Ligand-binding properties of substrate binding proteins of a maltose uptake system in Gardnerella swidsinskii

TL;DR

This study investigates how two substrate-binding proteins in Gardnerella swidsinskii interact with maltose and related sugars, revealing their binding affinities and expression patterns.

Contribution

The paper identifies and compares two substrate-binding proteins in G. swidsinskii, showing their similar ligand affinities and expression dynamics.

Findings

Both SBP genes are expressed as a polycistronic transcript, with musE1346 transcripts being more abundant.

Both SBPs show high affinity for maltose, maltotriose, and maltotetraose but lower affinity for longer malto-oligosaccharides.

Expression levels of the SBP genes do not change significantly in media with glycogen or maltotriose.

Abstract

Glycogen and its breakdown products, maltose and malto-oligosaccharides, are important carbon sources for vaginal bacteria including Gardnerella species. MusEFGKI transport systems for maltose and malto-oligosaccharides have been identified in all Gardnerella species; however, unlike in other species, the Gardnerella swidsinskii operon encodes two substrate-binding proteins (SBPs) (MusE1345, MusE1346, ~60% amino acid identity). Two SBPs could allow binding of additional ligands, providing a competitive advantage to G. swidsinskii relative to other species with only one SBP. Our objectives were to determine if both genes are expressed in G. swidsinskii and compare the specificity and affinity of G. swidsinskii MusE SBPs for glycogen breakdown products. Gene expression analysis showed the presence of a polycistronic transcript spanning both SBP encoding genes; however, musE1346 transcripts were more abundant, likely due to the presence of an additional promoter identified in the intergenic region. No difference in the relative expression of either gene was observed in isolates grown in media supplemented with glycogen or maltotriose. Predicted structures of both SBPs were highly similar and characteristic of previously characterized maltose-binding proteins. Both proteins had a high affinity for maltose, maltotriose and maltotetraose (Kd 10−6 to 10−7 M) and much lower affinities to maltopentaose and maltohexaose (Kd 10−3 to 10−4 M). Our results demonstrate that the affinities of G. swidsinskii MusE SBPs for maltose and malto-oligosaccharides are similar under the same experimental conditions.

Figures

Gene expression levels of G. swidsinskii isolates (NR016, NR020 and NR021) in two conditions, glycogen and maltotriose (M3) quantified by reverse transcription qPCR. Average fold change values were calculated using the Livak (∆∆Cq) method [18]. Mean Cq values are indicated by horizontal lines and error bars indicate standard deviation. Cq values from replicates of individual isolates are indicated by colour (and shape) according to the legend.

Predicted structures of G. swidsinskii MusE SBPs (PyMol v.2.5.2). (a) MusE1345 with N and C terminal domains coloured pink and green, respectively. (b) MusE1346 with N and C terminal domains coloured orange and cyan, respectively. (c) Structure alignment of both SBPs.

CD readings of MusE SBPs measured over 197–280 nm in mdeg. Each line represents the average of three CD experiments of the same SBP (blue, MusE1345; red, MusE1346). Concentrations used were 2.57 µM for MusE1345 and 2.66 µM for MusE1346. Composition of secondary structures was analysed using BeStSel v1.3.230210.

ITC exothermic saturation curves of MusE1345 and MusE1346 to maltose, maltotriose, maltotetraose, maltopentaose and maltohexaose in 20 injections (blue, MusE1345; red, MusE1346). Corrected heat is the heat signal after subtracting the average heat signal from the ligand-into-buffer run.

Tables

Table 1.. Quantification of protein-ligand interactions at 25 °C. Reported values are the average of triplicates±sd

	Ligand	(Ligand)	Kd (M)	n value	ΔH (kJ/mol)	ΔG (kJ/mol)	ΔS (J/mol×K)
MusE1345 ‘unique’
	Lactose	300 µM	N/Q*	–	–	–	–
	Glucose	300 µM	N/Q*	–	–	–	–
	Isomaltose	300 µM	N/Q*	–	–	–	–
	Maltose (M2)	300 µM	(1.07±0.22)×10−6†	0.10±0.00	−50.45±4.18	−34.14±0.49	−54.73±15.52
	Maltotriose (M3)	300 µM	(2.33±0.17)×10−6†	0.11±0.00	−100.00±0.00	−32.16±0.18	−227.57±0.62
	Maltotetraose (M4)	450 µM	(1.08±1.52)×10−7†	0.14±0.03	−35.73±12.09	−46.59±6.75	36.42±62.99
	Maltopentaose (M5)	15 mM	(1.15±0.10)×10−4	1	−4.38±0.04	−22.50±0.21	60.78±0.86
	Maltohexaose (M6)	15 mM	(8.97±5.49)×10−4	1	−6.16±0.75	−17.82±1.39	39.09±7.19
MusE1346 ‘shared’
	Lactose	300 µM	N/Q*	–	–	–	–
	Glucose	300 µM	N/Q*	–	–	–	–
	Isomaltose	300 µM	N/Q*	–	–	–	–
	Maltose (M2)	300 µM	(3.13±0.31)×10−7	0.92±0.05	−59.80±2.15	−37.14±0.25	−75.98±6.60
	Maltotriose (M3)	300 µM	(1.84±0.05)×10−7	1.08±0.04	−49.81±1.11	−38.44±0.06	−38.13±3.80
	Maltotetraose (M4)	450 µM	(9.20±0.11)×10−7	1.18±0.01	−31.06±0.56	−34.45±0.03	11.40±1.79
	Maltopentaose (M5)	15 mM	(4.41±0.04)×10−3	1.21±0.03	−96.09±5.52	−13.45±0.03	−277.20±18.60
	Maltohexaose (M6)	15 mM	(4.40±0.56)×10−3	1.03±0.37	−84.82±21.45	−13.47±0.31	−239.33±72.01

Full text

Introduction

Gardnerella bacteria were once thought to be the sole causal agent of bacterial vaginosis (BV), a vaginal condition in reproductive age women that can cause irritation, discomfort, malodor and increased vaginal discharge [1]. The recent recognition that multiple, phenotypically distinct Gardnerella species colonize the vaginal microbiome has further complicated the recognition of clinically significant dysbiosis and motivated the determination of factors influencing Gardnerella population structure and dynamics [2].

