The second messenger bis-(3′-5′)-cyclic-GMP and its PilZ domain-containing receptor Alg44 are required for alginate biosynthesis in Pseudomonas aeruginosa

Authors

  • Massimo Merighi,

    1. Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA, USA.
    Search for more papers by this author
  • Vincent T. Lee,

    1. Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA, USA.
    Search for more papers by this author
      Department of Cell and Molecular Genetics, University of Maryland at College Park, MD 20742, USA.
  • Mamoru Hyodo,

    1. Graduate School of Information Science/Human Informatics and CREST/JST, Nagoya University, Nagoya, Japan.
    Search for more papers by this author
      ‡Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA 02115, USA.
  • Yoshihiro Hayakawa,

    1. Graduate School of Information Science/Human Informatics and CREST/JST, Nagoya University, Nagoya, Japan.
    Search for more papers by this author
  • Stephen Lory

    Corresponding author
    1. Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA, USA.
      *E-mail stephen_lory@hms.harvard.edu; Tel. (+1) 617 432 5099; Fax (+1) 617 738 7664.
    Search for more papers by this author

  • Present addresses: Department of Cell and Molecular Genetics, University of Maryland at College Park, MD 20742, USA. Department of Microbiology and Molecular Genetics, Harvard Medical School, Boston, MA 02115, USA.

*E-mail stephen_lory@hms.harvard.edu; Tel. (+1) 617 432 5099; Fax (+1) 617 738 7664.

Summary

The ubiquitous bacterial second messenger c-di-GMP regulates the expression of various virulence determinants in a wide range of bacterial pathogens. Several studies have suggested that proteins with a PilZ domain function as c-di-GMP receptors. We have identified in the Pseudomonas aeruginosa genome eight genes encoding for PilZ orhologues and demonstrated binding of c-di-GMP to all but one of these proteins in a direct ligand binding assay. One protein with the PilZ domain, Alg44, is involved in biosynthesis of the extracellular polysaccharide alginate. We have shown that increasing c-di-GMP levels by overexpression of highly active diguanylate cyclases, or hydrolysis of c-di-GMP by phosphodiesterases, enhanced or reduced formation of alginate in mucoid strains, respectively. We have engineered substitutions in several conserved residues of the PilZ domain of Alg44 determined that they resulted in simultaneous loss of c-di-GMP binding and the ability to support production of alginate in P. aeruginosa. A 6xHis-tagged Alg44 fusion was also shown to localize in the membrane fraction of P. aeruginosa independently from its ability to bind c-di-GMP. Alg44 appears to be an essential component of the alginate biosynthetic apparatus, where, following binding of c-di-GMP, it controls polymerization or transport of the polysaccharide.

Introduction

Cyclic nucleotides 3′5′-cyclic-AMP (cAMP) and bis-(3′-5′)-cyclic dimeric GMP (c-di-GMP) are small molecule second messengers controlling a number of cellular processes in prokaryotic cells by functioning as cofactors for enzymes and regulatory proteins (Galperin, 2004; Romling et al., 2005). The broad range of stimuli that these small signalling molecules integrate into a variety regulatory targets is highlighted by the redundancy of enzymes that are responsible for their synthesis and degradation. In certain bacteria, dozens of adenylate cyclases (mostly belonging to the class III enzyme family) and multiple putative cAMP phosphodiesterases have been identified (Baker and Kelly, 2004), and the number of enzymes responsible for c-di-GMP metabolism diguanylate cyclases and phosphodiesterases) ranges from none in certain species to 99 in Vibrio vulnificus (Galperin, 2004). Signal transduction mediated by cAMP or c-di-GMP is initiated by activation of sensory recognition domains, usually fused to the catalytic domains of the cyclases, resulting in the production of the regulatory nucleotide which presumably binds to receptor proteins (Galperin, 2004). Phosphodiesterases are assumed to respond to different signals and their activation terminates the signalling event by hydrolysis of the signalling nucleotide.

A recent survey of the genome of Pseudomonas aeruginosa identified 17 genes encoding putative diguanylate cyclases, five phosphodiesterases and 16 proteins that contain both of these enzymatic domains. Enzymatic activity has been demonstrated in nearly half of these proteins (Kulasekara et al., 2006). Phosphodiesterase and diguanylate cyclase genes have also been reported to control phenotypes like biofilm formation and twitching motility (Hickman et al., 2005; Kazmierczak et al., 2006; Kulasekara et al., 2006). In general, high diguanylate cyclase activity promoted biofilm development including formation of multilayer bacterial mats (pellicles). On the other hand, high phosphodiesterase activity correlated with the loss of cytotoxicity. This implies that c-di-GMP levels control the ability of P. aeruginosa to form biofilm and to kill mammalian cell targets. These two key pathogenic mechanisms of P. aeruginosa have been shown previously to be reciprocally regulated by several regulatory networks involving two component regulatory systems (Goodman et al., 2004; Goodman and Lory, 2004; Ventre et al., 2006), and were suggested to play a role in controlling survival strategies of this organism in acute or chronic infections where formation of biofilm and cytoxicity play opposite roles (Furukawa et al., 2006). Therefore, c-di-GMP functions co-ordinately with transcriptional and post-transcriptional regulatory mechanisms in influencing the virulence characteristics of P. aeruginosa and other pathogens that produce this regulatory nucleotide.

In a variety of bacteria, conserved cAMP binding proteins have been identified (Botsford and Harman, 1992). Recently, Amikam and Galperin proposed that proteins containing a specific sequence motif, the so-called PilZ domain, are candidate receptors for c-di-GMP (Amikam and Galperin, 2006). Interestingly, multiple proteins with PilZ domains have been identified in different bacteria, however, until very recently these have not been shown experimentally to interact directly with c-di-GMP (Ryjenkov et al., 2006; Christen et al., 2007; Pratt et al., 2007; Ramelot et al., 2007). The original discovery of the regulatory role of c-di-GMP in the synthesis of cellulose in Gluconacetobacter xylinus (formerly Acetobacter xylinum) suggested that it may perform an analogous function in other polysaccharide biosynthetic pathways. In P. aeruginosa, one of the proteins with a PilZ domain is Alg44, encoded in the conserved cluster of genes responsible for the synthesis of the alginate exopolysacharide. Alginate plays an important role in P. aeruginosa chronic infections of cystic fibrosis patients (Gacesa, 1998; Yu and Head, 2002). Sequence analysis of Alg44 suggested that in this protein, the N-terminal PilZ domain is connected by a single predicted transmembrane segment to the a putative membrane fusion domain showing weak similarity to P. aeruginosa MexA, a membrane fusion component of the MexAB-OprM efflux pump (Poole et al., 1993). Furthermore, deletion of alg44 in P. aeruginosa and Azotobacter vinelandii abolishes alginate production (Maharaj et al., 1993; Mejia-Ruiz et al., 1997; Remminghorst and Rehm, 2006a). Based on these features, it has been suggested that Alg44 is a transmembrane protein with the C-terminal domain located in the P. aeruginosa periplasm, where it functions in facilitating the extrusion of polymerized alginate through the outer membrane pore AlgE. However, recent experiments with a PhoA fusion construct suggested that Alg44 is located entirely in the periplasm (Remminghorst and Rehm, 2006b).

