Increased self- and co-aggregation in the ΔrpoE mutant
The S. mutans cells from the stationary growth phase were used for the self-aggregation analysis. The ΔrpoE mutant showed higher self-aggregation than the wild type in PBS buffer (). The reduction in optical density (OD600) in the ΔrpoE mutant reached 38% after 2 hours, while the wild type obtained only 11% reduction. Proteinase K treatment had no obvious effect on the wild type under these conditions. By contrast, proteinase K treatment strongly reduced the self-aggregation of the ΔrpoE mutant, suggesting that surface proteins contribute to its self-aggregation. However, the self-aggregation of the Proteinase K treated ΔrpoE mutant was still higher than that of the wild type, thus a different surface structure would be expected in the mutant.
Figure 1 Self-aggregation of S. mutans wild type () and the ΔrpoE mutant (□).
Dental biofilms start with the adherence of the initial colonizers, e.g. Streptococci, and Actinomyces 
, to the exposed salivary pellicle. Then genetically distinct species of microorganisms coaggregate with these pioneer colonizers through specific receptors 
. The ability of S. mutans
to coaggregate with other oral microorganisms, e.g. the initial colonizers S. oralis
, S. sanguinis
, and A. naeslundii
was therefore investigated, as well as the coaggregation with the yeast C. albicans
. Following the procedure described by Periasamy et al. 
, cells grown in THBY medium until the exponential phase were collected and transferred into the coaggregation buffer to give a similar cell density (OD600
about 0.4) before each pair of two strains was mixed together. The result was recorded after 90 min of coaggregation, and the self-aggregation of each strain was used as a control. As shown in , the wild type had weak self- and co-aggregation, and no visible flocs were formed. By contrast, the ΔrpoE
mutant formed flocs in the self-aggregation test and coaggregated with S. sanguinis
, A. naeslundii
, and C. albicans
. This is in line with previous findings that S. mutans
had weak coaggregation with A. naeslundii 
, and with C. albicans 
. The secreted or surface associated proteins in S. mutans
strains contribute to the bacterial cell-cell interaction, e.g. the surface antigen SpaP 
, have been reported. As shown in , treatment of Proteinase K reduced the coaggregation capability of both strains, which was indicated by the increased optical density (OD600
) in the supernatant of Proteinase K treated samples. In addition, the flocs formation of the ΔrpoE
mutant was eliminated by Proteinase K (). Expression of the aggregation-mediating proteins depends on the growth phase and the extent of aggregation is also influenced by the buffer, since Proteinase K treatment had no obvious effect on the self-aggregation of the wild type for stationary phase cells suspended in PBS buffer ().
Coaggregation of S. mutans wild type (WT) and ΔrpoE mutant with oral microorganisms without (A)/with (B) Proteinase K treatment.
is known to metabolize sucrose and to produce polysaccharides (glucans and fructans) to promote adherence and aggregation 
. In agreement with this, supplementation of the THBY growth medium with sucrose resulted in very effective coaggregation for both the wild type and the ΔrpoE
mutant with all other microorganisms (), and Proteinase K only partially diminished the coaggregation effect (), suggesting not only proteins, but also polysaccharides are involved in the coaggregation reaction, which is in line with previous findings 
Coaggregation of sucrose grown S. mutans wild type (WT) and ΔrpoE mutant with oral microorganisms without (A)/with (B) Proteinase K treatment.
Although S. mutans
has multiple mechanisms to bind directly to the tooth pellicle, it is not an initial colonizer 
and needs to coaggregate with the pioneer species to adhere and build a spatially organized community in dental biofilms. The ability to self-aggregate allows the quicker accumulation of bacterial cells. Therefore, the higher self- and co-aggregation of the ΔrpoE
mutant may help to become dominant in the oral bacterial community.
