.bioinformatics.babraham.ac.uk). The Trimmomatic program was used to trim low-quality sections of reads [77,78]. Alignment of the reads to the complete genome of ATCC 27782 and the draft genome of DPC 6832 was carried out using Bowtie2 . HTSeq-count and DESeq were utilised to assess differential gene expression between stationary, swimming and swarming L. ruminis cells [80,81].
RT-PCR was used to confirm differential expression of selected genes. The SensiFAST™ SYBR® No-ROX One-Step Kit (Bioline, myBio, Ireland) was used to generate cDNA and RT-PCR was performed according to the manufacturer’s specifications. The amplification temperature for all reactions was 55°C. The expression data generated for each gene and condition (stationary, swimming and swarming) for L. ruminis strains ATCC 27782 and DPC 6832 were normalised using the housekeeping gene recA. Following normalisation, fold differences were calculated using the following formula: fold change = 2^(ΔΔCt). The standard deviation of the ΔCT was calculated from the standard deviations of the target and reference values using the formula: S.D. = (S1
2 + S2
2)^0.5. The resulting value was then added and subtracted to the ΔΔCT values to generate a range for the 2^(ΔΔCt) values.
This Whole Genome Shotgun sequences for L. ruminis strains, DPC 6832, S23, ATCC 27780 and DPC 6830 have been deposited at DDBJ/EMBL/GenBank under the accession numbers AWYA00000000, AWYB00000000, JHAJ00000000 and JHAB00000000, respectively. The versions described in this paper are versions AWYA01000000, AWYB01000000, JHAJ01000000 and JHAB01000000, respectively.