Here we report the results of experiments demonstrating c-di-GMP binding to the majority of P. aeruginosa proteins containing PilZ domains. In further experiments we focused on one of these proteins, Alg44. We initially observed that alginate production in P. aeruginosa is suppressed by overexpression of some of the P. aeruginosa phosphodiesterases and stimulated by overexpression of certain diguanylate cyclases, suggesting an important physiological role for c-di-GMP in the production of this exopolysaccharide. Based on alignments of orthologues and on structural predictions, a number of conserved amino acids in the PilZ domain of Alg44 have been identified and these have been mutated to alanine. These mutant forms of Alg44 showed a defect in alginate production and were unable to bind c-di-GMP. Cross-linking with radioactive dinucleotide resulted in the formation of covalently associated monomeric complexes. Quantitative in vitro studies defined affinity constant, specificity of binding and stability of the complex. The inability of Alg44 PilZ domain mutants to bind c-di-GMP in vitro correlated with a defect in their ability to support production of alginate in vivo.

This study represents the first report of c-di-GMP being required for alginate production in any bacterial species and further validates the predictions from computational analysis (Amikam and Galperin, 2006) that PilZ domain proteins represent a link in the signal transduction pathway controlled by cellular levels of c-di-GMP.

Results

Pseudomonas aeruginosa genome encodes several PilZ domain-containing proteins

The PilZ domain has been recently suggested to be a putative binding domain for c-di-GMP (Amikam and Galperin, 2006). Escherichia coli YcgR and G. xylinus BcsA, have been the first proteins experimentally shown to bind c-di-GMP (Ryjenkov et al., 2006). During the review of this manuscript, several PilZ domain proteins and their properties have been characterized to various extent in three additional species, P. aeruginosa (PA4608), Vibrio cholerae (several Plz proteins) and Caulobacter crescentus (DgrA) (Christen et al., 2007; Pratt et al., 2007; Ramelot et al., 2007). Searches of the Pfam database for additional members of this protein family (Pfam07238) identified eight PilZ domain-containing proteins encoded in the P. aeruginosa genome (Fig. 1). These include PilZ (PA2960) a protein involved in twitching motility (Alm et al., 1996), Alg44 (PA3542), a putative component of the alginate synthetase complex in P. aeruginosa (Maharaj, 1993), as well as six other proteins of unknown function (PA2799, PA4608, PA2989, PA4324, PA3353, PA0012). We obtained Mar2xT7 transposon mutants in PA2799, PA3353, PA4324, PA4608, PA2989 from the PA14 mutant collection (Liberati et al., 2006) and for those that were not in the transposon library (PA0012, PA3542, PA2960) we engineered deletion derivatives. None of the mutants showed any discernible alterations in colony morphology when grown on Luria–Bertani (LB) plates. In addition to the pilZ mutant, PA2989::MAR2xT7 also displayed a twitching motility defect and showed poor biofilm formation when evaluated following static growth in polystyrene tubes (data not shown). Mutant in this locus also have slower growth rate in liquid culture.

Figure 1.

Alignment of Pseudomonas aeruginosa PilZ proteins and modelling of the Alg44 PilZ domain.
A. Clustal X alignment of P. aeruginosa PilZ domains. The conserved residues mutagenized in this study are labelled (R17, R21, D44, S46).
B. The Alg44 PilZ domain primary sequence Ca trace was threaded into the NMR model structure of PA4608 and the protein backbone and side chain co-ordinates were calculated using the MaxSprout algorithm (Ignoring a VDPASR insertion in alg44 PilZ, which correspond to a loop between two b-strands). Conserved residues R17, R21, D44 and S46 are shown.
C. Consurf image of the Alg44 PilZ domain created by using the alignment of the eight PilZ domain paralogues from P. aeruginosa. Conserved residues are shown in dark pink, divergent residues are in dark cyan.

In vitro binding of c-di-GMP to P. aeruginosa PilZ domain-containing proteins

In order to determine whether the P. aeruginosa proteins with PilZ domains can function as receptors for c-di-GMP, we purified recombinant forms of these proteins and assessed their ability to bind c-di-GMP in an in vitro assay. All but one were expressed as full-length proteins fused at their amino-termini to maltose binding protein (MBP). The PA3542 (Alg44) PilZ domain construct contained the N-terminal MBP followed by the PilZ domain (amino acids 12–140) and lacked the putative transmembrane domain and its C-terminal NolF domain. The recombinant proteins were purified by amylose affinity chromatography to a greater than 90% purity (Fig. 2A). These hybrid proteins were then tested for their ability to bind to [32P]-c-di-GMP (Fig. 2B). All but one PilZ protein were able to bind [32P]-c-di-GMP to varying degrees. Only PA2960 (PilZ) showed no detectable binding of [32P]-c-di-GMP under the conditions of this assay. Although it is possible that PilZ (PA2960) interacts with c-di-GMP in vivo, it is conceivable that it represents an inactive form of the domain. For the remainder of this work, we have undertaken a detailed characterization of the alginate biosynthesis regulator Alg44.

Figure 2.

Binding of P. aeruginosa MBP-PilZ protein fusions to c-di-GMP.
A. Recombinant fusion proteins of MBP to the various P. aeruginosa PilZ domain-containing proteins were fused to MBP, they were purified by amylose affinity chromatography and analysed by SDS-PAGE.
B. Binding of [32P]-c-di-GMP to the MBP-PilZ domain fusions, measured as the amount of radiolabel retained on nitrocellulose filter. Nitrocellulose filters were exposed to phosphor imager screen and quantified in a phosphorimager.

