Glycerol metabolism supports oral commensal interactions

Abstract

During oral biofilm development, interspecies interactions drive species distribution and biofilm architecture. To understand what molecular mechanisms determine these interactions, we used information gained from recent biogeographical investigations demonstrating an association of corynebacteria with streptococci. We previously reported that Streptococcus sanguinis and Corynebacterium durum have a close relationship through the production of membrane vesicle and fatty acids leading to S. sanguinis chain elongation and overall increased fitness supporting their commensal state. Here we present the molecular mechanisms of this interspecies interaction. Coculture experiments for transcriptomic analysis identified several differentially expressed genes in S. sanguinis. Due to its connection to fatty acid synthesis, we focused on the glycerol-operon. We further explored the differentially expressed type IV pili genes due to their connection to motility and biofilm adhesion. Gene inactivation of the glycerol kinase glpK had a profound impact on the ability of S. sanguinis to metabolize C. durum secreted glycerol and impaired chain elongation important for their interaction. Investigations on the effect of type IV pili revealed a reduction of S. sanguinis twitching motility in the presence of C. durum, which was caused by a decrease in type IV pili abundance on the surface of S. sanguinis as determined by SEM. In conclusion, we identified that the ability to metabolize C. durum produced glycerol is crucial for the interaction of C. durum and S. sanguinis. Reduced twitching motility could lead to a closer interaction of both species, supporting niche development in the oral cavity and potentially shaping symbiotic health-associated biofilm communities.

Introduction

The distribution of species in space and time is of particular importance for the understanding of ecological niche development. While a comprehensive theoretical frame work has been developed to understand the pattern of niche occupation [1], molecular-level ecological interaction studies to explain why a specific occupation pattern occurs are scarce. Bacterial interspecies interaction studies have been particularly helpful to expand our understanding of (bacterial) ecology and niche development. For example, competence system regulated interspecies antagonism of oral streptococci is met with counteroffensive measures by co-inhabiting streptococci and therefore is directly influencing species distribution and evolution [2–4]. Another recent investigation into the role of metabolic interactions in niche expansion using a synthetic bacterial model system demonstrated that synergistic interactions can expand the niche that is occupied beyond the limits of the fundamental niche [5]. This particular experimental study emphasizes the value of microbial ecological studies [5], providing less complex but validated experimental outcomes able to expand our knowledge beyond theoretical frameworks.

Oral microbial genome sequencing and biogeographical studies provided the field with a unique position to evaluate ecologically relevant species interactions [6, 7]. Not only do we know what microbial species inhabit the oral cavity, we know how their relative abundance changes during dysbiosis when prevalent oral diseases like caries or periodontal disease develop [8, 9]. Together with information about the spatial organization gained with fluorescent in situ hybridization (FISH) studies on native plaque samples [7], we are now able to select specific species that inhabit a defined niche, for example the gingival margin, to characterize the molecular interactions that determine niche occupation patterns. Thus, we can address fundamental ecological questions, such as why a specific occupation of a niche occurs and how metabolic cross-feeding shapes interspecies interactions and affects community development. We have recently followed this experimental approach to learn how and why oral Corynebacterium species interact with streptococci [10]. The widespread Corynebacterium–Streptococcus association has been visualized with FISH at the subgingival margin and accompanying microbiome sequencing results identified both species as widely abundant oral commensal inhabitants [7, 11]. Their physical interaction in situ in so called corncob structures, with Corynebacterium forming long filaments originating from the gingival margin decorated with predominantly streptococci, is one of the few examples of microbial biogeography in the host (oral) setting [7]. This positioning of Corynebacterium in subgingival plague has led to the development of a new species-association map of this densely populated oral niche, which included both FISH biogeographical interaction information and certain significant metabolic interactions, such as the secretion of hydrogen peroxide [6, 7, 11, 12].

We previously identified specific molecular interactions between Corynebacterium durum and Streptococcus sanguinis, both prominent members of the Corynebacterium–Streptococcus association. C. durum is able to secrete membrane vesicles (MVs) that contain specific fatty acids. S. sanguinis is able to respond to the MVs with a severe chain elongation phenotype, increasing in length from about 8 to 160 µm. This response was specific for S. sanguinis and established MVs as mediators of such unique interspecies interactions [10]. No other oral Streptococcus species examined exhibited this elongation phenotype. Subsequent investigations revealed that S. sanguinis responds to C. durum lipids by decreasing the expression of key FASII genes involved in fatty acid synthesis. Several of these FASII genes were determined to be essential for the chain elongation phenotype. In addition, C. durum was found to enhance S. sanguinis fitness, induce cell aggregation, and interfere with S. sanguinis phagocytosis, revealing a complex association of these species that likely supports oral commensal colonization and survival [10]. Here, we further investigated the regulatory connection between lipid metabolism and chain elongation. Global gene expression profiling of dual species cultures revealed a complex association between both species that was later found to be mediated by the S. sanguinis glycerol kinase GlpK as well as its type IV pili. These results provide new insights into S. sanguinis–C. durum synergism, which may further reveal the molecular strategies used to support their niche occupation.

Materials and methods

Bacterial strains, plasmids, cell culture and media

Bacterial strains and plasmids are listed in Supplementary Table 1. Bacterial cells were routinely grown aerobically as static cultures (5% CO₂) at 37 °C in Brain Heart Infusion medium (Bacto BHI; Becton Dickinson & Co., Sparks, MD, USA) or on BHI agar plates as well as in an anaerobic chamber (90% N₂, 5% CO₂, 5% H₂) when indicated. BHI without glucose (United States Biological, Swampscott, MA, USA) or chemically defined artificial saliva solution (ASS) medium [13, 14] was used to culture bacterial strains when indicated. Saliva supplemented with 100 mM glucose [10] was freshly prepared when indicated. Individual and cocultured bacterial cell suspensions were prepared as described previously [10]. Escherichia coli (E. coli) DH10B was cultured aerobically at 37 °C in Luria-Bertani medium (LB; Lennox, Becton Dickinson & Co.) with agitation at 200 rpm. The following antibiotics were purchased from Sigma-Aldrich (St. Louis, MO, USA) and used when required: spectinomycin (500 μg/ml for S. sanguinis strain SK36 (SK36); 100 μg/ml for E. coli), erythromycin (5 μg/ml for SK36). The murine macrophage-like RAW 264.7 cells were purchased from the American Type Culture Collection (ATCC TIB-71TM) and maintained at 37 °C in Roswell Park Memorial Institute (RPMI) 1640 Medium supplemented with 10% (vol/vol) heat-inactivated fetal calf serum (GE Healthcare, WI, USA) in atmosphere of 5% CO₂ and 95% humidity. RAW 264.7 cells were continuously cultured until passage no. 15 [15].

