Biofilm production is thought to be an important step in many enterococcal infections. In several Gram-positive bacteria, membrane glycolipids have been implicated in biofilm formation. We constructed a non-polar deletion mutant of a putative glucosyltransferase designated biofilm-associated glycolipid synthesis A (bgsA) in Enterococcus faecalis 12030. Analysis of major extracted glycolipids by nuclear magnetic resonance spectroscopy revealed that the cell membrane of 12030ΔbgsA was devoid of diglucosyl–diacylglycerol (DGlcDAG), while monoglucosyl–diacylglycerol was overrepresented. The cell walls of 12030ΔbgsA contained longer lipoteichoic acid molecules and were less hydrophobic than wild-type bacteria. Inactivation of bgsA in E. faecalis 12030 and E. faecalis V583 led to an almost complete arrest of biofilm formation on plastic surfaces. Overexpression of bgsA, on the other hand, resulted in increased biofilm production. While initial adherence was not affected, bgsA-deficient bacteria did not accumulate in the growing biofilm. Also, adherence of E. faecalisΔbgsA to Caco-2 cells was impaired. In a mouse bacteraemia model, E. faecalis 12030ΔbgsA was cleared more rapidly from the bloodstream than the wild-type strain. In summary, BgsA is a glycosyltransferase synthetizing DGlcDAG, a glycolipid and lipoteichoic acid precursor involved in biofilm accumulation, adherence to host cells, and virulence in vivo.
Enterococcus faecalis is a commensal microorganism of the human biliary and gastrointestinal tract and a leading cause of wound, bloodstream and urinary tract infections. Enterococci are important nosocomial pathogens frequently related to foreign body infections (Fabretti and Huebner, 2005). According to a recent study, they are the third most common isolate in prosthetic valve endocarditis and account for 12% of cases (Wang et al., 2007). Biofilm formation is critical in the pathogenesis of many of these infections. The production of biofilm is a multistep process involving the adherence of bacteria to a foreign body or host tissue, microcolony formation and matrix production, maturation of attached bacteria into a differentiated biofilm, and detachment of bacteria to the planktonic phase (Costerton et al., 1999). Multiple mechanisms have been described for biofilm formation by E. faecalis including expression of the adhesion protein Esp (enterococcal surface protein), the protease GelE (gelatinase E), pili encoded by the epb locus, a cell wall-anchored protein (Bee), the autolysin Atn, lipoteichoic acid (LTA) alanine esterification (DltA), and a putative glycosyltransferase (Epa) involved in the synthesis of a cell wall-associated polysaccharide (Toledo-Arana et al., 2001; Hancock and Perego, 2004; Mohamed et al., 2004; Pillai et al., 2004; Tendolkar et al., 2004; 2006; Fabretti et al., 2006; Nallapareddy et al., 2006; Mohamed and Huang, 2007). Of those, the Esp and GelE have been best characterized. Esp is a 202 kDa adhesin protein containing a core region consisting of repeating units that make up 50% of the molecule (Shankar et al., 1999). The role of Esp in biofilm formation was confirmed by isogenic deletion and heterologous expression in esp-negative strains (Tendolkar et al., 2004). Esp probably mediates adherence to plastic surfaces by enhancing cell surface hydrophobicity (Tendolkar et al., 2004). The expression of Esp varies between strains, is growth condition dependent, and is quantitatively correlating with adhesion to polystyrene (Van Wamel et al., 2007). Although Esp-positive strains are preferentially isolated in biofilm infections, no clear correlation between its presence and the ability to make biofilm could be established (Toledo-Arana et al., 2001; Sandoe et al., 2003; Mohamed et al., 2004). The secreted metalloprotease GelE is also involved in biofilm formation by E. faecalis, and conditioned supernatant of gelE-positive strains enhances biofilm thickness of gelE-negative strains (Kristich et al., 2004). A recent publication suggests that GelE is an autolysin supporting biofilm formation by increasing the concentration of genomic, extracellular DNA (Rice et al., 2007; Thomas et al., 2008). GelE expression is regulated by the fsr quorum-sensing locus, a homologue of the global regulatory system agr in Staphylococcus aureus (Qin et al. 2000; 2001), and fsr controls biofilm development through the production of GelE (Hancock and Perego, 2004). Again, no straightforward correlation of GelE expression and biofilm formation of clinical isolates has been observed (Baldassarri et al., 2006; Di Rosa et al., 2006). A recent publication established the presence of pili in E. faecalis and demonstrated their role in attachment to plastic surfaces and in virulence of biofilm infections like endocarditis and urinary tract infections (Nallapareddy et al., 2006; Singh et al., 2007). Transposon mutagenesis to screen for genes involved in enterococcal biofilms confirmed the role of pili and other genetic loci previously described, but interestingly it failed to reveal any genes involved in polysaccharide biosynthesis (Kristich et al., 2008). This bias may reflect the limitation of the microtiter plate bioflm assay to screen for biofilm-defective mutants because it primarily reflects contributions from initial adherence and early stages of biofilm formation. In other Gram-positive biofilm producers, like S. aureus and Bacillus subtilis, membrane glycolipids and LTA content of the cell wall have been implicated in biofilm formation (Kiriukhin et al., 2001; Lazarevic et al., 2005; Fedtke et al., 2007). The aim of the current study was to examine the contribution of glycolipids and LTA to the formation of biofilm in E. faecalis in vivo and in vitro.
Deletion of EF2891
Enterococcus faecalis gene EF2891 shares high identity and similarity to ALdgs, a gene that encodes a diglucosyl–diacylglycerol (DGlcDAG) synthetase in Acheloplasma laidlawii and several other Gram-positive bacteria (see Table S1). It also shares high similarity with iagA, a gene required for anchoring of LTA to the cell membrane and for invasion across the blood–brain barrier by Streptococcus agalactiae (Doran et al., 2005). Because of its association with biofilm formation in E. faecalis (see below), we designated this gene biofilm-associated glycolipid synthesis A (bgsA). Basic local alignment search tool (blastp) analysis identified a second putative glucosyltransferase (TIGR number 2890), immediately downstream of bgsA (Fig. S1). To characterize the role of bgsA, a non-polar deletion mutant was created. Inactivation of this gene was accomplished by targeted mutagenesis (Fabretti et al., 2006). An internal fragment of 863 bp was deleted in the mutant 12030ΔbgsA (Fig. S1). Single gene reconstitution of bgsA completely restored the wild-type phenotype, including the electrophoretic mobility of LTA during SDS-PAGE (Fig. S2) and the ability to produce biofilm (see below).