Alginate biosynthesis requires c-di-GMP

Regulation of cell surface properties, including exopolysaccharides, is also a well establish paradigm for c-di-GMP signalling but there is no report of a link between alginate biosynthesis and c-di-GMP levels in the cell. We have previously identified a number of P. aeruginosa proteins with conserved EAL or GGDEF domains found in c-di-GMP phosphodiestereses (PDEs) and diguanylate cyclases (DGCs), respectively, and demonstrated enzymatic activities in a fraction of these proteins (Kulasekara et al., 2006). We therefore introduced an expression plasmid (pBTK27) carrying the gene for two of the most active enzymes (PA2133, a PDE, and PA1107, a DGC) into strain MM6267 (a mucoid variant of P. aeruginosa PA14) and assessed the consequences of altered cellular c-di-GMP levels on production of alginate, normalizing alginate production over cell density and total cell protein. As shown in Fig. 3A, overexpression of PA2133 (from plasmid pPA2133) in strain MM6267, leads to a significant reduction in detectable alginate by the uronic acid assay and these strains were non-mucoid on LB-glycerol agar plates (Fig. 3B). In contrast, raising c-di-GMP levels by overproduction of the DGC PA1107 (from pPA1107) significantly enhanced alginate production (approximately sixfold; Fig. 3A) and colony appearance (whitish and transparent mucoidy compared with the lesser mucoidy of the strain with the vector control; Fig. 3B). Phenotypes appeared stable: 8–10 independent colonies were re-isolated and upon restreaking, strain expressing PA1107 appeared hyper-mucoid while strains expressing PA2133 were non-mucoid. These qualitative results were reproduced also in the independently obtained strain MM6268 (data not shown). In a different mutant background producing approximately 10-fold higher amounts of alginate (a PAK mucA polar mutant isolated during an unrelated MAR2xT7 mariner transposon screening) the repression of alginate production by the phosphodiesterase enzymes were also reproducible but of a lesser magnitude, when using the two PDEs PA2133 and PA2200 (Fig. 3C). An enhancement in alginate accumulation was obtained using either of the DGCs PA1107 or PA1120, with effects comparable to what observed in the PA14 background (Fig. 3A–C). Collectively, these results demonstrate an important and novel role for c-di-GMP in alginate biosynthesis/export and correlate with the requirement of alg44 for alginate biosynthesis reported in the literature (Remminghorst and Rehm, 2006b).

Figure 3.

Effects of phosphodiesterases or diguanylate cyclases overexpression on alginate production in P. aeruginosa.
A. The P. aeruginosa PDE genes PA2133 and the DGC gene PA1107, in plasmid pBTK27, were introduced into the strain MM6267 (a mucoid PA14 mutant). Cells were grown on LB-glycerol plates with 1.5 mM IPTG for 24 h. Alginate production was quantified using the carbazole method for uronic acid determination and the data are shown as μg alginate per OD600 of the cell suspension, or per mg total bacterial cell protein (determined by a BCA assay in presence of 0.1% SDS). P. aeruginosa PA14 and MM6267/pBTK27 were used as controls. Data are from a representative experiment with three replicates.
B. The same strains described in A were visually evaluated for mucoid phenotype after growth in LB-glycerol with 1.5 mM IPTG for 30 h.
C. Strain MAR51, a mucoid P. aeruginosa PAK mucA::MAR2xT7 mutant carrying the diguanylate cyclases PA1107 or PA1120 or the phosphodiesterase PA2133 or PA2200 expressed from plasmid pBTK27, were plated on LB-glycerol agar containing 1.5 mM IPTG and grown for 24 h at 37°C. Alginate was quantified as described in A and normalized per optical density (at 600 nm) of the cells. Data are an average of four to six replicates in two separate experiments.

Alg44 sequesters c-di-GMP in vivo

We also examined the ability of Alg44 to sequester c-di-GMP away from a heterologous receptor. Certain isolates of E. coli produce extracellular cellulose in a process that requires AdrA, a diguanylate cyclase (Zogaj et al., 2001). Cellulose can be detected by calcofluor staining. For our experiments we used E. coli ZK57, a clinical isolate known to produce cellulose. In this strain, plasmid expression of a P. aeruginosa c-di-GMP phosphodiesterase (PA2200) abolishes the production of cellulose as evidenced by the lack of calcoflour staining of bacterial colonies to the level seen in an adrA mutant (Fig. 4). The full-length Alg44 protein was expressed in E. coli ZK57 and we examined its ability to sequester c-di-GMP from the cellulose synthase machinery by competing with BcsA, the c-di-GMP-binding subunit of the cellulose synthase complex (Ryjenkov et al., 2006). Upon induction of alg44 expression with IPTG, cellulose production was evaluated by appearance on calcofluor-containing agar plates exposed to UV light. Expression of Alg44 in E. coli ZK57 interfered with cellulose production (Fig. 4), while no differences were observed in the vector control. The reduction in fluorescence was comparable to that observed upon expression of the phosphodiesterase PA2200 (Fig. 4). Therefore, overexpression of the c-di-GMP binding protein Alg44 reduces cellulose production to a comparable extent as overexpression of an exogenous phosphodiesterase, likely reflecting titration and sequestration of cellular c-di-GMP away from the BcsAPilZ domain.

Figure 4.

Alg44 sequesters c-di-GMP in vivo. The cellulose producing E. coli strain ZK57 (WT) or the diguanylate cyclase (adrA) defective mutant ZK57 adrA::kan (adrA) were grown overnight in LB with or without plasmid pMMB-alg44 or with and without plasmid pPA2200, which expresses a P. aeruginosa phosphodiesterase. Drops of bacteria were placed on LB plates containing 200 μg ml−1 calcofluor and with or without 1.5 mM IPTG. The plates were incubate 48 h at 30°C and expose at 362 nm to detect blue fluorescence from cellulose-dependent binding of calcofluor.

