Glycerol metabolism contributes to competition by oral streptococci through production of hydrogen peroxide

ABSTRACT

As a biological byproduct from both humans and microbes, glycerol’s contribution to microbial homeostasis in the oral cavity remains understudied. In this study, we examined glycerol metabolism by Streptococcus sanguinis, a commensal associated with oral health. Genetic mutants of glucose-PTS enzyme II (manL), glycerol metabolism (glp and dha pathways), and transcriptional regulators were characterized with regard to glycerol catabolism, growth, production of hydrogen peroxide (H₂O₂), transcription, and competition with Streptococcus mutans. Biochemical assays identified the glp pathway as a novel source for H₂O₂ production by S. sanguinis that is independent of pyruvate oxidase (SpxB). Genetic analysis indicated that the glp pathway requires glycerol and a transcriptional regulator, GlpR, for expression and is negatively regulated by PTS, but not the catabolite control protein, CcpA. Conversely, deletion of either manL or ccpA increased the expression of spxB and a second, H₂O₂-non-producing glycerol metabolic pathway (dha), indicative of a mode of regulation consistent with conventional carbon catabolite repression (CCR). In a plate-based antagonism assay and competition assays performed with planktonic and biofilm-grown cells, glycerol greatly benefited the competitive fitness of S. sanguinis against S. mutans. The glp pathway appears to be conserved in several commensal streptococci and actively expressed in caries-free plaque samples. Our study suggests that glycerol metabolism plays a more significant role in the ecology of the oral cavity than previously understood. Commensal streptococci, though not able to use glycerol as a sole carbohydrate source for growth, benefit from the catabolism of glycerol through production of both ATP and H₂O₂.

IMPORTANCE

Glycerol is an abundant carbohydrate in the oral cavity. However, little is understood regarding the metabolism of glycerol by commensal streptococci, some of the most abundant oral bacteria. This was in part because most streptococci cannot grow on glycerol as the sole carbon source. In this study, we show that Streptococcus sanguinis, a commensal associated with dental health, can degrade glycerol for persistence and competition through two pathways, one of which generates hydrogen peroxide at levels capable of inhibiting Streptococcus mutans. Preliminary studies suggest that several additional commensal streptococci are also able to catabolize glycerol, and glycerol-related genes are actively expressed in human dental plaque samples. Our findings reveal the potential of glycerol to significantly impact microbial homeostasis, which warrants further exploration.

INTRODUCTION

Glycerol is a byproduct of microbial fermentation of carbohydrates, including what is produced by yeasts as a compatible solute to boost osmo-tolerance, and by certain oral commensal bacteria including Corynebacterium (1, 2). Glycerol is widely used as a food additive and is an important industrial ingredient for many healthcare products and cosmetics. In addition, glycerol and related compounds can be released from degradation of dietary or biological lipids, by the action of lipases or phospholipases, and are found in human blood as metabolic intermediates. As such, glycerol is an abundant carbon source for the oral microbiota and can potentially be utilized for both energy production and as a precursor to the production of many essential biomolecules such as phospholipids and lipoteichoic acids that are essential for bacterial envelope biogenesis (3, 4). An increasing body of research has indicated the significant contribution of glycerol metabolism to bacterial fitness and pathophysiology (3, 5–8). For example, glycerol oxidation by the bacterial genus Mycoplasma is considered critical to the virulence of these bacteria, in large part due to the resulting production of hydrogen peroxide (H₂O₂) (5).

Much of our understanding of glycerol metabolism by Gram-positive bacteria has been derived from research on the model organism Bacillus subtilis (9, 10), and in Enterococcus faecalis, a gut commensal and an opportunistic human pathogen (11). E. faecalis can utilize glycerol for growth, producing end products such as lactic acid, acetic acid, and ethanol (7). Regarding glycerol fermentation for acid production, a significant proportion of lactic acid bacteria (LAB) have been suggested to lack such capability, and notable intra-species variations are observed (12, 13). Two pathways are genetically encoded by enterococci and related LAB for the metabolism of glycerol, dehydrogenation, and phosphorylation, often with varying degrees of genomic reduction in glycerol-negative LAB (4). For the dehydrogenation pathway (encoded by gldA-dhaKLM), glycerol is converted into dihydroxyacetone (DHA) by a glycerol dehydrogenase (gldA) with a concomitant reduction of NAD⁺ to NADH and then into dihydroxyacetone phosphate (DHAP) by a DHA kinase (dhaKLM), which depends on bacterial PTS (phosphoenolpyruvate-dependent sugar: phosphotransferase system) (14) for supply of the phosphoryl group. For the phosphorylation pathway (glpKOF), glycerol is first phosphorylated by a glycerol kinase (GlpK; EC 2.7.1.30) into glycerol-3-phosphate (Gly-3-P) and then oxidized, in the presence of oxygen, into DHAP and H₂O₂ by the gene product of glpO (Gly-3-P oxidase; EC 1.1.3.21). Genes coding for both metabolic pathways in E. faecalis are important for glycerol metabolism and are expressed in a strain-dependent manner (13), and both pathways were required for virulence of E. faecalis in a mouse intraperitoneal model (8). The glycerol dehydrogenation pathway is considered by some to be indispensable for efficient glycerol metabolism in E. faecalis as glycerol phosphorylation creates H₂O₂ as a byproduct, accumulation of which could prove detrimental to the very bacterium that produces it (4). Conversely, it has been shown that oxygen is necessary for production of acids from glycerol by most LAB (12, 15), which differs from the metabolism of other common carbohydrates such as glucose, and that genes for the phosphorylation pathway appear to be better conserved overall in LAB than those of the dehydrogenation pathway (4). Furthermore, several species of the viridans streptococci that are considered important to oral health, including Streptococcus mutans, Streptococcus gordonii, Streptococcus sanguinis, Streptococcus mitis, and Streptococcus oralis, cannot ferment glycerol to lower the pH or support bacterial growth (12, 16). Nonetheless, a recent study has suggested that S. sanguinis can apparently utilize glycerol to achieve certain morphological phenotypes (17, 18). While some of these streptococcal species, including the major etiologic agent of dental caries, S. mutans, apparently lack critical genes required for either glycerol pathway, the S. sanguinis SK36 genome harbors the complete sets of genes for both (Fig. 1). The function and significance of these conserved genetic pathways in S. sanguinis, namely, glpKOF and gldA-dhaKLM, remain to be determined.

Fig 1.

Diagram depicting genes required for glycerol metabolism in several lactic acid bacteria. Genes annotated for the glycerol oxidation pathway (glp) are presented in blue, whereas those for the glycerol dehydrogenation pathway (dha) are in green. Putative transcription regulators associated with these genes are depicted in orange. Also denoted are catabolite response elements (cre) located in the intergenic regions, a formate acetyltransferase pfl2, and a transaldolase-like protein mipB.

With glycerol being a secondary carbohydrate, transcription of the genes encoding both catabolic pathways is often under carbon catabolite repression (CCR) (19), with a CcpA (catabolite control protein A)-binding motif (catabolite response element, cre) being identified in regions upstream of these catabolic operons (glp and dha) in B. subtilis and various LAB species including E. faecalis and S. sanguinis (4) (Fig. 1). Another CCR mechanism regulates the activity of the phosphorylation pathway at the enzymatic level, namely, the requirement of GlpK for PTS-mediated phosphorylation by Enzyme I (EI) and phospho-carrier protein HPr at a conserved histidine residue for maximal activity (20, 21). These mechanisms suggest that metabolism of glycerol is likely repressed by the presence of preferred sugars such as glucose. Additional regulators have been reported to modulate glycerol metabolism, including the RNA-binding anti-terminator GlpP identified in B. subtilis, which responds to Gly-3-P (9), and a positive regulator Ers in E. faecalis, which regulates metabolism of glycerol, citrate, and arginine (22). A putative transcriptional regulator (glpR, Fig. 1) exists as part of the glp locus in enterococci (13) and most streptococci, whose function remains to be characterized.

