Whole-Transcriptome Shotgun Sequencing (RNA-seq) Screen Reveals Upregulation of Cellobiose and Motility Operons of Lactobacillus ruminis L5 during Growth on Tetrasaccharides Derived from Barley β-Glucan

Abstract

Lactobacillus ruminis is an inhabitant of human bowels and bovine rumens. None of 10 isolates (three from bovine rumen, seven from human feces) of L. ruminis that were tested could utilize barley β-glucan for growth. Seven of the strains of L. ruminis were, however, able to utilize tetrasaccharides (3-O-β-cellotriosyl-d-glucose [LDP4] or 4-O-β-laminaribiosyl-d-cellobiose [CDP4]) present in β-glucan hydrolysates for growth. The tetrasaccharides were generated by the use of lichenase or cellulase, respectively. To learn more about the utilization of tetrasaccharides by L. ruminis, whole-transcriptome shotgun sequencing (RNA-seq) was tested as a transcriptional screen to detect altered gene expression when an autochthonous human strain (L5) was grown in medium containing CDP4. RNA-seq results were confirmed and extended by reverse transcription-quantitative PCR assays of selected genes in two upregulated operons when cells were grown as batch cultures in medium containing either CDP4 or LDP4. The cellobiose utilization operon had increased transcription, particularly in early growth phase, whereas the chemotaxis/motility operon was upregulated in late growth phase. Phenotypic changes were seen in relation to upregulation of chemotaxis/flagellar operons: flagella were rarely seen by electron microscopy on glucose-grown cells but cells cultured in tetrasaccharide medium were commonly flagellated. Chemotactic movement toward tetrasaccharides was demonstrated in capillary cultures. L. ruminis utilized 3-O-β-cellotriosyl-d-glucose released by β-glucan hydrolysis due to bowel commensal Coprococcus sp., indicating that cross feeding of tetrasaccharide between bacteria could occur. Therefore, the RNA-seq screen and subsequent experiments had utility in revealing foraging attributes of gut commensal Lactobacillus ruminis.

INTRODUCTION

Phylogenetically diverse, self-regulating communities of bacteria whose primary sources of nutrients are polymeric plant substances from the diet of the host, host secretions, and enterocytes sloughed from the mucosal surface inhabit the large bowels of humans (1, 2). Lactic acid bacteria belonging to the genus Lactobacillus are commonly detected in human feces, but most of these species seem likely to be transients rather than autochthonous inhabitants of the bowel (3). Lactobacillus ruminis, however, has been detected consistently in the stool of individuals during sampling conducted over periods of at least 18 months. Moreover, unlike other Lactobacillus species, whose population sizes are variable and low (<0.01%), L. ruminis predictably forms between 0.01 and 1.0% of the bacterial community (4). It therefore conforms to the definition of an autochthonous member of the bowel community (3).

Lactobacilli are genetically endowed to rapidly utilize simple carbohydrates that they ferment with the production of mostly lactic acid as the end product (5). They are not able to degrade polymeric substances such as those in the human diet that escape digestion in the stomach and small bowel and which become available for bacterial use in the large bowel (6). The lactobacilli must therefore rely on a supply of simpler carbohydrates released by the hydrolytic activities of other bacteria. The trophic characteristics of L. ruminis are therefore of interest because knowledge about them could provide examples of the interactive matrix that underpins the bacterial community of the human bowel. Without knowledge of the food webs and food chains, the regulatory mechanisms that maintain homeostasis of the human bowel microbiota will never be elucidated.

Annotated genome sequences of bacteria reveal the biochemical potential of their cells. Transcriptomic studies are helpful, however, to show how bacterial cells respond, especially in terms of the expression of operons, to the presence of particular growth substrates in their environment, and to identify the function of open reading frames (ORFs) that currently lack a functional association. DNA microarrays have traditionally been used as screening assays in transcriptomic studies, but whole-transcriptome shotgun sequencing (RNA-seq) is a recently developed approach (7). RNA-seq refers to the use of high-throughput sequencing (HTS) technologies to sequence cDNA generated from mRNA in order to measure gene transcription in bacteria and hence obtain information about gene expression under different ecological conditions. RNA-seq has an advantage over microarray analysis in that transcriptional responses can be determined for a particular bacterial strain even though its genome has not been sequenced. A recent comparative study of the transcriptome of Lactobacillus plantarum showed that similar information was obtained by cDNA sequencing methodology and microarray (8).

L. ruminis has been cultured from bovine rumen contents and human feces, indicating that the species can inhabit environments rich in plant polymeric material. In a previously reported study (9), we observed that, while unable to grow in medium containing plant β-glucans (1:3,1:4-β-d-glucans), L. ruminis cultures were able to utilize tetrasaccharide in β-glucan hydrolysate. β-Glucans are components of plant cells, especially of the endosperm of cereals such as barley, rye, sorghum, rice, and wheat. In the case of humans, they are not digested in the stomach and small bowel and thus pass to the large bowel, where they are hydrolyzed by bacterial metabolism. Since β-glucans are not hydrolyzed by L. ruminis, but tetrasaccharides derived from them are growth substrates for these bacteria, we reasoned that investigation of L. ruminis strains could assist in modeling of trophic interactions in the human bowel.