Biofilm structure and biofilm matrix assay
Differences in the structure of the biofilm surrounding matrix were revealed under the scanning electron microscope. The wild type biofilm matrix appeared to be more compact () than that of the ΔrpoE
mutant (). Moreover, unlike the smooth surface formed by the wild type (), the ΔrpoE
mutant produced dendrite-like extracellular components to attach on the surface (, arrow). This is consistent with our previous reports that the ΔrpoE
mutant formed a clumping inhomogeneous biofilm compared to the wild type 
Scanning electron microscopy of 16 h old biofilms of S. mutans strains.
The biofilm matrix is comprised of polysaccharides, proteins, and DNA together with other substances 
, thus the total amount of these three major components was quantified. The amount of extracellular insoluble polysaccharides and DNA was slightly less in the ΔrpoE
mutant than in the wild type when normalized to dry weight (). Lower yield of extracellular proteins was obtained in the ΔrpoE
mutant using the NaOH/EDTA extraction method, however, a similar amount of proteins was found in both strains by mild detergent triton extraction. Thus the ΔrpoE
mutant probably had a different extracellular protein composition compared to the wild type (). To investigate the effect of these components on biofilm formation, sodium meta
), Proteinase K, and DNase I were used as inhibitors for polysaccharides, proteins, and DNA, respectively. Treatment with Proteinase K caused partial detachment of 16 hour old biofilms of both strains, and moreover, adding the Proteinase K directly to the medium from the beginning of bacterial growth, strongly inhibited the adherence of both strains, thus no biofilm was formed (Figure S1
). Thus, the extracellular proteins are necessary for initial adhesion of S. mutans
to the surface.
Quantification of extracellular polysaccharides (A), proteins (B), and DNA (C) in S. mutans biofilms.
The proteins of the extracellular matrix were extracted and subjected to SDS-PAGE and the differentially expressed proteins were excised, digested with trypsin, and identified by MALDI-TOF (PMF and MS/MS) (Table S1
). As shown in , the ΔrpoE
mutant had a reduced expression of the cell surface antigen I/II SpaP, glucosyltransferases GtfB, and alcohol-acetaldehyde dehydrogenase AdhE; while increased expression of fructan hydrolase FruA was observed, all of which were consistent with our previous microarray data 
. AdhE is a bifunctional enzyme that is involved in carbon utilization and alcohol metabolism and its reduced expression has also been detected in our previous proteome study (Xue et al., submitted). The increased amount of fructan hydrolase FruA in the ΔrpoE
mutant could lead to quicker degradation of fructans, which are suggested as extracellular storage polysaccharides of S. mutans
. FruA has been shown to be necessary for cariogenicity of S. mutans 
and the FruA enzyme of S. salivarius
reduced the biofilm formation of S. mutans 
. Thus, the increased level of FruA in the ΔrpoE
mutant might contribute to its inhomogeneous biofilm structure. The glucosyltransferases GtfB is responsible for synthesis of water insoluble glucans 
, therefore the reduced amount of GtfB in the mutant strain explains its decreased extracellular insoluble polysaccharides as shown above (). The surface antigen SpaP in S. mutans
has been reported to contribute to biofilm formation in a glucan-binding independent way 
. Since both SpaP and GtfB displayed reduced expression in the ΔrpoE
mutant, the altered biofilm matrix structure and surface attachment shown above indicate a biofilm formation mechanism that is independent of SpaP and glucans. Two protein bands were clearly identified as GtfD, however, according to the molecular weight (about 160 kD), the second band could be truncated proteins. Since GtfD synthesizes water soluble glucans 
, the switch of GtfB to GtfD in the ΔrpoE
mutant could alter the glucan structure in a way that changes the biofilm matrix structure. Similar observations have been reported previously. The mutation of trigger factor (ropA
) caused reduced expression of the GtfB and GtfD enzymes, but the ΔropA
mutant had an increased biofilm formation compared to the wild type 
Extracellular matrix proteins of S. mutans biofilms.
