Staphylococcus aureus CidA and LrgA Proteins Exhibit Holin-Like Properties

Abstract

The Staphylococcus aureus cid and lrg operons are known to be involved in biofilm formation by controlling cell lysis and the release of genomic DNA, which ultimately becomes a structural component of the biofilm matrix. Although the molecular mechanisms controlling cell death and lysis are unknown, it has been hypothesized that the cidA and lrgA genes encode holin- and antiholin-like proteins and function to regulate these processes similarly to bacteriophage-induced death and lysis. In this study, we focused on the biochemical and molecular characterization of CidA and LrgA with the goal of testing the holin model. First, membrane fractionation and fluorescent protein fusion studies revealed that CidA and LrgA are membrane-associated proteins. Furthermore, similarly to holins, CidA and LrgA were found to oligomerize into high-molecular-mass complexes whose formation was dependent on disulfide bonds formed between cysteine residues. To determine the function of disulfide bond-dependent oligomerization of CidA, an S. aureus mutant in which the wild-type copy of the cidA gene was replaced with the cysteine mutant allele was generated. As determined by β-galactosidase release assays, this mutant exhibited increased cell lysis during stationary phase, suggesting that oligomerization has a negative impact on this process. When analyzed for biofilm development and maturation, this mutant displayed increased biofilm adhesion in a static assay and a greater amount of dead-cell accumulation during biofilm maturation. These studies support the model that CidA and LrgA proteins are bacterial holin-/antiholin-like proteins that function to control cell death and lysis during biofilm development.

INTRODUCTION

The control of cell death is a physiological process that has been characterized extensively in eukaryotic organisms but has only recently been studied in any detail in prokaryotes. Although several mechanisms mediating this process in bacteria have been proposed (28), our laboratory has focused on a novel family of cell death effectors of which the Staphylococcus aureus Cid and Lrg proteins are the prototypical members. This system comprises predicted membrane-associated proteins CidA, CidB, LrgA, and LrgB, consisting of multiple transmembrane domains encoded by the cid and lrg operons (8, 29). Based on secondary-structure similarities to bacteriophage-encoded holins/antiholins and the phenotypes of cid and lrg mutants, the cidA gene was proposed to encode a holin-like protein with a positive effect on murein hydrolase activity, and lrgA was proposed to encode an antiholin-like protein with an inhibitory effect on these enzymes (8, 29). Studies also revealed that mutations in the cid and lrg operons increased and decreased antibiotic tolerance, respectively (8, 27).

At least one biological function of the cid and lrg genes is the coordination of cell death and lysis during biofilm development, causing release of genomic DNA, which ultimately becomes a structural component of the biofilm matrix (18, 30). The S. aureus cidA mutant exhibited decreased lysis during biofilm formation (30), while the lrgAB mutant, as well as the lytSR mutant (which exhibits reduced lrgAB expression), exhibited increased lysis (18, 35). The consequence of decreased lysis was a decrease in genomic DNA release and biofilm adherence (30). In contrast, increased cell lysis during biofilm development resulted in increased biofilm adherence (18, 35). Based on their roles in controlling cell death and lysis during biofilm development, it was proposed that these proteins form the regulatory elements of bacterial programmed cell death (PCD) (5).

The biochemical functions of the CidA and LrgA proteins are based solely on characteristics of these proteins found to be in common with bacteriophage holins. These include their relatively small size, the presence of two to three putative transmembrane domains, and the presence of charge-rich N and C termini (27). Bacteriophage holins are well-studied, integral membrane proteins that control murein hydrolase activity by one of two mechanisms involving pore formation within the bacterial membrane (41). The first mechanism, utilized by the prototype λ S holin, serves as a gatekeeper of the bacteriophage-encoded endolysin, controlling access to the peptidoglycan by forming large, nonspecific holes in the cytoplasmic membrane (33, 41). The second mechanism is exhibited by bacteriophage P1, in which the P1-encoded endolysin possesses a signal-arrest-release (SAR) domain to control the activity of this enzyme. Unlike λ S endolysin, the P1 endolysin is recognized and transported via the cellular “Sec” machinery but remains in an inactive form anchored in the outer face of the cytoplasmic membrane until its cognate holin releases it (24, 40).

Despite the existence of differing mechanisms controlling murein hydrolase function, the underlying biochemical properties of holins that lead to hole formation and murein hydrolase activation are thought to be similar to those of the S holin/antiholin system of bacteriophage λ. First, S holins can exist in the oligomeric state in the inner membrane (42), and the liposomal studies carried out with purified holins have illustrated that holins are capable of forming aqueous pores in an artificial membrane (37). Moreover, studies with mutations that confer oligomerization deficiency displayed a loss of holin function (9), indicating that oligomerization is required for holin function. A search of the holin sequence revealed a unique cysteine at position 51 in the second transmembrane domain that causes the formation of disulfide-linked dimers (10). Although the substitution of cysteine with serine abolished dimerization, the studies indicated that cysteineless holin still oligomerizes and produces a functional holin (9) but one that triggered earlier lysis than the wild-type (wt) protein (10). These results suggest that cysteine-mediated dimerization has a negative effect on the timing of host cell lysis. The timing of lysis is also dictated by the ratio of holin and antiholin. Remarkably, the antiholin of the λ system (and many other bacteriophage systems) is encoded by the same gene encoding the holin. Due to the presence of a dual-start motif, the S antiholin (designated S107) contains a 2-amino-acid N-terminal extension, including the positively charged amino acid lysine (4). This subtle difference has a dramatic effect on the topology of this protein in the membrane, causing the N terminus of the antiholin to localize to the cytoplasmic face of the membrane (39). The consequence of this altered membrane topology is to inhibit oligomerization of the holin (S105) and the timing of hole formation.