Ecological factors that contribute to the BV-associated overgrowth of Gardnerella and which Gardnerella species dominate the microbiome in individuals with BV are not completely defined but competition for nutrients has been demonstrated to be important in determining Gardnerella population structure [3]. Similar to other vaginal bacteria, Gardnerella spp. depend on glycogen and its breakdown products as major nutrient and carbohydrate sources for their survival [4]. Glycogen molecules released into the vaginal lumen with the turnover of epithelial cells are digested into smaller molecules such as maltose and malto-oligosaccharides by host and bacterial enzymes and taken up by vaginal bacteria through transporter systems [57].

One of the most significant systems for the import of glycogen breakdown products in Gardnerella is the ATP-binding cassette (ABC) transporter superfamily. General components of an ABC importer for glycogen breakdown products like maltose and malto-oligosaccharides include two nucleotide-binding domains, two transmembrane domains and an extracellular substrate-binding protein (SBP), the latter of which determines the specificity and affinity of the transporter for different substrates [8]. In Gardnerella, six gene clusters encoding maltose and malto-oligosaccharide transporters of the ABC family have been annotated with RafEFGK and TMSP only found in Gardnerella vaginalis, MalEFG being unique to Gardnerella leopoldii and MalXFGK and MusEFGKI found in all Gardnerella spp. [9].

The MusEFGKI system was first identified in Corynebacterium glutamicum as a maltose and maltodextrin uptake system and includes an SBP (MusE), transmembrane domains (MusF, MusG), nucleotide-binding domain (MusK) and a membrane-associated protein with unknown function (MusI), with MusI being a novel, essential component that distinguished this system from others described at the time [10]. Henrich et al. identified preferred substrates of the MusEFGKI system, such as maltose, maltotriose and maltotetraose [10]. Although the analogous transporter system in Gardnerella species lacks MusI and MusK, it was classified as MusEFG based on sequence similarity to reference data in the Transporter Classification Database [9]. The operon encodes one SBP in all Gardnerella spp. except in Gardnerella swidsinskii, where it encodes two SBPs. The two SBP encoding genes occur in tandem within the operon. One of the G. swidsinskii SBPs shares ~90% amino acid sequence identity with the MusE SBP in other Gardnerella spp. (‘shared’ SBP), and the other is only ~60% identical and forms a separate clade (‘unique’ SBP) [9]. Since the ligand-binding affinities of SBPs provide specificity to transporters, dissimilarity in the sequences of the two G. swidsinskii MusE SBPs may indicate a difference in structure and function. We hypothesized that the two MusE SBPs differ in ligand binding properties, potentially providing G. swidsinskii with a competitive advantage in accessing glycogen breakdown products by broadening the range of transportable ligands. As a first step towards testing this hypothesis, the objectives of this study were to confirm that both MusE SBP encoding genes are expressed in G. swidsinskii and to compare the specificity and affinity of * G. swidsinskii* MusE SBPs for glycogen breakdown products.

Methods

Bacterial strains and culture

G. swidsinskii strains NR016, NR020 and NR021 were cultured on Columbia sheep blood agar plates for 48 h at 37 °C under anaerobic conditions (GasPak^™^ EZ Anaerobe Gas Generating Pouch System, BD). For broth cultures, modified NYC III medium (mNYC III) was prepared, supplemented with 1% filter-sterilized oyster glycogen (Sigma-Aldrich, cat#9005-79-2), maltotriose (Thermo Fisher, cat#J66491.03) or d-glucose (FisherChemical, cat#D16-500). mNYC III is NYC III (ATCC Medium 1685) with 10% (v/v) heat-inactivated FBS instead of horse serum and no additional glucose [7].

Gene expression analysis

Reference gene selection

Ten candidate reference genes were selected from the core genome of G. swidsinskii strains NR016, NR020 and NR021, representing a variety of functional categories (clusters of orthologous group categories). Primers were designed for candidate reference genes and genes of interest using Primer3Plus [11] (Table S1, available in the online version of this article). Validation of product-specific amplification was performed by conventional PCR assay with a thermal gradient (57.7–71.6 °C). Each reaction was 25 µl containing 1×PCR buffer, 2.5 mM MgCl_2_, 0.2 mM dNTPs, 0.4 µM forward primer, 0.4 µM reverse primer, 0.05 U µl^−1^ Taq DNA Polymerase (AccuStart II Taq DNA Polymerase, Avantor) and 20 ng of genomic DNA. The thermocycling parameters included an initial denaturation at 95 °C for 3 min, followed by 40 cycles of (95 °C for 30 s, T_anneal_ for 30 s and 72 °C for 15 s) and a final extension at 72 °C for 5 min. The products were visualized by agarose gel electrophoresis (1.5% w/v agarose) to select the best annealing temperature. Amplification efficiency for each primer set was calculated using real-time SYBR green PCR and a 10-fold dilution series of genomic DNA.