The Alg44 PilZ domain specifically binds c-di-GMP

Based on the Pfam annotation, Alg44 is a two-domain protein with an N-terminal PilZ domain and a C-terminal membrane fusion domain (Dinh et al., 1994) with weak similarity to AcrA/EmrA domains (COG0845/COG1566 respectively; but most similar to the AcrA domain of Acetinobacter baumanni NolF in blast searches) and separated by a potential single-pass transmembrane helix. The N-terminal portion of Alg44, containing the PilZ domain (amino acids 1–140, referred to as Alg44PilZ), was expressed as a His6-tagged protein and purified by nickel affinity chromatography. The His6-tagged protein was used in photo cross-linking with [32P]-labelled nucleotides to qualitatively assess the ability of the isolated His6-tagged domain to function as receptor for c-di-GMP. The Alg44PilZ-nucletotide mixtures were either incubated in the dark or they were exposed to UV light, followed by SDS-PAGE fractionation and autoradiography. A radiolabelled band of molecular weight corresponding to approximately 15 kDa was present when the proteins were incubated with [32P]-c-di-GMP (Fig. 5). Incubation of Alg44PilZ with radioactive GTP nor ATP did not yield a detectable cross-linked complex.

Figure 5.

Photo cross-linking of c-di-GMP to Alg44PilZ. Recombinant His6-Alg44PilZ (5 μm) was incubated with 100 μm [32P]-labelled nucleotides (ATP, GTP) or 2.5–500 μm [32P]-c-di-GMP (at ∼2 Ci Mmol−1) and irradiated at 254 nm (+UV) or incubated in the dark (–UV) for 5 min. Proteins were precipitated with trichloroacetic acid, fractionated by SDS-PAGE gel, stained with Coomassie Blue and exposed to a storage phosphor imager.

The half life of the complex between Alg44PilZ and c-di-GMP was estimated by a filter binding assay after incubating the protein with high concentration of [32P]-c-di-GMP and eliminating the excess unbound nucleotides by spin-column gel filtration (Fig. 6A).

Figure 6.

Specific binding of Alg44PilZ to c-di-GMP.
A. Determination of the dissociation constant for the Alg44PilZ : c-di-GMP complex by filter binding. His6-Alg44[1−140] was incubated with [32P]-c-di-GMP at saturating concentrations for 10 min, unbound ligand was separated by gel filtration over a G25 Sephadex spin column, and at various time points aliquots were applied to a nitrocellulose filter under vacuum. Teh amount of bound c-di-GMP was quantified by liquid scintillation counting and fitted to a first-order decay model with GraphPad.
B. Isothermal titration calorimetry: His6-Alg44[1−140] (2.5 μM) was titrated with c-di-GMP, concentration of 150 or 300 μM, in binding buffer, at 25°C. Power necessary to maintain constant temperature during the titration experiment (μcal/ s), reflection of the amount of heat released upon binding, was measured over a set of 30 injections of ligand (top graph) and integrated over time to obtain a specific enthalpy vs. ligand : protein molar ration plot (bottom). Background-subtracted titration isotherms are shown.
C. Effect of c-di-GMP, GTP and GMP on binding of [32P]-c-di-GMP to Alg44PilZ. Ligand-Alg44PilZ complexes (5 μM protein) were allowed to form in presence of constant concentrations of radiolabelled c-di-GMP in presence of increasing concentrations of non-radioactive ligand (c-di-GMP as the specific competitor or GMP, GDP as non-specific competitors). Bound c-di-GMP was estimated by filter binding over a time-course and data were fitted into an IC50 sigmoid model with GraphPad. Data are expressed as percentage ligand bound at various competitor concentrations compared with the no-competitor control (%B/Bo).

The bound c-di-GMP dissociated from the Alg44PilZ with a half-life of ∼5 min. Size exclusion chromatography using HPLC equipped with a GSW3000 column showed that the protein preparation remained as a monomer with the retention time of 7 min even in the presence of c-di-GMP (data not shown), suggesting that c-di-GMP binding is not causing multimerization of this protein.

The binding affinity of the ligand for its receptor protein was quantified by isothermal titration calorimetry for both His6-Alg44PilZ[1−140] (Fig. 6B) and MBP-Alg44[12−140] (data not shown) proteins. Proteins were titrated with an injectant solution of 300 μM c-di-GMP at 25°C and the heat released upon binding was measured by integration of power values over a range of ligand concentrations. The fitting of the binding isotherm data (plotted as specific binding enthalpy vs. receptor : ligand molar ratios) suggests binding of the ligand to Alg44PilZ with a modest affinity (KD = 8.4 μM for the His6-tagged protein, Fig. 6B and 5.2 μM for the MBP-tagged version, data not shown) and a 1:1 stoichiometry (Fig. 6B). To evaluate the relative specificity of the Alg44PilZ for the ligand, competition experiments were performed using filter binding and unlabelled c-di-GMP, GTP or GMP as competitor ligands, respectively. The Alg44PilZ c-di-GMP complexes were formed by incubating the protein with constant amount of radiolabelled ligand in the presence of increasing concentrations of unlabelled competitors. Measurements of the radioactivity retained on filters allowed to determine an apparent IC50 (half maximal inhibitory concentration) value of 45 μM for c-di-GMP, while increasing concentrations of GMP or GTP were unable to compete out [32P]-c-di-GMP effectively even at millimolar concentrations (Fig. 6C), confirming the specificity of the Alg44PilZ domain for cyclic diguanylate compared with other analogous guanidine base nucleotides. Similar experiments conducted using 1000-fold excess of the nucleotides GDP, cGMP, ATP, CTP, UTP also showed specificity of the Alg44PilZ domain for [32P]-c-di-GMP (data not shown).

