In dynamic environments, intracellular homeostasis is maintained by transport systems found in all cells. While bacterial influx systems for essential trace cations are known to contribute to pathogenesis, efflux systems have been characterized mainly in contaminated environmental sites. We describe that the high calcium concentrations in the normal human host were toxic to pneumococci and that bacterial survival in vivo depended on CaxP, the first Ca2+ exporter reported in bacteria. CaxP homologues were found in the eukaryotic sacroplasmic reticulum and in many bacterial genomes. A caxP− mutant accumulated intracellular calcium, a state that was used to reveal signalling networks responsive to changes in intracellular calcium concentration. Chemical inhibition of CaxP was bacteriostatic in physiological calcium concentrations, suggesting a new antibiotic target uncovered under conditions in the eukaryotic host.
Historically, the association of bacterial virulence and cation transport systems has emphasized the need for bacteria to scavenge scarce, essential resources, such as trace metals, to grow in the human host. On the opposite end of the spectrum, bacterial exporters are much more obscure. The efflux of trace elements, such as cobalt, copper, cadmium, gold, nickel, silver and zinc, has been found to be important in the context of heavy metal contamination of environmental sites (Silver, 1996; Nies, 2003; Moore and Helmann, 2006). In this circumstance, ABC transporters and P-type ATPases (Silver and Phung, 2005) prevent accumulation of potentially toxic concentrations of metals inside cells. Yet, a number of elements are also present in high concentrations in the human host and thus could potentially prove detrimental to bacterial growth during infection. For example, while the concentration of calcium in bacterial growth media is ∼20 μM, it is present in the millimolar range both on the mucosa and in the blood (Robinson et al., 1989; Vanthanouvong and Roomans, 2004). We hypothesized that during infection of the lung and blood, bacteria must use export systems to avoid accumulation of toxic intracellular levels of select, highly abundant cations, such as calcium.
Streptococcus pneumoniae commonly infects the lung and blood, acquiring nutrients such as carbohydrates and cations from the host (McAllister et al., 2004; Iyer and Camilli, 2007). The concentration of calcium in the pneumococcal cytoplasm is maintained between 2 and 14 μM (Chapuy-Regaud et al., 2001), a level at least 1000-fold less than that in the in vivo environment. Calcium efflux systems have been described in a number of bacteria, though the role of these transporters in pathogenesis remains unknown (Ivey et al., 1993; Ohyama et al., 1994; Norris et al., 1996; Tisa, 1998; Raeymaekers et al., 2002). Specialized import machineries are known to bring in calcium (Trombe, 1993) but a primary active efflux system has not been described in S. pneumoniae. We chose to investigate the mechanism of calcium homeostasis in face of the abundance of calcium in the host, as well as the key role calcium plays in a number of physiologic processes in S. pneumoniae, including genetic competence, autolysis and chaperone expression (Trombe, 1993; Trombe et al., 1994; Kwon et al., 2005; Johnston et al., 2006). We hypothesized that S. pneumoniae requires an active efflux system to maintain calcium homeostasis in a host environment that is highly enriched for this element. In this study, we describe the first primary calcium exporter in S. pnuemoniae and demonstrate that it is absolutely required for pneumococcal pathogenesis.