G. swidsinskii strains NR016, NR020 and NR021 were grown on Columbia sheep blood agar plates anaerobically at 37 °C for 48 h from isolates stored at −80 °C. Multiple colonies were then used to inoculate 5 ml of mNYC III media supplemented with 1% (w/v) d-glucose and incubated anaerobically at 37 °C for 48 h. Aliquots (200 µl) of the diluted culture of each strain were pipetted into separate flat-bottom 96-well plates with 44 replicate wells per condition (mNYC III supplemented with 1% maltotriose or 1% glycogen). Uninoculated mNYC III media was used as a negative control with four replicates per condition. The 96-well plates were then incubated anaerobically at 37 °C until the mid-exponential phase (OD_600_=0.3–0.6). Bacteria were harvested by pooling two wells (total volume of 400 µl) into pre-labelled microcentrifuge tubes and kept on ice while completing the harvesting of all bacteria. Bacteria were pelleted by centrifugation at 4 °C, the supernatant was removed and pellets were immediately stored at −80 °C until RNA isolation. RNA extraction was performed in triplicate per strain (NR016, NR020 and NR021) per condition (maltotriose and glycogen) using the PureLink RNA Mini Kit with modifications described in a detailed laboratory protocol published elsewhere [12]. An extraction control containing only kit reagents was included with each batch of extractions. RNA quality (A_260_/A_280_ and RIN) and quantity were examined using the Agilent TapeStation and Nanodrop 2000c. RNA extracts were stored at −80 °C overnight before first-strand cDNA synthesis using 50 ng of template RNA. cDNA synthesis was performed with the ThermoScientific Maxima First-Strand cDNA Synthesis Kit for reverse transcription quantitative PCR (qPCR), with dsDNase, and a reverse transcriptase minus (RT−) negative control and no-template control (NTC) included for each set of extractions.

qPCR was performed with two technical replicates for each 1 : 10 diluted cDNA sample. A total of 36 C_q_ values were obtained for each candidate reference gene (3 isolates×2 conditions×3 RNA/cDNA preps×2 PCR replicates). Each 10 µl qPCR reaction contained 1×PowerUp SYBR Green Master Mix (ThermoFisher Scientific, cat#A25742), 0.4 µM forward primer, 0.4 µM reverse primer and 2 µl template cDNA. The following amplification cycles were applied in a Bio-Rad CFX Connect real-time system: 1 cycle at 50 °C for 2 min, 1 cycle at 95 °C for 2 min, 40 cycles of 95 °C for 15 s and 60 °C for 15 s, followed by a melt curve, increasing the temperature by 0.5 °C increments from 65 to 95 °C. Reaction mixture with no template DNA was used as a negative control, and each biological replicate of G. swidsinskii in each condition had three technical replicates run in duplicate PCR reactions. The RT− control was tested using the ftsK PCR.

Candidate reference gene expression stability was assessed in media supplemented with M3 or glycogen and evaluated by analysing the raw C_q_ values from the reverse transcription qPCR data with RefFinder [13], which combines four algorithms: the standard deviations of the delta-C_q_ [14], BestKeeper [15], NormFinder [16] and geNorm [17]. A comprehensive ranking by RefFinder is provided by calculating the geometric mean of the rankings of all candidate reference genes by the four algorithms.

MusE gene expression

Primer design, validation and optimization

Multiple sequence alignment of genes encoding MusE SBPs of 3 G. swidsinskii isolates (NR016, NR020 and NR021) was performed using Geneious Prime version 2025.0.3. Primers amplifying MusE SBP encoding sequences were generated using the same software. Amplification efficiency of each primer pair was determined using a 10-fold serial dilution of G. swidsinskii genomic DNA in SYBR green qPCR assays. Primers JH0983 and JH0984 were designed to amplify a 357 bp sequence spanning the intergenic region between musE1345 and musE1346 (Table S1).

Real-time qPCR

Each qPCR reaction was 10 µl containing 1×PowerTrack SYBR Green Master Mix (Thermo Fisher Scientific, cat#A46109), 0.4 µM forward primer, 0.4 µM reverse primer and 2 µl template cDNA (1 : 10 dilution). The following amplification cycles were run on the QuantStudio 3 Real-Time PCR System (Thermo Fisher): 1 cycle at 50 °C for 2 min, 1 cycle at 95 °C for 2 min, 40 cycles at 95 °C for 15 s and 60 °C for 15 s, followed by a melt curve incrementally increasing the temperature by 0.15 °C starting from 65 to 95 °C. Reaction mixture with no template was used as a negative control and each biological replicate of G. swidsinskii in each condition had three technical replicates run in duplicate PCR reactions. The RT− control was tested using uppS primers. The average Cq values of qPCR technical duplicates were plotted using GraphPad Prism version 10.3.1. The average fold change of musE1345 and musE1346 gene expression was calculated using the Livak method [18].

Conventional PCR of the intergenic region

Each PCR reaction was 25 µl containing 1×PCR buffer (Quantabio, cat#84248, Beverly, MA, USA), 2.5 mM MgCl_2_ (Quantabio, cat#84030), 0.2 mM dNTPs, 0.4 µM forward primer (JH0983), 0.4 µM reverse primer (JH0984), 0.05 U µl^−1^ Taq DNA Polymerase (AccuStart II Taq DNA Polymerase, Quantabio, cat#84247) and 2 µl template cDNA. PCR protocol was as follows: 95 °C for 2 min, 40 cycles at 95 °C for 15 s, 60 °C for 30 s, 72 °C for 30 s, followed by 72 °C for 7 min, and hold at 4 °C. The PCR run included an NTC. Amplified PCR products were run on 2% agarose gel with ethidium bromide for 30 min at 110 V and visualized using the ChemiDoc XRS+ system (Bio-Rad).

Promoter prediction

Prediction of potential promoters within the MusE_2_FGK_2_I operon was performed using the BPROM tool of the Softberry bioinformatic website with the latest modification date of 24 October 2016 (Softberry Inc., Mount Kisco, NY, USA). BPROM predicts potential transcription start positions regulated by bacterial sigma70 promoters and assigns each of them a linear discriminant function (LDF) value. The higher the LDF value of the predicted promoter is, the higher the possibility of its existence. The threshold LDF value for consideration of potential promoters was set to 0.02.

Protein 3D structure prediction

Predicted structures of MusE1345 (Genbank accession WP_116995000.1) and MusE1346 (WP_012914535.1) from G. swidsinskii NR020 were generated using DMPFold 2.0 within the PSIPRED workbench (http://bioinf.cs.ucl.ac.uk/psipred/). PyMol version 2.5.2 (Schrödinger LLC, New York, NY, USA) was used to visualize and align 3D protein structures.