Recently, the holin/antiholin systems have been shown to differ dramatically in size of the hole that is eventually formed, reflecting the different mechanisms used for murein hydrolase activation. The hole formed by λ S is very large, allowing the release of preformed endolysin across the cytoplasmic membrane (33). In contrast, P1 holins form a “pinhole” just large enough to cause the release of small ions and the collapse of the proton motive force (PMF). The subsequent depolarized membrane then signals the release and activation of the membrane-associated SAR domain murein hydrolase (40). Thus, this system functions to stimulate the activity of murein hydrolases that have already been secreted and are poised to function. Consistent with this is the observation that endolysins associated with pinhole-type holins contain Sec-dependent secretion signals.

Although the regulatory processes controlling bacterium-encoded murein hydrolase activity associated with the cell wall remain largely unknown, the demonstrated impact of S. aureus cid and lrg mutations on Atl suggests the involvement of the putative CidA/LrgA holin/antiholin system in this process (8, 29). Furthermore, the primary S. aureus murein hydrolase, Atl, possesses a signal sequence (3), suggesting that if a holin/antiholin mechanism controls the activity of the Atl murein hydrolases, it likely functions in a manner more similar to that of pinhole-type holins. The objective of the current study was to conduct a biochemical characterization of CidA and LrgA, with the ultimate goal of determining the precise mechanism by which they function in the control of murein hydrolase activity.

MATERIALS AND METHODS

Bacterial strains, media, and growth conditions.

The descriptions of bacterial strains used in this study are listed in Table 1. Escherichia coli DH5α (11), used for cloning purposes, was grown in LB medium (Fisher Scientific). For membrane protein purification, E. coli C43 was grown in Terrific broth (12 mg/ml Bacto tryptone, 24 mg/ml Bacto yeast extract, and 0.4% [vol/vol] glycerol). All S. aureus stains were grown in tryptic soy broth (TSB) or tryptic soy agar (TSA) (Difco Laboratories). All planktonic cultures were grown at 37°C at a rotation speed of 250 rpm in volumes that did not exceed 20% of the flask volume. All antibiotics were purchased from Sigma Chemical Co. or Fisher Scientific and were used at the following concentrations: ampicillin, 100 μg/ml; chloramphenicol, 10 μg/ml; kanamycin, 50 μg/ml; and erythromycin, 10 μg/ml.

Table 1.

Strains and plasmids used in this study

Strain or plasmid	Genotype and description	Reference or source
E. coli strains
DH5α	Host strain for construction of recombinant plasmids	11
C43	Derivative of BL21(DE3) for expression of membrane proteins	21
S. aureus strains
RN4220	Highly transformable strain; restriction deficient	17
UAMS-1	Clinical osteomyelitis isolate	7
KB1050	UAMS-1 cidA::erm; Emr	31
KB1051	UAMS-1 mutant encoding CidA(C104S)	This study
KB1052	KB1051 repaired strain	This study
AJ22	RN4220 containing pAJ22; Cmr	23
Plasmids
pCN51	Cadmium-inducible S. aureus expression plasmid	6
pCM12	Plasmid encoding sGFP	26
pDR1	pCN51 with gene encoding sGFP	This study
pDR2	pCN51 with gene encoding AgrB-G	This study
pDR3	pCN51 with gene encoding CidA-G	This study
pDR4	pCN51 with gene encoding LrgA-G	This study
pET24b	IPTGa-inducible E. coli expression plasmid	Novagen
pDR7	pET24b with gene encoding CidA-H	This study
pDR8	pET24b with gene encoding LrgA-H	This study
pDR9	pET24b with gene encoding CidA-H(C104S)	This study
pDR10	pET24b with gene encoding LrgA-H(C55S C114S)	This study
pCL10	Temperature-sensitive shuttle vector	32
pDR150CS	pCL10-derived plasmid for the generation of KB1051	This study
pAJ22	β-Galactosidase reporter plasmid; Cmr	23

^aIPTG, isopropyl-β-d-thiogalactopyranoside.

Standard DNA manipulation, PCR, and DNA sequencing.

Plasmids and primers used in this study are listed in Tables 1 and 2, respectively. Standard DNA manipulation and PCR techniques were applied to create the plasmids used in this study. Preparation and transformation of E. coli DH5α were accomplished using the procedure described by Inoue et al. (14). Verification of the nucleotide sequences of plasmid constructs was performed using the High-Throughput DNA Sequencing and Genotyping Core Facility located at the University of Nebraska Medical Center. Transformation of confirmed plasmids into S. aureus was achieved by electroporation following a method described previously (34), and transformants were selected on agar medium containing erythromycin (for pCN51 and derivatives) or chloramphenicol (for pCL10 and derivatives).

Table 2.