The central role of carbohydrate metabolism in oral microbial ecology makes glycerol a novel subject when investigating metabolic interactions among constituents of the oral microbiome. For example, it is understood that oral commensals such as Candida (23) and Corynebacterium (17) can produce significant quantities (mM levels) of glycerol, and its metabolism and ecological implications remain largely unknown. We recently mapped the regulon of a glucose–PTS in S. sanguinis SK36 by deep RNA sequencing of a manL (EIIAB^Glc) deletion mutant (24), a spontaneous mutant that emerged due to enhanced fitness under acidic conditions in association with reduced glucose transport (25). As part of the glucose–PTS regulon, genes of both glycerol metabolic pathways showed significantly higher expression in the manL mutant, especially the phosphorylation pathway (glpKOF), which showed the greatest increase (fold-change >70) within the entire transcriptome (24). Conversely, a recent RNA-seq analysis performed on a ccpA mutant of SK36 indicated that none of these glycerol catabolic genes were transcriptionally affected by loss of CcpA (26). The genome of S. sanguinis lacks the homolog of the anti-terminator GlpP found in Bacillus. Therefore, PTS could be regulating glycerol metabolism in S. sanguinis and related LAB in a CcpA-independent manner that significantly deviates from that in Bacillus subtilis (9, 10). In this study, we conducted a systematic study of glycerol catabolism by SK36, its regulation in response to other carbohydrates, and the implications for the ecology of the oral microbiome.

RESULTS

Growth phenotypes of mutants deficient in glycerol metabolic enzymes and glucose-PTS

In contrast to previous findings suggesting a lack in glycerol fermentation by many lactic acid bacteria (12), recent research has shown significant effects of glycerol on the physiology of S. sanguinis in a manner dependent on the glycerol kinase (GlpK) (17). To understand the metabolism of glycerol and its effects on bacterial physiology, we constructed several genetic mutants (Fig. 1) in the background of S. sanguinis SK36 and tested their ability to grow in a synthetic medium (FMC) supplemented with glycerol as the sole carbohydrate source (Fig. 2A). Despite the conservation of both glpKOF and the gldA-dhaKLM genes in the genome, SK36 could only produce low levels of growth, which increased OD₆₀₀ by about 0.05 units, a result consistent with previous findings (12). However, deletion of glpK or glpR, but not deletion of glpF, abolished this consistent, albeit low level of, growth. Complementation of glpK and glpR deletion largely rescued the ability of the complemented strains to grow on glycerol (Fig. S1A). Compared to FMC supplemented with 20 mM glucose alone, the combination of glucose and glycerol (5 mM) did not significantly alter the growth for most strains. However, the manL mutant deficient in the major glucose-PTS permease (EIIAB^Man) showed a significantly reduced growth rate and yield in FMC supplemented with glucose and glycerol combined in comparison to FMC containing only glucose; the manL/glpK double mutant did not (Fig. 2B and C). These results suggested that genetic or biochemical constraints exist in S. sanguinis, which limit the utilization of glycerol as a carbohydrate. For example, previous research showed that deletion of manL altered the expression of more than 300 genes in SK36, with many involved in metabolism and fitness (24).

Fig 2.

Growth curves measured in FMC medium modified to contain (A) glycerol; (B) glucose; (C) glucose and glycerol; and (D) glucose, glycerol, and 5 µg/mL catalase. Strains were cultured to the mid-exponential phase (OD₆₀₀ = 0.5) in brain–heart infusion (BHI), before being diluted 100-fold into 200 µL of defined medium (FMC) and loaded into a 96-well plate and covered with 60 µL mineral oil. OD₆₀₀ was recorded using the Biotek Synergy 2 once every hour for 24 hours. The incubation was carried out at aerobic conditions at 37°C. Each sample was represented by at least four biological replicates, and error bars denote standard deviations.

Effects of glycerol metabolism on bacterial fitness through H₂O₂ production

To understand the cause of the reduced growth of the manL mutant in the presence of glycerol, we first measured the release of extracellular DNA (eDNA) in the culture medium as a simple way to assess bacterial autolysis or programmed cell death (27–29). Elevated levels (by about threefold) of eDNA were noted in cultures of the manL mutant in the presence of 20 mM glucose and 5 mM glycerol compared to cultures prepared with only 20 mM glucose, yet no such difference was observed in the cultures of the manL/glpK double mutant, the glpK mutant, or the wild-type parent (Fig. 3A). As oxidation of glycerol-3-phosphate by the glpKOF pathway likely involves generation of H₂O₂ and H₂O₂ has been known to induce DNA and membrane damage related to increased eDNA release (28), we then measured H₂O₂ levels in the same culture media. The results showed substantially higher levels of H₂O₂ in the supernatant of the SK36 cultures prepared with glucose and glycerol combined compared to those prepared with glucose alone (increased from 1 to 1.2 mM, Fig. 3B), despite a lack of statistical significance. Importantly, the manL mutant, which produced increased eDNA levels in the presence of glycerol, also produced significantly more H₂O₂ than the wild-type in cultures supported by glucose (1.4 mM) and especially in cultures containing both glucose and glycerol (1.9 mM). In support of the notion that the glp pathway might have contributed to this increase in H₂O₂ levels, deletion of glpK in the manL background abolished the glycerol-dependent increase. Deletion of glpK alone in the SK36 background also resulted in the loss of response to the addition of glycerol. Finally, to confirm that differential production of H₂O₂ was the main cause of these growth and eDNA phenotypes, 5 µg/mL catalase was added to the culture medium containing a combination of glucose and glycerol. As shown in Fig. 2D, relative to the same cultures prepared without catalase (Fig. 2C), addition of catalase largely rescued the growth of the manL mutant close to the level observed in the FMC containing only glucose. Notably, addition of catalase also benefited growth of the manL/glpK double mutant.

Fig 3.

Glycerol induced in the manL mutant increased release of eDNA and H₂O₂. The bacteria were cultivated in FMC constituted with 20 mM glucose and 0 or 5 mM of glycerol for 24 hours in an aerobic atmosphere with 5% CO₂. The supernatant of each culture was harvested by centrifugation, and (A) the relative levels of extracellular DNA were measured by reacting with a fluorescent DNA dye, and (B) H₂O₂ concentrations using a biochemical reaction and a standard curve. At least three biological replicates were included in each sample, and the results were normalized against the cell density represented by OD₆₀₀ of each bacterial culture. The bars represent the mean of the measurements, and error bars denote the standard deviations. Asterisks indicate statistical significance assessed by two-way ANOVA followed by Tukey’s multiple comparisons test (*, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001).

Therefore, H₂O₂ production from glycerol through the glp pathway is likely regulated by glucose and the glucose-PTS, the deletion of which increased the production of H₂O₂ to levels that triggered excessive bacterial death and release of eDNA.