Deletion of bgsA leads to a replacement of DGlcDAG by MGlcDAG as the major cell membrane glycolipid in E. faecalis
The most abundant glycolipid in E. faecalis is DGlcDAG, which accounts for 8–37% of all polar lipids of the cell membrane and for at least 68% of the total glycolipids (Fischer et al., 1978). The remaining glycolipids are mostly composed of α-monoglucosyl–diacylglycerol (MGlcDAG) (Fischer et al., 1978). Western blot analysis of cell wall-associated antigens of 12030ΔbgsA showed two major differences compared with the wild-type strain: First, the migration of LTA of the mutant in the gel was distinctly retarded (Fig. S2). This phenomenon was also observed by Grundling and Schneewind in mutants defective in the glycosylation of DAG in S. aureus (Gründling and Schneewind, 2007a). Second, a low-molecular-mass band (≈6 kDa) was missing in 12030ΔbgsA (Fig. S2). We hypothesized that this band represented a glycolipid containing α-kojibiosylglycerol, an epitope recognized by the antiserum used in the experiment. To further investigate the glycolipids in 12030ΔbgsA, polar lipids of the wild type and mutant strain were extracted according to the procedure described by Bligh and Dyer and separated on thin-layer chromatography (TLC) (Fig. 1). Staining with α-naphtol revealed one major (Rf 0.57) and four minor bands (Rf 0.80, 0.44, 0.35, 0.17; Fig. 1). Purification of the major glycolipid (Rf 0.57) was performed by preparative-layer chromatography (PLC), and the product was structurally characterized. Compositional analysis revealed the presence of glycerol (Gro), d-glucose (d-Glc), and the major fatty acids, octadecenoic acid and hexadecanoic acid, with traces of octadecanoic, hexadecenoic and cyclo-nonadecanoic acid also being detected. The structure of the glycolipid was established by nuclear magnetic resonance (NMR) spectroscopy. The 1H NMR spectrum revealed two doublets in the anomeric region at δ 5.01 (J1,2 3.2 Hz) and δ 4.96 (J1,2 3.6 Hz) corresponding to two α-linked Glcp residues (Fig. 2). A multiplet at δ 5.24 corresponded to H2 of glycerol, which was acylated, as indicated by its strong deshielding. A third signal, a triplet at δ 5.34 (δC 134.4), was assigned to the vinyl protons of an acyl chain. Further characteristic fatty acid signals were also identified. The detailed structural analysis of the glycolipid is presented in the Supporting information (Table S2). The structure of the glycolipid was 1,2-diacyl-3-[α-d-Glcp-(1→2)-α-d-Glcp]-sn-Gro (DGlcDAG, for structure also see Fig. 1).
In contrast, separation of glycolipids from 12030ΔbgsA by TLC revealed only a single glycolipid (Rf 0.80; Fig. 1). This band was also present in small amounts in the wild-type strain. Compositional analysis revealed the presence of Gro, d-Glc and fatty acids. NMR spectrometry of the compound was performed (see Supporting information for detailed analysis). The major fatty acids detected were comparable to those found in DGlcDAG. All 1H chemical shifts and 1H,1H coupling constants of MGlcDAG (native) isolated from strain E. faecalis were established from the 1H NMR spectrum (Table S3). This and data obtained on the deacylated sample (see Supporting information) identified the structure as 1,2-diacyl-3-α-d-Glcp-sn-Gro (MGlcDAG, for structure also see Fig. 1).
In summary, the major glycolipid in E. faecalis 12030 was DGlcDAG, while MGlcDAG represented only a minor proportion of extracted glycolipids. In contrast, DGlcDAG was absent in 12030ΔbgsA and replaced by MGlcDAG. Glycolipids with a lower mobility in TLC (Rf 0.44, 0.35 and 0.17) were not purified for structural analysis. According to Fischer, these glycolipids corresponded to phosphatidylglycolipids, glycerolphosphoglycolipids and glycerophosphophosphatidylglycolipids (Fischer, 1990); however, we did not confirm their structures in this study. Two of the three phosphoglycolipids (Rf 0.44 and 0.17) were also absent in the mutant E. faecalis 12030ΔbgsA (Fig. 1).
Structure of LTA
Previous studies have shown that alterations in the glycolipid anchor of LTA retard its migration in SDS-PAGE electrophoresis (Gründling and Schneewind, 2007a). To investigate structural changes of the LTA molecule, we isolated LTA from 12030 wild type and from 12030ΔbgsA.1H NMR analyses in D2O of the isolated LTA samples were performed (Fig. 3). Comparison of both spectra showed clearly the identity of the chemical structure in both samples to be a 1,3-poly(glycerol phosphate) non-stoichiometrically substituted at position C-2 of the glycerol residues with alanine and kojibiose, the later also non-stoichiometrically substituted at position C-6 by alanine in any of both glucose residues, as published previously (Theilacker et al., 2006). In order to compare the length of the polyglycerolphosphate chain, integration of isolated signals was undertaken based on the intensities of the typical fatty acid signals [α-proton at δ 2.33, β-proton at 2.018 (small signals), -CH2- at δ 1.26–1.29, and -CH3 at δ 0.88] as described previously (Morath et al., 2001). The signals corresponding to unsubstituted glycerol (Gro) (H2 at δ 4.044) and the Gro substituted by kojibiose (H2 at δ 4.197) are part of a multiplet; therefore, integration values could not be taken into account. This makes it impossible to calculate the total number of polyglycerolphosphate repeating units in the LTA samples. Nevertheless, it is clear that the overall chain length was longer in LTA from 12030ΔbgsA, as the proportion of carbohydrates to lipids in this case is higher than that in the case of the wild type (Fig. 3).