Primers used in this study

Primer name	Primer sequencea
lrgA-BamHI-F	5′-CCGGGGATCCATCAAACGTAGGAGGCAATG-3′
cidA-BamHI-F	5′-GCAGGATCCATATTTAGAAAGGGATGGCGCCATGCA-3′
agrB-TIR-BamHI-F	5′-GCGGATCCTGATTAACTTTATAAGGAGGAAAAACATATGTCGTATAATGACAG-3′
lrgA-sGFP-link-R	5′-GTTCTTCTCCTTTGCTATCATGAGCTTGTGCC-3′
cidA-sGFP-link-R	5′-GTTCTTCTCCTTTGCTTTCATAAGCGTCTACAC-3′
agrB-sGFP-link-R	5′-GTTCTTCTCCTTTGCTTTTTAAGTCCTCCTTAATAAAG-3′
sGFP-TIR-BamHI-F	5′-GGCGGATCCTGATTAACTTTATAAGGAGGAAAAACATATGAGCAAAGGAGAAGAAC-3′
sGFP-EcoRI-R	5′-GGCGAATTCTTATTTGTAGAGCTCATCCATGCC-3′
lrgA-NdeI-F	5′-GCGGCATATGGTCGTGAAACAACAAAAAGACGC-3′
cidA-NdeI-F	5′-GCGGCATATGCACAAAGTCCAATTAATAATCAAAC-3′
lrgA-XhoI-R	5′-CCGCTCGAGATCATGAGCTTGTCG-3′
cidA-XhoI-R	5′-CCGCTCGAGTTCATAAGCGTCTAC-3′
lrgA-G164C-R	5′-CAGCACCAGTAGATAATAATACAAATAATA-3′
lrgA-G164C-F	5′-ATTATTATCTACTGGTGCTGTTAAGTTAGG-3′
lrgA-G341C-R	5′-AGCCAGTAGAAATAAGTAATAGTATTGTTG-3′
lrgA-G341C-F	5′-ATTACTTATTTCTACTGGCTATGTCACACA-3′
cidA-G311C-R	5′-AACGATAGATGTTCCTATGATAATG-3′
cidA-G311C-F	5′-AACATCTATCGTTGCATTATCTTC-3′
cidA-up-EcoRI-F	5′-CGAATTCGAAGTCGTAAAGCAAGGAGGCATGAC-3′
cidB-dn-SalI-R	5′-CGTCGACGAAGCATTAGTTAAAGCATTACAAGCATGG-3′

^aUnderlining indicates restriction endonuclease recognition sites used for subsequent cloning steps.

Construction of cidA and lrgA expression vectors.

C-terminal translational fusions of CidA, LrgA, and AgrB with superfolder green fluorescence protein (sGFP) (26) were constructed using the “splicing by overlap extension” (SOE) technique (13). PCR-generated DNA fragments containing cidA, lrgA, and agrB were amplified from the chromosome of S. aureus strain RN6390 using forward primers incorporating BamHI restriction endonuclease recognition sites (primers cidA-BamHI-F, lrgA-BamHI-F, and agrB-TIR-BamHI-F, respectively) and reverse primers containing 16 bp upstream of the gene encoding sGFP (primers cidA-sGFP-link-R, lrgA-sGFP-link-R, and agrB-sGFP-link-R, respectively). The fusion construction was carried out using the SOE technique, in which the resulting cidA, lrgA, and agrB PCR products were mixed in equimolar amounts with pCM12 (encoding sGFP), which was used as a template for a second round of PCR. These PCR amplifications were carried out using the forward primers cidA-BamHI-F, lrgA-BamHI-F, and agrB-TIR-BamHI-F, respectively, and the reverse primer sGFP-EcoRI-R, containing an EcoRI site. For the cytoplasmic localization control, sGFP alone was amplified from pCM12 using the forward primer sGFP-TIR-BamHI-F (this primer incorporated a strong ribosome binding site [20]) and the reverse primer sGFP-EcoRI-R. Each of these PCR products generated by SOE was digested with BamHI and EcoRI and inserted between the BamHI and EcoRI sites of pCN51 (6), generating pDR1, pDR2, pDR3, and pDR4 (Table 1). By electroporation, the resulting plasmids were transformed into S. aureus RN4220 (17), and transformants were selected on agar medium containing erythromycin.

To purify the proteins from E. coli, the cidA and lrgA genes were ligated into pET24b (Novagen), generating C-terminal His tag fusions. PCR fragments of cidA and lrgA were amplified using forward primers containing an NdeI restriction site (primers cidA-NdeI-F and lrgA-NdeI-F, respectively) and reverse primers containing an XhoI site (primers cidA-XhoI-R and lrgA-XhoI-R, respectively). The PCR products were digested with NdeI and XhoI and inserted between the NdeI and XhoI sites of pET24b. For protein production, the resulting plasmids, pDR7 and pDR8 (Table 1), were transferred into E. coli strain C43, a mutant derivative of BL21(DE3) selected for optimal overproduction of membrane proteins (21).

CidA and LrgA mutagenesis.

Site-directed mutagenesis was used to convert the single cysteine of CidA and the two cysteines of LrgA to serines. For the LrgA construct, three separate mutagenic PCR fragments were generated using S. aureus RN6390 genomic DNA as a template. The first was a 164-bp DNA fragment generated using the 5′ primer lrgA-NdeI-F and the mutagenic 3′ primer lrgA-G164C-R, which introduced a G164C mutation in the PCR product. The second was a 176-bp DNA fragment generated using the mutagenic primers lrgA-G164C-F and lrgA-G341C-R, which introduced G164C and G341C mutations in the PCR product. Lastly, the third was a 104-bp DNA fragment generated using the 5′ mutagenic primer lrgA-G341C-F and the 3′ primer lrgA-XhoI-R, which introduced a G341C mutation in the PCR product. All three PCR products were gel purified, combined in equimolar ratios, and utilized as templates in a PCR with the primers lrgA-NdeI-F and lrgA-XhoI-R. The resulting PCR products contained mutated lrgA genes (containing either or both of the G164C and G341C mutations) and were flanked by NdeI sites at the 5′ ends and XhoI sites at the 3′ ends. These DNA fragments were digested with NdeI and XhoI and ligated into the corresponding sites of pET24b, generating the plasmid pDR10. A similar PCR-based site-directed mutagenesis strategy was used to change the single cysteine in CidA to serine using the primers cidA-NdeI-F, cidA-G311C-R, cidA-G311C-F, and cidA-XhoI-R. The resulting PCR product was ligated into pET24b to generate the plasmid pDR9. All of these plasmids were transferred into E. coli strain C43 and, upon expression, generated C-terminal His tag fusions for detection by Western blot analysis using anti-His tag antibodies (see below).

Cell fractionation and membrane protein purification.