Identification of putative calcium efflux system
Examination of the TIGR4 S. pneumoniae genome (http://www.ncbi.nlm.nih.gov/sites/entrez?db=genome&cmd=Retrieve&dopt=Overview&list_uids=184) revealed a candidate for calcium transport, sp1551, based on extensive sequence similarity (35% identity, 55% similarity) to the SERCA calcium transporter found in the sarcoplasmic reticulum of eukaryotes(Wuytack et al., 2002). SERCA is a member of the P-type ATPase transporter family that moves cytosolic Ca2+ into membrane-bound cellular compartments. Sequence analysis and hydrophobicity plots comparing SERCA and Sp1551 revealed highly conserved domain structures, including residues within the active site conferring metal selectivity (Fig. S1). Extension of this comparison to other bacterial genomes revealed potential SERCA homologues in many Gram-positive bacteria, including Streptococcus pyogenes, Clostridium bolulinum, Bacillus anthracis, Enterococcus faecalis, Lactobacillus acidophilus and Bacillus subtilis, though calcium transport has not been demonstrated (Raeymaekers et al., 2003). Consistent with this broad conservation, the sp1551 gene is a member of the core pneumococcal genome, indicating a strong maintenance selective pressure (Obert et al., 2006). To determine the contribution of this putative transporter to cation homeostasis, we generated a deletion mutation by gene replacement in the pathogenic TIGR4 strain. The mutant grew normally in C+Y, a defined semi-synthetic casein liquid media used as a standard media, and showed no discernible differences in either competence or autolysis.
Cation toxicity assays
To ascertain the substrate and direction of transport of Sp1551, growth phenotypes of the deletion mutant were examined in the presence of an array of cations: additional substrate rescues an influx system mutant and inhibits an exporter mutant (McAllister et al., 2004; Kloosterman et al., 2007). Lawns of TIGR4 and sp1551− grown on TSA+ 3% sheep blood were overlaid with discs saturated with solutions containing calcium, copper, manganese, magnesium, cobalt, nickel or zinc. The zone of growth inhibition was measured following overnight incubation (Fig. 1A).
Strong inhibition of growth of sp1551− compared with TIGR4 was observed with the addition of calcium and manganese (P < 0.01). No significant differences in growth were observed for copper, cobalt, magnesium, nickel or zinc. To confirm the toxicity was due to the cation, the cognate anion was switched from chloride to nitrate or sulphate with the same levels of toxicity being observed (P < 0.01) (Fig. 1A). The cation sensitivity profile of the deletion mutant was abrogated when sp1551 was complemented back on a plasmid (Fig. 1B). Sequence homology to SERCA predicted Sp1551 to be a member of the P-type ATPase transporter family that requires ATPase activity at the conserved D336 site for activity (Raeymaekers et al., 2003). Site-directed mutagenesis of the predicted ATPase residue, D336, to alanine resulted in a protein that was unable to complement the sp1551− deletion (Fig. 1B). Hence, it appears that sp1551 encodes a calcium/manganese transporter homologous to SERCA. The fact that Sp1551 may efflux both calcium and manganese is not surprising as purified SERCA proteins have been shown to transport both cations (Wei et al., 2000).
As the sp1551− mutant showed heightened calcium and manganese sensitivity, we next definitively verified that these cations accumulated intracellularly. The intracellular elemental composition was measured by inductively coupled plasma mass spectrometry (ICP-MS) in parental TIGR4, sp1551− mutant and the sp1551−/p1551 complemented strain. The sp1551− mutant accumulated approximately fivefold greater levels of calcium as compared with TIGR4 (Fig. 1C). Complementation of sp1551 on a plasmid reduced intracellular calcium to wild-type levels. No difference in intracellular concentrations of either manganese or zinc were observed in sp1551− (Fig. 1C). The fact the sp1551− mutant did not accumulate intracellular manganese could be due either to a downregulation of the PsaABC import system by the manganese-responsive PsaR regulator, the activity of a specific export system for manganese (J.W. Rosch et al., unpubl. results), or the fact that physiological levels of manganese were used in these experiments and higher concentrations would be required to observe manganese accumulation. To investigate the kinetics of this process, TIGR4 and sp1551− were grown in ThyB to an OD600 = 0.5 at which time the media was spiked with 10 mM calcium chloride. ICP-MS was used to measure calcium levels at 15 min intervals. At time 0, 15 and 30 min the parental TIGR4 showed calcium concentrations of 0.955, 1.22 and 0.95 p.p.m. respectively, indicating the wild-type cells were able to maintain homeostasis under high calcium concentrations. The sp1551− mutant had calcium concentrations of 1.15, 5.2 and 6.8 p.p.m. at the corresponding time points indicating a rapid accumulation of intracellular calcium. These data indicate that Sp1551 functions as a calcium efflux system in S. pnuemoniae and thus was named CaxP (calcium exporter of pneumococcus). To our knowledge, this is the first example of such a system identified in S. pneumoniae.