Biofilm formation is important for S. mutans
to survive inside the host oral cavity and on other tissues, e.g. heart valves. The ΔrpoE
mutant produced a reduced amount of the extracellular proteins SpaP and GtfB and formed biofilms with a looser extracellular matrix. However, the dendrite-like structure of the ΔrpoE
mutant biofilm extracellular matrix might allow firm attachment to surfaces. This could benefit its colonization inside the human organism or tissue, where the concentration of sucrose is low compared to the oral cavity and the expression of Gtfs and SpaP should be low to avoid inducing host immune defense mechanisms. Indeed, the S. mutans
strains isolated from infective endocarditis patients were lacking the intact Gtfs enzymes and had lower sucrose-dependent adhesion 
. Moreover, mutant strains defective for the surface antigen SpaP resulted in less phagocytosis by human polymorphonuclear leukocytes, thus had a higher survival rate and caused more severe systemic inflammation 
The ΔrpoE mutant strongly bound to the human extracellular matrix (ECM) components
The ability to bind ECM is one of the major mechanisms for streptococcal pathogenesis 
. As shown in , the wild type bound poorly to all of the ECM molecules under our experimental conditions. The weak binding of the wild type strain to the ECM components could be due to differences in S. mutans
strains. Although S. mutans
has been reported to bind to ECM components, e.g. fibronectin through surface antigen I/II SpaP 
, to the cell wall associated protein WapA 
, the PavA-like protein (SMU. 1449) 
, and AtlA 
, none of these experiments was carried out using the UA159 strain. Moreover, the expression and activity of these receptors is highly regulated by environmental factors 
, thus our experimental conditions could have been not suitable for induction of the high binding activity. By contrast, the ΔrpoE
mutant effectively bound to the ECM components collagen I, collagen II, tenascin, laminin, and most strongly, to fibronectin. However, the ΔrpoE
mutant had decreased expression of genes encoding the known ECM binding proteins as reported in S. mutans
, e.g. surface antigen I/II SpaP, AtlA, and no changed transcription level of WapA and PavA-like protein according to our transcriptome data (microarray GEO record GSE22333). The strong adherence of the ΔrpoE
mutant to the ECM components, especially fibronectin, must therefore be due to some other differentially expressed surface receptor or modified surface structures.
Adherence of S. mutans strains to human extracellular matrix (ECM) components.
The ΔrpoE mutant had a reduced adherence to HEp-2 cells
Because of the strong adherence of the ΔrpoE
mutant to the ECM components, its adherence to and invasion of host cells was tested using human epithelial cells HEp-2 and primary human large vascular endothelial cells HUVEC. Both the wild type and the ΔrpoE
mutant showed low adhesion to the endothelial cells HUVEC (data not shown). In comparison, they had much higher adhesion to the epithelial cell line HEp-2 (). This is consistent with previous reports that S. mutans
was found to adhere to oral epithelial cells 
, and its biofilm triggered complex host immune responses 
Adherence of S. mutans wild type and the ΔrpoE mutant to human epithelial cells HEp-2.
Although weakly bound to the ECM matrix, the wild type effectively bound to HEp-2 cells. It may be possible that S. mutans adhesion can be triggered by the presence of host cells, or there might be an alternative adhesion mechanism independent of ECM binding. By contrast, the ΔrpoE mutant had reduced adherence compared to the wild type, especially at the later time points (after 3 hours of incubation) when both strains started fast multiplication. The scanning electron microscope images show the adhesion of the wild type () and the ΔrpoE mutant () to HEp-2 cells. The ΔrpoE mutant tended to clump together when attached to the surface of HEp-2 cells. Thus, the attachment area was relatively small compared to the big mass of the aggregates, and this might have caused easier detachment. Indeed, according to the observation during the experiment, at the later time points, the ΔrpoE mutant started to form detached flocs which were easily washed away at the washing step (data not shown).
wild type and the ΔrpoE
mutant had a low frequency of invasion to both epithelial and endothelial cells (data not shown), indicating that S. mutans
UA159 derived strains are not strongly invasive. In addition, the lactate dehydrogenase (LDH) enzyme released upon HEp-2 cell lysis was quantified to determine the cytotoxicity upon bacterial adhesion. No obvious differences in the released LDH amount between autolyzed cells and cells to which bacteria adhered were found (data not shown). This suggests that there were no holes in the cell membrane upon S. mutans
adherence, consistent with the previous finding that serotype c S. mutans
strains (including strain UA159), although they are the most prevalent strains in dental plaque, are not invasive 
. However, rare examples of S. mutans
invading epithelial () and endothelial cells (Figure S2
) could occasionally be observed. It seems that UA159 can invade cells, albeit at a very low level.