To purify membrane proteins from S. aureus, cell fractionation experiments adapted from the procedure of Oku et al. (22) were performed. After overnight growth, S. aureus strains containing pDR1, pDR2, pDR3, and pDR4 were diluted into fresh TSB to an optical density at 600 nm (OD₆₀₀) of 0.1 and grown at 37°C until the OD₆₀₀ reached 1.0 (approximately 2 h). Cultures were supplemented with 5.0 μM CdCl₂ to activate transcription of the cadmium-inducible promoter, and incubation was continued for 4 h. Cells were harvested by centrifugation at 10,000 × g for 10 min using a Sorvall RC 5C Plus centrifuge and SLA 1500 rotor (Thermo Scientific). The pelleted cells were resuspended in phosphate-buffered saline (PBS) (Mediatch, Inc., VA) and treated with 100 U/ml of lysostaphin and 10 U/ml DNase I (Sigma), with a 10-min incubation at 37°C along with a final 1× protease inhibitor mixture (Complete minikit, EDTA free; Roche). The lysostaphin-treated cells were lysed with two passes through a high-pressure homogenizer (EmulsiFlex-C3; Avestin Inc., Germany) at 15,000 lb/in². Unbroken cells were removed by centrifugation at 10,000 × g at 4°C for 10 min. Crude membranes were pelleted by centrifugation at 100,000 × g for 1 h at 4°C using a Beckman L8 60M centrifuge and 45Ti rotor. The resultant supernatant was removed and taken as the cytoplasmic fraction. The pellet, composed of the crude membrane fraction, was resuspended in membrane buffer (100 mM Tris-HCl [pH 7.5], 100 mM NaCl, 10 mM MgCl₂, 10% glycerol, and 1% dodecyl maltoside [DDM]) to extract and solubilize the membrane proteins. This was again centrifuged at 100,000 × g as described above to remove the detergent-insoluble material. The concentration of proteins in the resultant supernatant, containing the detergent-soluble membrane proteins, was measured using a Bio-Rad protein assay kit.

Purification of His-tagged proteins.

His-tagged proteins were purified using a His SpinTrap column (GE Healthcare) following the instructions provided by the manufacturer. The membrane proteins, solubilized in the membrane buffer (see above), were prepared for the chromatographic purification by addition of imidazole to a final concentration of 20 mM. The column was equilibrated with the binding buffer (20 mM imidazole and 0.1% DDM in membrane buffer). Subsequently, the protein sample was loaded to the column and washed twice with 600 μl of binding buffer. The proteins were eluted from the column with a total of 400 μl of elution buffer (500 mM imidazole and 0.1% DDM in membrane buffer) and were analyzed by SDS-PAGE and Western blotting.

Western blot analysis.

To detect the recombinant CidA and LrgA proteins, samples were separated in a Tris-glycine SDS-polyacrylamide gel (12%). The separated proteins were electrophoretically transferred to a polyvinylidene difluoride (PVDF) membrane and blocked by incubation in 1% casein hydrolysate (Novagen) either for 1 h at room temperature or overnight at 4°C. The primary antibody used for detecting the sGFP fusions was anti-GFP (Living Colors Full-Length A.v. polyclonal antibody; Clontech), and that for the His fusions was anti-His tag monoclonal antibody (Novagen). The proteins bound to the PVDF membrane were incubated for 1 h at room temperature with a 1:1,000 dilution of the antibodies. For colorimetric detection, an alkaline phosphatase secondary antibody conjugate (Novagen) was used, and blots were developed using nitroblue tetrazolium and 5-bromo-4-chloro-3-indolylphosphate (NBT-BCIP) developing reagent (Sigma).

Confocal microscopy.

S. aureus RN4220 cells harboring plasmids containing sGFP fusion constructs were inoculated into fresh TSB containing erythromycin by diluting fresh overnight cultures to an OD₆₀₀ of 0.1, followed by growth until an OD₆₀₀ of 1.0 was reached. Expression of the sGFP fusion proteins was induced by adding CdCl₂, to a final concentration of 5.0 μM, to the growth medium (6). After an additional incubation period of 2 h, the cells were harvested by centrifugation in a Microfuge 18 centrifuge (Beckman Coulter) at 10,000 rpm for 10 min, washed with PBS, and observed by confocal laser scanning microscopy (CLSM) using a Zeiss LSM 710 microscope.

Allelic replacement of the cidA gene in UAMS-1.

To generate a cidA cysteine mutation in the chromosome of UAMS-1 (7), allelic replacement of the wt cidA gene with the cidA mutant allele containing a G311C mutation was carried out (29). First, a 1,391-bp DNA fragment spanning a region 5′ to cidA was PCR amplified using the 5′ primer cidA-up-EcoRI-F and the 3′ primer cidA-G311C-R and genomic DNA from S. aureus UAMS-1. The 3′ primer introduces a G311C mutation in the PCR product. Another PCR product spanning an 875-bp region 3′ to cidA was generated using the 5′ primer cidA-G311C-F and the 3′ primer cidB-dn-SalI-R. As described above, the 5′ primer introduces a G311C mutation in the PCR product. These mutagenic PCR products were gel purified, mixed in equimolar ratios, and used as a template for another round of PCR using the primers cidA-up-EcoRI-F and cidB-dn-SalI-R. The resulting DNA spanned the entire cidA gene and contained the G311C mutation. Next, the DNA was digested with EcoRI and SalI and inserted into the temperature-sensitive shuttle vector pCL10 (32) to generate the plasmid pDR150CS. After the nucleotide sequence of the inserted DNA was confirmed, the plasmid was transformed into the restriction-deficient S. aureus strain RN4220, reisolated, and then transformed into S. aureus UAMS-1. A resulting transformant was initially grown at a nonpermissive temperature (45°C) in the presence of chloramphenicol to select for cells in which the plasmid was integrated into the chromosome via homologous recombination. To promote a second recombination event, a single colony was inoculated into antibiotic-free TSB and grown at 30°C for 4 days with 1:1,000 dilutions into fresh antibiotic-free medium each day. After the fourth day, dilutions of the culture were spread on TSA to grow isolated colonies, which were subsequently screened for chloramphenicol sensitivity (loss of the plasmid). Verification of the presence of the cidA(G311C) mutation was carried out by PCR amplification and digestion of the product with the restriction endonuclease PciI, the recognition site of which was eliminated by the G311C mutation. The region carrying the entire cidA gene was PCR amplified and sequenced to confirm the G311C mutation in the chromosome. Final confirmation that this strain did not acquire any other major genetic changes was carried out by pulsed-field gel electrophoresis analysis. The confirmed mutant strain was designated KB1051 (Table 1). To reverse the G311C mutation back to the wt cidA allele, an allelic replacement strategy similar to that described above was utilized. This “repaired” cidA mutant strain was designated KB1052 (Table 1).