We obtained new information about L. ruminis using a culture-based screen of L. ruminis strains of rumen and human origins by determining the spectrum of 3-O-β-cellotriosyl-d-glucose (LDP4) and 4-O-β-laminaribiosyl-d-cellobiose (CDP4) utilization by 10 strains of this species. Subsequently, our study focused on strain L5 because it was previously found to be always detectable in the feces of a human subject over 18 months at population levels of ∼1% of the total microbiota and could therefore be considered an autochthonous strain (4). Additionally, this human isolate was able to use both 3-O-β-cellotriosyl-d-glucose and 4-O-β-laminaribiosyl-d-cellobiose. Therefore, we focused on this strain in our further investigations because of our interest in human bowel ecology with respect to the differential utilization of substrates by bowel bacteria. RNA-seq was used as a quick screen to detect altered gene expression when L. ruminis strain L5 was grown in medium containing tetrasaccharide. After further gene expression studies and phenotypic measurements, we found that L. ruminis strain L5 was capable of actively seeking out and utilizing tetrasaccharide liberated by hydrolysis of β-glucans by other bowel bacteria. Our study therefore contributes novel information about trophic specialties of this gut commensal.

MATERIALS AND METHODS

Preparation of β-glucan hydrolysates and purified tetrasaccharides.

To obtain tetrasaccharides for growth experiments, Glucagel, a barley extract containing approximately 75% (wt/wt) 1:3,1:4-β-d-glucan (β-glucan; a gift from GraceLinc Ltd., Christchurch, New Zealand), was dissolved in distilled water and precipitated with 4 volumes of ethanol. After centrifugation (3,500 × g, 20 min, 5°C), the precipitated material was dissolved in distilled water and freeze-dried. Analysis by ¹³C nuclear magnetic resonance (NMR) spectroscopy showed a spectrum consistent with β-glucan (10) and did not show the presence of any low-molecular-weight malto-oligosaccharides that are a minor component of Glucagel.

Purified β-glucan (10 g) was dissolved in distilled water (500 ml, 90°C, 15 min) and digested with either lichenase from Bacillus subtilis (250 units, 50°C, 5 h; EC 3.2.1.73; Megazyme, Ireland) or cellulase from Trichoderma reesei (∼500 units, 50°C, 20 h; EC 3.2.1.4; Celluclast 1.5L, Novozymes). The reactions were stopped by boiling for 5 min, and insoluble material was removed by centrifugation (3,500 × g, 20 min, 20°C). The supernatants were freeze-dried. To prepare tetrasaccharides from the hydrolysates, enzyme digests were loaded individually onto a charcoal-Celite column (1:1 [wt/wt], 5 by 30 cm) and washed with water until carbohydrate could no longer be detected (phenol-sulfuric acid assay [11]). Bound oligosaccharides were then eluted successively from the column with 2 liters each of 10, 12.5, and 15% (vol/vol) ethanol and 4 liters each of 20 and 25% (vol/vol) ethanol. The fractions were analyzed by gel filtration high-pressure liquid chromatography (HPLC) on two Superdex peptide columns (300 by 10 mm; GE Healthcare) connected in series. Samples (100 μg) dissolved in distilled water (1 mg ml⁻¹) were eluted with water (0.5 ml min⁻¹), and the eluent was monitored by differential refractometry (Waters 2410 refractive index monitor). The columns were calibrated with a series of malto-oligosaccharides (DP 2 to 7) and glucose. Fractions corresponding to tetrasaccharides were pooled to give the fractions LDP4 (from lichenase hydrolysate) and CDP4 (from cellulose hydrolysate), reduced in volume by rotary evaporation, and freeze-dried. High-performance anion-exchange chromatography (HPAEC) (see below) and ¹H NMR spectroscopic analysis showed that LDP4 and CDP4 contained >90% G4G4G3G (3-O-β-cellotriosyl-d-glucose) and G3G4G4G (4-O-β-laminaribiosyl-d-cellobiose), respectively.

Utilization of oligosaccharides in β-glucan hydrolysates, measured by HPAEC.

To show the utilization of oligosaccharides by L. ruminis during growth, the relative proportions of individual oligosaccharides in cell-free culture supernatant (spent medium) compared with those in uninoculated medium were detected. Samples were analyzed by HPAEC on a CarboPac PA-100 (4- by 250-mm) column equilibrated in 150 mM NaOH, using a Dionex ICS 3000 chromatography system (Dionex Corp., Sunnyvale, CA, USA). Samples (1 μg) were injected onto the column in distilled water (0.1 mg ml⁻¹) and eluted with a linear gradient of sodium acetate (NaOAc) (0 to 500 mM) in 150 mM NaOH from 5 to 25 min after injection. The eluent was monitored by pulsed amperometric detection.

L. ruminis strains.

Three isolates from bovine rumen and seven from human feces were used in the study (Table 1). The aim was to compare the use by the various bacterial isolates of 3-O-β-cellotriosyl-d-glucose and 4-O-β-laminaribiosyl-d-cellobiose. The species identity of the 10 strains was established/confirmed by sequencing the v1-to-v4 region (nucleotide positions 1 to 800) of their 16S rRNA genes and by alignment of the sequences with those in the NCBI database. Purified DNA from pure cultures was used as the template in a PCR with the bacterial universal primers 8fAll (GRGTTYGATYMTGGCTCAG) and 1492r (TACGGCTACCTTGTTACGACTT). The PCR product was cleaned with a Qiagen PCR cleanup column (Qiagen, Hilden, Germany) and submitted to the Massey Genome Service Centre (Palmerston North, New Zealand) for sequencing from the 8fAll primer using Life Technologies dGTP BigDye Terminator v3.0 chemistry. Sequence data were checked manually prior to species-level identification using the NCBI nucleotide BLAST application. The 16S rRNA gene sequences of isolates L5, L21, L23, L36, and L38 were 98.83 to 99.64% identical to that of the L. ruminis type culture (ATCC 27780^T) across the first 1 kbp of the gene. Cultures of ATCC 25644, ATCC 27780^T, ATCC 27781, and ATCC 27782 were obtained from the American Type Culture Collection (Manassas, VA, USA). SPM0211 was kindly supplied by Nam-Joo Ha, Sahmyook University, Seoul, South Korea. Cultures were maintained using lactobacillus de Man-Rogosa-Sharpe (MRS) medium at 37°C under anaerobic conditions.