Glycerol oxidation via GlpKO contributes to production of H₂O₂ independently of pyruvate oxidase

To further understand glycerol metabolism in S. sanguinis and the influence of the PTS, deletion mutants of glpK, glpO, glpF, gldA, and glpR were tested for their ability to catabolize glycerol along with mutants deficient in manL, ccpA, and spxB. We first carried out a Prussian blue (PB) plate assay to assess the release of H₂O₂ by some of these strains on TY agar plates supplemented solely with glucose or glycerol. As indicated in Fig. 4A, SK36 produced significantly more H₂O₂ when incubated on plates containing glycerol than on those containing glucose. Interestingly, while the spxB mutant largely lost the ability to produce H₂O₂ on glucose agar as expected, it continued to do so at levels comparable to the WT on plates containing solely glycerol. On the other hand, mutants deficient in glpK, or glpO (Fig. S2), produced significantly lower levels of H₂O₂ than the WT on glycerol agar plates, although they behaved much like the WT on glucose plates. The strain glpKCom, which had its glpK deletion corrected in the genome, produced WT levels of H₂O₂ on glycerol plates. These results suggested the ability of SK36 to produce H₂O₂ in an SpxB-independent, glpKO-dependent manner while catabolizing glycerol. Likely encoding a glycerol uptake facilitator protein, glpF is not essential to glycerol metabolism as its deletion did not result in a significant reduction in excretion of H₂O₂ on glycerol plates. In additional experiments (Fig. 4B) performed on glucose agar plates, mutants deficient in ccpA or manL (including manL/glpK) produced elevated levels of H₂O₂. This agrees with the theory that CcpA is the major regulator controlling the transcription of spxB when cells are catabolizing glucose, where loss of ManL can alleviate the CCR by reducing carbon flux (10). However, on glycerol agar plates, only the manL mutant produced higher levels of H₂O₂, and the manL/glpK double mutant behaved similarly to the glpK mutant, thus supporting the notion that glucose–PTS negatively regulates the glp pathway independently of CcpA.

Fig 4.

Prussian blue (PB) plate assay for measurement of H₂O₂ release. Overnight cultures of bacteria in brain–heart infusion (BHI) were washed and dropped onto PB plates containing 20 mM glucose, 40 mM glycerol, or 20 mM glucose plus 5 mM glycerol. After another day of incubation in an ambient incubator maintained with 5% CO₂, (A) the plates were photographed and (B) the width of the blue zone was measured to represent the relative amounts of H₂O₂ being released. At least three biological replicates were used for each sample, and the average size of the PB zone was used to plot the bar graph, with error bars representing standard deviations and asterisks denoting statistical significance relative to results of the parent strain SK36 assayed on the same carbohydrate (unless specified otherwise) (Student’s t-test; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001).

In addition, strain ΔgldA did not show any discernable difference relative to the wild-type in the production of H₂O₂ on either plate, and the gldA/glpK double mutant showed a phenotype similar to that of ΔglpK. Strain ΔglpR, on the other hand, behaved similarly to ΔglpK by not producing significant amounts of H₂O₂ on glycerol. Also, like ΔglpKCom, complementation of glpR rescued the production of H₂O_2. Considered together with the growth phenotype of ΔglpR on glycerol alone (Fig. 2), it is plausible that the glpR gene product is required for the expression of the glp operon.

We also tested the same strains using tryptone yeast (TY) agar plates containing a combination of 20 mM glucose and 5 mM glycerol. In the presence of both carbohydrates, most of the strains tested generated a PB zone similar in size to that produced on glucose-only plates. However, several strains, including SK36, ΔmanL, ΔmanL/glpK, ΔgldA/glpK, and ΔglpRcom, did consistently produce slightly more H₂O₂ than in TY agar containing glucose alone.

Transcription analysis delineates the contributions of CcpA, PTS, and GlpR in regulating glycerol metabolism

A previous RNA-seq analysis of the manL mutant of S. sanguinis SK36 identified the glp operon genes as the most highly upregulated among the entire transcriptome (24) (Table S1). Also increased in the expression in the manL mutant were the dhaKLM genes, though to a lesser degree. Unlike in E. faecalis, the first gene of the glycerol dehydrogenase pathway, gldA (Fig. 1) is encoded separately from the dha operon in S. sanguinis. To begin unraveling the influence of both glucose–PTS and CcpA on transcriptional regulation of glycerol metabolic genes, we performed qRT-PCR assays on several representative genes from both glycerol branches by first analyzing strains SK36, ΔmanL, ΔccpA, and ΔmanL/ΔccpA that were grown to the exponential phase, with TY media supplemented with glucose or a combination of glucose and glycerol.

We first compared all the cultures prepared with TY-glucose (Table 1). Deletion of ccpA resulted in a markedly higher expression of genes dhaL and spxB, whereas deletion of manL led to a similarly drastic increase in glpK mRNA levels. Deletion of manL also resulted in a significant increase, though to a lesser degree than in ΔccpA, in the expression of dhaL and spxB. By contrast, deletion of ccpA produced little to no change in glpK expression. Deletion of both ccpA and manL however resulted in the highest increase in expression by each of these three genes, although this further increase, relative to their changes in either single mutant, was modest (less than twofold). These results were consistent with the notion that the CCR of glpK, and likely also glpO and glpF, is mediated by a mechanism independent of CcpA, whereas the CCR of dhaL and spxB is mediated directly by CcpA since the deficiency in glucose–PTS should also alleviate CCR mediated by CcpA (10). On the other hand, transcription of gldA appeared to be uncoupled from the rest of the dehydrogenation pathway, as it showed little change in the ΔccpA background and moderately reduced transcription in the manL mutant, confirming the findings of the RNA-seq analyses in both ΔccpA and ΔmanL (24, 26). Last, deleting glpR resulted in little to no change to the transcript levels of all four genes analyzed in cultures grown with glucose alone.

TABLE 1.

Relative abundance of mRNA levels in SK36 and various deletion mutants^a

Media		Strains	gldA	glpK	dhaL	spxB
TY	Glc	SK36	1.01	1.01	1.00	1.01
		ΔglpR	0.86	1.65	1.04	0.95
		ΔccpA	1.48	1.55	68.73****	10.25**
		ΔmanL	0.45	65.50****	9.04**	3.68*
		ΔccpA/ΔmanL	1.12	79.09****	91.59****	13.43***
	Glc +Gly	SK36	1.46	1.93	0.90	1.53
		ΔglpR	0.91	3.17*	0.91	1.66
	Fru +Gly	SK36	2.30*	0.76	1.11	0.36*
		ΔmanL	2.11	0.64	1.03	0.38*
	Gal	SK36	1.15	164.62****	31.68***	1.33
		ΔglpR	1.76	2.09	46.06***	0.76
		ΔglpRCom	0.65	218.96****	21.06***	3.60*
	Gal + Gly	SK36	0.43	161.89****	47.82***	0.98
		ΔglpR	0.88	2.37*	44.37***	0.71
		ΔglpRCom	0.65	218.59****	20.65***	3.08*
FMC	Glc	SK36	1.01	1.01	1.01	1.00
		ΔglpR	1.36	1.58	1.23	0.88
	Glc +Gly	SK36	2.28*	3.58*	1.07	3.17*
		ΔglpR	2.02	5.84**	0.82	4.53**
	Gal	SK36	0.44*	2.93*	12.02***	17.42**
		ΔglpR	0.33*	1.91	18.15***	8.77*
	Gal + Gly	SK36	0.32**	69.75***	5.44**	12.58***
		ΔglpR	0.29**	2.55	14.55***	13.83***

^a Cultures (n = 3) of SK36 and isogenic mutants were prepared with TY or FMC containing 20 mM glucose (Glc), fructose (Fru), or galactose (Gal), with or without 5 mM glycerol (Gly). Asterisks denote statistical significance in mRNA abundance relative to results in SK36 grown under the glucose condition in the same type of medium (assessed by Student’s t-test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001).

To explore the influence of PTS other than glucose–PTS on glycerol metabolism, RT-qPCR was carried out using SK36 and ΔmanL cultures prepared with TY containing a combination of 20 mM fructose and 5 mM glycerol. As the glucose–PTS does not transport fructose (25, 30), this experiment should determine if PTS EIIs other than EII^Man may trigger CCR on the glp operon. The results (Table 1) showed that just like glucose, the presence of fructose inhibited the expression of glpK—an effect not removed by the deletion of manL. Notably, compared to glucose, fructose showed no effect on dhaL expression yet reduced mRNA levels of spxB by close to threefold.