Measurement of β-galactosidase activity.

Overnight cultures of UAMS-1, KB1050 (31), and KB1051, each harboring the plasmid pAJ22 (23), were washed twice with TSB, inoculated into TSB (no antibiotic) to an OD₆₀₀ of 0.1, and grown for 48 h. At various time points, the OD₆₀₀ of the growing culture was measured, and supernatants from strains were harvested by centrifugation. β-Galactosidase activity in the supernatant was determined as described previously (38), using o-nitrophenyl-d-galactopyranoside as the substrate (19). Results were recorded in triplicate, representing three independent experiments. Means were calculated and analyzed by analysis of variance (ANOVA) with heterogeneous variances, followed by the Tukey-Kramer method for multiple comparisons (SAS version 9.2).

Biofilm assays.

Static biofilms were grown in 96-well Costar 3596 plates (Corning Life Sciences, Acton, MA) precoated with 20% human plasma (Sigma Chemical Co., St. Louis, MO) in bicarbonate buffer at 4°C overnight. Isolated colonies of S. aureus strains were grown overnight in TSB supplemented with 3% NaCl and 0.5% glucose, which was diluted to an OD₆₀₀ of 0.05 in the same medium, and 200 μl of diluted culture was transferred into the precoated wells and incubated at 37°C for 24 h. Adherence of the resulting biofilm was analyzed as described previously (18) by washing twice with 200 μl PBS, fixing with 100 μl of ethanol for 1 min, and staining with 100 μl crystal violet (0.41% in 12% ethanol) for 1 min. Quantification of the biofilm was carried out by measuring the amount of crystal violet retained in the wells (at an absorbance of 595 nm) with a Victor multilabel counter (Perkin Elmer, Waltham, MA). Results were recorded in triplicate, representing three independent experiments. Means were calculated and analyzed by ANOVA with heterogeneous variances, followed by the Tukey-Kramer method for multiple comparisons (SAS version 9.2).

To assess biofilm grown in the presence of hydrodynamic shear forces, an FC271 flow cell system (Biosurfaces Technology Inc., Bozeman, MT) containing a polycarbonate coupon was grown as described previously (18) by diluting a fresh overnight culture into TSB to an OD₆₀₀ of 0.002 and aseptically injecting 3.0 ml of this cell suspension into the chamber. After 2 h of static incubation to allow the bacteria to attach to the surface of the coupon, 5% TSB (vol/vol) supplemented with 0.125% glucose was then constantly pumped into the bacterium-laden flow cell chamber using a Rainin RP-1 peristaltic pump (Rainin Instrument LLC, Woburn, MA) at a flow rate of 0.25 ml/min, and bacteria were allowed to grow for 72 h.

To analyze the flow cell biofilms by CLSM, the medium flow was stopped, and the fluorescent dyes Syto-9 (1.3 mM final concentration), which stains viable cells, and Toto-3 (2.0 mM final concentration), which stains dead cells and extracellular DNA (eDNA), were applied. The biofilm images were visualized using a Zeiss 510 Meta CLSM with an Achroplan 40× (0.8-numerical-aperture) water-dipping objective. An argon laser at 488 nm was used to excite Syto-9, whereas excitation of Toto-3 was achieved using a HeNe 633-nm laser, and emissions were collected using a 680 (±30)-nm laser. The confocal parameters used in these experiments were set using wild-type (UAMS-1) biofilm and utilized as standard settings for comparison to the biofilm produced by KB1051 and KB1052. Regions of interest within the biofilm were selected from similar areas of each flow cell chamber, and each confocal experiment was repeated a minimum of three times. To generate three-dimensional renderings, z-stacks were collected at 1.0-μm intervals, and the images were compiled using both the Zeiss ZEN LE software package (Carl Zeiss, Jena, Germany) and Volocity software (Improvision, Lexington, MA). Statistical analyses of the biofilms produced were performed using the COMSTAT software package (12), calculating the biomass, thickness, and roughness coefficients of the biofilm data accumulated for at least three separate z-stack images.

RESULTS

Subcellular localization of CidA and LrgA.

The S. aureus CidA and LrgA proteins have been hypothesized to control murein hydrolase activity by functioning as a membrane-associated holin and antiholin, respectively. This hypothesis is based solely on phenotypic analyses of cidA and lrgA mutants and has not been corroborated by any biochemical studies to date. In an initial attempt to characterize the properties of CidA and LrgA, recombinant genes encoding C-terminal sGFP (26) fusions of these proteins (designated CidA-G and LrgA-G, respectively) were expressed from the cadmium-inducible plasmid pCN51 and analyzed by confocal microscopy. As shown in Fig. 1, S. aureus cells expressing the CidA-G and LrgA-G proteins produced ring-like images (Fig. 1C and D, respectively) comparable to those produced by constructs expressing AgrB-G, a known membrane protein similarly fused to sGFP (Fig. 1B). In contrast, a control strain expressing unfused sGFP produced cells that contained evenly distributed fluorescence throughout the cell (Fig. 1A). These observations indicate that CidA and LrgA are localized to the surface of S. aureus cells.