DNA manipulations

Standard recombinant DNA manipulations were used [16]. Restriction enzymes and other molecular biology reagents were purchased from New England Biolabs (Beverly, MA, USA), Life Technologies (Grand Island, NY, USA), Promega Life Science (Madison, WI, USA), and used according to the manufacturer’s instructions. Primers are listed in Supplementary Table 2. PCR products were purified using the Wizard SV gel and PCR clean-up system (Promega). All plasmids were extracted using the Wizard plus SV minipreps DNA purification system (Promega).

Construction of mutant strains

For the construction of mutant strains via allelic exchange, an overlapping PCR strategy was used as previously described [17]. Complementation was achieved using the respective glpK open reading frame with native promoter cloned into plasmid pDL278 [18, 19]. Details can be found in Supplementary Information.

RNA isolation, cDNA synthesis, and RT-PCR

RNA extraction was performed as described previously [10, 17]. RNA library construction and sequencing reactions were conducted at GENEWIZ Inc (South Plainfield, NJ). cDNA synthesis and RT-PCRs were carried out for further analysis. Details can be found in Supplementary Information.

Glycerol assay analysis

EnzyChrom Glycerol Assay Kit (BioAssay Systems, Hayward, CA) was used to assess the amount of glycerol under various conditions. Supernatant samples were prepared as previously explained [10]. In brief, bacterial supernatants collected from overnight cultures were filtered through 0.45-μm–pore size cellulose acetate filters (VWR International). The supernatants were concentrated using Amicon Ultra Centrifugal Filters (30 kDa MWCO, Sigma-Aldrich) and then proceeded to the enzymatic assay according to the manufacturer’s instructions. Data from biological triplicates were then analyzed.

Surface swarming/twitching motility assays

Bacterial motility was macroscopically observed on agar plates as previously described with minor modifications [20, 21]. C. durum JJ1 (Cd) supernatant was prepared [10] prior to being mixed with agar for preparing 1% (final concentration) agar plates. To further investigate bacterial swarming/twitching motility under iron stimulation [22], defibrinated sheep blood (Thermo Fisher) was used to substitute for Cd supernatant in order to prepare 1% (final concentration) agar plates. Bacterial strains were cultured overnight, washed with PBS, and inoculated on 1% agar plates. The plates were then incubated in anaerobic conditions for 10 days. The “twitching/swarming” motility of the bacteria was examined and imaged.

Phagocytosis assay in murine macrophage-like RAW 264.7 cells

The assay essentially followed a previously described protocol [10]. In brief, murine macrophage-like RAW 264.7 cells were plated at 1.0 × 10⁵ cells/well in 24-well plates and preincubated overnight at 37 °C in 5% CO₂. Overnight bacterial cells were pelleted and washed with phosphate-buffered saline (PBS) prior to adjusting the OD₆₀₀ to 0.6 (equivalent to approximately 10⁸ CFU/ml) using RPMI 1640 supplemented with 10% (v/v) heat-inactivated fetal bovine serum as a diluent. Cocultured bacterial cell suspensions were prepared by coculturing Cd with the wild-type SK36, ΔglpK, and ΔglpKc strains in BHI medium at 37 °C in 5% CO₂ until the OD₆₀₀ was approximately 0.6. For investigating the effect of glycerol on phagocytosis, bacterial strains were cultured in BHI without glucose supplemented with 10% (v/v) sterile glycerol (+ Glycerol) at 37 °C in 5% CO₂ until the OD₆₀₀ was approximately 0.6. BHI without glucose and glycerol (− Glycerol) was used as control. Cells were then pelleted, washed, and resuspended as described earlier [10]. Bacterial uptake was subsequently allowed for one hour prior to being extensively washed with PBS and then inoculated into fresh medium supplemented with penicillin–streptomycin (MP Biomedicals, CA) for 2 h at 37 °C in 5% CO₂ in order to kill extracellular bacteria. RAW 264.7 cells were lysed with 0.1% Triton X-100 and lysates dispersed by vigorous pipetting and vortexing. Serial dilutions of each sample were plated for counting viable bacteria (CFU/ml).

Scanning electron microscopy (SEM)

Bacterial samples were prepared and fixed as published previously with some modifications [10]. In brief, bacterial strains were cultured as both single and mixed cultures with Cd on ITO coated cover slips overnight anaerobically for visualizing pilus structures [17, 23]. Samples were fixed and then sputter coated with 10-nm thick carbon (ACE600 coater). Imaging was performed using a Helios Nanolab 660 dual-beam scanning electron microscope (FEI).

Statistical analysis

The statistical significance of the difference between experimental groups was determined by Student’s t-test (2-tailed). p values less than 0.05 were considered significant.