To understand the functions of GlpR, further transcriptional analysis was carried out in SK36 and ΔglpR cultivated in TY media supplemented with a combination of glucose and glycerol, galactose alone, or a combination of galactose and glycerol (Table 1). Addition of 5 mM glycerol to TY-glucose medium failed to induce in SK36 either the glp or dha pathway to a notable degree, a finding that is consistent with the strong CCR effect exerted by glucose. At the same time, deletion of glpR slightly increased the expression of glpK, but not dhaL, in the presence of both glucose and glycerol. Several studies in streptococci have suggested that relative to glucose, catabolism of galactose via the tagatose pathway results in a minimum level of CCR since few CCR-inducing metabolites such as fructose-1,6-bisphosphate are produced (31–33). In SK36 cultures prepared with galactose alone, glpK showed the highest expression so far, relative to cultures prepared with glucose, and dhaL also presented a considerable increase in mRNA levels. Deletion of glpR under galactose conditions abolished much of the increase seen with glpK, but not dhaL. This loss of glpK expression in ΔglpR was rescued by complementation in strain ΔglpRCom. These findings in the galactose medium supported the notion of GlpR acting as an activator of the glp operon; however, they also suggested that glycerol may not be needed to induce the expression. When SK36 and ΔglpR were each cultured with a combination of galactose and glycerol, the expression of both glpK and dhaL mirrored that in cultures prepared with galactose alone.

To rule out the possibility that TY medium possesses at least low levels of glycerol, we repeated this experiment using the chemically defined medium FMC supplemented with galactose or a combination of galactose and glycerol. The results (Table 1) now showed that the expression of glpK in FMC constituted with galactose alone was only slightly higher than levels measured in FMC-glucose; however, addition of glycerol to FMC-galactose drastically enhanced its expression. This result suggested that there likely was glycerol, or other inducing substrate, present in the TY–galactose medium at levels sufficient to induce the expression of glpK. Moreover, the expression of dhaL increased in FMC-galactose even without the addition of glycerol. We also repeated RT-qPCR using FMC supplemented with glucose +/-glycerol (Table 1), which largely supported findings observed in TY media. Conversely, transcription of gldA in galactose-containing media was two to threefold lower than in glucose-based media. We subsequently measured glycerol levels in all the media available to us, including BHI, TY, TV, T-broth, FMC, and BM (34), and found similar levels of glycerol (2–3 mM) in TY, TV, and T-broth and slightly lower levels in BHI as well (Fig. S3). Considering this information, we have tried to conduct all the critical experiments using the FMC and BM and by washing cultures prepared with BHI before diluting into FMC or BM. Notably, we were not able to use FMC as the base of the Prussian blue agar due to chemical incompatibility.

Glycerol catabolism impacts growth on galactose by SK36 and by related oral streptococci

In light of the reduced CCR by galactose on glycerol catabolic genes, we conducted a growth analysis on strains SK36, ΔglpR, ΔglpRCom, ΔglpK, and ΔglpKCom by incubating them in FMC constituted with 20 mM galactose, with or without 5 mM of glycerol, in an aerobic environment supplemented with 5% CO₂. While all strains grew relatively normally on galactose alone, addition of 5 mM glycerol markedly reduced the growth rate and yield of SK36, ΔglpRCom, and ΔglpKCom (Fig. 5A and B). Interestingly, strains ΔglpR and ΔglpK each produced a growth in FMC containing a combination of galactose and glycerol, which was comparable to growth in FMC with galactose alone (Fig. 5B). Considering the lack of CCR by galactose on the expression of glpK in SK36 (Table 1), this GlpR- and GlpK-dependent growth reduction by glycerol can be explained as due to the over-production of H₂O₂, similar to the phenotype of the manL mutant in Fig. 2C. Indeed, when we added 5 µg/mL catalase to FMC containing a combination of galactose and glycerol, growth of SK36, ΔglpRCom, and ΔglpKCom was largely rescued to levels comparable to those of ΔglpR and ΔglpK (Fig. 5C). This finding not only demonstrated the drastic effect of glp-mediated H₂O₂ production on the physiology of SK36, but it also provided us a simple test to explore if a similar response to glycerol existed in other oral streptococcus isolates.

Fig 5.

Growth curves constructed using S. sanguinis strains SK36 and its mutant derivatives (A, B, C), S. gordonii strain DL1 (D), S. oralis strain BCC11 (E), and S. mitis strain BCC36 (F). The media were based on FMC that was modified to contain 20 mM galactose, with or without 5 mM glycerol, and 5 µg/mL catalase (C). Each strain was represented by at least three biological replicates, with their means and standard deviations (error bars) in OD₆₀₀ values being presented.

A BLAST search using the protein sequences of GlpK and GlpO identified respective homologs in most of the species belong to the mitis group streptococci. Two isolates each of S. sanguinis, S. gordonii, S. oralis, S. cristatus, and S. mitis were inoculated into FMC with galactose +/-glycerol as mentioned previously. One strain each of S. dentisani and S. intermedius was also included as no homologs of GlpKO were identified in either species. Most of these strains have previously been analyzed by whole-genome sequencing (35, 36). As shown in Fig. 5D through F (and Fig. S1D~K), at least one strain each of S. sanguinis, S. gordonii, S. oralis, and S. mitis showed reduced growth in the presence of glycerol, relative to the galactose-only condition. Some strains failed to grow on either medium, whereas S. intermedius BCC01 and two S. cristatus stains grew normally, yet did not show a notable response to glycerol. Therefore, glycerol-mediated H₂O₂ production appears to be relatively well-conserved in a group of streptococci that are associated with oral health, although further research is needed to confirm this conclusion.

Influence of glycerol metabolism on streptococcal competition

We then performed plate and liquid-based competition assays to explore the effect of glycerol on antagonism between S. sanguinis and S. mutans. First, strains SK36, ΔmanL, ΔglpK, and ΔmanL/glpK were each inoculated on FMC-based agar plates supplemented with 20 mM glucose as the sole carbohydrate source. FMC was selected in favor of TY to avoid contaminating glycerol, and plates were incubated in an aerobic incubator with 5% CO₂. S. mutans UA159 was inoculated to the right of S. sanguinis colonies after 24 hours of incubation, followed by another day of incubation. Compared to the wild-type SK36 (Fig. 6), deletion of manL resulted in a minor increase in the inhibition of S. mutans. When 5 mM of glycerol was included in addition to glucose, however, inhibition of S. mutans by strain ΔmanL was significantly enhanced. At the same time, deletion of glpK in the background of manL reversed this effect, although deletion of glpK in the background of SK36 showed little impact. Interestingly, when glycerol was supplied as the sole carbohydrate source, S. sanguinis, but not S. mutans, formed a visible colony on the agar plates (Fig. 6), echoing the limited growth of SK36 observed on glycerol alone (Fig. 2). When this experiment was repeated in galactose-based medium, with or without glycerol, SK36 and all its mutants greatly inhibited the growth of UA159 in comparison to UA159 inoculated on the plate alone, consistent with previous observations that both spxB and glpK were highly expressed under these conditions (Table 1) (37). Nonetheless, the glpK mutant did show less antagonism than SK36 in the presence of glycerol. These experiments confirmed glycerol metabolism by S. sanguinis (via function of GlpKO) as an ecologically significant source of H₂O₂ and the role of glucose–PTS in regulating this activity.

Fig 6.

Deletion of manL enhanced the antagonism of S. mutans in the presence of glycerol. To avoid contaminating carbohydrates, FMC was used as the base medium and was modified to contain 20 mM glucose, 20 mM glucose and 5 mM glycerol, 40 mM glycerol, 20 mM galactose, or 20 mM galactose and 5 mM glycerol. Each S. sanguinis strain was inoculated on the agar plates first. After overnight incubation, a fresh BHI culture of S. mutans UA159 (SMU) was inoculated to the immediate right of the colony, followed by another day of incubation before photography. Each interaction was tested at least twice on separate days, with a representative set of results being presented here.