Table 1.

L. ruminis strains

L. ruminis strain	Origin	Reference or source
L5	Human feces	4
L21	Human feces	4
L23	Human feces	4
L36	Human feces	4
L38	Human feces	4
SPM0211	Human feces	Nam-Joo Ha
ATCC 25644	Human feces	American Type Culture Collection (Manassas, VA, USA)
ATCC 27780T	Bovine rumen	American Type Culture Collection (Manassas, VA, USA)
ATCC 27781	Bovine rumen	American Type Culture Collection (Manassas, VA, USA)
ATCC 27782	Bovine rumen	American Type Culture Collection (Manassas, VA, USA)

Screening L. ruminis strains for utilization of tetrasaccharides.

To determine whether the L. ruminis strains were able to use the tetrasaccharides for growth, cultures of the 10 strains were used to inoculate medium containing 0.2% (wt/vol) of the tetrasaccharide fraction prepared from β-glucan hydrolysates (LDP4, 3-O-β-cellotriosyl-d-glucose; CDP4, 4-O-β-laminaribiosyl-d-cellobiose). The tetrasaccharide substrate was added to MRS medium made from scratch (per liter: proteose peptone, 10.0 g; beef extract, 10.0 g; yeast extract, 5.0 g; Tween 80, 1.0 ml; ammonium citrate, 2.0 g; sodium acetate, 5.0 g; magnesium sulfate, 0.1 g; manganese sulfate, 0.05 g; dipotassium phosphate, 2.0 g) in which glucose was replaced with purified tetrasaccharide. Sterilization was by filtration (0.45 μm). The media were inoculated with 1% (vol/vol) bacterial culture (∼1 × 10⁹ bacteria per ml) that had been incubated anaerobically for 18 h at 37°C. The tetrasaccharide cultures were incubated under these same conditions, and optical density (A₆₀₀) readings were made after 24 h of incubation. Three biological replicates of each strain/tetrasaccharide culture were measured. MRS–no-carbohydrate controls were below an A₆₀₀ of 0.2 for all L. ruminis strains.

Growth of strain L5 in spent medium from Coprococcus sp. culture.

To show that tetrasaccharides liberated upon hydrolysis of β-glucans by another bowel inhabitant could be utilized by L. ruminis L5, Coprococcus sp. strain 801 (which had been isolated previously from the same human host as that for L5) was grown in MRS medium containing 0.5% (wt/vol) β-glucan for 6 h, anaerobically, at 37°C. The supernatant from this culture was filter sterilized and used as growth medium for strain L5 in comparison with previously unused β-glucan medium. The media, in triplicate, were inoculated with a 1% (vol/vol) 18-hour culture of L5 and were incubated for 5 h, anaerobically, at 37°C. The cultures were diluted in sterile water and plated on MRS agar plates to determine the CFU/ml. Supernatants from the cultures and uninoculated medium were examined by HPAEC to determine hydrolysis products from β-glucan due to coprococcal growth and utilization of oligosaccharides by strain L5.

RNA-seq screen of the transcriptome of L. ruminis L5 grown in CDP4.

A transcriptional screen of strain L5 when grown in CDP4 relative to glucose (both at 0.5% [wt/vol]) was carried out to discover differential responses to growth in media with different fermentable substrates. To accomplish this, RNA was extracted from cells grown to an A₆₀₀ of ∼1.0 in the presence of glucose or CDP4. rRNA was depleted with the Ambion MICROBEnrich kit (Life Technologies) according to the manufacturer's instructions. Enriched mRNA was then amplified using the Ambion MessageAmp II bacterial RNA amplification kit (Life Technologies) according to the manufacturer's instructions. Amplified RNA was sent to Macrogen Inc. (South Korea) where 75-base-paired-end sequencing using the Illumina GAIIx platform was performed. Paired-end sequences were mapped to publically available L. ruminis genomes using either maq, version 0.7.1 (12), or bowtie, version 0.12.7 (13). Reads were normalized using the RPKM approach (14). Gene expression when cells were grown in CDP4 medium was compared with that of glucose-grown cells.

Extended transcriptional analysis measured by reverse transcription-quantitative PCR (RT-qPCR).