Results

Effect of C. durum on the transcriptome of S. sanguinis

The metabolite-driven interspecies interaction between S. sanguinis SK36 (SK36) and C. durum JJ1 (Cd) has been previously established [10]. To confirm and visualize the interaction between Cd and SK36 fluorescent microscopy was performed. Intermingled aggregates suggest direct physical contact between Cd and SK36 (Fig. 1, merged; white arrows) verifying interspecies interactions in vitro. To further explore the interactions between these core oral commensals, we investigated the Cd influenced transcriptional profile of SK36 using RNA‐seq. Hierarchical clustering analyses revealed a total of 30 genes as being differentially expressed (adjusted p < 0.05) when SK36 was grown with Cd supernatant (Fig. 1C). Cd supernatant is able to induce metabolite-driven phenotypic responses in SK36, identical to what we reported for cocultures [10]. Of those 30 genes, genes involved in transcriptional mechanisms and RNA/DNA biosynthetic network were found to be upregulated (Fig. 1C; Supplementary Fig. 1A). Genes involved in carbohydrate/sugar metabolic pathways, particularly, phosphoenolpyruvate (PEP): sugar phosphotransferase system (PTS), were also upregulated (Fig. 1C; Supplementary Fig. 1B). Consistent with our previous work [10], the expression of SK36 gldA (SSA_0287) and its neighboring gene fructose-6-phosphate aldolase (SSA_0286) was significantly increased in Cd supernatant-treated cells, confirming a role for gldA in SK36-Cd interspecies interaction (Fig. 1D). Similarly, glycerol kinase (glpK; SSA_1826) and its neighboring downstream genes, type 1 glycerol-3-phosphate oxidase (glpO; SSA_1827) and a putative glycerol transporter from the aquaporin family (glpF; SSA_1828) were substantially upregulated after SK36 was grown with Cd supernatant (Fig. 1D). In contrast, the putative pilin gene cluster of SK36 (from SSA_2307 to SSA_2318) was significantly downregulated in the Cd supernatant-treated condition (Fig. 1E). All of the genes in this cluster are predicted to encode the characteristic features of type IV pilin proteins, including a class III leader peptide and a conserved N-terminal motif, which form an α-helix structure [21, 22]. Aside from this, SK36 SSA_0642, a gene predicted to function as pre-pilin signal peptidase during cell motility, intracellular trafficking, and vesicular transport [21, 24], was found to be downregulated when grown in Cd supernatant (data not shown). Taken together, the expression changes in key genes associated with glycerol metabolism are in agreement with our previous published results indicating a major role of glycerol and its associated metabolism in the dual interspecies relationship between both species. Further, changes in the expression of pilin genes suggest that C. durum can directly influence S. sanguinis attachment or motility, thus potentially influencing its localization inside the oral biofilm.

Fig. 1. Interspecies interaction of S. sanguinis and C. durum on the phenotypic and genotypic level.

Representative immunofluorescence microphotographs of SK36 (GFP) cocultured with (A) DAPI or (B) CellTracker Red CMTPX fluorescent dye-labeled C. durum (Cd) in saliva supplemented with 100 mM glucose. Interspecies coaggregation was shown in the merged images by white arrows. Scale bars indicate A 20 μm and B 10 μm. C Volcano plot demonstrates the overall transcriptional changes across Cd supernatant-treated SK36 (red) in comparison to the medium control (blue) condition. Each data point represents one gene. The X-axis represents the log₂ fold change of each gene and the Y-axis represents the log₁₀ of its adjusted p value. Genes with an adjusted p < 0.05 and log₂ fold change > 1 (indicated in red dots) show upregulation. Genes with p < 0.05 and log₂ fold change < −1 (indicated in blue dots) show downregulation. Hierarchical clustering analysis and heatmap of SK36 genes transcriptomic profiles responses to Cd supernatant in comparison to the medium control (n = 3). The top genes involved in D lipid and glycerol metabolism and E pilus biosynthesis were upregulated and downregulated, respectively, in the presence of Cd supernatants. The color gradient in D and E indicates the range of gene expression levels within a given sample. Red represents low expression levels, yellow intermediate, and green represents high expression levels.

Glycerol consumption of S. sanguinis via GlpK pathway

In Bacillota species (Firmicutes), GlpK has been extensively studied as an essential enzyme used for both glycolysis and lipid biosynthesis [25–28]. Nonetheless, to our knowledge, a role of GlpK in bacterial interspecies interactions has never been reported. Analysis of our RNA-seq data (Fig. 1D) led us to speculate that glpK (SSA_1826) and genes in the cluster (SSA_1825, glpO (SSA_1827) and glpF (SSA_1828)) could play an important role in the C. durum – S. sanguinis interspecies interaction, possibly in a similar manner to gldA (SSA_0287) as previously demonstrated [10]. We first examined if Cd supernatant contains glycerol prior to investigating if SK36 was able to utilize glycerol when grown in Cd supernatant. Cd supernatant was prepared and SK36 was treated with the supernatant overnight as described previously [10], and the glycerol concentration was measured. The glycerol concentration in Cd supernatant with SK36 (Cd sup + SK36) was significantly reduced when compared to Cd supernatant control without SK36 (Cd sup), (5.91 mM vs 9.47 mM, p < 0.005; Fig. 2A). Further, strains SK408, SK1056, and VMC66, were examined regarding their glycerol utilization. Similar to SK36, all were capable of utilizing glycerol content in Cd supernatant at various degrees (5.91, 2.82, and 2.93 mM, respectively, p < 0.005; Fig. 2A). Next, we generated a model of the S. sanguinis GlpK protein structure by threading the sequence onto staphylococcal GlpK (PDB: 3G25) using SWISS-MODEL (100% coverage, 63% sequence identity) and observed an identical overall topology, including the conserved phosphorylation site histidine 232 (H232) and surrounding residues (Fig. 2B; Supplementary Fig. S2) [26, 28, 29].

Fig. 2. Glycerol consumption of S. sanguinis.

A Total glycerol production of Cd and the unique ability of S. sanguinis SK36, SK408, SK1056, and VMC66 strains to utilize the glycerol being produced by Cd. Cd was cultured in BHI overnight and its supernatant was collected and filtered through 0.45 μm membrane. SK36 was then cultured in the Cd supernatant overnight. Both samples were subsequently measured for total glycerol concentration. Results shown are the average of biological triplicates. Error bars represent the standard deviation. B Ribbon diagram of SK36 glycerol kinase (GlpK) with the β-strands shown as arrowed ribbons, α-helices, and the connecting loops. Colors indicate the N-terminus (blue) to the C-terminus (red). The histidine 232 (H232) phosphorylation site is marked in magenta.