In consideration of the oxygen-limited condition in the oral cavity, a competition assay was carried out in planktonic cultures to mimic such a condition, by mixing S. sanguinis and S. mutans at an approximately 1:1 ratio and cultivating overnight in an aerobic chamber maintained with 5% CO₂. When these mixed cultures were diluted and allowed to compete in TY containing only 20 mM glucose, S. mutans UA159 was drastically more competitive over S. sanguinis SK36, with the competitive index (SSA/SMU) nearing 10⁻⁸. Little difference in the competition was noted when 5 mM glycerol was added to the medium containing 20 mM glucose. However, when 40 mM glycerol was used as the only carbohydratesource in the TY medium, SK36 was approaching being equal in competitiveness with UA159, and this competitiveness was reduced by about 12-fold when the glpK gene was deleted (Fig. 7A). In each of these 20-hour two-species cultures prepared with TY-glycerol, the total CFU was around 10⁸ to 10⁹ CFU/mL. Meanwhile, SK36, glpK and UA159 each yielded 10⁸ to 10⁹ CFU/mL viable cells after overnight incubation on TY-glycerol as single-species cultures. These results demonstrated the ability of glycerol to greatly enhance the competitive fitness of SK36 against UA159, mostly in the absence of glucose, a capacity that is not entirely dependent on an intact glpK gene. Interestingly, when the gldA mutant was tested similarly, it behaved much like the WT parent under all three conditions, showing no reduced competitiveness in TY containing glycerol.

Fig 7.

Glycerol greatly benefits S. sanguinis (SSA) against S. mutans (SMU) in a planktonic competition assay. Exponentially growing BHI cultures of S. sanguinis SK36, ΔglpK, ΔgldA, and ΔdhaKL were each mixed with equal amounts of S. mutans strain UA159, inoculated at a 100-fold dilution rate, and allowed to grow overnight in TY (A) or FMC (B) medium supplemented with 20 mM glucose, 40 mM glycerol, or a combination of 20 mM glucose and 5 mM glycerol. The CFU of each competing species before and after the competition was enumerated to calculate the competition indices, with values greater than 1 indicative of an advantage for S. sanguinis. Each strain was represented by four separate cultures, from which the means and standard deviations (error bars) were calculated and presented. Statistical analysis was carried out using two-way ANOVA, followed by Tukey’s multiple comparisons (*, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001).

Results obtained thus far suggested that the ORF annotated as gldA may not be contributing significantly to glycerol metabolism. Therefore, a new mutant deficient in both dhaK and dhal, ΔdhaKL, was constructed to represent the dha pathway. Strain ΔdhaKL showed no change in the ability to grow on glycerol (Fig. S1O) or produce H₂O₂ (Fig. S2) in comparison to its WT parent. We then repeated the planktonic competition assay by replacing ΔgldA with ΔdhaKL and by using FMC as the base medium to avoid glycerol contamination. Unlike S. sanguinis, S. mutans UA159 lacks homologs of the glpKOF and dhaKLM operons and does not produce detectable growth in FMC containing only glycerol (Fig. S1P). Overall, the S. sanguinis wild-type was significantly more competitive when growing in FMC-based media than in TY-based media, likely owing to superior buffering capacity of the former. Relative to glucose, glycerol alone enhanced the competitiveness of S. sanguinis against S. mutans by two logs, and disruption of either the glp or dha pathway significantly reduced its competitiveness (Fig. 7B). Specifically, deletion of glpK resulted in a 34-fold reduction (P < 0.0001) in the competitive index in FMC–glycerol, as well as a sixfold reduction in FMC containing glycerol and glucose combined, although the latter did not reach statistical significance. On the other hand, deletion of dhaKL also reduced the competitive index of SK36 by sevenfold on glycerol alone (P < 0.0001) and by threefold on glycerol + glucose (P > 0.05). Therefore, it appears that both the glp and dha branches, glp especially, are required for competitive fitness of S. sanguinis under planktonic condition in the presence of glycerol.

Last, the effect of glycerol on bacterial competition was tested in a biofilm setting. A dual-species biofilm model was established using S. sanguinis and S. mutans UA159 on the surface of a hydroxyapatite disk submerged in a BM medium (34) supplemented with 25% saliva, 2 mM sucrose, and 18 mM glucose. After 1 day of incubation, the medium of the biofilm was replaced with BM containing glucose, BM with glycerol, or BM base without any carbohydrate. About 24 h later, the biofilm was harvested, and CFUs of both species were enumerated. This experiment was meant to simulate a mature dental biofilm being exposed to a dose of glucose or glycerol. The results (Fig. 8) showed a strong negative effect of glycerol on the persistence of S. mutans by eliminating S. mutans cells from the biofilm. This effect of glycerol was dependent on the integrity of glpK, yet not that of dhaKL. This difference from the planktonic model is likely because no substantial growth is expected from either S. mutans or S. sanguinis when glycerol was applied to the biofilm. When the same experiment was repeated in an anaerobic atmosphere, no significant effect was observed in association with glycerol (Fig. S4). Interestingly, deletion of gldA in SK36 resulted in a significant loss of the CFU of S. sanguinis under all three conditions, including when treated with BM-glycerol where both bacteria returned no CFU from the biofilms. Furthermore, when different batches of pooled saliva were used in these assays, substantial variation was noted in the effectiveness of glycerol in inhibiting S. mutans in a glpK-dependent manner (Fig. S4). Considered together with the planktonic competition assay, both branches of the glycerol metabolism could benefit S. sanguinis during its interaction with S. mutans, with their relative contributions being dependent on experimental settings and/or certain environmental factors that are yet to be clarified.

Fig 8.

Glycerol promotes S. sanguinis (SSA) competition against S. mutans (SMU) in a two-species biofilm in a glpK-dependent manner. SK36 and its isogenic mutants ΔglpK, ΔdhaKL, and ΔgldA were each mixed with an equal volume of S. mutans UA159 culture and inoculated, at a 1:100 ratio, into a BMGS medium held in a 24-well plate with a hydroxyapatite disk and kept in an aerobic atmosphere with 5% CO₂. After 24 hours of incubation, the culture supernatant was replaced with BM base, BM containing 18 mM glucose (Glc), or BM with 36 mM glycerol (Gly). After another day of incubation, the biofilm was washed and harvested by sonication, followed by serial dilution and CFU enumeration. Each strain was represented by four separate cultures and each condition four biofilm samples. The average CFU and standard deviation (error bars) of each species were used to plot the graph and for statistics (Students’ t-test; *, P < 0.05).