Invasion of human epithelial HEp-2 cells by S. mutans wild type cells.
The strong adherence of the ΔrpoE mutant to the ECM components, especially fibronectin, suggests that it might have a higher potential to bind to host cells. However, probably due to the clumping effect, the ΔrpoE mutant could not adhere as well as the wild type on the epithelial HEp-2 cells. Nevertheless, the invasion of human cells could be seen in both the wild type and the ΔrpoE mutant occasionally. With its reduced expression of Gtfs and SpaP, the ΔrpoE mutant probably has a lower antigenicity. Thus it might survive longer once it invades the host.
Characterization of the ΔrpoE mutant by phenotypic microarray (PM) assays
The PM technology is with 1,920 testable traits the most comprehensive approach for high throughput phenotyping 
. The system detects the conversion of carbon, nitrogen, phosphate and sulfate sources, but it also monitors the sensitivity for osmotic stress, various heavy metal ions, the pH and inhibitory chemicals. PM assays are performed in microtiter plate format and record the respiration of living cells by the NADH-dependend reduction of a tetrazolium redox dye. The formation of the purple color reflects both the import as well as the metabolic conversion of a specific substrate. The absence of enzymes, e.g. induced by gene knock-outs, results in lack of color formation. Measurement intervals of 15 minutes of the CCD camera are the prerequisite for the generation of PM-kinetics, which provide information about timing and strength of the cell's metabolic activity. The assay is more sensitive than traditional phenotypic growth tests on minimal medium because it also allows to monitor the usage of substrates that are not sufficient for growth 
Freshly grown S. mutans
wild type and mutant cells were inoculated within the complete set of all twenty PM plates (PM 1–PM 20). We established PM data from two biological replicates, e.g. two independent experiments, of both strains in order to investigate the reproducibility of these experiments. Among all twenty plates, the reproducibility was generally high for PM 1, PM 2, and PM 9 to PM 20, as shown in the comprehensive overview in Figure S3
. The respiration curves of PM 10 are shown as examples of good reproducibility for both the wild type and the ΔrpoE
mutant (Figure S3
The results from PM 3 to PM 8 exhibited a very low reproducibility. The poor metabolic response of S. mutans
in PM 3 to PM 8 assays has previously been reported 
. We improved it by modifying the pH of tricarballylic acid (see methods
section), but the results were still not satisfactory, thus, the results from PM 3 to PM 8 will not be discussed in this study. Although Biolog PM plates and protocols were conceived to be applicable to diverse bacterial lineages, a further improvement of specific assay conditions, including the provision of supplementary ingredients, is required to obtain optimal results for S. mutans
The PM data from PM 1, 2 and PM 9 to PM 20 were further analyzed by comparing the ΔrpoE
mutant with the wild type, and a general overview of both biological replicates is shown in the Figure S4
. The ΔrpoE
mutant and the wild type kinetics are colored in green and red, respectively. The predominant occurrence of green signals reflect a conspicuous gain of functions to metabolize sugars (PM 1 & PM 2) and an enhanced resistance against various antibiotics and toxic compounds (PM 9 to PM 20). A complete list of deviating phenotypes of the ΔrpoE
mutant including the plate type, the well position of each chemical compound and its mode of action is given in the Table S2
The comparison of the ΔrpoE
mutant with the wild type respiratory activity indicates that the mutant strain metabolized 20 additional carbon sources (). The 20 sugars that were metabolized by the ΔrpoE
mutant but not by the wild type are highlighted with black boxes, they include mono- (e.g. D-galactose, PM 01, A06), di- (e.g. sucrose, PM 01, D11), tri- (e.g. D-raffinose, PM 01, D 01), and tetra-saccharides (e.g. stachyose, PM 02, D 05) and the sugar derivates (e.g. D-mannitol, PM 01, B 11). The major involved pathways are the galactose and sucrose metabolism (). S. mutans