A role of glpK in the C. durum-S. sanguinis interspecies interaction

The unique effect of C. durum fatty acid-containing membrane vesicles on streptococcal chain length has been previously reported [10]. Here, we investigated whether we could repeat this unique phenotype with glycerol given that glycerol appears to be the backbone of relevant lipid molecules [30–32]. In order to prevent any glucose interference on GlpK activities [26, 28], BHI without glucose and glycerol (- Glycerol) was used as the base culture medium, and 10% (v/v) sterile glycerol (Supplementary Fig. S7) was supplemented in the base medium as the + Glycerol condition. Glycerol was able to induce S. sanguinis chain elongation to the same degree (Fig. 3A), approximately 10 times longer (p < 0.0005; Fig. 3B), as Cd supernatant [10]. Due to our RNA-seq data (Fig. 1D), we next investigated if glpK is involved in the observed SK36 chain elongation phenotype. A deletion mutant of SK36 glpK (ΔglpK) was generated (Supplementary Fig. S3) and its chain morphology was examined as described previously [10]. Unlike the wild-type SK36, the SK36 ΔglpK chain length was almost identical irrespective of glycerol supplementation (Fig. 3). However, Cd supernatant (+ Cd sup) remained able to induce the ΔglpK chain elongation, suggesting that the glycerol pathway can be bypassed via consumption of fatty acids (Supplementary Fig. S4C). We next verified the results of the SK36 ΔglpK strain by generating the complemented strain (SK36 ΔglpKc). The SK36 ΔglpKc chain length was similar to the wild-type strain when grown under the + Glycerol conditions (p < 0.0001; Fig. 3A, B).

Fig. 3. Effect of glycerol on SK36 chain morphological alteration via the glpK metabolic pathway.

SK36 wild-type (SK36), ΔglpK, the complemented mutant (ΔglpKc), and its single point mutant H232E (glpKHE) strains were treated with the–Glycerol and + Glycerol conditions. A Chain morphologies of the SK36, ΔglpK, glpKHE, and ΔglpKc were then examined and imaged using an Olympus IX73 inverted microscope. The pictures are representative of three independent experiments. Scale bars indicate 10 μm. B For quantification, bacterial chain length was measured from 100 bacterial chains per biological replicate using ImageJ software. Results shown are the average of biological triplicates. Error bars represent the standard deviation.

GlpK contains a range of conserved binding sites responsible for specific roles of the enzyme [25, 27, 33]. Of all of the multiple conserved residues, histidine 232 (H232) is known as the phosphorylation site located in its activation loop (Fig. 2B; Supplementary Fig. S2) [26, 28], indicating a key role of the H232 site in GlpK enzyme activity [28, 34, 35]. Amino acid substitution, in particular, from histidine to glutamate (H232 → H232E) has been shown to negatively affect the GlpK functions, similar to the deletion mutation [25]. Hence, to further characterize, site-specific mutagenesis was conducted to create a GlpK H232E mutant (glpKHE; Supplementary Fig. S3). Subsequently, the SK36 glpKHE strain was grown under the + Glycerol and compared to the–Glycerol conditions, and bacterial chain length was examined. Similar to the ΔglpK, the glpKHE strain did not exhibit chain elongation in the presence of glycerol (Fig. 3A, B). This finding verified that the H232 residue is responsible for SK36 GlpK activity. Aside from glpK, the genes in the same cluster SSA_1825, glpO (SSA_1827), and glpF (SSA_1828) (Fig. 1D) were further characterized in the same manner as the ΔglpK strain. While the + Glycerol condition had no impact on the ΔSSA_1825 chain morphology as the mutant remained able to elongate its chain length (p < 0.005), both ΔglpO and ΔglpF strains were significantly affected (Supplementary Fig. S4A, B), approximately to the same degree as the ΔglpK strain. This result suggested an involvement of both downstream genes in the glycerol-induced chain elongation in SK36.

Glycerol kinase directly functions in SK36 glycerol utilization

To directly confirm GlpK dependent glycerol catabolism in SK36, we determined glycerol consumption in BHI medium without glucose. This medium was chosen to avoid any potential effect of catabolite repression on glycerol metabolism as reported for other related species [25, 28, 35–38]. Glycerol and glucose are both carbon sources that feed into the central metabolism and fatty acid synthesis [39–41]. Glucose is considered a preferred carbohydrate source for S. sanguinis and exerts carbon catabolite repression [18, 42, 43] and a potential CcpA binding side (cre) is present upstream of glpK (data not shown). BHI without glucose (- Glucose) was used as a base medium to grow the SK36 wild-type, ΔglpK, and ΔglpKc strains overnight, and the glycerol concentration (mM) was then measured as an indicator of bacterial glycerol consumption. The SK36 wild-type strain was able to utilize glycerol in the - Glucose condition (p < 0.0001; Fig. 4). In contrast, the ΔglpK strain was unable to significantly reduce the glycerol concentration, confirming an involvement of glpK in bacterial glycerol consumption (p < 0.0005; Fig. 4). In order to verify the result in the ΔglpK strain, the complemented strain (ΔglpKc) was tested for its ability to utilize glycerol during growth. Similar to the wild-type strain, the ΔglpKc strain showed a reduction of glycerol content to almost to the same degree as the wild-type strain (p < 0.0001; Fig. 4). While comparing the consumption of glycerol from cells grown in − Glucose vs − Glucose supplemented with glucose (+ Glucose), we only detected a moderate catabolite repression effect of glucose. Compared to the depletion of glycerol in − Glucose (approximately sevenfold; p < 0.0001; Fig. 4), the reduction in + Glucose was still approximately fourfold (p < 0.0005; Fig. 4). Taken together, these findings confirmed the direct involvement of GlpK and a weak catabolite repression of SK36 glycerol utilization.