Exclusion of pleiotropic effects
Wild type and mutant 12030ΔbgsA displayed equivalent growth kinetics and stationary phase survival in tryptic soy broth (TSB) and transmission electron microscopy revealed no major differences in cell morphology (data not shown). SDS-PAGE analysis of cell wall-associated proteins was also similar for wild type and mutant strain (data not shown). Analysis of phospholipids and aminophospholipids in the cell membrane by semiquantitative analysis by TLC was comparable as well (data not shown). The rate of autolysis was about 10% after 180 min in hypotonic phosphate buffer for both 12030ΔbgsA and wild-type strain (data not shown). Sensitivity to cationic antimicrobial peptides including polymyxin B, colistin and nisin did not differ between wild type and mutant (data not shown).
Quantification of cell wall-associated LTA and hydrophobicity
To determine the impact of changes in glycolipid composition of the cell membrane on its LTA content, we quantified surface-bound LTA by enzyme-linked immunosorbent assay (ELISA). Deletion of bgsA resulted in a 3.3-fold increase in cell wall-associated LTA in the mutant (Fig. S3). As mentioned above, the LTA molecules of 12030ΔbgsA were also longer compared with wild-type bacteria. Increased chain length or increased kojibiose substitution (see above) may therefore account for higher amount of LTA measured in the 12030ΔbgsA by our immunoassay. Levels of LTA shed into the growth medium did not differ significantly between 12030 wild type and 12030ΔbgsA (data not shown). We also evaluated the hydrophobicity of bacterial cells using distribution into the organic or aqueous phase in a dodecane/buffer mixture. Compared with wild-type bacteria, significantly less 12030ΔbgsA diffused into the organic phase (Fig. 4).
In S. aureus and B. subtilis, deletion mutants with defective glycolipid synthesis showing a decreased ability to form biofilm have been reported (Kiriukhin et al., 2001; Doran et al., 2005; Lazarevic et al., 2005; Fedtke et al., 2007). To quantify the biofilm in 12030ΔbgsA, bacteria were grown for 18 h in polystyrene microtiter plates and stained with crystal violet. Deletion of bgsA in E. faecalis 12030 caused a significant reduction in biofilm production (Fig. 5). Reconstitution of the gene restored the ability to form biofilm (Fig. 5). Strain 12030 is a strong biofilm producer compared with other strains frequently used to study E. faecalis biofilm formation (Fig. 5). In order to examine whether bgsA is also involved in biofilm production in another strain background, we inactivated bgsA in E. faecalis strain V583 by chromosomal insertion. Strain V583 elaborates less biofilm than 12030 on plastic surfaces (Fig. 5). The mutant V583ΔbgsA was also highly impaired in biofilm production confirming our findings in strain 12030 (Fig. 5).
Overexpression of bgsA increases biofilm formation by E. faecalis 12030
To further corroborate the role of bgsA in biofilm formation in E. faecalis, we evaluated the effect of overexpression of this gene on biofilm formation. Additional copies of bgsA were introduced into E. faecalis 12030 wild type on a Gram-positive expression plasmid (pEU327) under control of a constitutively expressed xylose promoter. For maintenance of the plasmid in vitro the biofilm assays had to be performed in TSA plus spectinomycin. To exclude an effect of this antibiotic on biofilm production, wild-type bacteria containing pEU327 without the bgsA gene were used as control (Fig. 6). Biofilm production of pEU327-complemented E. faecalis 12030 was significantly reduced in the presence of spectinomycin (Fig. 6). Bacteria containing pEU327/bgsA, however, produced significantly more biofilm compared with bacteria containing only pEU327 and with wild-type bacteria without the plasmid (Fig. 6). Our data imply that the number of copies of the bgsA gene correlates with the increased accretion of biofilm.
Initial adherence and biofilm accumulation
The inability to produce biofilm may be caused by an impaired initial adherence or by defective accumulation of biofilm mass in the later stages. To differentiate between these mechanisms, we monitored the production of biofilm over time (Fig. 7). During the first 2 h of incubation no difference in adherence between wild-type bacteria and the 12030ΔbgsA mutant was observed. In the later stages of biofilm development, wild-type 12030 continued to accumulate more biofilm while 12030ΔbgsA did not increase its biofilm density beyond the level found at 2 h. The increasing difference in biofilm mass between the strains indicates a defective biofilm accumulation in 12030ΔbgsA.
Biofilm in mixed culture
Biofilm accumulation may be mediated by components of the cell wall interacting with other cells or by extracellular macromolecules of the matrix, which anchor bacteria within the biofilm. In a mixed culture, wild-type bacteria could provide compounds that enable the accumulation of bgsA-deficient bacteria in the growing biofilm. To test this hypothesis, we cocultured equal numbers of V583 wild type and V583ΔbgsA on polystyrene, detached adherent bacteria after 18 h of incubation, and quantified the wild-type and mutant bacteria utilizing the antibiotic resistance marker of the chromosomal insertion of V583ΔbgsA for selection (Fig. 8). Compared with the pure culture of V583 wild type and V583ΔbgsA, no difference in relative numbers of bgsA-deficient bacteria in the mixed biofilm was found, indicating that even in multicellular communities with wild-type bacteria V583ΔbgsA is unable to accumulate efficiently within the growing biofilm.
Adhesion to Caco-2 cells
While adhesion to plastic material such as polystyrene may play a role in certain settings (i.e. infections related to intravenous or urinary catheters), we also wanted to assess if biofilm formation was reduced on clinically relevant epithelial cells. As enterococci are normal commensals of the gastrointestinal tract, adhesion to Caco-2 cells (a human colon carcinoma cell line) was investigated. Figure 9A shows that attachment of 12030ΔbgsA to Caco-2 cells was also modestly reduced compared with the wild type (50.3% of wild-type level). Neither E. faecalis 12030 wild type nor 12030ΔbgsA invaded Caco-2 cells in significant numbers, confirming the results obtained with HEp-2 cells by Fabretti et al. (2006).