Fig. 1.

Subcellular localization of CidA-G and LrgA-G fusion proteins. S. aureus cells expressing sGFP fusion proteins were observed by confocal microscopy using a 63× objective and ×4 magnification. Expression of sGFP alone (control for cytoplasmic localization) (A), AgrB-G (control for membrane localization) (B), CidA-G (C), and LrgA-G (D).

To confirm that the CidA and LrgA proteins are membrane proteins, S. aureus cells expressing the CidA-G and LrgA-G proteins were separated into cytoplasmic and membrane fractions, and the membrane proteins were examined by Western blot analysis, using anti-GFP antibodies as a probe. As shown in Fig. 2, the membrane proteins of cells expressing CidA-G and LrgA-G produced 41-kDa and 42-kDa proteins, respectively, reacting with the anti-GFP antibodies. Similarly, the AgrB membrane protein control also produced a reactive band in the membrane fraction consistent with the predicted size (49 kDa) of the AgrB-G fusion. In contrast, the construct expressing unfused sGFP produced a reactive protein in the cytoplasmic fraction only. To rule out the possibility that the GFP fusion caused anomalous localization to the membrane fraction, we performed similar fractionation and Western blot experiments of S. aureus cells expressing C-terminal His-tagged fusions of CidA and LrgA (designated CidA-H and LrgA-H, respectively) and detected similar localization of these proteins in the membrane fraction using anti-His antibodies (data not shown). In addition, anti-SarA antibodies were used, and SarA-reactive bands were identified in the cytosolic fractions of all tested strains but not in the membrane fractions (Fig. 2B). Combined, the results of the confocal and Western blot analyses confirmed that CidA and LrgA are membrane-associated proteins.

Fig. 2.

Membrane localization of CidA-G and LrgA-G. Whole-cell extracts (A) and cytosolic and membrane fractions (B) from S. aureus cells expressing sGFP (lanes 2), AgrB-G (lanes 3), LrgA-G (lanes 4), and CidA-G (lanes 5) were isolated and analyzed under reducing conditions by SDS-PAGE. Recombinant proteins were detected by Western blotting using anti-GFP antibodies. Lane 1 contains proteins from the vector-only control, and lane M contains size markers, with molecular masses (in kilodaltons) indicated to the left of the blots. The bottom image in panel B is a Western blot of the cytosolic and membrane fractions probed using anti-SarA antibodies.

CidA and LrgA oligomerization.

Another characteristic of holins is that they form self-assembling, high-molecular-mass oligomeric complexes. We examined the propensity of CidA-H and LrgA-H to oligomerize by expressing these proteins in E. coli strain C43 (21), purifying them using nickel affinity chromatography, and then subjecting them to Western blot analyses. As with holin proteins, high-molecular-mass oligomeric bands, ranging from 16 kDa to approximately 180 kDa in size, were observed for both LrgA-H (Fig. 3 and 4) and CidA-H (Fig. 4, lane 5). Inspection of the protein bands suggests the presence of monomeric and dimeric forms of these proteins, as well as other potential oligomeric forms of unknown composition. Interestingly, the addition of the reducing agent dithiothreitol (DTT) to the LrgA-H sample resulted in the conversion of the higher-molecular-mass forms to the monomeric form, indicating the dependence of oligomerization on the formation of disulfide bonds. In contrast, the addition of DTT to the CidA-H sample had only a minimal effect on the oligomerization of this protein (data not shown).

Fig. 3.

DTT-sensitive oligomerization of LrgA-H. Membrane fractions from E. coli cells expressing LrgA-G were isolated and analyzed by Western blotting using anti-His tag antibodies. Whole-cell lysates of induced E. coli C43 cells containing the vector control (lane 1) or expressing LrgA-H (lane 2). Expressed LrgA-H was purified as described in Materials and Methods and analyzed in the absence (lane 3) or presence (lane 4) of 100 mM DTT. Lane M contains size markers, with molecular masses (in kilodaltons) indicated to the left of the blots.

Fig. 4.

Cysteine-dependent oligomerization. CidA-H and LrgA-H were expressed in E. coli and analyzed by Western blotting. Whole-cell lysates of induced E. coli C43 cells expressing LrgA-H (lane 3), LrgA-H(C55S C114S) (lane 4), CidA-H (lane 5), and CidA-H(C104S) (lane 6). Lane 1 contains size markers, with molecular masses (in kilodaltons) indicated to the left of the blots, and lane 2 contains proteins from the vector-only control.

Previous analysis of the S holin of bacteriophage λ has revealed a role for intermolecular disulfide bond formation in the dimerization of these proteins (10). Indeed, sequence analysis of CidA and LrgA confirms that there is one cysteine present in CidA, at position 104, and two cysteines in LrgA, at positions 55 and 114. To determine the potential roles of the cysteines of CidA and LrgA, each cysteine was replaced with a serine using site-directed mutagenesis, and these proteins were expressed in E. coli as described above. The Western blot shown in Fig. 4 revealed that in contrast to the CidA-H and LrgA-H proteins, which produced high-molecular-mass oligomers, the Cys mutations completely abolished oligomer formation by both proteins. Interestingly, both of the single Cys mutant derivatives of LrgA oligomerized to the same extent as LrgA-H, containing both cysteines (data not shown), indicating that both of these residues must be mutated to prevent oligomerization of LrgA-H. Furthermore, despite the observation that reducing conditions failed to eliminate CidA-H oligomerization, the single Cys of this protein appears to be required for the oligomerization of this protein.