Calcium efflux mutant as a probe for calcium-responsive gene networks
Gene expression in a number of regulatory pathways responds to metal availability (Kwon et al., 2005; Johnston et al., 2006). This can occur by sensing changes in extracellular signals, such as for two-component regulators, or by actually changing intracellular concentrations of key cofactors. While mutants in two-component systems have been widely studied, the latter situation has not. The accumulation of intracellular calcium in the caxP− deletion mutant suggested that loss of this protein function in bacteria could constitute a probe for pathways sensing intracellular calcium concentration. As calcium flux has been implicated in a number of cellular processes of pneumococci including competence, stress response and autolysis (Trombe, 1993; Azoulay-Dupuis et al., 1998; Dominguez, 2004; Kwon et al., 2005), we examined the global transcriptional profile of caxP− versus TIGR4 by microarray (Table S2).
Loci that showed significant upregulation in the absence of caxP generally were involved in stress response, including those encoding the Clp protease subunits, a heat-inducible repressor HrcR, DnaK and other putative stress proteins (sp1996) (Table S2). The MerR/NmlR transcriptional regulator, known to protect against oxidative stress, was induced 10-fold as was the downstream alcohol dehydrogenase which may bind calcium (Herbaud et al., 1998). Consistent with these changes in gene expression, the caxP− strain demonstrated 76 ± 13% greater lethality when exposed to nitric oxide stress than the parental control (Fig. S3). These findings suggest that an increase in the absolute intracellular concentration of calcium may induce oxidative defences of S. pneumoniae.
A number of genes considered to be important in pathogenesis were upregulated in the caxP− mutant, including neuraminidase and the pilus locus (Table S2). The pilus subunits and associated sortases were consistently upregulated threefold in the absence of the calcium efflux system as confirmed by qRT-PCR (Fig. S2). This did not appear to involve increased expression of the cognate transcriptional regulator RlrA as assessed by qRT-PCR or Northern blot (Fig. S2). Thus, in response to accumulation of intracellular calcium, a situation that could arise as the bacteria enter the high calcium environment of the lung, mucosal pathogenesis elements, such as pili and neuraminidase, appear to be induced.
Calcium export required for pathogenesis
Calcium is present at ∼5 mM on the respiratory mucosa (Robinson et al., 1989). To determine whether the calcium efflux defect in caxP− was relevant at physiological calcium concentrations, standard ThyB media (calcium concentration ∼20 μM) was supplemented with calcium chloride to final concentrations ranging from 2.5 to 10 mM and the growth of TIGR4 and the caxP− mutant was assayed. Both strains grew well in standard medium when calcium concentrations were lower than 1 mM. As the calcium concentration was increased, TIGR4 showed no growth defects, but the caxP− strain showed severe growth limitations at 2.5–5 mM calcium and total growth inhibition at 10 mM (Fig. 2A, 100% equivalent to OD = 1.0). No growth inhibition or benefit was observed with the addition of manganese or magnesium at physiological concentrations (data not shown). This phenotype was recapitulated in blood (calcium concentration ∼1 mM). TIGR4 displayed robust growth in blood while the caxP− mutant slowly died over 2 h (Fig. 2B) and no growth recovery was observed even after 24 h. Addition of EGTA to blood rescued the lethality of blood for caxP−, but inhibited growth of both TIGR4 and caxP−, likely due to depletion of an essential cation (data not shown). The same phenotype was observed when strains were cultured both in sterile human plasma (Fig. 2C) and serum (Fig. 2D) demonstrating relevance in human infection. These data indicate that CaxP is required for bacterial growth at physiologically relevant calcium concentrations, and the phenotype is not evident in standard bacterial growth medium.