Fig. 4. Total glycerol production under the influence of glucose.

SK36, ΔglpK, and the complemented mutant (ΔglpKc) were cultured overnight in BHI with glucose (+ Glucose) compared to BHI without glucose (− Glucose). The supernatants were collected, filtered through a 0.45 μm membrane, and then measured for total glycerol concentration (mM). Results shown are the average of biological triplicates. Error bars represent the standard deviation.

Involvement of SK36 glpK in macrophage phagocytosis

Given that glpK exhibited an obvious change in the RNA-seq results (Fig. 1D) and directly impacted SK36 chain morphology, we further investigated if glpK could also be involved in interfering with host phagocytosis as we had previously observed for SK36 grown in the presence of Cd supernatant [10]. RAW 264.7 cells were prepared and subsequently challenged with the mixed cultures of SK36 and Cd in comparison to the single cultures. A bacterial uptake assay was then conducted in order to enumerate the number of internalized bacteria (CFU/ml) as described previously [10]. In the absence of Cd (− Cd), the number of internalized SK36 wild-type was significantly lower when compared to the ΔglpK strain (p < 0.05; Fig. 5). Further, the number of phagocytosed ΔglpK cells increased approximately 5 times when compared to the SK36 wild-type strain in the presence of C. durum (+ Cd) ((p < 0.001; Fig. 5). Similar to what we reported previously [10], cocultures of the SK36 wild-type with Cd (+ Cd) led to a significant drop in internalized SK36 cells, approximately 4 times, when compared to the single cultured SK36 (− Cd). To verify these findings, the complemented ΔglpKc strain, as a single (− Cd) and mixed culture (+ Cd) with Cd, was used to challenge RAW 264.7 cells. The ΔglpKc strain exhibited similar numbers of internalized bacteria (CFU/ml) when compared to the wild-type strain in both conditions (Fig. 5). These findings revealed a strong connection between S. sanguinis glycerol metabolism via glycerol kinase and inhibition of macrophage phagocytosis. To further examine if glycerol could potentially mimic the effect of C. durum on interfering with bacterial uptake and phagocytosis, the + Glycerol condition was used to culture SK36 wild-type and ΔglpK strains with—Glycerol as control. Subsequently, all the bacterial samples were used to challenge RAW 264.7 cells, and internalized bacteria were enumerated. The SK36 wild-type number (CFU/ml) in the + Glycerol condition revealed a significant decrease, approximately 6 times, in comparison to the–Glycerol condition (p < 0.0001; Supplementary Fig. S8). However, there was no substantial difference in the internalized ΔglpK strain irrespective of glycerol supplementation (Supplementary Fig. S8). Furthermore, the fold difference of internalized bacteria between the SK36 wild-type and ΔglpK strains in the + Glycerol condition was approximately 6 time while it was twice in the – Glycerol control condition (p < 0.0005; Supplementary Fig. S8). These findings suggested that glycerol, similar to C. durum, can interfere with bacterial phagocytosis to a certain extent, possibly via the GlpK pathway.

Fig. 5. Effect of glpK in bacteria-phagocytic interactions.

Bacterial uptake was determined by challenging murine macrophage-like RAW 264.7 cells with the wild-type SK36, ΔglpK, and the complemented ΔglpK strains with (+ Cd) and without (− Cd) Cd. The number of internalized bacteria (CFU/ml) in RAW 264.7 cells then enumerated. Data are presented as the means of biological triplicates. Error bars denote standard deviations.

C. durum affects the twitching/swarming motility of S. sanguinis

Our RNA-seq data showed a significant downregulation of the S. sanguinis type IV pilus biosynthetic and functional gene cluster in the presence of C. durum (Fig. 1E). Type IV pili are known to be important in a wide range of bacterial functions including adhesion, biofilm formation, host cell interactions and translocation mechanisms such as twitching motility and swarming [44–46]. While Gram-negative bacterial pili have been extensively studied over decades [45], information about pilus gene clusters in Gram-positive bacteria remains scarce [20–22, 24, 47]. Strains in the Mitis group, including S. sanguinis and S. pneumoniae, seem to harbor at least one type of pilus islets, called pilus islet 1 (PI-1) and/or 2 (PI-2) [48, 49]. Recently, the unusual characteristics of S. sanguinis type IV pili have been reported to play a role in bacterial twitching motility, bacterial adhesion, and biofilm formation in a strain-specific manner [21, 22, 50, 51]. Two pilus proteins PilE1 and PilE2 appear to play a key role in processing and assembling bacterial filaments of S. sanguinis [24]. Our RNA-seq data suggested a putative role of SK36 type IV pilus genes during their interspecies interaction with Cd. Hence, it was of interest to investigate if the presence of Cd could affect SK36 twitching/swarming motility phenotypes. Bacterial suspensions were tested on 1% agar plates containing Cd supernatant, and the plates were cultured as described previously [21, 24, 51]. Since twitching/swarming motility has been reported to be strain-dependent [49, 52, 53], we also tested other S. sanguinis strains in parallel [54–57]. No visible difference was observed for S. sanguinis SK36 motility appearance on both Cd supernatant (+ Cd; Fig. 6B) and BHI negative control plates (− Cd; Fig. 6A). Of all the S. sanguinis strains, only strain SK408 showed a substantial reduction of twitching/swarming motility appearance on the + Cd supernatant plate (Fig. 6B).

Fig. 6. Differences in surface twitching/swarming motility phenotypes of S. sanguinis SK strains under the influence of Cd.

All S. sanguinis SK strains were prepared and spotted on 1% agar plates containing (B) Cd supernatant (+ Cd) in comparison to (A) BHI control (− Cd) plates. C Two key SK strains (SK36 and SK408) were further investigated twitching/swarming motility phenotypes on 1% agar plates supplemented with 5% defibrinated sheep blood. The plates were then incubated anaerobically. Images are representative of three independent experiments. White asterisks indicate a visible halo or hazy zone of bacteria that have twitched/swarmed across the plate. Full images of the plates are exhibited in Supplementary Fig. S5.