To confirm and extend the transcriptional analysis of strain L5, RNA was extracted from cells harvested from CDP4-, LDP4-, and glucose-grown cultures as follows. Total RNA was extracted from cultures of strain L5 grown in MRS medium containing 0.5% (wt/vol) CDP4, LDP4, or glucose. Bacterial cells were harvested at two points during anaerobic growth at 37°C: when the culture had attained an A₆₀₀ of ∼0.1 and at an A₆₀₀ of ∼1.0. Cells were pelleted by brief centrifugation at 12,000 × g and washed with 750 microliters of RNAprotect reagent (Qiagen). The bacterial cells were disrupted by bead beating (5,000 rpm, twice for 40 s each) in TRIzol reagent (Invitrogen) and extracted with chloroform. RNA was precipitated with iso-propanol and dried after removal of iso-propanol by careful pipetting. The dried RNA was then dissolved in nuclease-free water. The RNA was further purified using the RNeasy minikit (Qiagen). RNA quality was assessed using a Bioanalyzer 2100 instrument (Agilent Technologies) and quantified with a NanoDrop ND-1000 spectrophotometer (Thermo Fisher). RNAs from 3 biological replicates of each culture (glucose, CDP4, and LDP4) were prepared.

Reverse transcription-quantitative PCR (RT-qPCR) was carried out using 384-well plates (4titude Ltd. Surrey, United Kingdom) with Optical adhesive film (Life Technologies, Foster City, CA, USA) and an ABI Viia-7 Fast system. Primers were designed using Primer 3 software (15) and were purchased from Invitrogen. Primer details are shown in Table 2. Primers were tested for equivalent efficiency using six-point, 10-fold-dilution standard curves utilizing purified L. ruminis (ATCC 25644) genomic DNA as the template. All reactions were carried out in a final volume of 10 microliters containing 1× Fast SYBR Green PCR master mix (Life Technologies) and 300 nM (each) primer. Template RNA was obtained as described above from glucose, LDP4, and CDP4 cultures. RNA (500 ng per sample) was converted to cDNA using the Invitrogen Superscript IV kit in a final reaction volume of 20 microliters. Two microliters of cDNA template was added to the master mix. The thermocycling profile consisted of an initial activation of the polymerase at 95°C for 30 s, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s. Fluorescence levels were measured after a 60°C annealing/extension step. In all cases, a melt curve was generated to analyze product specificity. Three biological replicates were tested, and reactions were carried out in duplicate. Statistical analyses, including tests for normalcy and Mann-Whitney nonparametric tests, were carried out using GraphPad Prism v. 5.0a (GraphPad Software Inc., CA).

Table 2.

Sequences of primers used for RT-qPCR

Primer name	Primer sequence	ORF targeta
10224_balH_F	CGGGTACGTCGAAAATGAGT	10224
10224_balH_R	AGGGAAAATCGGAAGCTGAT	10224
10225_celB_F	CAGCTATCCTTTGCGGTCTC	10225
10225_celB_R	CCAGTTCAAAAGGTCCGTGT	10225
10227_celC_F	CTTGAAGAAGCCAACCAAGC	10227
10227_celC_R	TGGCAAGATCGACAAACGTA	10227
10230_bglB2_F	ACGCGGTCGATTTTTACAAC	10230
10230_bglB2_R	AATCCGCCGTATTTATGCTG	10230
10409_che2_F	ACCGGAATTGCTGACGTTAC	10409
10409_che2_R	TTTCTTCAACGCTGTGCATC	10409
12063_flgC_F	GGCGTCAAGGTAACGAACAT	12063
12063_flgC_R	TTGCTTCAGTTGACGAAACG	12063
12047_FliR_F	GGCTTGCAATCATGGTCTTT	12047
12047_FliR_R	TCTTCACGTTCAACGACAGC	12047
12059_sec_F	AGCCGTACACGGTGAAAATC	12059
12059_sec_R	TCATGCTCATTCTGCTCACC	12059

^aL. ruminis ATCC 25644 genome HMPREF0542 locus tag number.

Genomic comparisons.

Genomic comparisons were made in the expectation of understanding differential utilization of tetrasaccharides by L. ruminis strains. Paired-end reads were mapped to the publically available L. ruminis genomes: ATCC 25644 (did not utilize tetrasaccharides), SPM0211 (did not utilize tetrasaccharides), and ATCC 27782 (utilized LDP4). Mapping was carried out using bowtie (version 0.12.7) following quality and length trimming (average score, ≥30; read length, ≥50) using SolexaQA (16). Reads that did not map to each genome were identified and extracted using bam2fastq (www.hudsonalpha.org/gsl/information/software/bam2fastq; Genomic Services Lab, HudsonAlpha) and were assembled into contigs using the Trinity RNA-seq de novo Assembly package (17). Assembled contigs were ranked by RPKM (the number of reads contributing to the transcript, normalized by transcript length and total reads per sample). Highly expressed contigs were searched against the NCBI protein database using BLASTX (18). To further compare the L5 genome with those of other L. ruminis strains, transcript sequences from tetrasaccharide- and glucose-grown cells were pooled and assembled into contigs using the Trinity Assembly package. Reference genomes and assembled contigs were compared using the online version of VISTA (19).

Microscopy.

Microscopy of the cells of strain L5 was carried out to determine whether upregulation of the chemotaxis/motility operon influenced motility. For light microscopy, a drop of tetrasaccharide- or glucose-grown cultures (A₆₀₀, ∼1.0) was placed on a glass slide and movement of bacterial cells was observed using a Zeiss Axiostar Plus microscope (Gottingen, Germany). The number of cells showing active movement was counted per 100 bacteria. For transmission electron microscopy (TEM), 10-microliter aliquots of tetrasaccharide- or glucose-grown cultures (A₆₀₀, ∼1.0) were placed onto plasma glow discharge-treated, carbon-coated copper mesh TEM grids for 60 s and then blotted with filter paper to leave a thin film of liquid on the carbon support. The bacterial cells were fixed by placing 10 microliters of 2.5% glutaraldehyde in 0.1 M sodium cacodylate onto the grids for 3 min and then blotted. The grids were washed three times using 10 microliters of double-distilled water per grid and blotted after each wash. Contrast was obtained by adding 10 microliters of 0.25% phosphotungstic acid, pH 6.8. This was applied to the grids and removed immediately by blotting. The grids were then dried and viewed using a Philips CM100 BioTWIN transmission electron microscope (Philips/FEI Corporation, Eindhoven, Netherlands). Images were captured with a MegaView III digital camera (Olympus Soft Imaging Solutions GmbH, Muenster, Germany) (20). The number of bacterial cells with a flagellum was counted per 100 cells.