Conserved residues in the PilZ domain of Alg44 are required for c-di-GMP binding

From the alignment of the various PilZ domains of P. aeruginosa, several charged or polar residues in the N-terminal portion are highly conserved (Fig. 1A, black boxes). We modelled the structure of Alg44PilZ with MaxSprout (http://www.ebi.ac.uk/maxsprout/) using NMR structural data obtained for PA4608 (Ramelot et al., 2007) (Fig. 1B) and determined the predicted surface localization of several of the conserved amino acids of the PilZ domain, namely the R17, R21, D44 and S46 residues. These amino acids residues are on the same side of the protein, potentially forming a charged/polar binding surface, as visualized in the model structure created with Consurf (http://consurf.tau.ac.il/) (Fig. 1C). Clusters of charged (especially arginine) residues were identified in allosteric and catalytic c-di-GMP binding sites in diguanylate cyclases (Chan et al., 2004; Christen et al., 2006). Therefore, the corresponding residues in Alg44 represent potential ligand binding contact residues. These residues also appear to be in a solvent-exposed region of the protein (an unstructured N-terminal chain and a β-strand), increasing the likelihood that substitution mutations at these locations may affect function rather than disrupt protein folding. We created mutants of Alg44PilZ where codons for R17, R21, D44 and S46 were replaced with codons for alanine using polymerase chain reaction (PCR) mutagenesis. Proteins were expressed as fusion proteins with N-terminal His6-tags and purified (Fig. 7; bottom panel). When expressed in E. coli, their yield and stability during storage was comparable to that of the wild-type Alg44PilZ (data not shown) suggesting that the substitutions introduced in the polypeptide did not affect the stability of the mutated proteins. The proteins with the various substitutions were used to quantify the c-di-GMP binding activity compared with wild-type His6-Alg44PilZ by filter binding assays. Fig. 7 (top panel) shows the results of a typical experiment with the data expressed as per cent binding of the saturated Alg44PilZ::c-di-GMP complex (Bmax) across a 5-log ligand concentration range. The wild-type protein showed an apparent KD of 98 μΜ (roughly comparable to the apparent IC50, 45 μM). In contrast, the mutant proteins showed little binding of c-di-GMP below ∼500 micromolar range. After fitting to a single binding site model using GraphPad, the R21A variant had a KD of 2.4 mM, the D44A variant a KD of 0.89 mM, the S46A variant showed a KD of 495 μM and the double mutant R17AR21A had an estimated KD of 4.98 mM. These apparent KD values were calculated by fitting the binding data to a one-site binding isotherm model and they are useful only for relative comparisons not as estimates of the actual KD value.

Figure 7.

Binding of c-di-GMP to the Alg44 PilZ domain and the various mutants with alanine substitutions in the conserved residues. Binding of [32P]-c-di-GMP to wild-type and mutant His6-Alg44PilZ was determined by filter binding using increasing concentration of c-di-GMP at constant specific activity (0.2 Ci mmol−1) in the micromolar to millimolar range. Data for the wild-type protein are presented as per cent of maximal fractional binding at a concentration of c-di-GMP corresponding to ∼100-fold over the KD. Mutant proteins binding is shown as per cent relative to the wild-type binding (%Bwt) at various ligand concentrations. Data fitting was performed using GraphPad Prism 3.02 with a non-linear regression model. Data points are averages of two to four replicates and standard error bars are shown. The His6-tagged proteins used in this experiment were expressed in E. coli BL21(DE3) and purified by immobilized metal affinity and ion exchange chromatography. WT is His6-Alg44PilZ[1−140] (lane 2) and R21A (lane 3), R17AR21A (lane 4), D44A (lane 5) and S46A (lane 6) are the various substitutions in His6-Alg44PilZ.

Phenotypes of alg44 deletion and point mutants

To demonstrate that alginate production requires Alg44 to posses a functional c-di-GMP-binding activity, we generated a number of strains with deletions and point mutations in alg44 and measured alginate production in a strain where the alginate biosynthetic operon was ectopically induced by AlgU. Multiple alleles for the PilZ domain deletion were constructed with the objective of testing various boundaries for the PilZ domain, but only one of these is shown here because their effects were comparable (Fig. 8A). We also created one deletion of the short hydrophobic segment in the middle of the proteins (amino acids 475–534) and deleted the entire C-terminal NolF domain while leaving intact the putative transmembrane domain (Fig. 8A). Finally, the alginate phenotype of the point mutations which abolished c-di-GMP binding in vitro to various degrees were also determined. All of the mutations were introduced into the P. aeruginosa chromosome by allelic exchange. As a control for complete loss of alginate production, a deletion of the alg8 gene encoding for the alginate polymerase was also constructed. Because P. aeruginosa PA14 is not mucoid and does not produce alginate when grown in LB (Fig. 3A), we introduced plasmid pMMB67-algU into each mutant strain. Overexpression of AlgU overcame the inhibitory effect of the MucA anti-sigma factor and allowed synthesis of alginate. The mucoid (Alg+) phenotype and alginate accumulation by strains grown in the presence of IPTG on LB + glycerol was visually evaluated and quantified with a chemical assay for uronic acid. The analysis of the various mutations and their effect on alginate production is shown in Fig. 8A. Deletion of alg44 resulted in the non-mucoid phenotype and a significant (50-fold) reduction of alginate as measured by the uronic acid assay. This level of alginate was comparable to that observed in a strain with the deleted alginate polymerase gene (alg8). Deletion of the N-terminal PilZ domain (in strain MM6314) or the C-terminal NolF domain (strain MM6432) severely impaired alginate production (Fig. 8A) and the colonies became non-mucoid not shown). In-frame deletion of the hydrophobic segment between these domain (in strain MM6438) did abolish mucoidy, however, low amounts (approximately 17% of the wild-type level) of alginate were detectable by the uronic acid assay. Point mutations that abolished binding of c-di-GMP also caused an impairment of alginate production. On the other hand, strain MM6434, which carries an alg44 allele with an S46A mutation, was somewhat leaky, producing amounts of alginate significantly above the background (∼15% of the wild type level). However, apparently insufficient amount of alginate was produced to convert it to mucoidy. As shown earlier, this variant of Alg44 bound c-di-GMP very poorly (KD of 495), however, it still had the highest binding affinity among all of the Alg44 point mutants tested. This residual binding activity, especially when combined with higher steady state level of the protein (see below), may have contributed to the leaky synthesis of alginate.

Figure 8.