As caxP− is deficient in the efflux of calcium that would be encountered at high concentrations in the host, we next sought to determine the role of this transporter in the pathogenesis of pneumococcal infection. Mice were infected intranasally and nasal and blood titres were taken daily to monitor disease progression. The overall survival data indicated that the caxP− strain was completely attenuated when compared with TIGR4 (Fig. 3A; P < 0.0001). A dramatic 5-log decrease in nasal colonization was observed for the caxP− strain after 24 h (Fig. 3B) and no bacteria were detected in blood at any time. To ensure this was not due to differences in the adhesive capability of these strains, adherence to nasopharygeal cell lines was confirmed to be unaffected by the caxP− mutation (Fig. S3). Likewise, no discernable effect on adherence was observed when bacteria were exposed to various concentration of calcium either immediately prior to or during the adhesion assay (data not shown). The attenuation of the caxP− strain was also observed in an intraperitoneal infection model with the caxP− mutant being rapidly cleared from the host within 18 h (Fig. 3D) and all mice surviving a lethal challenge (Fig. 3C). Due to rapid clearance of bacteria, we were unable to harvest enough bacteria to determine whether bacteria accumulated high levels of intracellular calcium during host infection. These data indicate that the calcium efflux system CaxP plays a vital role in the survival of the pneumococci in the host environment.
Five inhibitors were tested at various concentrations up to maximum solubility to determine whether they caused calcium-dependent growth inhibition of S. pneumoniae, effectively mimicking the phenotype of loss of function of CaxP. As the caxP− mutant is unable to grow under these high calcium conditions, we were unable to determine the effects of these inhibitors on the mutant. 2,5-di(tert-butyl)hydroquinone (DBHQ) and clotrimazole target residues in the drug binding site conserved between CaxP and SERCA and both showed calcium-specific bacteriostatic activity (Fig. 4A and B). For the three ineffective inhibitors, the residues involved in inhibitor binding were not conserved between the mammalian and bacterial homologues (Fig. 4A). This calcium-dependent growth inhibition supports the notion that calcium homeostasis achieved by CaxP is a viable and previously unrecognized antibacterial target.
The ability to both sense and respond to environmental cation concentration is a vital aspect of bacterial pathogenesis. Long understood to be a cornerstone of signalling in eukaryotes, calcium signalling in prokaryotes has attracted attention recently as bacteria also sense and respond to calcium levels (Dominguez, 2004). Calcium concentration is one cue for the expression of virulence factors by Yersinia species (Straley et al., 1993). Conceptually, such sensing could be accomplished by two general mechanisms. The bacteria could sense changes in the extracellular cation concentration, such as through two-component signal transduction systems (Gryllos et al., 2003). Less well recognized, bacteria could theoretically respond to increased intracellular calcium concentrations that accumulate if the rate of cation export cannot compensate for influx from high extracellular concentrations. This latter possibility is now clearly demonstrated to be operative by the alterations in gene transcription shown in the caxP− mutant. Thus, cation exporters could play a role in bacterial-host cell signalling similar to that of two-component systems.
The essential nature of the CaxP export system was evident only in high concentrations of cation found in eukaryotic environments. This explains in part why calcium exporters have not been identified previously even though it has been established that streptococci utilize ATP to transport calcium across the membrane (Ambudkar et al., 1986). The fact that disrupting calcium export shows profound defects in host colonization indicates a vital role in maintaining homeostasis in the host environment. Little is known regarding the expression of these efflux systems in the host, though caxP transcription is significantly increased in the cerebrospinal fluid (Orihuela et al., 2004). Further study of the interplay between these regulatory pathways in response to cation levels both in vitro and at various host tissues will contribute to our understanding of the specific responses to S. pneumoniae to these signals.