Opsonophagocytosis and in vivo virulence
Compared with wild-type bacteria, E. faecalis 12030ΔbgsA was more susceptible to opsonophagocytic killing mediated by polyclonal rabbit antiserum raised against whole bacterial cells of the same strain (96% versus 87%, P = 0.03, Mann–Whitney t-test). However, when using an antiserum specific for LTA of 12030 no significant difference was found (data not shown). Neither E. faecalis 12030 nor 12030ΔbgsA was sensitive to complement-mediated killing at the chosen complement concentration (data not shown). We have shown previously that intravenous challenge with a high inoculum of E. faecalis strain 12030 is not lethal to mice but causes a bacteraemia lasting several days before clearance (Huebner et al., 2000). While Swiss-Webster mice challenged with E. faecalis 12030 wild type were still bacteraemic 72 h after challenge, almost all mice infected with 12030ΔbgsA had cleared the infection from their bloodstream (Fig. 9B). The level of bacteraemia after 72 h of E. faecalis 12030ΔbgsA was also significantly lower in this infection model using BALB/c mice (data not shown).
A variety of mechanisms have been described for the biofilm formation in E. faecalis (Mohamed and Huang, 2007). Except for a putative glycosyltransferase of the epa operon (Mohamed et al., 2004), which is thought to be involved in the biosynthesis of a cell wall-associated polysaccharide, all known genes involved in biofilm production encode for proteins. In other Gram-positive pathogens, however, mutations of genes involved in the synthesis of precursors of the glycolipid and LTA synthesis also affect biofilm production (Branda et al., 2004; Lazarevic et al., 2005; Fedtke et al., 2007). A blast search (blastp) revealed that a homologue of a glycosyltransferases of A. laidlawii is also present in E. faecalis (EF2891, 48% sequence identity with Aldgs from A. laidlawii). We designated the homologue of Aldgs in E. faecalis, EF2891, bgsA and constructed a deletion mutant of bgsA in E. faecalis 12030.
The deletion of bgsA had profound consequences on the phenotype of E. faecalis 12030. Most notably, it led to an almost complete loss in the ability to form biofilms on plastic surfaces. In addition, adherence to colonic epithelial cells and virulence in a mouse bacteraemia model were significantly reduced. On the contrary, many other aspects of basic bacterial physiology like cell morphology, growth rate, expression of cell wall-associated proteins, autolysis and sensitivity to antimicrobial peptides were not affected in the 12030ΔbgsA mutant. As inactivation of bgsA primarily affected biofilm formation we studied this process in more detail.
Because E. faecalis 12030 is a strong biofilm producer, we wanted to confirm our findings in a strain that elaborates less biofilm. We inactivated bgsA in E. faecalis V583, a fully sequenced vancomycin-resistant clinical isolate, by chromosomal insertion (Paulsen et al., 2003). Regardless of the strain background and method of genetic manipulation used, inactivation of bgsA resulted in loss of biofilm formation. To further corroborate the role of bgsA in biofilm formation we overexpressed this gene by introducing additional copies of the bgsA gene under the control of a constitutively expressed xylose promoter into wild-type 12030. We could demonstrate that overexpression of bgsA significantly enhanced biofilm production when compared with wild-type bacteria or controls containing the empty plasmid. Our data indicate that the number of gene copies of bgsA correlates with the biofilm mass, underlining the importance of this gene in the establishment of multicellular communities of E. faecalis.
Closer analysis of biofilm formation revealed that the accumulation of bacteria in the growing biofilm, but not initial adherence to polystyrene, was disrupted in the 12030ΔbgsA mutant. Even when grown in a mixed culture with wild-type bacteria, bgsA-deficient bacterial cells did not accumulate in the biofilm, suggesting that secreted matrix polymers or cell wall components provided by wild-type cells cannot anchor mutant cells within the growing biofilm. Adherence and accumulation as different stages of biofilm development are still incompletely understood in E. faecalis. Mohamed and coworkers examined mutants in biofilm-associated gene loci (fsr, gelE, atn and epa) and found that these mutations primarily affect initial adherence (Mohamed et al., 2004). Disruption of esp in E. faecium also comprised initial adherence (Heikens et al., 2007). To our knowledge, bgsA is the first gene locus that seems to be primarily involved in the accumulation of E. faecalis in biofilms.
To elucidate a potential mechanism for the biofilm-negative phenotype of E. faecalis 12030ΔbgsA we characterized glycolipids and LTA of wild-type and mutant cells in more detail. DGlcDAG synthesis was completely abolished in 12030ΔbgsA and this glycolipid was replaced by its putative precursor MGlcDAG. This indicates that BgsA is a (1→2) glucosyl transferase synthesizing the glucosylation of MGlcDAG to yield DGlcDAG, a function also designated to its homologue in A. laidlawii, alDGS (Edman et al., 2003). As DGlcDAG is the major bilayer-forming glycolipid in E. faecalis, we expected compensational changes in the concentration of other polar lipids of the cell membrane. Analysis of total phospholipids and aminophospholipids by TLC, however, did not reveal major differences compared with wild-type cells. Deletion of bgsA did, nonetheless, affect LTA synthesis in E. faecalis. The polyglycerolphosphate moiety LTA from 12030ΔbgsA was longer and contained more kojibiose and alanine substituents at the C2 of glycerol.
The multiple structural changes of the cell envelope make it difficult to delineate the specific function of glycolipids and LTA in enterococcal biofilms. Separate mutants to segregate LTA and glycolipid biosynthesis would be helpful to better define their individual role. LTA, however, is essential for viability of Gram-positive bacteria making it impossible to generate mutants devoid of LTA (Gründling and Schneewind, 2007b; Jerga et al., 2007). Hence, evidence for the role of LTA in biofilm formation is only indirect. In B. subtilis, a gene encoding an α-phosphoglucomutase producing precursors for the synthesis of glycolipids, LTA and wall teichoic acid was found to be involved in biofilm formation (Pooley et al., 1987; Branda et al., 2004; Lazarevic et al., 2005). Mutations of the glycosyltranferase YpfP, an enzyme that adds consecutively two glucose residues onto DAG in S. aureus and B. subtilis, have a variable effect on the biofilm phenotype and LTA expression depending on Gram-positive pathogen and strain background used (Kiriukhin et al., 2001; Lazarevic et al., 2005; Fedtke et al., 2007). YpfP mutants can no longer produce glycolipids, and the glycolipid anchor of LTA is replaced by DAG (Fedtke et al., 2007; Gründling and Schneewind, 2007a). In S. aureus RN4220, a poor biofilm producer, the inactivation of ypfP increases LTA turnover and the overall content of LTA in the cell wall (Kiriukhin et al., 2001). In a different strain, S. aureus SA113, deletion of the same gene resulted in reduced amounts of cellular LTA and a highly impaired ability to form biofilms on plastic and glass surfaces (Fedtke et al., 2007). This finding is in contrast to our results in E. faecalis 12030ΔbgsA, a strain also defective in a glycosyltransferase that provides precursors for LTA synthesis. Increased polymer length, not only increased LTA concentration in the cell wall, however, may account for these discrepancies. Also, the different structural features of our mutant LTA may alter the ability of this compound to interact with plastic surfaces by themselves.