DNA cloning and cell transformation

The amino acid sequences of the two SBPs were examined for the presence of signal peptides and sortase motifs (e.g. LPxTG) or potential membrane anchor sequences using InterPro version 107.0 (https://www.ebi.ac.uk/interpro/), and the DNA sequences encoding these features were excluded from constructs for protein expression. The resulting DNA sequences encoding MusE1345 (aa 36–471) and MusE1346 (aa 36–447) were codon-optimized for expression in Escherichia coli, and restriction sites for BamHI (5′-end) and HindIII (3′-end) were added. Sequences were synthesized and inserted into the cloning vector pUCIDT-AMP by IDT (Coralville, IA, USA). Vectors containing sequences encoding either MusE1345 or MusE1346 SBP were restriction digested with BamHI and HindIII, and the protein-encoding sequences were ligated into pET41a^+^ (Novagen, MilliporeSigma, cat#70556–3, Burlington, MA, USA) or pQE-80L (QIAGEN, cat#32923, Germantown, MD, USA), respectively. Chemically competent E. coli DH5α was transformed with either pET41a^+^+MusE1345 or pQE-80L+MusE1346. Plasmid constructs were sequenced to confirm that inserts were in-frame with the N-terminal glutathione-S-transferase (GST) tag (MusE1345) or 6xHistidine (His6) tag (MusE1346) (Plasmidsaurus, Eugene, OR, USA). Chemically competent E. coli BL21(DE3) was transformed with pET41a^+^+MusE1345 for protein expression.

Protein expression and purification

Cell culturing and induction

Frozen stocks of E. coli carrying the expression vector [either E. coli BL21(DE3) with pET41a^+^+MusE1345 or E. coli DH5α with pQE-80L+MusE1346] were inoculated into 30 ml of Luria broth (LB) with kanamycin (Kan, 30 µg ml^−1^) or ampicillin (Amp, 100 µg ml^−1^) depending on the vector and grown overnight at 37 °C with shaking at 200 r.p.m. Overnight culture (25 ml) was added to 2,475 ml of LB with 30 µg ml^−1^ Kan (or 100 µg ml^−1^ Amp) making a total volume of 2.5 l. The inoculated culture was incubated at 37 °C, 200 r.p.m. until OD_600_=0.4–0.6. Isopropyl β-d-1-thiogalactopyranoside (IPTG) [final (IPTG) = 0.1 mM] was added to induce protein expression. The induction condition was either 25 °C, 200 r.p.m., 24 h for MusE1345 or 37 °C, 200 r.p.m., 4 h for MusE1346. Cultures were pelleted by centrifugation in 1 l aliquots.

Cell lysis

Each 1 l cell pellet was resuspended in 100 ml of lysis buffer (50 mM Tris, 150 mM NaCl, 0.1% Triton X-100, 10% glycerol, pH 8.0) for MusE1345 or (20 mM Tris, 2 mM NaH_2_PO_4_, 300 mM NaCl, 20 mM imidazole, 10% glycerol, pH 8.0) for MusE1346 and lysed using sonication for 8 rounds of 20 s of sonication and 10 s of resting on ice or using a cell disruptor. The lysate was clarified by centrifugation at 18,549 g for 30 min using an Avanti J-26 XP centrifuge with JA25.50 rotor (Beckman Coulter, Brea, CA, USA).

Protein purification

Purification of GST-tagged MusE1345 SBP

Purification using the GST tag on MusE1345 started with 100 ml of clarified lysate being incubated with 10 ml of glutathione sepharose beads (MilliporeSigma, cat#G4510) overnight at 4 °C with constant rotation to allow binding of GST-tagged MusE1345 to glutathione beads. The bead mixture was washed once in 1×PBS (pH 7.4) with 1% Triton X-100, once with 1×PBS (pH 7.4) and twice with 20 mM Tris, 2 mM NaH_2_PO_4_, 300 mM NaCl, 10% glycerol, pH 8.0. MusE1345 was released from the bead-bound GST tag by overnight incubation of the washed beads with 32 units of thrombin (MilliporeSigma, cat#696713) in 20 mM Tris, 2 mM NaH_2_PO_4_, 300 mM NaCl, 10% glycerol, pH 8.0 at 4 °C with constant rotation. Thrombin was removed from the supernatant by incubation with 5 ml of benzamidine sepharose slurry (Cytiva, cat#17512310, Wilmington, DE, USA) for 2 h at 4 °C with constant rotation to allow binding of thrombin to the sepharose beads. The mixture was then filtered through an empty PD-10 column to remove the beads. The flow-through containing MusE1345 was concentrated using an Amicon Ultra-4 centrifugal filter unit (MilliporeSigma, cat#UFC803024) by centrifuging at 2,500 g. Purified, concentrated products were assessed on 12% SDS-PAGE. Purified MusE1345 was dialysed (1 : 1,000 vol ratio) against buffer containing 20 mM Tris, 2 mM NaH_2_PO_4_, 300 mM NaCl, 10% glycerol, pH 8.0 (same buffer as wash 3 and 4).

Purification of His6-tagged MusE1346 SBP

The clarified lysate was loaded into an NGC Plus FPLC system (Bio-Rad) with a Ni-NTA affinity column to purify the His6-tagged MusE1346 SBP. Protein was eluted from the affinity column using an elution buffer with high imidazole concentration (20 mM Tris, 2 mM NaH_2_PO_4_, 300 mM NaCl, 500 mM imidazole, 10% glycerol, pH 8.0). Eluted fractions from FPLC were analysed on 12% SDS-PAGE to confirm the presence of MusE1346 SBP (predicted mass 45 kDa). Based on the results of SDS-PAGE analysis, appropriate FPLC fractions were pooled and dialysed (1 : 1,000 vol ratio) against lysis buffer without imidazole (20 mM Tris, 2 mM NaH_2_PO_4_, 300 mM NaCl, 10% glycerol, pH 8.0).