Expression of glycerol metabolic genes in dental plaque

To explore the contribution of glycerol to microbial homeostasis in vivo, a bioinformatic analysis was conducted using metatranscriptomic data sets generated previously in a deep sequencing project (24, 38). Seventy human plaque samples were used in that study, including 34 caries-free (PF) and 36 caries-active samples, with the latter comprising 11 samples from enamel lesions (PE) and 25 from dentin lesions (PD). After normalization for total reads per plaque, transcript abundance for individual genes was calculated and tabulated into pathways of known bacterial functions. Subsequent comparative analysis (Fig. 9) indicated an increase in overall transcript levels of most S. sanguinis pathways in PF samples relative to both caries-active groups. This was likely in large part due to the elevated abundance of S. sanguinis as a species in healthy biofilms compared to dysbiotic ones (39). We then identified two pathways associated with glycerol metabolism, one named “super-pathway for glycerol degradation to 1,3-propanediol” and the other “CDP-diacylglycerol biosynthesis pathway”, neither of which included the glycerol degradation pathways studied here (4). Nonetheless, as shown in Fig. 9, the “super-pathway for glycerol degradation to 1,3-propanediol” was among the top 10 functions in terms of fold-change in PF vs PE samples or PF vs PD samples. As a critical function in lipid metabolism (40), CDP-diacylglycerol biosynthetic genes also showed an above-average PF/PE ratio. We then analyzed the transcript levels of specific genes of glp and dha pathways. Expression of the dehydrogenation pathway by S. sanguinis was above average levels in PF samples, with gldA at ~2.6 fold that of the median of fold-changes calculated over the entire genome, both against PE and PD samples (Table S2). Expression of glpK in PF samples was slightly above the median of the genome, at 1.3-fold, and so was a glycerophosphodiester phosphodiesterase (twofold) with the purported function of hydrolyzing glycerophospholipid and liberating Gly-3-P. When we expanded the analysis to include other mitis group streptococci, a similar trend in expression of these glycerol metabolic genes was noted (Table S2), especially in the Streptococcus_mitis_oralis_pneumoniae subgroup, followed by S. australis and S. gordonii. Therefore, genes required for glycerol metabolism in S. sanguinis appeared to be actively expressed in dental plaque in association with oral health.

Fig 9.

Glycerol metabolic pathway of S. sanguinis shows elevated expression in caries-free dental plaque samples. Metatranscriptomic analysis was performed on 70 plaque samples, 34 caries-free (PF) and 36 caries-active [11 from enamel sites (PE) and 25 from dentin sites (PD)]. Scatter plots show fold changes in transcript abundance in (A) caries-free vs enamel lesions and (B) caries-free vs dentin lesions. The red lines denote the value of 1, and the blue lines show the median of the population. The red circle represents the “super-pathway of glycerol degradation to 1,3-propanediol,” and the orange diamond represents the “CDP-diacylglycerol biosynthesis pathway.”

DISCUSSION

The main findings of this study included the following: i) While S. sanguinis SK36 cannot grow optimally on glycerol alone, a consistent, low-level of growth was achieved on mM levels of glycerol in a glpK-, glpR-dependent manner. ii) the glp pathway can produce H₂O₂ independently of SpxB by catabolizing glycerol in the presence of oxygen, and the level of H₂O₂ derived from glycerol readily resulted in both reduced growth in S. sanguinis and inhibition of S. mutans in co-cultures. iii) Expression of the glycerol dehydrogenation (dha) branch is primarily regulated by CCR mediated by CcpA, whereas the glycerol phosphorylation branch (glp) requires GlpR and glycerol for induction and is subjected to CCR by the PTS. iv) Metabolism of glycerol enhanced the competitiveness of S. sanguinis against S. mutans primarily in a GlpK-dependent manner, especially when tested in a biofilm model. The dha branch also contributed to the persistence of S. sanguinis under either planktonic or biofilm setting. With glycerol being an abundant carbohydrate in the oral cavity and released by other microbiomes, these novel findings open the door to further exploring the influence of glycerol in human health and diseases.

Oral streptococci have long been considered incapable of catabolizing glycerol for growth or acid production (12). Our study on one hand partly confirmed these earlier findings but on the other hand revealed a new dimension in the physiology of these abundant oral bacteria concerning their metabolic interactions with the rest of the microbiota. Since SK36 harbors all the known genes of both glp and dha branches, we can predict that their gene products can catabolize each glycerol molecule to obtain one molecule of pyruvate and one net gain of ATP molecule, in addition to two NADH molecules by way of dehydrogenation (dha) or one NADH and one H₂O₂ through the phosphorylation (glp) branch. One more ATP can be obtained if pyruvate is further oxidized, e.g., by SpxB, PFL, or PDH branches, into acetate, although NADH/NAD⁺ ratio must be maintained either by NADH oxidase (Nox) or other pyruvate-reducing enzymes such as lactate dehydrogenases (LDH) or alcohol dehydrogenases (ADH). There are reasons to believe that the phosphorylation branch is favored by SK36 under most of our test conditions. According to RT-qPCR, the transcript levels of the gldA gene in SK36 grown with glucose were about half that of glpK and remained unchanged in most conditions or mutants we tested; expression of dhaL was not inducible by glycerol and was largely independent of GlpR. Another supporting evidence is the substantial amounts of H₂O₂ produced by the spxB mutant on glycerol. If so, lower amounts of pyruvate were produced by the dehydrogenation branch, and this inference is supported by the lack of substantial H₂O₂ yield by the glpK mutant on glycerol, with the spxB gene being intact (Fig. 4); and by the importance of glpK to interbacterial interaction as revealed in our dual-species competition assays (Fig. 7 and 8). Furthermore, S. gordonii, which is closely related to S. sanguinis in the glycerol phenotype (Fig. 5), lacks all three dha genes (Fig. 1). Nonetheless, results of the competition assays (Fig. 7 and 8) indicated that the dehydrogenation branch was indispensable in the competitive fitness of SK36, both with and without glycerol. Further research is needed to understand if environmental factors and additional genetic mechanisms contributed to such differences in our model systems.

Another question raised during our study was why glycerol alone cannot support SK36 to grow to the levels similar to glucose, especially since all necessary genes to support gluconeogenesis and pentose phosphate pathway are present in its genome (41). This remains unclear even as we try to exhaust the test conditions. Aside from the fact that gluconeogenesis is energetically costly and glycerol dehydrogenation appears constrained by gldA-dhaKLM expression patterns, production of H₂O₂ by glycerol phosphorylation could prove to be a double-edged sword. However, addition of catalase only resulted in a slight enhancement in growth by SK36 on glycerol, as seen in Fig. 2A and Fig. S1M, whereas incubating it in an anaerobic chamber abolished all indications of any increase in optical density (Fig. S1L, M and N), arguing against the notion that either oxygen or production of H₂O₂ was the cause of the lack of substantial growth on glycerol. To appreciate the significance of this unique metabolic activity, perhaps it is more appropriate to consider it in the context of intermicrobial interactions. Recent work on interactions between S. sanguinis and Corynebacterium durum, an abundant oral commensal, has provided direct support to the availability of mM levels of glycerol from another bacterium and its effects on S. sanguinis (17). Another important source of glycerol is the commensal yeasts such as Candida. When we cultivated a C. albicans strain SC5314 in FMC supplemented with various carbohydrates, as much as 1 mM glycerol was detected in the spent media (Fig. S3). Research has shown that C. albicans significantly increases its activity of glycerol release when it transitions into a biofilm state (23), that C. albicans readily co-aggregates with commensal oral streptococci (42), and that mixing with C. albicans cultures for 30 minutes induced the transcription of glycerol-metabolic genes in S. gordonii (43). Given the dual role of Candida species as both commensals and pathobionts, we posit that bacterial metabolism of glycerol released by Candida could also impact health in a nuanced manner dependent on the genomic makeup of its microbial niche.