Effects of mutations in Alg44 on alginate production and protein stability.
A. Deletion and point mutations in alg44 were constructed and they were introduced into the chromosome of P. aeruginosa PA14 by allelic exchange (see gene maps). Alginate synthesis was induced in the various strains by expressing algU from plasmid pMMB67EH-algU. A deletion mutant in the alginate synthase gene alg8 was used as negative control. Cells were grown overnight in LB and 50 μl were plated on LB-glycerol agar and incubated for 18–20 h in presence of 10 μM IPTG. Cells were collected, washed in saline and alginate was prepared by isopropanol precipitation of the cell supernatant. Alginate determinations were performed by the carbazole method and seaweed alginate was used as standard. Normalization was performed using cell density, with results comparable to the use of dry cell weight or total protein.
B. P. aeruginosa Δalg44 cell pellets expressing AlgU from pPSV35 and C-terminal His6-tagged Alg44 proteins from pMMB67EH were boiled in sample buffer and total cell proteins were run on an SDS-PAGE gel, transferred to PVDF and probed with anti-His6 antibodies. The same samples were visually evaluated for alginate production (Alg+ or Alg–). Lane 1, PA14 Δalg44 (pMMB67EH-alg44-His6, pPSV35-algU); lane 2, PA14 Δalg44 (pMMB67 EH, pPSV35-algU); lane 3, PA14 Δalg44 (pMMB67 EH-Δalg44[475534]-His6, pPSV35-algU); lane 4, PA14 Δalg44 (pMMB67 EH-alg44[S46A]-His6, pPSV35-algU); lane 5, PA14 Δalg44 (pMMB67 EH-alg44[R21A]-His6, pPSV35-algU); lane 6, PA14 Δalg44 (pMMB67 EH-alg44[R17A,R21A]-His6, pPSV35-algU); lane 7, PA14 Δalg44 (pMMB67 EH-alg44[D44A]-His6, pPSV35-algU).

To determine if the mutant proteins were stably expressed in P. aeruginosa, we cloned all the mutant alleles as C-terminal His6-tags into pMMB67EH. We then ectopically express these constructs in MM6225 (an alg44 deletion strain) carrying algU on the compatible plasmid pPSV35 (pPSV35-algU). Following IPTG induction, the mucoid phenotype was assessed and the level of Alg44 protein in each strain was determined by Western blot analysis using anti-His6 antibodies(Fig. 8B). All of the mutant proteins with amino acid substitutions were stably expressed, as already expected from the fact that the analogous recombinant proteins containing the mutated domains could be purified in E. coli for the biochemical experiments. Of the deletion mutants only the product of the Δalg44[475534] allele could be detected, although it was expressed at significantly reduced levels. The remaining deletion derivatives were either unstable as His-tagged proteins or were structurally impaired and degraded. The Alg44[S46A] was highly overexpressed when compared with the other point mutants (Fig. 8B lane 4). This observation could in part explain the elevated levels of alginate seen in P. aeruginosa carrying this ectopically expressed mutant giving rise to mucoid (Alg+) colonies.

Alg44 preferentially localizes in the total membrane fraction

It has been recently proposed that an ectopically expressed Alg44-PhoA fusion is soluble and localized in the periplasm of P. aeruginosa (Remminghorst and Rehm, 2006b). Periplasmic localization of Alg44 would raise several issues in view of our results, including the need for a mechanism for c-di-GMP secretion. Therefore, to confirm these results using an alternative approach we determined the physical localization of recombinant Alg44-His6 in cell fractionation experiments of non-mucoid cells as well as mucoid cells where the alginate bisynthetic machinery was induced by overexpression of algU.

The alg44 gene including its ribosome binding site was cloned into plasmid pMMB67EH following PCR amplification of the DNA coding sequence using primers that add a His6-tag to the C-terminus of the expressed protein. The Alg44-His6 was fully functional, and complemented the Alg, non-mucoid phenotype in PA14 Δalg44/pPSV35-algU, while N-terminal fusion of Alg44 to the His6 tag could not complement a Δalg44 mutant, as assessed by mucoid colony morphology of plate-grown cells (data not shown). Following induction of alg44 expression we prepared periplasmic, total membrane and cytoplasmic fractions from non-mucoid cells as well as from cells where the alginate biosynthetic machinery was induced by ectopic expression of algU. The fractions were analysed by SDS-PAGE and Western blotting with anti-His6, anti-σ70 and anti-β-lactamase antibodies to detect Alg44-His6, the cytoplasmic marker σ70 and the periplasmic marker β-lactamase, respectively (Fig. 9A). Alg44-His6 was found in the membrane fraction, with very modest amounts sometime also present in the cytoplasm when AlgU was not induced (data not shown). The mutant of Alg44 lacking the hydrophobic segment (ΔTM; Fig. 9A) was synthesized at reduced amounts as previously observed (Fig. 8B), however, it was still localized to the membrane fraction (Fig. 9A).Interestingly, all the point mutants showed identical membrane localization as the wilt-type Alg44, implying that localization in this compartment did not require c-di-GMP binding.

Figure 9.

Localization and stability of Alg44.
A. Cell fractionation. P. aeruginosa strains were grown in LB to OD600 of approximately 0.15–0.2 and induced with 0.25 mM IPTG for 40 min Cells were fractionated into cytoplasmic fraction, total membranes (100 000 g fraction) and periplasm. Fractions were analysed by SDS-PAGE loading same number of cells per lane using OD600 as reference. Western immunobloting was done using monoclonal antibodies against hexa-histidine (His6), β-lactamase (Bla; periplasmic marker) and the σ70 subunit of RNA polymerase (σ70; cytoplasmic marker); Alg+ indicates ectopic expression of AlgU: 1, PA14 Δalg44 (pMMB67EH, pPSV35-algU); 2, PA14 Δalg44 (pMMB-alg44His6/pPSV35-algU); 3, PA14 Δalg44 (pMMB-Δalg44[473–534]-His6, pPSV35-algU); 4, PA14 Δalg44 (pMMB-alg44[S46A]-His6, pPSV35-algU); 5, PA14 Δalg44 (pMMB-alg44[R17AR21A]-His6, pPSV35-algU); 6, PA14 Δalg44 (pMMB-alg44-His6, pPSV35).
B. Steady-state levels of Alg44 in the presence or absence of AlgU (Alg±) was evaluated by inducing strains PA14 Δalg44 (pMMB-alg44His6, pPSV35) and PA14 Δalg44 (pMMB-alg44-His6, pPSV35-algU) by 0.25 mM IPTG in a 0–120 min time-course. Alg44-His6 was quantified by storage phosphor analysis and densitometry following Western blot with anti-His6 antibody.
C. Stability of Alg44 was evaluated by inducing the strains described in A for 45 min and by blocking translation with 300 μg ml−1 tetracycline. Total cell protein was separated by SDS-PAGE and Alg44-His6 was quantified by as described in B.

We consistently observed lower amounts of Alg44 protein in membrane fractions purified from cells that were not induced for alginate production by ectopic expression of algU (Fig. 8A, lanes 2 vs. lane 6). This implies that AlgU-dependent factors may be important for Alg44 stability. We therefore determined the kinetics of Alg44 synthesis in whole cells. As shown in Fig. 9B, in the absence of the alginate biosynthesis, Alg44 production was delayed and accumulated at reduced levels during the course of the experiment. Furthermore, following inhibition of translation with tetracycline a shorter half-life for Alg44 was observed in absence of algU expression (Fig. 9C), suggesting that AlgU affects Alg44 protein stability rather than its expression.