Another characteristic of 12030ΔbgsA that could explain its decreased ability to grow in a biofilm is the reduced hydrophobicity of mutant bacteria. Hydrophobic interaction has been shown to be a major determinant in biofilm formation (Bos et al., 1999) and has been suggested as a mechanism how Esp mediates biofilm production (Tendolkar et al., 2004). Secondary effects not specifically addressed in this study may also be involved in the impaired biofilm formation of 12030ΔbgsA. Changes in glycolipid composition, for example, lead to an altered cell membrane elasticity and fluidity (Edman et al., 2003) and may affect membrane protein folding and stability (Hong and Tamm, 2004).
Apart from the diminished ability to form biofilm on plastic surfaces, 12030ΔbgsA also adhered less to Caco-2 cells than wild-type bacteria. As a gut commensal, adherence to colonic epithelium is the first step in translocation across the intestinal barrier. Our data indicate that bgsA may also be critical for adherence to epithelial cells of the gastrointestinal tract. Two protein adhesins, Esp and Ace, and pili have been previously implicated in the adherence of enterococci to host cells (Rich et al., 1999; Shankar et al., 1999; Nallapareddy et al., 2006; Singh et al., 2007). Our results may point to additional mechanisms in the adhesion of enterococci to host cells. Reduced invasion of brain microvascular endothelial cells has also been described for the iagA mutant in group B Streptococcus (Doran et al., 2005). However, adherence to a brain microvascular endothelial cell line was not affected. As E. faecalis invades neither Caco-2 nor HEp-2 cells efficiently, it is difficult to compare our results with the study by Doran. Further studies must address the role of bgsA and its substrates in adhesion to host tissues.
Aside from reduced biofilm formation and attachment to colonic epithelium, 12030ΔbgsA was also attenuated in a mouse bacteraemia model. Although this model does not imitate a classical biofilm infection, biofilm-defective mutants in E. faecalis (bopD, dltA) and E. faecium (I. Sava, unpublished personal observation) were also unable to cause prolonged bacteraemia indicating that this model may be of some value in the characterization of biofilm-associated genes (Hufnagel et al., 2004a; Fabretti et al., 2006). Further studies, however, are needed to confirm these findings in true biofilm infections.
In summary, inactivation of bgsA leads to abolished biosynthesis of DGlcDAG, the major cell membrane glycolipid in E. faecalis, resulting in a profound impairment of accumulation of bacterial cells in biofilms, reduced adhesion to the gastrointestinal mucosa, and a shorter duration of bacteraemia in vivo. Future studies need to address the mechanism how glycolipid biosynthesis affects biofilm formation in more detail.
Bacterial strains, plasmids and growth conditions
The bacterial strains and plasmids used in this study are shown in Table S4. E. faecalis strain 12030 is a clinical isolate obtained from Cleveland, OH. It is a strong biofilm producer and is opsonized by antibodies against LTA. The strain has been further characterized in several publications (Huebner et al., 1999; 2000; Hufnagel et al., 2004a,b; Theilacker et al., 2006; McBride et al., 2007). Enterococci were grown at 37°C without agitation in TSB (or CASO-broth; Merck) or M17 broth (Difco Laboratories), and addition of 1% glucose as indicated (TSBG), or on tryptic soy agar or M17 agar respectively. Escherichia coli DH5α and TOP10 (Invitrogen) were cultivated aerobically in Luria–Bertani broth, at 37°C. Kanamycin was added for enterococci (1 mg ml−1) and for E. coli (50 μg ml−1); tetracyclin was used at 12.5 μg ml−1 for E. coli and 10 μg ml−1 for enterococci. All antibiotics were from Sigma Chemical.
Chromosomal DNA from enterococci was prepared using the DNeasy Tissue Kit (Qiagen) according to the manufacturer's instructions. Plasmid DNA was prepared from enterococci or E. coli using the Wizard Plus SV Miniprep System or PureYield Plasmid Midiprep System (Promega). DNA was purified from polymerase chain reactions (PCRs) or from agarose gels using the QIAquick PCR purification Kit or the Gel Extraction Kit (Qiagen) according to the manufacturer's instructions. Restriction and modifying enzymes were obtained from Invitrogen and New England Biolabs. Custom primers were manufactured by Invitrogen. Electrocompetent enterococci were prepared according to the method of Fiedler and Wirth (1991), and electroporation was performed in a Bio-Rad Gene Pulser Xcell Electroporation System using the parameters given by Fiedler and Wirth (1991). All the other methods (DNA ligations, electrophoresis, transformation of competent E. coli, and SDS-PAGE) were performed using standard techniques (Sambrook and Russell, 2001).