CidA(C104S) mutation effects autolysis.

Previous work in our laboratory indicated that disruption of the cid operon enhances murein hydrolase activity and autolysis, while mutation of the lrg operon decreases these processes (8, 18, 29, 31). Additionally, studies of a cidA null mutant demonstrated that this strain exhibited decreased stationary-phase lysis, as measured by the absence of a decline in optical density (25), as well as by the reduced release of a constitutively produced β-galactosidase (30). To investigate the biological role of cysteine-dependent oligomerization in S. aureus, a chromosomal mutation in the cidA gene of UAMS-1 was generated by allelic replacement of the wild-type copy of this gene with the cidA mutant allele encoding CidA(C104S), giving rise to strain KB1051. Importantly, optical density measurements during growth revealed that this strain grew at a rate similar to that of the wild-type strain (data not shown). However, as shown in Fig. 5, the KB1051 strain demonstrated an increase in the lysis-dependent release of β-galactosidase in stationary phase compared to that for UAMS-1, and this phenotype was restored in the KB1051 derivative (strain KB1052) in which the cidA mutant allele was repaired such that it was returned back to the sequence of the wild-type allele. These results indicate that the cysteine mutant derivative of CidA has an enhanced ability to cause lysis in stationary phase.

Fig. 5.

β-Galactosidase release assays. Cell lysis of UAMS-1, KB1051 (cidA point mutant), and KB1052 (repaired KB1051), each constitutively expressing β-galactosidase, was monitored by measuring β-galactosidase release into the culture supernatant. White bars correspond to UAMS-1, dark-gray bars to KB1051, and light-gray bars to KB1052; values represent means from three independent experiments ± standard errors of the means (SEM). Significant differences were seen between UAMS-1 and KB1051 at both time points (P = 0.05 at 52 h and P < 0.001 at 72 h) as well as between KB1051 and KB1052 at both time points (P = 0.02 at 52 h and P = 0.007 at 72 h); however, no significant difference was observed between UAMS-1 and KB1052 at either time point.

Effect of CidA(C104S) on biofilm development.

The role of S. aureus cid and lrg operons in the development of biofilm was previously demonstrated in our laboratory, where we observed biofilm with altered structures and growth characteristics (18, 30, 35). To test the importance of Cys104 during biofilm development, we first examined biofilms grown under static growth conditions. As shown in Fig. 6, the KB1051 strain, encoding CidA(C104S), exhibited a higher capacity for biofilm adherence than the parental and KB1052 strains. The biofilm adherence was previously demonstrated to be dependent on the amount of extracellular DNA released upon lysis of the cells during growth (30). However, quantitative real-time PCR assays failed to reveal a significant difference in the amounts of extracellular DNA associated with these biofilms (data not shown). We also examined the later stages of biofilm development by performing a flow cell experiment in which the biofilms were allowed to develop over a period of 3 days. Confocal laser scanning microscopy (CLSM) of biofilms stained with Syto-9 (stains viable cells) and Toto-3 (stains dead cells and extracellular DNA) revealed similar biofilm architectures produced by all three strains (Fig. 7). However, in contrast to the wild-type and KB1052 strains, the KB1051 strain produced an unusual staining pattern in which large patches of Toto-3 staining were observed, indicating large aggregates of dead cells or extracellular DNA accumulation. These results suggest that CidA(C104S) has a significant effect on the distribution of dead cells and/or extracellular DNA on the staphylococcal biofilm structure during growth under flow cell conditions.

Fig. 6.

Static biofilm assays. Cultures of UAMS-1, KB1051 (cidA point mutant), KB1052 (repaired KB1051), and KB1050 (cidA null mutant) were grown statically as described in Materials and Methods; biofilms attached to the bottom of wells were stained with crystal violet and scanned at 595 nm. Values represent means from three independent experiments ± SEM. Significant differences were seen between KB1050 and each of the other groups: UAMS-1 (P < 0.0001), KB1051 (P < 0.0001), and KB1052 (P = 0.0008). Significant differences were also seen between KB1051 and UAMS-1 (P = 0.01). However, no significant difference was observed between UAMS-1 and KB1052 or between KB1051 and KB1052.

Fig. 7.

Effect of CidA(C104S) on biofilm maturation. UAMS-1 (A), KB1051 (B), and KB1052 (C) were grown on polycarbonate coupons enclosed in a flow cell chamber for 3 days and stained with Syto-9 and Toto-3 as described in Materials and Methods. The distributions of populations within the biofilm are depicted by green stain for live cells and red stain for dead cells. These CLSM images are representative of three independent experiments.

DISCUSSION

Studies of the S. aureus cid and lrg operons have led to the model that these operons encode the molecular components controlling bacterial programmed cell death (5). Although the CidA and LrgA proteins are proposed to function in a way that is analogous to bacteriophage holins and antiholins, the biochemical characterization of these proteins has not previously been reported. Thus, the current study was designed to fill this void and, ultimately, to elucidate the molecular mechanisms affecting cell death and lysis in S. aureus and other bacterial species harboring cid and lrg homologues (5).

Computational analyses of the CidA and LrgA amino acid sequences suggested that the CidA and LrgA proteins are integral membrane proteins containing four membrane-spanning domains (8, 29). To demonstrate membrane localization experimentally, we utilized two different but complementary approaches. Examination of S. aureus cells expressing fluorescent fusion derivatives of CidA and LrgA (Fig. 1) indicated the surface localization of these proteins, a conclusion that is supported by the observation that an sGFP fusion to the well-characterized S. aureus membrane protein AgrB (43) was similarly localized to the cell surface. Confirmation of this subcellular localization was achieved using cell fractionation experiments, which clearly demonstrated the association of both CidA and LrgA with the membrane (Fig. 2B). This is an important finding since previous studies proposed that CidA and LrgA are functional analogues of the bacteriophage-encoded λ S holin and antiholin, respectively (29), whose functions are associated with the alteration of the cytoplasmic membrane.