Like most secondary catabolic genes, both branches of glycerol metabolism are subjected to negative regulations by CCR. The novelty of our finding resides in the fact that CcpA appears to be directly controlling the transcription of only the dha branch, whereas the PTS regulates the glp branch. Our initial findings suggest that glucose-PTS may be the chief mechanism controlling glp expression. However, as fructose similarly confers strong CCR on the same circuit (Table 1; Fig. S1B and C), it stands to reason that a mechanism shared by both glucose and fructose is likely responsible. Such phenotypes echo what have been reported in other streptococci, in particular S. mutans, where HPr has been suggested to work in concert with the glucose-PTS in controlling catabolic genes responsible for several secondary carbohydrate sources (44–46). The physiological significance of this bifurcated strategy in regulating two branches of the same metabolic pathway warrants detailed investigation, but as we have suggested previously concerning the regulation of a fructanase operon (fruAB) by both CcpA-dependent and PTS-dependent mechanisms, difference in the nature of the signals for these two systems allows for thresholded responses: PTS could recognize and respond to lower levels (tens of μM to low mM) of carbohydrates with specific structures, whereas CcpA primarily responds to changes in intracellular energy status triggered by higher amounts (>3 mM) of carbohydrates (47, 48). Glucose can be found in μM levels in saliva outside of mealtimes (49), whereas levels above mM can be expected in human blood or in the oral cavity during feeding. With GlpR being required for glp expression and the genome of SK36 apparently lacking an anti-terminator-style regulator (GlpP) that is required for CcpA-independent CCR of the glp pathway in B. subtilis (9), it seems reasonable to hypothesize that such a PTS-mediated regulation could be conducted via the ability of EI/HPr to phosphorylate and activate GlpK (17), whose product Gly-3-P in turn can stimulate GlpR’s DNA-binding activity. In addition, glycolytic intermediate fructose-1,6-bisphosphate has been reported to inhibit GlpK activities (20), which may also contribute to PTS-mediated regulation. Notably, we have preliminary evidence (Fig. 5) suggesting that additional peroxigenic streptococci likely also possess the ability to catabolize glycerol for H₂O₂ production, and similar operons containing glpR homologs appear to be conserved in at least some of these bacteria, including S. pneumoniae and S. gordonii (Fig. 1). It would be interesting to explore if similar metabolic strategies or CCR mechanisms exist in these related bacteria and if so, how that contributes to microbial homeostasis or host interactions in relevant microbiomes. Furthermore, we cannot rule out the possibility that glpK is also subjected to the conventional CCR mediated by CcpA. After all, a near perfect cre sequence exists (Fig. S5) in the intergenic region upstream of the glpK sequence, and mutants defective in both manL and ccpA showed slightly higher glpK expression than that in either single mutant (Table 1). If so, CcpA could provide another layer of negative control on top of PTS on the catabolic operon, like the CCR of the fruAB operon in S. mutans, which involves both CcpA and the PTS (47). Lastly, a recent study reported that glycerol can be readily catabolized by SK36 both with and without addition of glucose, suggesting a moderate catabolite repression by glucose as opposed to the tight CCR observed here (Table 1) (17). While we continue to unravel the complex nature of this phenomenon, an undefined medium used in that study could be at least partly responsible for the different findings.

It is perhaps worth noting that in most of our assays testing the functionality of the glp pathway, a glpK rather than a glpO genetic mutant was used. This was out of the concern that loss of glycerol oxidase GlpO alone, with the glycerol kinase GlpK being intact, would result in accumulation of Gly-3-P intracellularly, which has been known to cause cytotoxicity effect (50). Given the presence of significant amounts of glycerol in several common media including BHI, the effects of Gly-3-P accumulation could incur a fitness cost and complicate interpretation of the data involving ΔglpO. Nonetheless, we have performed the PB plate assay on ΔglpO, which showed a similar phenotype as that of ΔglpK (Fig. S2). Regarding the genetic uncoupling of gldA from the rest of the dehydrogenation branch, it could be the result of an evolution that favored H₂O₂ production by shunting glycerol toward the phosphorylation branch. A recent study on SK36 revealed two putative transcriptional regulators, SSA_0278 and SSA_0279, which were required for gldA expression in response to certain fatty acids during its interaction with C. durum (18), indicative of a possible specialization in function from or in addition to glycerol metabolism. The biofilm deficiency of strain ΔgldA revealed in Fig. 8 suggests certain novel functions of GldA in biofilm attachment by S. sanguinis that may or may not be related to catabolism of glycerol. Alternatively, there could exist a not-yet-identified isozyme that complements or replaces the function of GldA in dehydrogenating glycerol under certain conditions. Our finding that deletion of dhaKL genes, but not deletion of gldA, significantly impacted the fitness of SK36 in the presence of glycerol (Fig. 7) supports this notion. The gene product of gldA in Escherichia coli has been suggested to primarily catalyze the reverse reaction, by converting DHA into glycerol to avoid DHA-derived cytotoxicity; DHA or its metabolite being a reactive electrophile species capable of glycating various biomolecules (51). In support of this notion, a likely transaldolase is encoded by an ORF (mipB) immediately upstream of gldA in SK36 (and in S. gordonii DL1), with the predicted function of converting DHA and glyceraldehyde 3-phosphate into fructose-6-phosphate. S. mutans, which lacks both glp and dha operons, also maintains the gldA and mipB homologs (Fig. 1) in its core genome (52), whose functions remain to be determined.

In conclusion, our research into glycerol metabolism by S. sanguinis revealed a novel pathway for generation of H₂O₂, an activity that appears uniquely significant regarding the competitiveness of the commensals and the health of the microbiome, rather than its contribution to bacterial growth. Further studies focusing on glycerol pathways in abundant oral bacteria, complex biofilms, and in vivo settings could help identifying potentially novel genetic mechanisms that can be targeted to enhance the homeostasis of the oral microbiome and oral health.

MATERIALS AND METHODS

Bacterial strains and culture conditions

Strains (Table 2) of S. sanguinis SK36 and its isogenic mutants, isolates of S. mutans, S. gordonii, and other oral streptococcal species were routinely maintained on BHI (Difco Laboratories, Detroit, MI) agar plates supplemented with 50 mM potassium phosphate (pH 7.2), and antibiotics (kanamycin at 1 mg/mL; erythromycin at 5 µg/mL) when necessary. Colonies were inoculated in liquid BHI medium and incubated overnight, before being dropped onto tryptone–yeast extract medium (TY, 30 g of tryptone and 5 g of yeast extract per liter) agar plates with or without Prussian blue reagents (53), or agar plates based on synthetic FMC (54), or diluted into fresh TY or FMC medium that was modified to contain various specified carbohydrates. Glucose was generally used at 20 mM and glycerol at 40 mM due to a smaller number of carbons in each glycerol molecule, and 5 mM glycerol was used in combination with 20 mM glucose to simulate the in vivo situation where glycerol is a secondary carbohydrate (17). Unless specified otherwise, all cultures were incubated at 37°C in an aerobic atmosphere supplemented with 5% CO₂.

TABLE 2.

Strains used in this study

Strains	Relevant characteristicsa	Source or reference
SK36	S. sanguinis wild-type	Kitten laboratory
ΔglpK	SK36 glpK::Em	From SK36
ΔglpKCom	ΔglpK complementation (Km)	From ΔglpK
ΔglpR	SK36 glpR::Em	From SK36
ΔglpRCom	ΔglpR complementation (Km)	From ΔglpR
ΔglpF	SK36 glpF::Em	From SK36
ΔgldA	SK36 gldA::Em	From SK36
ΔgldA/glpK	SK36 gldA::Em glpK::Km	From ΔgldA
ΔdhaKL	SK36 dhaKL::Em	From SK36
ΔspxB	SK36 spxB::Em	From SK36
ΔmanL/glpK	SK36 manL::Km glpK::Em	From ΔmanL
MMZ1945	SK36 gtfP::Em	(24)
ΔmanL	SK36 manL::Km	(25)
MMZ1905	SK36 manLComp::Em	(25)
MMZ1913	SK36 ccpA::Km	(25)
MMZ1910	SK36 ccpA::Em manL::Km	(25)
UA159	S. mutans wild type, perR+	ATCC 700610
MMZ1939	UA159/pDL278::gfp (Sp)	From UA159

^a Km, kanamycin resistance cassette; Em, erythromycin; Sp, spectinomycin.