Construction of a non-polar deletion mutant
A non-polar deletion of a portion of the gene designated bgsA (EF2891, in the E. faecalis V583 genome, GenBank ID accession number AAO82580) was created using the method described by Cieslewicz et al. (2001) with the following modifications: primers 1 and 2 (Table S5) were used to amplify a 599 bp fragment from the region upstream into the gene bgsA, and primers 3 and 4 to amplify a 526 bp fragment from the gene bgsA end to the gene EF2890 (Fig. S1). Primers 2 and 3 contain a 21 bp complementary sequence (underlined in Table S5). Overlap extension PCR was used to create a PCR product lacking a portion of EF2891 gene. The resulting fragment was cloned into the Gram-negative cloning vector pCRII-TOPO (Invitrogen) and cut with the restriction enzyme EcoRI (Invitrogen); the fragment was then inserted into the shuttle vector pCASPER containing a temperature-sensitive origin of replication. The resulting plasmid, pCASPER/ΔbgsA, was transformed into E. faecalis 12030 by electroporation, and integrants were selected at the non-permissive temperature (42°C) on TSA plates with kanamycin. A single colony was picked, and insertion of plasmid into the chromosome was confirmed by PCR. The integrant was passaged 10 times in liquid culture without antibiotic at the permissive temperature (30°C), and colonies were replica-plated to screen for loss of kanamycin resistance. The excision of the plasmid either creates a reconstituted wild-type strain or leads to an allelic replacement with the deleted sequence in the chromosome. The deletion mutant created was designated E. faecalis 12030ΔbgsA, and the genotype was confirmed by PCR and automated sequencing.
Single gene reconstitution of 12030ΔbgsA
For reconstitution of the 12030ΔbgsA mutant, a DNA fragment of 1988 bp containing the entire bgsA gene as well as upstream (526 bp) and downstream (457 bp) sequences was amplified by PCR using primers 1 and 4 (Table S5). The resulting PCR product was cloned into the Gram-negative cloning vector pCRII-TOPO (Invitrogen) and cut with the restriction enzyme EcoRI (Invitrogen). The fragment was further cloned into the Gram-positive vector pCASPER, which contains a temperature-sensitive origin of replication. The recombinant plasmid was electroporated into E. faecalis 12030ΔbgsA competent cells. Integrants were selected at the non-permissive temperature on TSA plates with kanamycin. Colony PCR was done to confirm the insertion of the plasmid into the chromosome. The integrants were passaged between 6 and 10 times in broth without antibiotic at the permissive temperature and screened by replica plating for loss of kanamycin resistance. Kanamycin-susceptible clones were analysed by PCR for the presence of the intact bgsA gene.
Insertional inactivation of bgsA
The inactivation of bgsA by chromosomal insertion was done using a method based on the conditional replication of the pORI19-1/pG+host3 system (Law et al., 1995). A 622 bp sequence of the bgsA gene was amplified using primers EF2891 FS and EF2891 RA (Table S5). The PCR product was digested with SbfI and AvrII and ligated into the NsiI/SpeI-digested RepA vector pVE14218. Plasmid pG+host3 is a thermosensitive plasmid which provides the RepA protein to pVE14218 at the permissive temperature (28°C). E. faecalis V583ΔABC is a strain cured of all plasmids and was subsequently transformed with the helper plasmid pG+host3 and with the pVE14218 plasmid containing the 622 bp chromosomal DNA fragment. Transformants obtained at the permissive temperature were transferred to 42°C in order to select for insertional mutants. At 42°C, pG+host3 is cured and only integrants resulting from integration of pVE14218 into the chromosome are selected by plating bacteria at 42°C on media containing tetracycline. The transformants were analysed for the presence of the integrated DNA fragment by PCR and Southern Blot.
Overexpression of bgsA gene
To evaluate the effect of overexpression of bgsA on biofilm formation, additional copies of the gene were introduced into E. faecalis 12030 on a pEU327 plasmid under control of a constitutively expressed xylose promoter (Hancock and Gilmore, 2002). A 1005 bp PCR product that contains the entire bgsA gene was generated (primers bgsA HINDIII and bgsA HINDIII rev, Table S5) and cloned into pEU327. The proper orientation of the insert was determined by restriction analysis using BglI and HincII. The resulting plasmid designated pEU327/bgsA was electroporated into E. faecalis 12030 and integrants were selected on spectinomycin 500 μg ml−1.
Biofilm plate assay
Enterococci were tested for production of biofilm using the protocol described by Baldassarri et al. (2001). Briefly, bacteria were grown at 37°C in TSB for 18 h. Polystyrene tissue-culture plates (Greiner, Nürtingen, Germany) were filled with 180 μl of TSBG and 20 μl of this culture, and the plates were then incubated at 37°C for 18 h. The plates were read in an ELISA reader (Bio-Rad Microplate reader) at an optical density of 630 nm. The culture medium was then discarded and the wells were washed three times with 200 μl of phosphate-buffered saline (PBS) without disturbing the biofilm on the bottom of the wells. The plates were dried at 60°C for 1 h and then stained with 2% Hucker's crystal violet for 2 min. Excess stain was removed by rinsing the plates under tap water, and the plates were then dried at 60°C for 10 min. The optical density at 630 nm was determined.
Biofilm in mixed culture
After 18 h of culture in TSB at 37°C, E. faecalis V583 and V583ΔbgsA were adjusted A630 of 0.4 and then diluted 1:100 in TSB + 1% glucose. Equal numbers of bacteria in the coculture were confirmed by quantitative bacterial counts on TSA and TSA plus tetracycline. The biofilm assay with the mixed culture was performed as described above. After 18 h of incubation at 37°C either duplicate plates were stained with crystal violet as described above or bacterial cells were detached from the plates as described previously (Kristian et al., 2008). Briefly, microtiter plates were washed three times to remove non-adherent bacteria, and wells were filled with 100 μl of Hank's balanced salt solution (HBSS plus 1 mmol l−1 calcium, 0.05% glucose, and 5 mmol l−1 HEPES (HBSS- HEPES) and bacteria were mobilized by repeated pipetting. The supernatant was collected and residual bacteria were detached with 0.5% trypsin and 0.2% EDTA in saline for 5 min and combined with the supernatant. The complete removal of the biofilm was confirmed by staining with crystal violet. The combined detached biofilm was diluted in 0.9% sodium chloride, 0.15% EDTA, and 0.1% Triton X-100, vortexed vigorously and sonicated for 1 min at 120 W. Bacterial counts were quantified on TSA and TSA plus tetracyclin.