An effect of iron on S. sanguinis SK36 twitching motility has been recently demonstrated, indicating a regulatory signal of iron on type IV pilus function [22]. Thus, given the findings above, any potential effect of iron on SK36 and SK408 twitching/swarming motility was further examined. Twitching/swarming motility on 1% agar plates supplemented with defibrinated sheep blood (ThermoFisher) to a final concentration of 5% were conducted as described previously [22]. Similar to the Cd supernatant condition (+ Cd; Fig. 6B), SK36 showed minimal evidence of motility on 5% blood agar plates (Fig. 6C; Supplementary Fig. S5). In contrast, a single culture of the SK408 stain (SK408) exhibited an outward expanding colony phenotype (white asterisks; Fig. 6C; Supplementary Fig. S5) consistent with its phenotype on BHI plates (− Cd; Fig. 6A). This motility phenotype was significantly affected by Cd during simultaneous coculture with Ss and when Cd was added after Ss. (Fig. 6C; Supplementary Fig. S5). While we did confirm the inhibition of SK408 motility by Cd, we could not confirm any role of iron on twitching/swarming motility under the here used conditions.

Effect of C. durum on type IV pilus gene expression

Following the surface motility results, in particular, of strain SK408, we next investigated and compared the type IV pilus gene cluster between several phylogenetically separated S. sanguinis strains. S. sanguinis strain SK36 is commonly used as reference S. sanguinis strain to investigate the ecology and physiology of S. sanguinis [19, 58]. Nonetheless, given S. sanguinis high variability in the motility assay, SK36 is unlikely to encompass the entire genotypic and/or phenotypic spectrum. Type IV pilus genes in all examined strains revealed significant similarities in terms of gene alignment and nucleotide sequence identities (Fig. 7A), thus four core pilus genes were chosen for further investigation, which showed nucleotide sequence identities between 85% to 100% (Supplementary Table 4). All strains were cultured in the presence and absence of Cd for subsequent cDNA synthesis as described previously [10, 17]. Afterwards, qRT-PCR was conducted using primers specific for the 4 core pilus genes (pilE1, pilE2, pilT, and pilB). Although SK36 lacked surface motility in all of the tested conditions, the expression of pilE2 (20-fold), pilE1 (15-fold), pilT (5-fold), and pilB (7-fold) was significantly decreased in the presence of Cd (P < 0.005; Fig. 7B). This result verified the RNA-seq data (Fig. 1E). In contrast to SK36, the expression of the SK408 orthologous pilus genes were increased when SK408 was cocultured with Cd (P < 0.05; Fig. 7C); while significant, the fold increase was less than twofold. Similar to SK36, the presence of Cd significantly downregulated the 4 core pilus genes in VMC66 (P < 0.001; Fig. 7E), while SK1056 was more comparable to SK408 with a 1.5-fold increase in pilE2 expression, but slightly lower expression of the other tested pilus genes (P < 0.05; Fig. 7D). These findings confirmed S. sanguinis strain heterogeneity, contributing to various phenotypic characterization even within the same S. sanguinis species. To further confirm differences in gene expression of the here tested pilus genes, we compared the expression among the strains. Pilus gene expression varies considerable suggesting that pilus genes have distinct regulation and heterogenous expression on the strain level of the species S. sanguinis (Supplementary Fig. S10).

Fig. 7. Comparison of type IV pilus gene expression in S. sanguinis SK strains ± Cd.

A Single-nucleotide polymorphism (SNP)-based neighbor-joining phylogram illustrating variations within core gene clusters that lack paralogs (modified from [54]). Core pilus genes of the selected SK strains (**) with their gene IDs and sequence identities (%) were listed in Supplementary Table 4. Relative core type IV pilus gene expression (pilE2, pilE1, pilT, and pilB) in B SK36, C SK408, D SK1056, and E VMC66 strains in the presence (+ Cd) versus the medium control (− Cd). Gene expression data are presented relative to the medium control (− Cd), which was arbitrarily assigned a value of 1. Data represent the means of biological triplicates. Error bars denote standard deviations.

Visualization of type IV pili on the surface of SK36 and SK408 in the presence compared to the absence of C. durum

Both RNA-seq and the relative pilus gene expression data from SK36 suggested that bacterial pili should be affected to some extent by the presence of Cd. However, there is a discrepancy between the observed pilus gene expression of SK408, which was slightly higher in the presence of Cd, and the lack of motility. Hence, to further investigate, scanning electron microscopy (SEM) was used to visualize the pili of SK36 and SK408 when they were cocultured with Cd (+ Cd) in comparison to their single culture controls (− Cd). Bacterial cells were grown and then prepared for SEM imaging as described previously [10, 17]. Both SK36 and SK408 showed the presence of a meshwork of abundant thin fibrillar structures protruding from the bacterial surface and networking with the neighboring cells (white arrows; Fig. 8; Supplementary Fig. S6A). In correlation to the RNA-seq data, a substantial reduction of pilus structure was observed when SK36 was cocultured with Cd (Fig. 8; Supplementary Fig. S6A). Similarly, Cd had a significant impact on the meshwork of pili of the SK408 strain (Fig. 8; Supplementary Fig. S6A). To further investigate an involvement of the core pilus genes (pilE1/E2, pilT, and pilB) in pilus formation [20, 21, 47, 51], the ΔpilE1/E2, ΔpilT, and ΔpilB deletions in SK36 and SK408 were constructed. Each of the mutants was then cocultured with Cd compared to their single culture controls prior to proceeding to SEM visualization along with the wild-type strains. All of the mutants in SK36 (Supplementary Fig. S6B) and in SK408 (Supplementary Fig. S6C) revealed no pilus formation in the presence or absence of Cd, indicating that the 4 core pilus genes in both SK36 and SK408 strains are responsible for the here observed pili formation on the surface of S. sanguinis. Thus, these findings confirmed an impact of Cd on pili formation of, at least, SK36 and SK408.