Upon integration into the membrane, holins oligomerize into a structure essential for their function as effectors of cell death and lysis during the lytic cycle of a bacteriophage infection (36, 42). However, only recently has it been shown using cryoelectron microscopy that these proteins actually form a defined hole (33). Remarkably, these hole structures were shown to be quite large, having outer and inner diameters of 23 nm and 8.5 nm, respectively (33). Similarly, Bax, a mammalian protein, is believed to form high-molecular-mass oligomeric complexes in the mitochondrial membrane of apoptotic cells (2), and such oligomerization of Bax is required to trigger apoptosis (1). Thus, it is evident that oligomerization of membrane proteins is a well-conserved biochemical process used to induce cell death. Indeed, the functional similarities between holins/antiholins and the Bax/Bcl-2 family of proteins were key observations that led to our current model for bacterial PCD (5, 28). As such, we examined the propensity of the CidA and LrgA proteins to oligomerize and found that they also exhibit the ability to assemble into high-molecular-mass complexes (Fig. 3 and 4). Interestingly, the ability of the reducing agent DTT to resolve LrgA-H into its monomeric form (Fig. 3, lane 4) suggests that disulfide bonds are important for stabilizing these protein-protein interactions, an observation that was confirmed by the analysis of CidA and LrgA mutant proteins in which the cysteines had been changed to serines (Fig. 4). These results are similar to the finding that cysteine-mediated disulfide bond formation is important in the dimerization of λ S holin (10). However, unlike the result with an S holin cysteine mutant, which was still able to form oligomers (9), cysteine-mediated dimer formation of CidA and LrgA appeared to be a prerequisite for the oligomerization of these proteins.

Interestingly, the dimerization of the S holin was proposed to be an early step along the pathway to hole formation (10, 42). Based on the observation that the S holin cysteine mutant produced an early-lysis phenotype, it was reasoned that the formation of a holin-antiholin dimer had an inhibitory effect on the timing of lysis (10). Therefore, to understand the biological significance of cysteine in CidA, a cysteineless CidA mutant (strain KB1051) in which the wild-type copy of the cidA gene was replaced with the cysteine mutant allele, converting CidA cysteine into a serine [CidA(C104S)], was generated from UAMS-1. Although the growth characteristics of KB1051 were similar to those of UAMS-1, increased lysis of the KB1051 strain relative to that of UAMS-1 and the repaired strain (KB1052) became evident in late stationary phase. In addition, noticeable effects of the cysteine mutation were observed during biofilm growth. Under static growth conditions, biofilm generated by the KB1051 strain exhibited a small but reproducibly increased capacity to adhere to the surface of microtiter wells. Under these conditions, a corresponding increase in eDNA could not be detected, as was observed with the lrgAB and lytS mutant strains (18, 35). Under flow cell conditions, large areas of intense Toto-3-stained regions of the KB1051 biofilm were observed, suggesting the presence of dead cell clumps and/or areas rich in eDNA. Since these phenotypes were not observed in the wild type or the repaired strain (KB1052), we conclude that the loss of the cysteine-dependent oligomerization of CidA is responsible for the altered biofilm phenotype. This phenotype is in stark contrast to that produced by the cidA null mutant KB1050, which exhibited reduced cell lysis (30). Unlike S holin, whose function is dependent on oligomerization, these results suggest that CidA maintains functionality (albeit altered) in the absence of the ability to oligomerize. However, in light of the observation that an S holin cysteine mutant was still able to oligomerize and actually demonstrated an early-lysis phenotype (10), we speculate that the CidA(C104S) mutant KB1051 exhibits a similar S holin-like function in vivo, in that it is unable to dimerize but can still form functional oligomers. As with S holin, the dimerization of CidA may play a regulatory role in the function of this protein.

Interestingly, the predicted topology of CidA and LrgA in the bacterial membrane positions the cysteines in the hydrophobic region of the membrane bilayer (Fig. 8). Thus, the cysteines of CidA and LrgA may be important because the redox status of the membrane can affect the oxidative state of this amino acid. It has been implicated that redox sensing and signaling utilize sulfur switches, specifically cysteine residues, which are sensitive to reversible oxidation, nitrosylation, glutathionylation, acylation, sulfhydration, or metal binding (15, 16). Although the modification of these cysteines could alter the normal response of these proteins to the redox status of the cells, the molecular details of how the CidA(C104S) mutation affects biofilm formation remain to be elucidated. One possibility involves a scenario where modification of C104 leads to altered hydrophobicity of the transmembrane domain spanning this amino acid, which could change the topology of this protein in the membrane and, thus, alter its function. Alteration of Cys104 could then disrupt the normal adaptation of this protein to the varied redox conditions found in a developing biofilm.

Fig. 8.

Sequences and predicted topologies of CidA and LrgA, based on TMHMM server v2.0 (http://www.cbs.dtu.dk/services/TMHMM/). According to this prediction, CidA has one cysteine, located at middle of the fourth transmembrane domain, whereas LrgA has two cysteines, situated in the core of the second and fourth transmembrane domains (filled circles).

In this report, we examined the biochemical properties of CidA and LrgA of S. aureus, which have been proposed to function similarly to bacteriophage-encoded holins and antiholins, respectively. The results demonstrate common biochemical features shared between these proteins, supporting the model that the CidA and LrgA proteins are bacterial members of the holin family. Furthermore, these results provide the first glimpse of the molecular mechanism controlling bacterial PCD.