Two areas of bacterial metabolism appeared to be significantly affected by CaxP function: response to oxidative stress and pilus expression. The fact that calcium appears to play a role in the oxidative stress response is not surprising. Previous studies have found that intracellular calcium levels fluctuate in response to hydrogen peroxide (Herbaud et al., 1998). Depletion of calcium from media has also been shown to induce overexpression of antioxidant proteins including catalase and AhpC (Herbaud et al., 1998). AhpC appears to be a calcium-binding protein in both B. subtilis and Bordetella pertussis (Dominguez, 2004). Proteomics analysis of Pseudomonas aeruginosa revealed that five oxidative stress proteins were differentially regulated in response to calcium, including AhpC and catalase (Patrauchan et al., 2007). Of considerable interest is the observation that the activity of the mammalian homlogues of caxP can be modulated by selective oxidation of specific cysteine residues (Adachi et al., 2004; Dremina et al., 2007). One potential scenario is that under oxidative stress conditions, CaxP activity would be modulated via oxidation, resulting in an increase in intracellular calcium levels, signalling the bacteria to upregulate antioxidant proteins. Sensing an increase in cellular calcium levels in the high calcium environment of the lung may also play a role in virulence gene regulation, particularly for pilus and neuraminidase. Pili are involved specifically in infection of the lung (Rosch et al., 2008) and one signal for inducing pilus expression appears to be high intracellular calcium, a condition that could arise as bacteria transition into the particularly high calcium environment of the lung.
The cation efflux systems themselves may present a new target for future antimicrobial agents. As the active efflux of calcium by S. pneumoniae is required for host pathogenesis, one could envision inhibitors that specifically block the bacterial SERCA homologue that are unable to inhibit the eukaryotic counterparts. Calcium efflux plays a vital role in spore germination in both Bacillus anthracis and Clotridium perfringens, both of which encode proteins with remarkable homology to SERCA (Alvarez and Abel-Sanots, 2007). It is interesting to note that deletion of the SERCA homologue in Bacillus resulted in a specific defect in spore viability (Raeymaekers et al., 2003). Drugs targeting calcium efflux may prove effective in prevention of spore germination as well as host infection. Our data indicate that caxP is a vital aspect of pneumococcal survival in the host and that targeting this exporter is a viable antimicrobial strategy. This indicates new classes of antimicrobial compounds may be discovered through methodology that more closely mimics the host environment rather than standard laboratory conditions.
Media and growth conditions
Streptococcus pneumoniae was grown on tryptic soy agar (EMD Chemicals, NJ) supplemented with 3% sheep blood or in C+Y, a defined semi-synthetic casein liquid media (Lacks and Hotchkiss, 1960) supplemented with 0.5% yeast extract. Cultures of S. pneumoniae were inoculated from frozen stock and incubated at 37°C in 5% CO2. Mutant construction is detailed in supplementary information. Todd Hewitt Broth (ThyB) and ThyB depleted of cations (ThyC) using Chelex resin (Bio-Rad) was performed as previously described (Hanks et al., 2006). When appropriate, pH was adjusted using concentrated hydrochloric acid and subsequently sterile filtered. Sterile, defibrinated sheep blood was obtained from Enova Medical Technologies (St Paul, MN). Sterile filtered human plasma containing heparin as anticoagulant and sterile filtered human serum was obtained from Equitech-bio (Kerrville, TX).
Mutants in TIGR4 were made by PCR-based overlap extension. Briefly, flanking regions 5′ and 3′ to the target gene were amplified by PCR and spliced to an antibiotic cassette. The final PCR product was transformed into the pneumococcus by conventional methods, replacing the targeted gene with the antibiotic cassette. To confirm transformation, primers outside the transformed region were used for PCR and the region was subsequently sequenced. To construct sp1551− the gene was replaced with erythromycin cassette. For p1551 the coding region for sp1551 was cloned into the EcoRI/PstI site of the pABG5 expression vector. The oligos used are listed in Table S1.