Determination of doubling times.

In order to obtain kinetic information about growth in different tetrasaccharide media, the doubling times of L. ruminis strain L5 were determined by culture of the lactobacilli in MRS medium in which 2% glucose was replaced by either 0.5% (wt/vol) LDP4 fraction, 0.5% (wt/vol) CDP4 fraction, or 0.5% (wt/vol) glucose, at 37°C under anaerobic conditions. Single aliquots of triplicate cultures were removed at hourly intervals over an 8-hour incubation period, and the A₆₀₀ was determined. The A₆₀₀ values were converted to natural logarithms and plotted against sampling times. Doubling times were calculated by log(2)/slope of the linear part of the graph.

Chemotaxis.

We looked for evidence of chemotactic behavior by strain L5 to confirm that transcriptional upregulation of the chemotaxis/motility operon resulted in an appropriate phenotypic change. The methodology was based on the classic chemotaxis experiments of Freter and O'Brien (21). Strain L5 was cultured for 18 h in 3 ml of MRS medium in which 2% glucose was replaced by either 0.5% (wt/vol) LDP4, 0.5% (wt/vol) CDP4, or 0.5% (wt/vol) glucose. Glass capillary tubes were filled (about 60 microliters per capillary) with these same media. Combinations of cultures (LDP4, CDP4, and glucose as growth substrates) and attractants (LDP4, CDP4, and glucose in capillary tubes) were prepared in triplicate. Control capillary tubes contained MRS medium without carbohydrate. Capillary tubes were placed vertically in tubes containing 18-hour growth and incubated anaerobically at 37°C for 6 h. The outside of each capillary tube was rinsed with sterile water, and the contents were expelled into a sterile petri dish. Ten microliters of contents was used to prepare 10-fold dilutions to 1 × 10⁻⁶, and 10 microliters of each dilution was spread on the surface of sectored MRS agar plates. The plates were incubated anaerobically for 48 h at 37°C when colony counts were made and CFU per ml of capillary contents was calculated.

RESULTS

Screening tetrasaccharides as growth substrates for L. ruminis strains.

Previous work (9) had shown that some L. ruminis strains used tetrasaccharide fractions of β-glucan for growth. Therefore, we tested the 10 strains for growth in media containing tetrasaccharide fractions of lichenase and cellulase hydrolysates. Seven human isolates (L5, L21, L23, L36, L38, ATCC 25644, and SPM0211) and three bovine rumen isolates (ATCC 27780^T, ATCC 27781, and ATCC 27782) were used. Trophic variations were shown between strains grown in medium containing LDP4 and those grown in medium containing CDP4 (Fig. 1). Three human strains (L36, SPM0211, and ATCC 25644) did not grow to any extent in tetrasaccharide media. The rumen strains ATCC 27780^T, ATCC 27781, and ATCC 27782 grew only on LDP4. Both LDP4 and CDP4 supported the growth of human strains L5, L21, L23, and L38.

Fig 1.

Growth of L. ruminis strains in medium containing tetrasaccharide fractions from β-glucan hydrolysates. n = 3 for each strain. MRS–no-carbohydrate controls had an A₆₀₀ of <0.2 (dashed line).

Strain L5 used 3-O-β-cellotriosyl-d-glucose (LDP4) released by Coprococcus sp. from β-glucan.

Growth of Coprococcus sp. strain 801 resulted in hydrolysis of β-glucan with release of tri- and tetrasaccharide (3-O-β-cellotriosyl-d-glucose [LDP4]) (Fig. 2a). These oligosaccharides were used by L5 when cultured in the spent medium. In contrast, growth of L5 in β-glucan medium was negligible (Fig. 2b). This result showed that cross feeding of tetrasaccharide between β-glucan-hydrolyzing bacteria and L. ruminis in the human large bowel was a feasible proposition.

Fig 2.

Hydrolysis of β-glucan by Coprococcus sp. strain 801 and utilization of oligosaccharides by L. ruminis strain L5. (a) Line A, β-glucan medium inoculated with L. ruminis L5; line B, spent β-glucan medium after Coprococcus sp. strain 801 growth and then inoculation with L. ruminis L5; line C, spent β-glucan medium after Coprococcus sp. strain 801 growth. Oligosaccharides G4G3G and G4G4G3G (3-O-β-cellotriosyl-d-glucose) were identified by cochromatography with lichenase hydrolysate of β-glucan. PAD, pulsed amperometric detection. (b) Growth of L. ruminis L5 in β-glucan medium or spent medium (coprococcal culture). Mean values and SEMs for three biological replicates are shown.

RNA-seq screen of strain L5 grown in CDP4 or glucose medium.