Discussion

The regulatory functions of second messenger cAMP is mediated by specific cellular receptors whose function is altered following binding (or release) of the nucleotide ligand (Botsford and Harman, 1992; Kolb et al., 1993) Recent discovery that a significant fraction of prokaryotic cells use c-di-GMP as a small regulatory molecule to regulate various functions involved in biofilm formation, motility, expression of virulence factors or biosynthesis of polysaccharide intensified the search for components of the signalling pathway that use c-di-GMP as a signalling intermediate. A novel class of proteins of unknown function carrying so called PilZ domain (Pfam07238) was initially implicated in c-di-GMP binding by bioinformatic analysis and by direct demonstration of interactions of an E. coli and G. xylinus PilZ domain-containing proteins with c-di-GMP (Amikam and Galperin, 2006; Ryjenkov et al., 2006). We have therefore carried out a systematic analysis of all PilZ domain-containing proteins encoded in the P. aeruginosa genome. Eight proteins that carry this domain were identified. These were expressed as fusions with MBP and following affinity purification, they were tested for c-di-GMP binding. All but one fusion protein bound radioactive c-di-GMP, ranging from approximately four- to 99-fold over the control MBP protein. Interestingly, the only protein unable to bind c-di-GMP was PilZ (PA2960), a protein required for type IV pilus-mediated twitching motility, after which the PilZ motif was named. It is conceivable that PilZ binds c-di-GMP with an affinity that is not detectable in our assay or it may have a special requirement for c-di-GMP binding in vivo. However, PilZ could represent an evolutionary relic of a conserved family of nucleotide binding proteins that accumulated mutations, resulting in its loss of ability to bind c-di-GMP without affecting its role in biogenesis or function of type IV pili. Certain key residues are also missing in PA2989 (e.g. the residue corresponding to Alg44 R17 and D44), but they are either flanked by residues with similar charge properties (e.g. K) or substituted by polar residues (e.g. S instead of D), which may perform the same function in c-di-GMP binding. It must be noted also that in different PilZ proteins the requirement for the serine residue seems to be variable, and in one case substitution to alanine lead to a slight increase in binding affinity (Ryjenkov et al., 2006; Pratt et al., 2007).

One of the members of the P. aeruginosa PilZ family, Alg44, was studied in greater details. Alg44 is required for alginate production in P. aeruginosa although its precise role in synthesis, export or modification of this polysaccharide is not known. Alg44 is a bipartite modular protein, containing a PilZ domain at its amino terminus, separated from the C-terminal NolF domain by a short hydrophobic stretch of amino acids. The NolF-AcrA-EmrA domain is found in various Gram-negative ABC transport systems where it has been suggested to function as a coupling protein between the cytoplasmic and outer membrane components of the transport apparatus (Binet et al., 1997). We have shown that Alg44PilZ binds c-di-GMP specifically, and its nucleotide binding activity requires several amino acids located on the exposed loop of the PilZ domain, recently shown to be one of the two c-di-GMP-binding surfaces of PilZ domains (Christen et al., 2007). These data, especially the critical role of arginine residues for c-di-GMP binding, are supported also by similar site-directed mutagenesis experiments performed with E. coli YcgR and Vibrio cholerae PilZ domain proteins (Pratt et al., 2007). On the other hand, less certain is the importance of the conserved serine residue in various PilZ domains. A similar mutation (S147A) in E. coli YcgR does not negatively affect c-di-GMP binding (Ryjenkov et al., 2006), while in Vibrio cholerae PlzD (S164A) caused a sharp reduction in binding of the dinucleotide, indistinguishable from other substitutions in the conserved residues (Pratt et al., 2007). In P. aeruginosa, the Alg44, the S146A substitution caused a strong reduction in c-di-GMP binding but much less than other mutations, while the mutant phenotype is somewhat leaky, especially upon overexpression.

We have also shown that substitutions in these same amino acids do not affect protein localization in the membranes but lead to a severe impairment of alginate production by mutant P. aeruginosa expressing these Alg44 variants. Therefore, c-di-GMP binding to Alg44 appears to be essential for alginate production by P. aeruginosa and represents a new post-translational regulatory mechanism controlling the synthesis of this important virulence factor. This is also consistent with the observation that overexpression of P. aeruginosa phosphodiesterases previously shown to have among the highest enzymatic activities (Kulasekara et al., 2006) was able to suppress alginate production to a level comparable to null alg44 or alg8 mutants in PA14-derivative backgrounds. The different magnitude of this effect observed in another P. aeruginosa strain, PAK, may be due to the different level of basal alginate production or on the different basal levels of c-di-GMP in the two strains.

Alginate expression in infected hosts, such as those with cystic fibrosis, is induced as a consequence of mutations in the anti-sigma factor MucA, allowing the AlgU (AlgT) sigma factor unimpeded access to the core RNA polymerase and transcription of alginate biosynthetic genes (Deretic et al., 1994). Once induced, alginate is produced from activated mannuronic acid precursors (in form of GTP-mannuronic acid) polymerization and transfer across the cytoplasmic membrane followed by epimerization of a fraction of mannuronic acids into guluronic acids and irregular O-acetylation of mannuronates in the periplasmic compartment. The final alginate polymer is exported from the periplasm through a channel formed by the outer membrane protein AlgE (Jain and Ohman, 2004). Given the predicted domain organization of Alg44, particularly the presence of the NolF membrane fusion domain associated with various Gram-negative multi-subunit transporters, it is conceivable that it functions in alginate export. The role of the c-di-GMP-binding N-terminal PilZ, domain, however, it is less clear. The predicted topology suggested that Alg44 is a single pass membrane protein, with N-terminal Pil44 domain and a periplasmic NolF domain. The function of Alg44 in alginate production would therefore involve binding of c-di-GMP in the cytoplasm, which would then affect the as yet unknown activity of the NolF domain. However, this model for Alg44 topology has been recently challenged by Remminghorst and Rehm (Remminghorst and Rehm, 2006b). Using an Alg44-PhoA fusion, they showed that Alg44 was periplasmic. Therefore, for c-di-GMP to assert its regulatory role, it would have to be also exported from the cytoplasm where it is synthesized by one of the P. aeruginosa diguanylate cyclases. Our data suggest that Alg44 is not free in the periplasm, and it is found largely associated with membranes. In particular, the observation that during the cell fractionation procedure Alg44-His6 tends to degrade in the membrane into fragments ranging from ∼38 (Fig. 9A) to ∼15 kDa (data not shown), yet remains associated to this cell fraction, may suggest interaction of the C-terminal-tagged portion of these protein fragments (still carrying portions of the NolF membrane fusion domain) with outer or inner membrane components without involvement of the hydrophobic interdomain segment. Whether Alg44 is a single pass transmembrane protein as the topology predictions suggest or whether is merely associated to the membrane is still unclear. Alternatively, Alg44 could be secreted into the periplasm and then remain tightly associated with the cytoplasmic membrane without being released under our experimental conditions. Deletion of the putative transmembrane motif does affect the stability of the protein and Its steady-state accumulation, but it is less clear if the reduction in alginate production is rather the result of altered function. Results from localization experiments suggest that this protein is still located in the membrane fraction implying that the deleted hydrophobic stretch is not essential for targeting Alg44 this cellular compartment.