RNA isolation and microarray analysis
Bacterial RNA was harvested from mid-log phase cultures using Qiagen RNAeasy minikit. Microarray experiments were performed as described previously (Orihuela et al., 2004). Briefly, whole-genome S. pnemoniae cDNA microarrays representing segments of all 2131 open reading frames of TIGR4 were obtained from TIGR (PFGRC). Microarray experiments were performed by the Functional Genomics Laboratory (Hartwell Center for Bioinformatics and Biotechnology, St Jude Children's Research Hospital) using standard protocols provided by PFGRC (http://pfgrc.jcvi.org/index.php/microarray/protocols/htm). Genes were considered differentially regulated for a minimum of twofold deviation from wild-type standards from three independent replicates, including a dye swap to negate staining bias.
Cells were grown in 24-well plates at 37°C in 5% CO2 to 80% confluency and activated for 2 h with TNF-α (10 ng ml−1). Pneumococcal cultures were grown to an OD620 of 0.5, washed and then added to eukaryotic cells at 1 × 107 cfu per well. Three wells were used for each mutant and the assays were repeated 3–4 times. For adherence assays, cells were incubated 30 min with bacteria, a time chosen to minimize internalization of adherent bacteria. After washing 3× in dPBS, the cells were released from the plate with trypsin but not lysed before plating on blood agar plates. Colonies grown overnight were counted as bacteria adherent to cells.
To determine cation sensitivity, 1 M solutions of calcium chloride, calcium nitrate, magnesium chloride, manganese chloride, manganese sulphate, nickel chloride, zinc chloride, cobalt (II) chloride (Sigma) were prepared in milli-Q water and sterile filtered. A freshly plated bacterial lawn on TSA + 3% sheep blood was overlaid with a sterile, 5 mm filter disc with 10 μl of the cation solutions. Plates were incubated overnight and zone of inhibition was measured for each disc. Statistical significance was determined using the Friedman test.
Intracellular elemental composition was measured by ICP-MS as previously described (Outten and O'Halloran, 2001). Cells were grown in Todd Hewitt Broth treated with Chelex (Sigma, St Louis, MO) to remove cations and supplemented with 5 mM calcium chloride, 35 μM manganese chloride, 1 mM magnesium chloride and 200 μM zinc chloride. Bacteria were collected by centrifugation at OD600 = 0.500 and washed four times in PBS + 10 mM EDTA. Pellet was dried overnight and analysed by the Bodycote Testing Group (Santa Fe Springs, California). Bacterial CFUs and dry weight were enumerated prior to sample analysis. Data were representative of three independent experiments.
Female BALB/cJ mice (Jackson Laboratory, Bar Harbor, ME) aged 6 weeks were maintained in BSL2 facilities. All experiments were done under inhaled isoflurane (2.5%). Bacteria were introduced by intranasal administration of 107 cfu in 25 μl of PBS. Nasal passages were lavaged and blood extracted from the tail vein at 24 and 72 h post infection, diluted, and plated on blood agar to ascertain bacterial colonization and bacteremia. In addition, mice were also challenged intraperitoneally with 105 cfu in 100 μl of PBS via injection. Statistical analysis was performed using Kaplan–Meier survival estimates.
Artemisinin, clotrimazole, cyclopiazonic acid, thapsigargin and DBHQ were purchased from Sigma and dissolved in DMSO (Sigma). Inhibitors were added to 10 ml of ThyB or ThyB supplemented with 10 mM calcium chloride (Sigma). A total of 105 cfu of TIGR4 was added to the cultures and growth was measured during incubation at 37°C in 5% CO2. Dose was set at a level fivefold lower than MIC in ThyB or saturation, whichever was lower. Inhibitors were considered effective if no growth was observed under high calcium concentrations when the control ThyB culture reached stationary phase.
We thank Dr Kip Guy for helpful scientific discussion, Morgann C. Reilly for critical editing of the manuscript and Granger L. Ridout for assistance in microarray experiments. We thank TIGR for providing microarrays. We thank Caroline Obert for assistance with the statistical analysis. This work was support in part by NIH R01AI27913, U54HL070590 and the American Lebanese Syrian Associated Charities.