The intention was to employ RNA-seq to identify genes or operons that were differentially expressed under glucose and tetrasaccharide growth conditions and then use RT-qPCR to confirm transcriptional changes with biological replicates, consistent with traditional microarray studies. Paired-end reads were initially mapped to the ATCC 25644 L. ruminis (human isolate) reference genome (accession number NZ_ACGS00000000). Between 29.3 × 10⁶ (glucose) and 48.9 × 10⁶ (CDP4) reads mapped to the reference genome, providing genome coverage of 80.6% (glucose) to 85.7% (CDP4). The average sequence coverage for each open reading frame (ORF) was 1,098× (glucose) to 1,833× (CDP4). Gene expression was normalized by calculating RPKM values for each ORF, and ORFs were ranked by RPKM value to ascertain which genes were highly and weakly expressed under each of the growth conditions. An island of chemotaxis and flagellum biosynthesis genes was highly expressed in CDP4-grown cells compared with growth in glucose (Fig. 3a). Additionally, a cluster of genes involved in cellobiose transport and utilization was highly expressed when cells were grown in CDP4 (Fig. 3b). Several transport gene clusters, predominantly ABC superfamily members, were more highly expressed during growth in CDP4 (see Table S1 in the supplemental material). Other genes showing high expression during growth on CDP4 included several carbohydrate utilization genes, transcription regulators, and signal transduction proteins. Genes downregulated during growth in CDP4 compared with growth in glucose were predominantly hypothetical proteins, but some genes encoding transporters, membrane proteins, and transcription regulators were also downregulated (see Table S1).

Fig 3.

Expression of genes involved in flagellum biosynthesis/chemotaxis (a) and cellobiose transport/utilization (b). Expression levels, shown as RPKM normalized values, were derived from RNA-seq data obtained from cells grown in medium containing CDP4 or glucose. PTS, phosphotransferase.

Extended transcriptional analysis measured by reverse transcription-quantitative PCR.

Confirmation and extension of the RNA-seq results to both CDP4 and LDP4 were carried out using RT-qPCR on selected genes. L5 cultures were grown to an A₆₀₀ of ∼0.1 or ∼1.0 to test for “early” and “late” transcriptional effects. Based on RNA-seq data, four genes involved in cellobiose utilization (balH, celB, celC, and bglB2) and four genes involved in chemotaxis and flagellum biosynthesis (che2, flgC, fliR, and sec) were chosen for RT-qPCR analysis. Both RNA-seq and RT-qPCR showed the same trends for the expression of all gene targets at comparable growth points (A₆₀₀ of ∼1.0) (Table 3).

Table 3.

Comparison of gene expression levels measured by RNA-seq and RT-qPCR and normalized to glucose-grown cells (A₆₀₀ of 1.0)

Gene	Function	Gene expression level measured bya:
		RNA-seq	RT-qPCRb
balH	β-Glucosidase	8.41	17.46 (2.47)
celB	Cellobiose transporter	3.46	2.14 (0.28)
celC	Cellobiose transporter	13.16	11.71 (1.33)
bglB2	β-Glucosidase	4.40	5.26 (1.41)
che2	Chemotaxis	11.54	4.67 (1.22)
flgC	Flagellar basal body rod	16.33	2.75 (0.26)
fliR	Flagellar biosynthesis	3.45	2.94 (0.55)
sec	Flagellar secretory pathway	26.82	15.54 (5.62)

^aFold increase in expression when grown in CDP4 compared to growth in glucose.

^bValues are shown as mean (SEM). RNAs from 3 biological replicates of each culture (glucose, CDP4, and LDP4) were prepared.

Genes involved in cellobiose utilization were more highly expressed during early growth than during late growth when both CDP4 and LDP4 were used as the substrate. The opposite was observed when cells were grown in glucose (Fig. 4a to c). Genes involved in chemotaxis and flagellum biosynthesis were more highly expressed during late-phase growth for all substrates (Fig. 4d to f). However, in late-phase growth the glucose-grown cells tended to express chemotaxis and flagellum biosynthesis at lower levels than did tetrasaccharide-grown cells. When using growth on glucose as a normalizer (threshold cycle [ΔΔC_T] values), cellobiose utilization genes were more highly expressed during early growth than late growth on CDP4 or LDP4 (Fig. 4g and h). Overall, the results confirmed the RNA-seq screen using CDP4 and extended the observations to LDP4-grown cells.

Fig 4.

(a to f) Normalized expression (2^−ΔCT) of genes involved in cellobiose transport/utilization (a to c) and flagellum biosynthesis/chemotaxis (d to f). Expression during growth in the presence of CDP4 (a and d), LDP4 (b and e), and glucose (c and f). The normalizer target for generating ΔC_T values was 16S rRNA. (g and h) Normalized relative expression (2^−ΔΔCT) of genes involved in cellobiose transport/utilization. (g) CDP4 compared with glucose; (h) LDP4 compared with glucose. Mean values and SEMs for three biological replicates analyzed in duplicate are shown. Statistical analysis was by two-way analysis of variance. OD, optical density.

Genomic comparisons.