Protein conformation and molecular weight

The conformation and molecular weights of purified proteins were analysed using size exclusion chromatography with multi-angle static light scattering (SEC-MALS). The apparatus includes the NGC Plus FPLC system (Bio-Rad) connected to Wyatt MALS-RI detectors (Wyatt Technology, Goleta, CA, USA). The output was visualized on ASTRA software version 7.3.2.19.

Protein quantification

Protein concentration was determined using a NanoDrop2000c spectrophotometer, ProteinA280 function, adjusted for molecular weight (50.56 kDa for MusE1345, 44.38 kDa for MusE1346, obtained from SEC-MALS) and extinction coefficient [96260 for MusE1345, 88810 for MusE1346, obtained from amino acid sequences analysis by ProtParam (Expasy, https://web.expasy.org/protparam/, SIB Swiss Institute of Bioinformatics, Écublens, Vaud, Switzerland)].

Protein secondary structure analysis

Circular dichroism (CD) was performed using the Chirascan Plus CD Spectrometer (Applied Photophysics, Leatherhead, Surrey, UK). The buffer used for CD was 20 mM Tris, 2 mM NaH_2_PO_4_, 100 mM NaCl, pH 8.0, filtered with a 0.2 µm membrane and degassed before use. Concentrations used for CD were 2.57 µM for MusE1345 and 2.66 µM for MusE1346. Data were collected every 0.5 nm within the 197–280 nm wavelength range (1 s per data point) using a 1-mm path-length cuvette at 20 °C. Each CD dataset represents an average of three scans per SBP. Spectrum analysis was performed with Beta Structure Selection (BeStSel) online tool version 1.3.230210 (ELTE Eötvös Loránd University, Budapest, Hungary) and visualized in GraphPad Prism version 10.3.1 (Boston, MA, USA).

Affinity quantification by isothermal titration calorimetry

Isothermal titration calorimetry (ITC) experiments were performed using an Affinity-ITC apparatus (TA Instruments, New Castle, DE, USA). Experimental setup: 25 °C temperature set point, 200 r.p.m. stirring rate, medium peak height to width ratio, 20 incremental injections of the ligand (2.5 µl per injection every 200 s), auto-equilibrated with medium expected heats, data collection starting 300 s before the first injection and ending 300 s after the last injection. The affinity of each ligand and SBP pair was measured in triplicate. Proteins were diluted to ~45 µM in the same buffer used for dialysis (20 mM Tris, 2 mM NaH_2_PO_4_, 300 mM NaCl, 10% glycerol, pH 8.0, filtered with a 0.2 µm membrane and degassed prior to use). Carbohydrate solutions were prepared using the same buffer as the proteins [lactose, glucose, isomaltose, maltose (M2), maltotriose (M3), maltotetraose (M4), maltopentaose (M5) and maltohexaose (M6)]. Carbohydrates were sourced from several manufacturers: lactose (MilliporeSigma, cat#L-2643), glucose (Fisher Chemical, Thermo Fisher, cat#D16-500, Waltham, MA, USA), isomaltose (TCI America, cat#I0231, Portland, OR, USA), M2 (Fisher Chemical, cat#M75-100), M3 (Thermo Fisher, cat#J66491.03), M4 (Biosynth, cat#OM02796, Compton, Berkshire, UK), M5 (Megazyme, Neogen, cat#O-MAL5, Lansing, MI, USA or MilliporeSigma, cat#SMB01321) and M6 (Biosynth, cat#OM06869). ITC data was analysed using NanoAnalyze Data Analysis software version 4.0.0.4 (TA Instruments) and its independent model for one-site binding. Integration peaks were 60 s for MusE1345 and 45 s for MusE1346. Each set of triplicates was normalized using a control file (average area value) from the corresponding ligand-into-buffer run. Saturation curves were graphed using GraphPad Prism version 10.3.1.

Results

musE1346 is more highly expressed than musE1345

Reverse transcription qPCR was performed to determine if both musE genes are transcribed and compare musE1345 and musE1356 transcript levels. A necessary first step was to identify a suitable reference gene since none are currently established for G. swidsinskii.

The expression stability of candidate reference genes (ftsK, gatA, gatC, hrdB, pgm, ppgK, rpoB and uppS) reverse transcription qPCRs had Cq values ranging from 15.63 to 22.92 (Fig. S1). Stability was quantified using RefFinder (Fig. S2), and uppS was the top-ranked reference gene, followed by gatA, gatC, ftsK, rpoB, hrdB and ppgK (Fig. S3).

Once a suitable reference gene was identified, transcripts corresponding to MusE SBP encoding genes were quantified with reverse transcription qPCR. RNA extracts contained 1.1–4.1 µg of RNA (mean 2.3 µg), with A260/A280 ratios of 2.06–2.24 (median 2.13) and RIN values 7.5–8.9 (median 8.05). MusE1346 (shared SBP) transcripts were more abundant than MusE1345 (unique SBP) transcripts in both conditions (glycogen or M3). No difference in relative gene expression was observed for either gene in the two conditions tested, with average fold change values for MusE1345 and MusE1346 of 1.21 and 1.18, respectively (Fig. 1).

Gene expression levels of G. swidsinskii isolates (NR016, NR020 and NR021) in two conditions, glycogen and maltotriose (M3) quantified by reverse transcription qPCR. Average fold change values were calculated using the Livak (∆∆Cq) method [18]. Mean Cq values are indicated by horizontal lines and error bars indicate standard deviation. Cq values from replicates of individual isolates are indicated by colour (and shape) according to the legend.