Adherence and invasion assay
A human cell line, Caco-2, derived from a colon carcinoma, was used in this study. Cells were cultivated in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum and 1% nonessential amino acids in a 5% CO2 atmosphere. All the experiments were performed on cells between the 15th and 25th passages. Caco-2 cells were cultivated in 24 well plates to a density of 1 × 105 cells/well for 3–5 days. Bacteria were grown to mid log phase at 37°C without agitation in TSB. Prior to inoculation, bacteria were washed and resuspended in DMEM. Caco-2 cells were incubated with bacteria for 2 h at a bacteria-to-cell ratio of 100:1. After infection of the monolayer, epithelial cells were washed and lysed with 0.25% Triton-X at 37°C for 20 min. Bacteria attached and internalized by the Caco-2 cells were quantified by cultivation of serial dilutions of the cell–culture lysates. Pilot experiments confirmed the proportion of non-adherent bacteria to be < 1% of the total colony-forming unit (cfu) in the lysates. Internalized bacteria were quantified using a multiple of infection of 1000:1. Before lysis of the cell monolayer, adherent extracellular bacteria were killed by incubation with DMEM supplemented with 0.25% gentamicin for 2 h.
Isolation of glycolipids
For isolation of glycolipids, E. faecalis strains 12030 and 12030ΔbgsA were grown in TSB at 37°C to stationary phase (18 h). Bacterial cells were collected by centrifugation and washed once in PBS. The bacteria were resuspended in 0.1 M Na-Citrate buffer pH 4.7 and cell walls were disintegrated by vibration with glass beads (Beadbeater, Glenn Mills, Clifton, NJ) followed by stirring with an equal volume of n-butanol for 30 min as described previously (Theilacker et al., 2006). After phase separation by centrifugation, the aqueous layer was removed, dialysed against 0.1 M ammonium acetate (pH 4.7) and lyophilized. The butanol phase was dried by rotary evaporation and extracted according to the method of Bligh and Dyer (1959). Briefly, butanol extracts were resuspended in distilled water, and CHCl3/MeOH was added (CHCl3/MeOH/H2O 2:1:0.8; v/v/v) and mixed thoroughly. For phase separation, more water was added to obtain a final CHCl3/MeOH/H2O ratio of 2:1:1.8. The organic phase was dried under a stream of nitrogen and further characterized by TLC and chemical analysis.
Glycolipid analysis by TLC
Glycolipids were separated by TLC using aluminium silica sheets (0.2 mm Silica gel 60 F254 Merk, Darmstadt) using a solvent system of CHCl3/MeOH/H2O (65:25:4, v/v/v). Ten to thirty micrograms of the sample was loaded on the TLC plates. For detection of glycolipids, TLC plates were sprayed with α-naphthol (3.2%) in MeOH/H2SO4/H2O (25:3:2, v/v/v) and heated at 110°C for 5–10 min. For detection of phospholipids, TLC plates were stained with molybdenum blue, and aminophospholipids were stained with ninhydrin, as previously described (Peschel et al., 2001).
Purification of glycolipids
Glycolipids were isolated by PLC. Plates (20 × 20 cm, 2 mm thickness, Silica gel 60 F254, Merk) were loaded with Bligh/Dyer extracts and were run in CHCl3/MeOH/H2O (65:25:4, v/v/v). For detection of fractions, the margin of the plate was cut off, and bands were visualized with α-naphtol-MeOH/H2SO4/H2O as described above. Two major fractions were detected, scraped off, and eluted with CHCl3/MeOH in different ratios (1:1; 2:1, 3:1, v/v). The eluates were combined, filtered with a 0.2 μm PTFE filter and dried under a stream of nitrogen. Purity of the fraction was evaluated by TLC as described above.
General and analytical chemical methods
Compositional analyses were carried out first by methanolysis of the samples with 0.5 M HCl/MeOH at 85°C for 45 min, followed by acetylation using acetic anhydride and pyridine (1:1, v/v), and detection by gas–liquid chromatography (GLC) and GLC-mass spectrometry. The absolute configuration of the sugar was determined by GLC, by comparison with authentic standards, of the acetylated (S)-2-butanol glycoside derivatives after butanolysis (2 M HCl/(S)-2-butanol at 85°C for 2 h) and acetylation [acetic anhydride and pyridine (1:1, v/v)] (Gerwig et al., 1979).
Nuclear magnetic resonance spectroscopy
Nuclear magnetic resonance experiments were carried out in CDCl3/CD3OD (2:1, v/v). All one-dimensional (1H and 13C) and two-dimensional homonuclear (COSY, TOCSY and ROESY) and heteronuclear (1H,13C HSQC-DEPT and HMBC) experiments were recorded at 295 K and/or 300 K for DGlcDAG and MGlcDAG, with a Bruker DRX Avance 600 MHz spectrometer (operating frequencies of 600.17 MHz for 1H NMR and 150.92 MHz for 13C NMR) using standard Bruker software. Chemical shifts were reported according to an internal standard of tetramethylsilan (δH 0.00, δC 0.00). COSY, TOCSY and ROESY experiments were recorded using data sets (t1 by t2) of 2048 by 512 points, and 16 scans were acquired for each t1 value for COSY and TOCSY, and 2048 by 256 (32 scans for each t1 value) for ROESY. The TOCSY experiments were carried out in the phase-sensitive mode with mixing times of 100 ms. The 1H,13C correlations were measured in the 1H-detected mode via HSQC-DEPT with proton decoupling in the 13C domain, and HMBC spectra were acquired using data sets of 2048 by 512 points, and 56 scans were acquired for each t1 value.
Isolation and characterization of LTA
Lipoteichoic acid of E. faecalis strain 12030ΔbgsA was isolated and analysed by NMR spectroscopy as described previously (Theilacker et al., 2006). The degree of substitution of the poly Gro-P chain was calculated by the method of Morath et al. (2001).
Autolysis assay and sensitivity to antimicrobial peptides
Cell autolysis was determined as described by Qin et al. (1998). In brief, bacteria were grown to exponential phase in Todd Hewitt broth supplemented with 2% glucose, chilled on ice, washed three times with distilled water at 4°C, and resuspended in 6 ml of 10 mM sodium phosphate buffer (pH 7.0). The suspension was then incubated at 37°C, and the OD675 was measured at 15 min intervals for up to 3 h. The MIC of polymyxin B, nisin, and colistin against wild type and 12030ΔbgsA were determined by a modified NCCLS broth dilution method as described previously (Fabretti et al., 2006).