Genome coverage (how many ORFs in the reference genome were accounted for by transcripts from the L5 RNA-seq data) was about 85% for all reference strains. A comparison of reference genomes using the online wgVISTA program showed considerable differences between each strain. In comparison with LDP4-utilizing strain ATCC 27782, strain ATCC 25644 lacked 218 ORFs and SPM0211 lacked 246 ORFs. There was an overlap of 124 ORFs missing from both ATCC 25644 and SPM0211. If no reads from the L5 transcriptome sequence mapped to an ORF, it was assumed that this ORF was absent in the mapping strain. According to this criterion, 232 ORFs present in ATCC 27782 were absent in L5. As ATCC 27782 and L5 are capable of utilizing LDP4, it is possible that ORFs present in these two strains but absent in ATCC 25644 and SPM0211 may be required for growth on LDP4. The 19 ORFs conforming to this criterion are shown in Table 4. The majority of these ORFs were annotated as hypothetical proteins, so it is not possible to associate them with a role in the utilization of LDP4. A single ABC transporter ATP-binding protein was present in ATCC 27782 and L5 but absent in ATCC 25644 and SPM0211.

Table 4.

ORFs present in ATCC 27782 and L5 but absent in ATCC 25644 and SPM0211

Locus taga	Product
LRC_02310	ABC transporter ATP-binding protein
LRC_05290	Hypothetical protein
LRC_05310	SNF2 family DNA/RNA helicase
LRC_06350	Hypothetical protein
LRC_07160	Hypothetical protein
LRC_07210	Putative transposase
LRC_07220	Hypothetical protein
LRC_08810	Type III restriction-modification system
LRC_08820	Type III restriction-modification system enzyme
LRC_08830	Hypothetical protein
LRC_08840	SNF2 family helicase
LRC_09920	CRISPR-associated endoribonuclease Cas2
LRC_09930	CRISPR-associated endonuclease Cas1
LRC_13830	Arginine/ornithine antiporter
LRC_17090	Hypothetical protein
LRC_17100	Hypothetical protein
LRC_17620	Hypothetical protein
LRC_18070	Hypothetical protein
LRC_18090	Hypothetical protein

^aLocus tag identifier from the ATCC 27782 genome sequence.

Pooled sequences from cells grown in glucose and tetrasaccharide were used to generate 1,857 contigs, ranging from 13,299 to 201 bp (mean ± standard error of the mean [SEM], 938 ± 27 bp). Each contig was compared to the three available genome sequences using BLAST. Contigs that did not match any of the L. ruminis genome sequences were compared to the NCBI nr database. Of these contigs, 51 could be matched with bacterial sequences other than L. ruminis (see Table S2 in the supplemental material). However, none of these contigs were predicted to code for proteins that might explain the trophic differences between strain L5 and the fully sequenced L. ruminis strains.

Microscopy.

Light microscopy showed that about 84/100, 72/100, and 0/100 bacterial cells had slow motility in preparations from LDP4-, CDP4-, and glucose-grown cultures, respectively. Transmission electron microscopy showed that 99/100, 90/100, and 12/100 cells in preparations prepared from L5 culture (late growth) in LDP4, CDP4, and glucose medium, respectively, had a single subterminal flagellum. These results showed that upregulation of the chemotaxis/motility operon resulted in increased production of flagella and motility.

Doubling times of strain L5 with different growth substrates.

Doubling times for L. ruminis strain L5 were 78 min for CDP4 medium, 64 min for LDP4 medium, and 35 min for glucose medium.

Chemotaxis.

Larger numbers of bacterial cells entered the capillary tubes containing attractants (glucose, CDP4, and LDP4) than entered tubes without carbohydrate. This indicated chemotactic behavior by L. ruminis L5. Larger numbers of bacterial cells were recovered from capillary tubes containing attractants when L. ruminis L5 was grown in CDP4 or LDP4 than when it was grown in glucose. This indicated that more cells were actively motile in tetrasaccharide-containing cultures than in glucose-grown cultures (Fig. 5).

Fig 5.

Chemotaxis in relation to growth substrate and attractant. L. ruminis strain L5 was grown in MRS medium containing CDP4 (a), glucose (b), or LDP4 (c). The CFU/ml of L. ruminis in capillary tubes containing glucose, CDP4, and LDP4 was greater than that in control tubes that did not contain an attractant (MRS without fermentable carbohydrate, NO CHO). Mean values and SEMs for three biological replicates are shown. Unpaired t test with Welch's correction for glucose-grown cells, P < 0.05; LDP4- and CDP4-grown cells, P < 0.01.

DISCUSSION

The large bowel of humans is home to hundreds of bacterial species that, collectively, form communities comprised of trillions of cells (22). Obligately anaerobic bacteria predominate in the bowel microbiota. An important function of the gut microbiota is to begin the recycling of dietary waste inside the large bowel by degrading plant-derived polymers and fermenting the hydrolysis products to short-chain fatty acids. These fatty acids provide a source of calories for the human host because they are taken up by the intestinal mucosa and incorporated into biochemical pathways (23).

High-throughput sequencing (HTS) technologies have greatly advanced knowledge of the compositions of bowel microbiotas of humans in health and disease (24). Knowing “who is there” in the ecosystem is, however, just the first step to understanding this biological system. Bacterial communities are natural assemblies of organisms whose members fill numerous ecological niches. How these relatively stable, self-regulating communities are formed is still poorly understood. This is probably because the culture and physiological study of bowel bacteria have been neglected. “What are they doing” and “how are they doing it” can sometimes be inferred from genomic data of bowel inhabitants, but the interactive capacity and interactive mechanisms of bowel bacteria will be understood only by revealing the food webs and food chains that underpin the functioning assemblage. HTS of fecal microbiotas does not permit identification of strain-to-strain variation within particular species with regard to the utilization of substrates released through hydrolytic reactions carried out by other members of the microbiota.