Construction of genetic mutants

Genetic mutations were engineered by following a protocol based on allelic exchange using mutator DNA molecules generated using PCR amplification followed by ligation by Gibson assembly (25). Briefly, two homologous DNA fragments, DNA_1 and DNA_2, flanking the target gene X were amplified using primers engineered to contain short DNA sequences that allowed an overlap between DNA_1 and the 5’ of the antibiotic marker and an overlap between the 3’ of the antibiotic marker and DNA_2. These three DNA fragments were subsequently fused together in the same Gibson assembly reaction to yield the mutator DNA. Wild-type S. sanguinis strain SK36 was cultured in BHI till early exponential phase (OD₆₀₀ = 0.05–0.08), when competence stimulating peptide (synthesized at ICBR protein core, University of Florida) and the mutator DNA were added to the culture, followed by 3 hours of incubation before plating on selective agar plates containing specific antibiotics (25). Colonies obtained this way were validated by PCR coupled with Sanger sequencing targeting the region of mutagenesis. All DNA oligos (listed in Table S3) were synthesized by Integrated DNA technologies (Coralville, IA).

Genetic complementation was carried out using a “knock-in” strategy detailed in our previous publication (25). Briefly, to correct the allelic replacement in the deletion mutant of gene X (e.g., ΔX::Em), the upstream flanking fragment DNA_1 was extended by using a different primer (Comp-2GA, Table S3) to include the entire coding sequence of the wild-type gene X. A mutator DNA was then created by ligating together this new DNA_1 fragment, the original downstream flanking fragment DNA_2, and an alternative antibiotic marker (e.g., Km). After transforming ΔX::Em with this new mutator DNA, Km-resistant colonies were picked, confirmed for sensitivity to erythromycin, and validated by sequencing, similarly as above.

Measurement of the relative abundance of eDNA in bacterial cultures

Concentrations of extracellular DNA (eDNA) in FMC culture supernatants were assessed by mixing them with a fluorescent DNA dye, SYTOX Green (Ex/Em 504/523 nm; Thermo Fisher Scientific, Waltham, MA) and reading the intensity of light emission at 523 nm on a 96-well, Synergy H1 multimode plate reader (Agilent Technologies, Santa Clara, CA). The relative abundance of eDNA was presented as RFU normalized against OD₆₀₀ of the bacterial culture. The validity of this assay was confirmed using a DNA standard curve, as detailed in our previous publication (27).

Measurement of H₂O₂ released by bacterial cultures

Prussian blue (PB) plate assay: Each bacterial strain was cultivated overnight in liquid BHI medium, harvested by centrifugation, washed, and resuspended with sterile PBS. Ten microliters of the cell suspension was then inoculated onto TY agar plates supplemented with the specified carbohydrates. After overnight incubation in an aerobic incubator containing 5% CO₂, the plates were photographed, and the width of the PB zone was quantified using ImageJ software.

Measurement of H₂O₂ levels in liquid cultures was carried out using an enzymatic assay according to a previously published protocol detailed elsewhere (36).

Plate-based antagonism assay

S. sanguinis strains were cultivated till the exponential phase (OD₆₀₀ = 0.5), from which a 5-µL aliquot was placed on an FMC agar plate modified to contain specified carbohydrates and incubated for 24 hours to form a colony. An overnight culture of S. mutans UA159 prepared in BHI was then inoculated to the right of the colony. The plates were incubated for another day before being photographed.

Mixed-species competition in planktonic cultures

For interspecies competition in liquid cultures, differentially marked strains of S. sanguinis (Em, MMZ1945) and S. mutans (Sp, MMZ1939) were each cultured in BHI to the exponential phase (OD₆₀₀ = 0.5). An inoculum of approximately 10⁶ CFU/mL of SK36, or its otherwise isogenic mutants, together with an inoculum of similar numbers of S. mutans were added to a TY medium or FMC supplemented with 20 mM glucose, 40 mM glycerol, or a combination of 20 mM glucose and 5 mM glycerol and then cultured for 20 hours in a 5% CO₂ environment at 37°C. At both the start and the end of the experiment, cultures were sonicated for 15 seconds, serially diluted, and plated on respective antibiotic plates to enumerate the CFU of both species. The competitive index was calculated using the following formula: [SSA(t_end)/SMU(t_end)] / [SSA(t_start)/SMU(t_start)], with values > 1 indicating SSA (S. sanguinis) being more competitive than SMU (S. mutans), and vice versa.

Dual-species biofilm assay

Pooled donor (n > 4) saliva was collected according to the IRB protocol (University of Florida IRB201500497), heat-inactivated by incubating at 60°C for 60 minutes, clarified by centrifugation, and filter-sterilized. Twenty-five percent (vol/vol) of saliva was then added to a biofilm base medium (BM) (34) supplemented with 2 mM sucrose and 18 mM glucose (BMGS). About 500 µL of BMGS and a sterile hydroxyapatite (HA) disk were placed in each well of a 24-well microtiter plate. Differentially marked S. sanguinis (Em) and S. mutans (Sp) strains were each cultured to the exponential phase and diluted at 100-fold into the microtiter plate, followed by incubation at 37°C in an anaerobic chamber supplied with 5% CO_2, 10% H_2, and 85% N₂ or in an aerobic incubator maintained with 5% CO₂. After 24 hours, the culture supernatant was removed gently and replaced with BM-glucose (18 mM), BM-glycerol (36 mM), or BM with no carbohydrate, followed by another day of incubation in the same environment. Subsequently, the HA disks were removed and washed three times by dipping in sterile PBS, from which bacterial biomass was dispersed into 1 mL PBS by sonication (FB120 water bath sonicator, Fisher Scientific) at 100% power, twice, for 15 seconds. Each sample was then serially diluted and plated onto selective BHI agar plates for CFU enumeration. Each strain was represented by four biological repeats and each condition by four biofilm samples.

RNA extraction and RT-qPCR

Bacterial strains were each inoculated into BHI for an overnight culture, which was diluted 20-fold into TY or FMC supplemented with carbohydrates as specified. To eliminate glycerol contamination, the overnight BHI cultures were collected by centrifugation and washed twice with sterile PBS before being diluted into the FMC media. Cells were harvested when the OD₆₀₀ reached 0.5–0.6, from which total RNA was extracted by following an established protocol using the RNeasy mini-kit and an RNase-free DNase I solution (Qiagen, Germantown, MD) for in-column gDNA removal (24). The total RNA was used in generation of cDNA using a reverse transcription kit (Bio-Rad, Hercules, CA) with gene-specific anti-sense primers, followed by quantitative PCR analysis using the CFX96 system and SYBR Green Supermix (Bio-Rad). The relative abundance of mRNA levels of each target gene was calculated using a ΔΔC_t method relative to a house-keeping gene (gyrA) used as the internal control.

Bioinformatic analysis

All bioinformatics analyses wwere carried out using the HiperGator cluster computer at the University of Florida. On all paired-end reads across all samples, quality control of sequencing data was performed with FASTQC (55), and adapter contamination and quality trimming (cutoff = 30) were done with Cutadapt (56). Subsequently, reads were mapped against reference genomes via BWA-mem (57) by removing those derived from human host (GRCh38) and 16S rRNA (Silva database) (58). We used Samtools (59) and FASTQ for handling and sorting of BAM files, and Bedtools (60) for recovery of unmapped reads. Finally, pathway profiling for each data set was carried out using the Humann2 (61) pipeline, and the subsequent differential expression analysis was done in R Statistical Language (62) with base and edgeR (63) packages.

Statistical analysis and data availability

Statistical analysis of the data was carried out using the software of Prism (GraphPad of Dotmatics, San Diego, CA). Any data, strains, and materials generated by this study will be available upon request from the authors for research or validation purposes.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/jb.00227-24.

Supplemental figures and tables. jb.00227-24-s0001.pdf.

Figures S1 to S5; Tables S1 and S3.

Supplemental table. jb.00227-24-s0002.xlsx.

Table S2.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.