To determine if polycistronic transcripts covering both MusE SBP encoding genes were produced, we performed conventional PCR using primers spanning the intergenic region on cDNA from all isolates in both M3 and glycogen culturing conditions. The expected PCR amplicon was produced in all cases. Since we also observed a difference in relative gene expression between musE1345 and musE1346, we examined the surrounding operon sequence for the presence of promoters that might explain the observations. Within the MusE_2_FGK_2_I operon of NR020 isolate, promoters were identified upstream of each of the MusE SBP encoding genes and upstream of an adjacent gene encoding a predicted fructose 1,6-biphosphatase (Fig. S4).

MusE SBPs are lipoproteins with a typical ABC transporter SBP fold

The amino acid sequence of MusE1345 shared 62% identity with MusE1346. N-terminal signal peptides and cysteine residues important for type II signalling were identified in both proteins, indicating that the SBPs are lipoproteins. No LPxTG motifs were detected. The 3D structures of both G. swidsinskii MusE SBPs, excluding signal peptides, showed distinct N- and C-terminal domains comprising mainly α-helices and a hollow cleft between the domains (Fig. 2). Alignment of the predicted structures demonstrated similar overall spatial conformation. To identify potential ligand-binding site residues in MusE1345 and MusE1346, we performed a structural alignment of the models of MusE1345 and MusE1346 with the published structure of MalX from Streptococcus pneumoniae complexed with maltoheptaose (2XD3) [19]. Residues making up the ‘aromatic cradle’ of MalX, Y197 (subsite 2), W273 (subsite 1) and W384 (subsite 3), are conserved in MusE1345 as Y200, W266 and W385 and in MusE1346 as Y184, W250 and W370. Other subsites identified in the MalX-maltoheptaose structure were not discernible in the MusE1345 and MusE1346 models, and the extended binding cleft reported for MalX that accommodates up to maltooctaose is apparently occluded in both MusE1345 and MusE1346 by a conserved SSDWRF motif.

Predicted structures of G. swidsinskii MusE SBPs (PyMol v.2.5.2). (a) MusE1345 with N and C terminal domains coloured pink and green, respectively. (b) MusE1346 with N and C terminal domains coloured orange and cyan, respectively. (c) Structure alignment of both SBPs.

Initial attempts to express MusE1345 SBP in E. coli DH5α cells with pQE-80L as the vector were unsuccessful; however, protein expression was successful in E. coli BL21(DE3) using pET41a^+^ as the expression vector. The expressed protein has N-terminal His6 and GST tags (Fig. S5), and thus, the protein size before cleaving off the tags was ~75 kDa. SEC-MALS results confirmed that the purified MusE1345 (no GST tag) was monomeric with a molar mass of 50.56 kDa (Fig. S6 and Fig. S8). MusE1346 SBP was successfully expressed in E. coli DH5α using pQE-80L as the expression vector. The His6 tag was not removed after purification. SEC-MALS results confirmed the purified protein was monomeric with a molar mass of 44.38 kDa (Figs S7 and S8).

Data from CD experiments were analysed using BeStSel to identify protein secondary structure composition of purified SBPs. CD spectra of the proteins were similar (following the standard spectrum of α-helix) and indicated a low representation of misfolding/random coil (Fig. 3). MusE1345 contained 42% α-helix (23% regular, 19% distorted), 4% antiparallel (right-twisted only), no parallel, 16% turn and 39% others. MusE1346 contained 33% α-helix (16% regular, 17% distorted), 11% antiparallel (10% relaxed, 1% right-twisted), 5% parallel, 12% turn and 40% others. BeStSel defines the ‘Others’ category as having 3_10_-helix, π-helix, β-bridge, bend, loop/irregular and invisible regions in the structure. Based on composition analysis, both SBPs contained more α-helix structures than β-sheets.

CD readings of MusE SBPs measured over 197–280 nm in mdeg. Each line represents the average of three CD experiments of the same SBP (blue, MusE1345; red, MusE1346). Concentrations used were 2.57 µM for MusE1345 and 2.66 µM for MusE1346. Composition of secondary structures was analysed using BeStSel v1.3.230210.

MusE SBPs had the highest affinities to maltose, maltotriose and maltotetraose

Purified SBPs MusE1345 and MusE1346 were measured for their specificity and affinity to seven possible glycogen breakdown products [glucose, isomaltose, maltose (M2), maltotriose (M3), maltotetraose (M4), maltopentaose (M5) and maltohexaose (M6)] and lactose. All tested glycogen breakdown products have one to six monomers of glucose linked by an α-1,4 glycosidic bond, except isomaltose (two glucose monomers linked by an α-1,6 glycosidic bond). Lactose, a disaccharide of glucose and galactose with β-1,4 linkage, was included to investigate the importance of the linkage. Ligand binding properties of purified MusE1345 and MusE1346 to the chosen ligands were determined using ITC (Table 1).

Both MusE1345 and MusE1346 had the highest affinities (lowest dissociation constant, Kd) for the α-1,4-linked oligosaccharides M2, M3 and M4, with all Kd at µM level (10^−6^ to 10^−7^ M) though the ligand/protein ratio (n) was <1 for MusE1345. MusE1345 exhibited lower affinities for longer oligosaccharides, M5 and M6, two to three logs different from affinities for M2–M4 (Kd=10^−4^ M). Affinities of MusE1346 for these ligands were also lower, about four orders of magnitude different (Kd=10^−3^ to 10^−4^ M) with n=1, despite ligand concentrations of M5 and M6 being 50 times higher than those of the shorter oligosaccharides (15 mM vs. 300–450 µM). Standard deviation values of the average of triplicate reactions of M4 and M6 and MusE1345 were high because there was one outlier in each triplicate. There was no quantifiable binding of lactose, glucose or isomaltose to MusE1345 and MusE1346. The heat signal in response to the injection of these ligands into the reaction cell containing MusE1345 or MusE1346 was similar to that of the ligand-into-buffer run (used as control files).