is known to be able to grow on all the carbon sources that were only metabolized by the RpoE mutant in the PM assays. Apparently, the respective enzymes were not synthesized by the wild type under the conditions of the PM assay. The regulatory defect of the RpoE mutant, i.e. its lack of transcriptional specificity, caused it to express a number of enzymes for sugar metabolism which are normally tightly regulated. This is confirmed by our previous transcriptome analysis 
, which suggested an alternative carbon metabolism in the ΔrpoE
mutant. We showed that the multiple sugar metabolism (MSM) system, which transports and metabolizes various sugars (e.g. raffinose, sucrose, and melibiose) and plays a major role in galactose and sucrose metabolism 
, was highly induced in the ΔrpoE
. In addition, the upregulation of other genes in the sucrose metabolism pathway, e.g. the sucrose specific transporter (scrA
) and the sucrose-6-phosphate hydrolase (scrB
) was also found in our microarray data (microarray GEO data GSE22333). Reproducible weaker metabolic activity of the mutant was only observed on four aldopentoses (PM 1: A02 L-arabinose, B08 D-xylose, C04 D-ribose, H06 L-xylose) that are involved in the pentose phosphate pathway. The difference between wild type and mutant was not above the threshold, but the data are supported by our proteome data showing that the mutant had reduced expression of phosphopentomutase DeoB, which catalyzes the intramolecular transfer of the phosphate group between Ribose-1P and Ribose-5P, an important step in the pentose phosphate pathway (Xue et al., submitted). It would have to be tested if the use of a broader spectrum of carbon sources by the ΔrpoE
mutant could provide a growth advantage under certain conditions.
Phenotype Microarray comparison of carbon source utilization by the ΔrpoE mutant compared to the wild type.
Gained and lost metabolic activity in carbon sources utilization in the S. mutans ΔrpoE mutant compared to the wild type.
The ability of the ΔrpoE
mutant to maintain respiratory activity in such a large number of assays for chemical sensitivity (PM 9 to PM 20) in contrast to the wild type was unexpected (, Figure S5
). The mutant showed an enhanced resistance against 142 different antibiotics or toxic compounds. For example, metabolic activity of the ΔrpoE
mutant was observed in the presence of chemicals which affect DNA synthesis, unwinding, and replication. Many inhibitors that block protein synthesis generated positive results for the ΔrpoE
mutant. The ΔrpoE
mutant was resistant to many toxic anions and cations, and chemicals that interfere with the tRNA synthetase, cell wall and membrane synthesis. Similar findings were also reported for the mutation of the histidine kinase gene liaS
in S. mutans
, which resulted in resistance to antibiotics that are targeting cell-wall biosynthesis, as well as antibiotics inhibiting protein and DNA synthesis 
. Since antibiotic resistance mechanisms are normally very specific, we speculate that the increased resistance to such a large number of antibiotics could result from general changes in the ΔrpoE
mutant. The modified surface structure may block the entry of antibiotics and the loosened transcriptional specificity might generate a larger variety of proteins allowing the cell to be readily prepared for facing toxic compounds. An alternative explanation might be the occurrence of secondary mutations in the mutant to compensate for the defects caused by the lack of RpoE. This is a general problem in bacteria. The careful investigation of various phenotypic traits in the ΔrpoE
mutant and its genetically complemented strain (
and unpublished data) did not show indications of secondary mutations. To minimize the possible impact of strain instability, all experiments were carried out with fresh glycerols from the same culture of the mutant.
Acquired resistance to antibiotics and toxic compounds in the S. mutans ΔrpoE mutant compared to the wild type.