We reasoned that a study of L. ruminis in relation to oligosaccharides derived from β-glucans would generate information about how these members of the microbiota cope in a nutritionally competitive ecosystem. The results show that, unable to use β-glucans for growth, some strains of L. ruminis have arisen that can grow on tetrasaccharides resulting from the hydrolysis of the polymers by other members of the microbiota. Tetrasaccharide utilization by every strain of L. ruminis in our study would have indicated that this was an essential factor for their relative fitness in the bowel. The three rumen isolates were able to utilize one of the tetrasaccharides (LDP4) for growth, but three of seven human isolates could not grow using either of the tetrasaccharides. While apparently not essential for relative fitness in the human bowel, tetrasaccharide utilization can nevertheless be seen as an important trophic characteristic. It would allow an L. ruminis strain to share ecosystem resources with β-glucan degraders (9). The nutritional interaction demonstrated between Coprococcus sp. and L. ruminis L5 supports this contention. Availability of tetrasaccharide for L. ruminis might be low (“crumbs” from the figurative commensal “table”), so the advantage of increased transcription of genes associated with foraging is evident. Search-and-locate mechanisms (motility and chemotaxis) would be essential for L. ruminis L5 ecological fitness. Temporal differences in the expression of the cellobiose operon (early growth) relative to the chemotaxis/motility operon (late growth) probably relate to the different availabilities of tetrasaccharide as the bacterial culture developed. In studies such as the one that we have carried out, it would be ideal to perform mutagenic and complementation experiments involving cellobiose and chemotaxis/motility genes to directly link expression of genes and phenotypic characters. We do not have methods for the genetic modification of strain L5, but the obvious alterations in gene expression linked to predictable phenotypic changes make this unnecessary.

The differentiation between L. ruminis strains that use one or both kinds of tetrasaccharides for growth remains unexplained at the molecular level since genomic comparisons did not reveal a physiological explanation. The increased transcription of genes encoding “hypothetical proteins” and an ABC transporter gene may be involved in the phenomenon. A recent analysis of the transcriptional response of Lactobacillus acidophilus to growth on stachyose (a tetrasaccharide) showed increased transcription of ABC transporters (25). Crystal structure studies have shown ABC transporters binding tetrasaccharide with the strongest interactions at the nonreducing end of the carbohydrate (26). The ABC transporter ATP-binding protein that we detected in strain L5 may therefore be a candidate molecule in the uptake of tetrasaccharide by L. ruminis L5 and ATCC 27782. There is clearly much work to be done to discover the phenotypic traits encoded by the genes that had increased transcription in strain L5. This is, however, difficult work for the following reasons. The rationale behind homology-based annotation is that, if two sequences have a high degree of similarity, then they have evolved from a common ancestor, and thus, they should have similar, if not identical, functions. However, with increasing numbers of sequences in databases as well as the effects of gene duplications, which might be followed by divergence of function, the power of homology-based annotation has lessened. Adding to this is the problem of errors in annotation which spread misannotations when homology-based approaches are used. Additionally, most of the newly identified proteins do not show significant sequence similarity with experimentally characterized proteins (27). Some genes have multiple annotations reflecting functions of gene products in different experimental systems (28).

The differential utilization of tetrasaccharides by L. ruminis strains might be due to mutations involving a single nucleotide change that could result in altered binding/transport/hydrolysis of specific oligosaccharides. The “periodic selection model” of microbial evolution infers that beneficial mutations with large effects in fitness arise rarely but that, once a clone contains a large-effect single nucleotide polymorphism (SNP), this rapidly rises to fixation (29, 30). As a result, clonal ecotypes occupy particular niches. Mutations involving a single nucleotide change might alter binding/transport/hydrolysis of specific oligosaccharides by bacterial cells. Hence, the presumptive clonal ecotypes of L. ruminis that we have detected might have been selected in bowel communities in which there was either a predominance or a paucity of β-glucan degraders. The microbiota of humans varies in composition between individuals, although details regarding β-glucan degraders are lacking (31).

Doubling times of L. ruminis L5 in tetrasaccharide media were about twice as long as those in glucose medium, indicating a more complex pathway of utilization of tetrasaccharides. L. ruminis L5 may transport tetrasaccharide into the cell before hydrolysis or hydrolyze the substrates extracellularly and transport simpler molecules into the cell. The latter mechanism would expend less energy. Both types of tetrasaccharide upon hydrolysis yield cellobiose, which is transformed by the products of genes in the cellobiose operon (32). However, tetrasaccharide produced by lichenase digestion yields a molecule with a β-1:3 linkage at the reducing end, whereas cellulase digestion results in the β-1:3 linkage at the nonreducing end of the molecule. Fascinatingly, our data imply that such a difference in molecular structure can affect utilization of tetrasaccharide by L. ruminis.

The results that are reported here show the need to learn about the growth requirements of specific bacterial strains rather than to generalize at species level. Pursuit of these kinds of studies provides knowledge that may lead to increased comprehension of the evolution of bowel commensals. Importantly, our study showed that RNA-seq was a useful platform with which to begin investigations of trophic phenomena in the human bowel. Similarly to a microarray experiment, it provided a quick screen of the transcriptome of strain L5 under different growth conditions. These preliminary results were confirmed and extended by RT-qPCR of selected genes of interest. When these data were coupled with those from bacterial growth experiments, foraging attributes of gut commensal L. ruminis L5 were revealed in relation to tetrasaccharides sourced from polymeric substances that it is itself unable to degrade.