Fig. 8. Scanning electron micrographs (SEM) of type IV pilus morphology in the SK strains.

(Top left) SK36 and (Top right) SK408 single species cultures (− Cd). (Bottom left) SK36 and (Bottom right) SK408 dual species cocultures with Cd (+ Cd). Top left scale bar indicates 1 µm. Top right scale bar indicates 2 µm. Bottom left and right scale bars indicate 1 µm. White arrows indicate putative pili. The pictures are representative of three independent experiments. Full images are exhibited in Supplementary Fig. S6A.

Discussion

The human oral microbiome comprises a vastly complex and dynamic microbial ecosystem represented by hundreds of microbial taxa colonizing different habitats inside the oral cavity [7, 59]. Not only are these multispecies communities of oral microbes crucial in maintaining the dynamic balance of oral health and immunity, they are also important for systemic health homeostasis [59–61]. Due to the fact that we have a detailed understanding of species biogeographical distribution and that a vast number of members of this ecological system are cultivable, the oral biofilm can function to address important ecological questions. Here, we sought to further uncover the molecular principles that determine niche occupation of oral commensals to illuminate what governs species distribution and interspecies interaction. We tried to approach this question at least partially by looking at oral Corynebacteria. In general, Corynebacteria are associated with the human microbiome with clearly distinguishable behavior towards other microbiome members. For example, nasopharyngeal non-diphtheriae Corynebacteria can inhibit the growth of S. pneumoniae [62], while oral C. durum can support the growth and survival of oral commensal S. sanguinis [10]. The diverse streptococcal-corynebacterial interspecies interactions and the relative abundance of both species as part of the (oral) microbiome suggest a major role in microbiome ecology and specific niche development. Given that C. durum generates membrane vesicle associated with lipids during growth that influence S. sanguinis behavior [10, 63], we speculated that genes involved in S. sanguinis lipid metabolism, aside from the previously described gldA [10], play a major role in their interactions. Similar to SK36 gldA [10], three core genes in the glp operon, glpK, glpO, and glpF, were upregulated in the presence of C. durum JJ1. In many bacteria, these genes have been found to be crucial during lipid and glycerol catabolism and transport [29, 64–66]. Our glpK inactivation and glycerol consumption studies confirm that the glp operon in S. sanguinis is responsible for glycerol catabolism, but in addition it is playing a previously unknown role in the interspecies interaction between C. durum and S. sanguinis. The upstream gene (SSA_1825) directly preceding glpK (Fig. 1D), did not show any significant changes in the transcriptomic profile. However, genomic context and transcriptional orientation suggest that it might belong to the glp operon. The protein encoded by SSA_1825 includes a helix-turn-helix motif, suggesting a regulatory role for the glp-operon. This is speculative at this time, but interestingly, a dedicated transcriptional regulator like GlpR in Gram-negative bacteria has not been identified in Gram-positive bacteria [41], and future genetic studies will determine whether SSA_1825 serves that function.

Despite the fact that both glpK and gldA operons are upregulated in SK36-Cd cocultures, unlike the ΔgldA mutant [10], the ΔglpK strain was still able to elongate its chain length when grown in Cd supernatant (Supplementary Fig. S4C). Contrarily, when the ΔglpK strain was tested with glycerol as carbon source, the ΔglpK strain did not show the chain elongation phenotype like the wild-type, suggesting an entirely glycerol-dependent phenotype. This is further supported by our previous findings regarding the ΔgldA phenotype under Cd supernatant condition (Supplementary Table 3) [10]. Referring to Treerat et al. [10], the ΔgldA strain can only use glycerol via the GlpK pathway. However, GlpK is allosterically inhibited by fructose 1,6-bisphosphate under glycolytic conditions [25, 26, 38], resulting in preventing glycerol from internal metabolization and then enhancing its internal accumulation. Such glucose-free, glycerol growth condition appears to be similar to the one being tested with the ΔglpK strain in this study. GlpK has been shown to be conserved throughout evolution [34, 67–72]. In this study, the S. sanguinis GlpK (GenBank: ABN45208.1) amino acid sequence alignment and the predicted structure revealed the presence of the highly conserved H232 phosphorylation site, with an overall conserved surrounding protein topology. Both have been reported as key characteristics of bacterial GlpK [25–27, 34, 35, 69, 70]. Site-specific mutagenesis for replacing histidine with glutamate (H232 → H232E) of Enterococcus casseliflavus GlpK was shown to inhibit GlpK activities [25]. In this study, the response of the SK36 GlpK H232E mutant to glycerol was identical to the SK36 ΔglpK mutant, suggesting that phosphorylation of the H232 residue in SK36 was required. In other species, the phosphorylation of conserved H232 is dependent on enzyme I and Hpr of the phosphorenol-pyruvate:carbohydrate phosphotransferase system (PTS). Hpr-dependent phosphorylation of GlpK in Enterococcus faecalis increases its activity about tenfold [36, 38]. Thus, in the absence of PTS carbohydrates with relaxed catabolite repression, phosphor-Hpr not only predominates over Hpr, but can also stimulate GlpK activity. While we observed only a moderate catabolite repression of glycerol consumption in our study, consistent with other reports of S. sanguinis catabolite repression to be somewhat differently regulated [18, 43], a detailed molecular dissection of the various catabolite repression mechanism is required to fully understand its role in glycerol metabolism of S. sanguinis and how it interferes with interspecies interaction.