For the ligands with quantifiable Kd values, all of the ITC heat signals indicated exothermic reactions for both MusE SBPs (negative ΔH). The saturation curves of MusE1345 with M2–M6 had a similar trend, with most protein molecules being saturated within the first few ligand injections of ITC runs. MusE1346 interacting with M5 and M6 showed more gradual, non-sigmoidal saturation curves than MusE1345, although this protein exhibited sigmoidal saturation curves for M2, M3 and M4. Nineteen different combinations of MusE1345 concentration (2–144 µM) and ligand concentrations (0.002–15 mM) were tested during ITC optimization; however, there was no improvement to MusE1345 saturation curves, with all of them being non-sigmoidal. Nineteen different combinations of MusE1346 concentration (10–100 µM) and ligand concentration (0.25–15 mM) were also tested with the only combination showing improvement to the saturation curve being 3 mM M5 into 20 µM of MusE1346. Protein concentration was kept at 45 µM to facilitate direct comparison of binding affinities to different ligands of the two MusE SBPs.

We compared the interactions of MusE1345 and MusE1346 with individual ligands that had Kd values (Fig. 4). At the same concentrations of protein and ligand, MusE1346 produced a higher heat change for four out of five ligands, except M3, where the first injection of ligand into MusE1345 released more heat than for MusE1346. The heat signals (saturation curves) of the two proteins for each ligand were different despite having similar Kd values.

ITC exothermic saturation curves of MusE1345 and MusE1346 to maltose, maltotriose, maltotetraose, maltopentaose and maltohexaose in 20 injections (blue, MusE1345; red, MusE1346). Corrected heat is the heat signal after subtracting the average heat signal from the ligand-into-buffer run.

Discussion

Overgrowth of Gardnerella species is associated with dysbiosis in the human vaginal microbiome, leading to bacterial vaginosis in reproductive-aged women [2021]. Co-colonization of different Gardnerella spp. in individuals is common, though the number of species and their relative abundances vary, with G. vaginalis and G. swidsinskii being the most abundant [2]. Resource-based scramble competition has been identified as a contributor to Gardnerella population dynamics [3]. Since glycogen and its breakdown products are important carbon sources for vaginal bacteria, including Gardnerella, learning how they utilize these nutrients helps broaden knowledge of bacterial survival and competition mechanisms in the vaginal microbiome [2224]. The MusEFGKI system for carbohydrate uptake encodes two SBPs in G. swidsinskii (MusE1345 and MusE1346), while it encodes only one SBP (orthologous to MusE1346) in other Gardnerella spp*.* [9]. Operons encoding components of maltose and malto-oligosaccharide uptake systems usually have one SBP encoding gene, so it is unusual for G. swidsinskii to retain two SBP encoding genes [91025]. SBP provides specificity to maltose and malto-oligosaccharide transporters by binding to certain ligands in the environment; thus, having two distinct SBP encoding genes may provide G. swidsinskii with two SBPs that differ in ligand binding properties and broaden the range of ligands to the species, conferring competitive advantages over other Gardnerella spp. in accessing glycogen breakdown products. Another feature of the G. swidsinskii operon is the lack of a nucleotide-binding domain encoding gene (Fig. S4). This has been observed in ABC transporter operons in other species, including Bacillus subtilis and S. pneumoniae, where an ATPase encoded by a distant locus acts as the nucleotide-binding domain for maltodextrin importers [2627].

Transcripts from both G. swidsinskii SBP encoding genes, including a polycistronic transcript spanning both ORFs, were detected by PCR, confirming that both genes are expressed. Interestingly, the expression level of musE1346 was higher than that of musE1345 in G. swidsinskii isolates cultured in media supplemented with glycogen or M3 (Fig. 1), which was the opposite of what might be expected based on gene order within the operon. Further examination of the operon sequence showed the presence of multiple promoters, including an internal promoter between musE1345 and musE1346 (Fig. S4). The presence of primary and internal promoters within operons is common in bacteria and likely allows for compensation for premature terminations of transcription and flexibility in regulation and coordination of gene expression [28]. There was no difference in expression levels of the musE genes between glycogen and M3 culturing conditions (Fig. 1), which was perhaps not surprising since G. swidsinskii, like all Gardnerella spp., produces amylase enzymes for the release of maltose and malto-oligosaccharides from glycogen [7]. It is noteworthy that we identified uppS as a suitable reference gene for G. swidsinskii, which will be valuable in future gene expression studies in this organism. Further investigation of gene expression in additional conditions will be needed to determine mechanisms of the regulation of SBP gene expression in G. swidsinskii and whether it is governed by carbon catabolite repression as has been demonstrated for genes encoding glycolytic enzymes and transporter components in other species [2930].

The amino acid sequences of MusE1345 and MusE1346 are only 62% identical; however, the predicted structures were very similar, with two distinct domains (N and C domains) linked by a ligand-binding cleft, similar to previously characterized maltose and malto-oligosaccharides SBPs (Fig. 2) [3132]. A structural alignment between the two MusE SBPs and a well-characterized MalX from S. pneumoniae complexed with maltoheptaose (2XD3) [19] revealed three potentially important conserved ligand-binding site residues in MusE1345 and MusE1346, which make up the ‘aromatic cradle’ in S. pneumoniae MalX. The possession of a conserved SSDWRF motif occluding additional subsites identified in MalX may impose an upper limit on the size of oligosaccharide that can bind to MusE1345 and MusE1346 and may contribute to the reduced binding affinity for M5 and M6 (Table 1).

In Gram-positive bacteria like Gardnerella, SBPs are often tethered to the cell wall, unlike in Gram-negative bacteria, where the SBPs are free-floating in the periplasm [3334]. Tethering of proteins to the Gram-positive cell wall can occur through several mechanisms. Sortase enzymes recognize sorting motifs (e.g. LPxTG) in the C-terminus of a protein and covalently link the exported protein to peptidoglycan [3435]. Proteins containing distinct N-terminal signal peptides, including a characteristic terminal cysteine residue, are recognized by type II signal peptidases and covalently linked to lipids [36]. Alternatively, exported proteins may be anchored by terminal hydrophobic domains that embed in the cell membrane or non-covalent interactions such as electrostatic attraction [3738]. Both MusE proteins were predicted to be lipoproteins based on the identification of type II signal peptides.