Our previous data showed that the ΔrpoE
mutant was more sensitive to antibiotics that target protein synthesis, such as tetracycline and kanamycin 
. However, the PM assays showed that the mutant strain could maintain an active metabolism in contrast to the wild type in the presence of different concentrations of kanamycin (PM 11, H 05–H 08) and tetracycline (PM 12, A 05, 06) (Figure S5
). We therefore conducted independent growth tests which showed that the growth conditions strongly influenced the growth and antibiotic resistance of S. mutans
, and caused these variations. As shown in Figure S6
, when grown in 96-well microtiter plates at 37°C without enriched CO2
(similar growth condition as in the PM assays), the ΔrpoE
mutant had growth advantages compared to the wild type in the presence of 100 µg/ml kanamycin, while both strains could overcome the inhibitory effect of 1 µg/ml tetracycline after 48 hours of growth. However, when grown in 96-well microtiter plates at 37°C enriched with 5% CO2
, the wild type could overcome the inhibitory effect of kanamycin; by contrast, the growth of both strains was inhibited by tetracycline (Figure S6
). Our previous antibiotic sensitivity test was carried out at 37°C enriched with 5% CO2
, but growth was in falcon tubes rather than microtiter plates. As shown in Table S3
, under this condition the wild type strain had no difficulty to grow in the presence of both antibiotics, while the ΔrpoE
mutant failed to grow after the first 20 h, confirming our previous results. Better growth was also found in the closed system with a larger volume of cultures, which left less free space for air than that of smaller volume and thus less oxygen stress occurred. These different cultivation conditions, e.g. the extent of oxygen stress and CO2
supplementation, affect the growth and metabolism of both strains in different ways, and thus indirectly result in altered antibiotic sensitivity. In contrast to the drastic effect of growth conditions, the effects of growth media were less pronounced (data not shown). Since antibiotic treatment is still the most important means of disease control 
, the resistance of the ΔrpoE
mutant to such a large spectrum of antibiotics would make it potentially more difficult to be removed by conventional antibiotic treatment if it were able to colonize a host.
The results from this study show the multi-dimensional influence of RpoE on virulence related traits of S. mutans. The investigated traits are related to plaque formation potentially resulting in damage to host teeth (caries) as well as traits related to adherence, invasion, and survival in host cells and tissues. Both aspects are important for its pathogenesis. Our data are derived from in vitro experiments. They do not allow to predict the virulence or competitiveness of the RpoE mutant in vivo.
We previously showed that loss of RpoE resulted in massive changes in the transcriptome, and that these changes caused impaired growth, reduced stress tolerance, inhomogeneous biofilm structure, and decreased resistance to tetracycline and kanamycin 
. The slower growth and weaker resistance to hydrogen peroxide might impair the survival of the mutant in dental plaque, since many of the commensal oral streptococci produce hydrogen peroxide during aerobic growth.In contrast to these findings, we report in this study that the mutant has increased virulence related traits and broader metabolic functions, which however depended strongly on the cultivation conditions. The data show that the transcriptional specificity provided by RpoE restrains the physiology of S. mutans
. The more or less uncoordinated transcription of a large number of genes in the RpoE mutant results in the synthesis of diverse functional proteins. The work presented here shows that some of the newly expressed traits might increase the virulence of the ΔrpoE
mutant, e.g. by enhancing its resistance to antibiotics and toxic compounds, reducing its immunogenic surface properties, and altering its carbon metabolism, all of which could result in better survival in the host. Our data also show that many of the virulence related traits essentially depend on cultivation conditions, and thus extrapolation from laboratory data to in vivo
processes requires extreme caution.
From a genetic perspective, the observed differences of the phenotypes between the S. mutans wild type and ΔrpoE mutant are striking, especially since their genomes are identical, with the sole exception of the rpoE gene replaced by an erythromycin antibiotic resistance gene. The acquired metabolic capacities of the ΔrpoE mutant, which occurred simply due to the release of transcriptional specificity, indicate the physiological potential of S. mutans UA159. It is conceivable that processes occurring in nature, such as genetic mutations, horizontal gene transfer or environmental changes, could similarly trigger a release of genetic control, thus resulting in comparable phenotypic changes in microbes in the environment. Our data show that we generally see only a fraction of the physiological capabilities of a microorganism by the standard conditions in the laboratory.