Glycerol and glycerol-containing compounds, such as lipids, are among the most abundant organic compounds that can be utilized as an alternative carbon/energy sources for many host-adapted bacteria [73, 74]. It is also an essential precursor for the synthesis of other key molecules in many Gram-positive bacteria, including (lipo)teichoic acids [75, 76]. Several bacterial pathogens can utilize host glycerol and phospholipids for their colonization and survival [36, 73, 77]. However, investigations into the production and utilization of glycerol molecules for mutual interspecies interactions of bacterial commensals are limited. We previously identified a key role of C. durum JJ1 fatty acids in beneficial interactions with S. sanguinis [10]. Here we demonstrate that C. durum is able to produce glycerol during growth, which shapes the S. sanguinis/C. durum interspecies interaction and is possibly a crucial ecological factor in health-associated subgingival niche development. While we observed chain elongation to occur at lower concentrations (Supplementary Fig. S7), we selected 10% (1.37 M) for our experimental set-up, since it most resembled the phenotype of C. durum supernatant-induced S. sanguinis chain elongation. The actual glycerol concentration determined in C. durum supernatant was considerably lower with 8 mM. We would like to point out that the phenotype can also be induced by reconstituted fatty acids (oleic acid, palmitic acid, and stearic acid) [10]. Therefore, the observed chain elongation of S. sanguinis in C. durum supernatant is most-likely caused by a mixture of these compounds. Moreover, glycerol in C. durum supernatants is primarily supplied as part of the fatty acids contained in its excreted membrane vesicles (MV) [10]. It is conceivable that MV-derived glycerol could be more efficient and effective at modulating cell-cell interactions than exogenously supplied free glycerol. It is worth noting that glycerol has been reported to remodel the host oral mucosa and maintain tissue homeostasis [78]. Thus, it is possible that C. durum, as a prominent oral commensal, could promote symbiosis and oral health by producing glycerol as well as fatty acid-containing membrane vesicles.

Multiple lines of evidence reveal significant similarities between the two prevalent oral corynebacterial commensals C. durum and C. matruchotii, including their ability to induce S. sanguinis chain elongation [10, 63, 79, 80]. Although both belong to the same genus, species-related differences in gene regulation can contribute to distinct phenotypes [63, 79, 80]. In addition, it is important to note that the initial RNA-seq data in this study solely focused on the interspecies interaction between C. durum and S. sanguinis. Our findings so far have only revealed a unique C. durum ability to produce certain fatty acids during growth for interacting with S. sanguinis via gldA and/or glpK mechanisms [10], which was supported by a previous study [63]. It remains unclear as to what strategies/mechanisms C. matruchotii would utilize in order to mutualistically interact in the oral cavity. Hence, it is of interest to conduct further investigations in order to compare with C. durum findings in this study.

Bacterial type IV pili are known to be essential for various functions influencing microbial ecology, including adhesion, twitching/gliding motility, as well as bacterial-bacterial and bacterial-host interactions [21, 22, 44, 51–53, 81, 82]. While the role of type IV pili in Gram-negatives is very well investigated, their role in Gram-positives and in the mutual interspecies interactions of bacterial commensals is poorly understood. Surprisingly, the SK36 gene cluster involved in type IV pilus biogenesis was found to be significantly downregulated in the presence of C. durum JJ1. Type IV pili in S. sanguinis have been implicated in gliding/twitching motility in some strains [20, 21, 24, 47]. Therefore, we further characterized the effect of C. durum JJ1 on type IV pilus dependent motility using several S. sanguinis strains. Interestingly, only strain SK408 exhibited a substantial reduction in twitching/gliding motility in the presence of C. durum JJ1, while none of the other S. sanguinis strains showed any significant motility under the tested conditions. S. sanguinis SK36 has been described before as lacking type IV dependent motility [20, 21, 24, 51], thus we also investigated type IV pili directly with SEM. For both, SK36 and SK408, long filamentous pili could be observed when grown alone. However, a significant visual reduction in those filaments was obvious when the cells were grown in C. durum supernatants. How the reduction in pili abundance is regulated might differ between strains. The expression of the pili gene cluster in SK36 and VMC66 decreased while in SK408 and SK1056 the expression was, unexpectedly, 1.5-fold elevated or only downregulated slightly. If this differential regulation is associated with the phylogenetic relationship between the strains warrants further investigation. At the moment, we do not have any experimental evidence about the role of twitching/gilding motility in the oral biofilm. Type IV pili dependent adhesion of SK36 to host cell components has been confirmed to be important for vegetation development in a native valve infective endocarditis rabbit model. Thus, a C. durum dependent repression of pilus gene expression seem to favor the commensal state of S. sanguinis. It might also decrease its ability to adhere to other species thus aiding in the close interaction of both providing the ideal commensal relationship.

In conclusion, this study provides an in-depth analysis of S. sanguinis and C. durum synergism. It also provides an important example of the factors mediating microbial distribution in the oral biofilm. In this case, the metabolic interaction between S. sanguinis and C. durum is directly dependent on glycerol and fatty acid production by C. durum. We also observed a strong effect on type IV pili production during co-cultivation. The association of both species is therefore dependent on metabolic as well as a potentially physical interaction as suggested by biogeographical studies [7]. Interestingly, while some cross-feeding between microbes leads to an obligate mutual relationship between species, S. sanguinis and C. durum are not dependent on each other for growth in vitro. How might such synergism evolve? Viability of both species is clearly increased during co-cultivation and S. sanguinis might metabolize the fatty acids and glycerol of C. durum, but we have thus far failed to observe any significant benefit for C. durum. A potential protection of C. durum from other competitors could be supported by the H₂O₂ production ability of S. sanguinis, similar to what has been described for the relationship between Streptococcus gordonii and Aggregatibacter actinomycetemcomitans. This will be investigated in future experiments, which will require an expansion of our current model to include other species. Moreover, it will lead to an enhanced understanding of more complex ecological interactions that are central for maintaining symbiosis with the oral microbiota and ultimately oral health.

Author contributions

PT, DA, JM and JK designed and conceptualized the research project. PT and DA performed experimental work and PT, DA, RAG and JK performed data analysis. All authors were involved in drafting, revising and finalizing the manuscript.

Data availability

All data are available in the main text and the supplementary materials. Raw and processed sequencing data can be found under GSE230560 in the NCBI GEO data repository.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41396-023-01426-9.