Both G. swidsinskii MusE SBPs purified from E. coli were monomers, consistent with previously characterized prokaryotic SBPs, and suggesting that the ligand-to-protein binding ratio (n) should be n=1 [3941]. Secondary structure analysis using CD was performed to ensure that the purified proteins were not disordered, which could interfere with the assessment and comparison of ligand binding properties. CD results for MusE1345 and MusE1346 were consistent with the 3D predicted structures (Figs23). The CD spectra of both proteins followed a standard α-helix spectrum with a lack of disordered domains [4243], similar to previously characterized maltose and malto-oligosaccharide SBPs [4445].

We conducted ITC experiments with MusE1345 and MusE1346 at the same concentration (45 µM) in the same buffer to facilitate direct comparisons of their binding to lactose, glucose, isomaltose, maltose (M2), maltotriose (M3), maltotetraose (M4), maltopentaose (M5) and maltohexaose (M6). All selected ligands (except for lactose) are products of glycogen breakdown in the vaginal environment. M2–M6 are oligosaccharides with glucose subunits linked by α-1,4-glycosidic bonds. Lactose (glucose and galactose linked by a β-1,4-glycosidic bond) and isomaltose (two glucose subunits linked by an α-1,6-glycosidic bond) were included to test if the nature of the linkage affected binding to disaccharides. The combined observations of MusE SBPs strongly binding to M2 while having no binding to isomaltose or lactose indicated that the nature of the linkage (α-1,4-glycosidic bond) influences the binding ability of the proteins. Other studies on maltose and malto-oligosaccharide SBPs have also reported a lack of binding to these ligands [4647]. Certain members of the major facilitator superfamily or phosphoenolpyruvate-dependent phosphotransferase system families and RafEFGK of the ABC superfamily have been reported to uptake glucose and isomaltose, respectively, which may provide the binding function that the MusE system lacks [948].

MusE SBPs had the highest affinity (lowest Kd) for M2–M4, two of which (M2 and M3) are the most abundant oligosaccharides in vaginal fluid and the primary products released from glycogen by Gardnerella amylase enzymes [749]. The lowest order of magnitude of Kd was 10^−7^ M for M4 (MusE1345) and M2–M4 (MusE1346), which is consistent with previous reports for maltose and malto-oligosaccharides binding proteins [1950]. This sub-µM Kd has been described as ‘high’ by other studies quantifying affinities of SBPs, although the exact criteria for this designation are not well-defined [195152].

Despite the overall similarity in Kd values across the ligands tested, the saturation curves obtained for protein–ligand interaction in the ITC experiments varied substantially. The assumption of n=1 for monomeric protein did not apply to MusE1345 interacting with M2–M4 as the analysis software failed to calculate Kd at n=1 but was able to do so at n=0.1–0.2 (Table 1). It is important to note that the unusual n values required for curve fitting are likely not biologically meaningful in terms of stoichiometry. Having Kd at 10^−6^ M and n=0.1–0.2 resulted in a non-sigmoidal curve as the Wiseman value (c) <10 and outside the ideal range of 10<c<100 (Fig. 4). A possible reason for n<1 is overestimating the concentration of ‘active’ protein in the reaction cell. Both active and non-active proteins can exhibit the same CD spectrum in the far-UV region because of similarities in secondary structures, even though non-active proteins may lack higher-order conformation (e.g. tertiary) necessary for ligand binding [5354]. An overestimation of available protein for binding could also explain why most MusE1345 molecules were saturated in just a few ligand injections (Fig. 4). To achieve c within the ideal range for lower affinity interactions (Kd~10^−4^ M), protein concentration in the reference cell must be ~1 mM, which is more than 20 times higher than what we were able to achieve. There was little improvement in the shape of the saturation curves, with most of the proteins in the cell being saturated in the first few ligand injections, despite many ITC experiments involving different combinations of protein and ligand concentrations attempted. Regardless of the non-sigmoidal saturation curves observed in our study, our primary objective was to have the same ITC conditions (protein and ligand concentrations) to facilitate direct comparison between MusE1345 and MusE1346, and thus, further testing of different protein and ligand concentrations was not pursued.

The final ITC conditions that we selected for comparing ligand binding properties of the two SBPs were based on MusE1346 binding to M2–M4, which were the only reactions that resulted in sigmoidal curves (Fig. 4). Thus, protein concentration (45 µM), ligand concentration, injection numbers, time intervals between injections and other experimental parameters from the MusE1346 experiments were applied in the MusE1345 ITC runs. The interaction of MusE1346 with M5 and M6 at 300–450 µM produced little heat change, so ligand concentrations for M5 and M6 were increased to 15 mM, at which level Kd values were calculable (Table 1). The observation that saturation curves of the two MusE proteins with individual ligands differ from each other despite being assayed under the same in vitro conditions may be a result of differences in the binding site configuration or amino acids that directly interact with the ligands [5556]. Despite the different saturation curves, Kd values for all ligands were similar between MusE1345 and MusE1346, indicating that in these in vitro conditions, the affinities of the two SBPs do not differ substantially.

Conclusion

Based on our observations, genes encoding two distinct MusE SBPs are expressed at different levels in G. swidsinskii. The two G. swidsinskii MusE SBPs expressed and purified from E. coli have similar binding affinities to M2–M6 oligosaccharides despite having different saturation curves. Both SBPs have the highest affinities for M2 and M3, the most abundant glycogen-derived oligosaccharides in vaginal fluid. While our results do not support the hypothesis that the two SBPs have different preferred ligands, having two SBPs with similar affinities to the same oligosaccharides may still be advantageous since G. swidsinskii may uptake nutrients at a faster rate and have a competitive advantage over other Gardnerella spp.

Supplementary material

10.1099/mic.0.001685Supplementary Material 1.