Microbial affinity to organic solvents
The hydrophobicity of bacterial cells was analysed by comparing distribution in hydrophilic and hydrophobic solvents as described previously (Reid et al., 1992). Briefly, bacteria were grown to logarithmic phase, washed two times in sodium phosphate and resuspended in the same buffer up to an OD600 of 0.4–0.5. The bacterial suspension was then mixed with the same volume of dodecane, and solvents were mixed vigorously for 1 min. After 10 min of phase separation, absorbance of the water-phase was measured. The proportion of cells in the dodecane phase was calculated according to the formula: % hydrophobicity = [1 - (A/A0)] × 100.
Mouse bacteraemia model
The virulence of E. faecalis strain 12030ΔbgsA was evaluated in a mouse bacteraemia model as described previously (Hufnagel et al., 2004a). In summary, eight 6–8-week-old female Swiss-Webster mice were challenged by intravenous injection of E. faecalis 12030ΔbgsA (1.9 × 109 cfu) or the wild-type 12030 (5.4 × 108 cfu) via the tail vein. Three days after infection, the mice were sacrificed and exsanguinated, and bacterial counts in the blood were enumerated by serial dilutions. The experiment was repeated in 6–8-week-old female BALB/c mice using a challenge dose 4.5 × 108 cfu ml−1 (12030 wild type) and 2 × 109 cfu ml−1 (12030ΔbgsA).
SDS-PAGE and immunoblotting
Enterococcus faecalis sonicate was prepared by treatment of enterococci at 4°C for 6 min (1 min cycles with 1 min cooling intervals) with a VCX 400 W ultrasonic disintegrator. Recombinant 12030ΔbgsA and E. faecalis 12030 wild-type sonicates were subjected to SDS-PAGE in gradient gels containing acrylamide (4/12% w/v, Invitrogen), and the proteins were visualized by staining with Coomassie blue. Separated proteins were transferred onto polyvinylidene difluoride membrane and blocked at 4°C in PBS containing skim milk (5% w/v, Sigma) for 18 h, then incubated at 20–22°C for 2 h with the respective primary rabbit antibody diluted 1–500 in Tris-buffered saline (TBS). After washing in TTBS (Tween 20 0.05% v/v in TBS), the sheets were incubated at 20–22°C for 1 h with goat anti-rabbit immunoglobulin G (IgG) (whole cell) alkaline phosphatase conjugate (Sigma), diluted 1:1000 with TBS and then washed again in TTBS. Binding of the enzyme-conjugated antibodies was detected with the NBI/BCIP (Roche).
Transmission electron microscopy
Bacterial cells were prepared for transmission electron microscopy as described previously (Fabretti et al., 2006).
An opsonophagocytic assay was used as previously described (Kropec et al., 2005; Theilacker et al., 2006). White blood cells (WBCs) were prepared from fresh human blood collected from healthy adult volunteers. Twenty-five millilitres was mixed with an equal volume of dextran–heparin buffer and incubated at 37°C for 1 h. The upper layer containing the leucocytes was collected, the cells were pelleted by centrifugation, and hypotonic lysis of the remaining erythrocytes was accomplished by resuspension of the cell pellet in 1% NH4Cl and incubation for 10 min at 20–22°C. WBCs were then washed three times and resuspended with RPMI with 15% fetal bovine serum (RPMI-FBS). Using trypan blue staining to differentiate dead from live leucocytes, the final WBC count was adjusted to 2.5 × 107 WBCs per ml. Baby rabbit serum (Cedarlane Laboratories, Hornby, Ontario, Canada), diluted 1:10 in RPMI-FBS, was used as complement source. The bacterial strains to be evaluated for phagocyte-dependent killing activities of antibody were adjusted to an OD650 of 0.1 with fresh TSB and allowed to grow to an OD650 of 0.4. A 1:100 dilution was then made in RPMI-FBS for use in the killing assay. The opsonophagocytic assay was performed with 100 μl of leucocytes, 100 μl of bacteria, 100 μl of the complement solution, and 100 μl of heat-inactivated rabbit immune serum at various dilutions. Controls were made by replacing the antibody, complement, or polymorphonuclear leucocytes with RPMI-FBS. The reaction mixture was incubated on a rotor rack at 37°C for 90 min. The tubes were vortexed for 15 s and diluted in TSB with 0.25% Tween to prevent bacterial aggregation, and samples were plated onto tryptic soy agar plates. The percentage of killing was calculated by determining the ratio of the cfu surviving in the tubes with bacteria, leucocytes, complement, and antibody to the cfu surviving in the tubes with all these components but lacking antibody. For antibody-independent phagocytosis, we calculated the percentage of killing of bacteria by leucocytes and complement only, comparing the surviving counts with those in tubes lacking either leucocytes or complement.
Quantification of LTA
The LTA content of bacterial cell walls was measured according to the method of Fedtke et al. (2007). Briefly, wild-type and mutant bacteria were grown for 18 h in TSB, adjusted to an equal OD600, and bacteria from equal volumes were collected by centrifugation. LTA was extracted from the cell walls by butanol extraction. The aqueous phase of the extract was dialysed and lyophilized. ELISA plates (Costar) were coated with the LTA extract after resuspension in phosphate buffer (pH 7.0). LTA was quantified by ELISA using a rabbit antiserum specific for E. faecalis LTA as primary antibody, and alkaline phosphatase conjugate of a goat anti-rabbit IgG whole molecule served as secondary antibody. LTA purified from E. faecalis 12030 was used as standard.
If not stated otherwise, observations were confirmed by at least one independent experiment. Comparisons were made by unpaired t-test (parametric data) or Mann–Whitney t-test (non-parametric data) using the Prism Graphpad 4 software package. A P-value of < 0.05 was considered statistically significant.
The authors thank Dr Feuerhake for help with electron microscopy and Dr Serror, Dr Hartke and Dr Benachour for providing the plasmids and strains used in mutagenesis of V583. J.H. was supported by a grant from the NIH (R01 AI050667) and by a grant of the German Ministry of Science and Education (ERA-Net PathoGenoMics 0313933).