Interestingly, Alg44 point mutants at the two extremes of the binding affinity range (Alg44[S46A and Alg44[R17AR21A]) still were membrane associated (Fig. 8A lanes 3 and 4), implying that localization of this protein into this compartment is independent of the levels of c-di-GMP in the cell. A more careful analysis of Alg44 topology, perhaps by protease protection assays, is needed before c-di-GMP can be given an extracellular regulatory role.

In addition to alginate biosynthesis in P. aeruginosa, c-di-GMP regulates the production of a number of extracellular polysaccharides, namely cellulose in Gluconacetobacter xylinum, E. coli, Salmonella (Ross et al., 1990; Garcia et al. 2004) and the PEL polysaccharide of P. aeruginosa (V.T. Lee and S. Lory, manuscript submitted). The most likely role for c-di-GMP binding proteins is regulation of polysaccharide biosynthesis. However, the presence of the NolF membrane fusion domain at the Alg44 C-terminus suggests that it may also interact with outer membrane components, perhaps with AlgE, and regulate the export of alginate, which is usually coupled to its synthesis. The localization of Alg44 in membrane fractions is consistent with its role as regulator of the alginate polymerase (Alg8), which is an integral cytoplasmic membrane protein with substantial domains in the cytoplasm (Remminghorst and Rehm, 2006a). The challenge remains to determine the precise function of the various c-di-GMP proteins, such as Alg44, in the production of polysaccharides as well as other macromolecules and their structure-function relationships in relation to other biosynthetic components. For this, complete in vitro reconstruction of the various biosynthetic machineries may be necessary to determine the precise site of activity of these proteins.

Finally, the redundancy of various DGCs and PDEs with precise functions in different cellular responses indicates that the action of c-di-GMP is local, as suggested by experiments in C. crescentus (Paul et al. 2004). This model implies that the c-di-GMP biosynthetic enzymes may colocalize with the c-di-GMP binding proteins, such as those with the PilZ domain, to a specific bacterial compartments in the cytoplasm. Reversal of c-di-GMP action presumable involves hydrolysis of the nucleotide by phosphodiesterases, and this could also occur in a complex with specific c-di-GMP binding proteins.

Experimental procedures

Bacterial strains, plasmids and growth conditions

Bacterial strains and plasmids are described in Table 1. When not otherwise stated, bacteria were grown at 37°C in LB medium (Ausubel et al., 1987). Glycerol was added at 0.5% when indicated. For positive selection in allelic exchange experiments, 6% sucrose was added to LB agar plates. When required, calcofluor (fluorescent brightener 28) was added to the agar plates to 200 μg ml−1. Antibiotics were added at the following concentrations: carbenicillin 50 μg ml−1 (E. coli) or 150 μg ml−1 (P. aeruginosa and E. coli ZK57); gentamicin 15 μg ml−1 (E. coli) or 75 μg ml−1 (P. aeruginosa); chloramphenicol 34 μg ml−1; kanamycin 50 μg ml−1, and for Psudomonas selection media Irgasan® at 25 μg ml−1 was used. When required, IPTG was added at the concentrations described in the figure legends.

Table 1.  Bacterial strains and plasmids.
Strain or plasmidRelevant phenotypes and genotypesSource or reference
Escherichia coli
 TOP10F′F′[lacIq Tn10(tetR)]mcrAΔ(mrr-hsdRMS-mcrBC) ϕ80lacZΔM15 ΔlacX74 deoR nupG recA1 araD139Δ(ara-leu)7697 galU galK rpsL (StrR) endA1λInvitrogen
 BL21(DE3)/pLysSFompT hsdSB (rBmB) gal dcm (DE3) carrying pLysSNovagen
 DH10BFmcrAΔ(mrr-hsdRMS-mcrBC) φ80 dLacZΔM15 ΔlacX74 endA1 recA1 deoRΔ(ara,leu)7697 araD139 galU galK nupG rpsL λInvitrogen
 HB101Fthi-1 hsd20 (rBmB) sup E44 recA13 ara-14 leuB6 proA2 lacY1 rpsL20 (Smr)xyl-thr-1Boyer and Roulland-Dussoix (1969)
 SM10thi-1 leuB6 supE44 tonA21 lacY1 recA::RP4-2-Tc::Mu KmrTaylor et al. (1989)
 ZK57Cel+ clinical isolate from urinary tractRoberto Kolter
 ZK57 adrA::kanCel, adrA::kanrThis study
Pseudomonas aeruginosa
 PA14Wild-type, Alg+Rahme et al. (1995)
 48398PA14 PA2989::Mar2xT7 (GenR)Liberati et al. (2006)
 54152PA14 PA2799::Mar2xT7 (GenR)Liberati et al. (2006)
 48100PA14 PA2960::Mar2xT7 (GenR)Liberati et al. (2006)
 28843PA14 PA4608::Mar2xT7 (GenR)Liberati et al. (2006)
 27409PA14 PA4334::Mar2xT7 (GenR)Liberati et al. (2006)
 54293PA14 PA3353::Mar2xT7 (GenR)Liberati et al. (2006)
 23859PA14 PA3542::Mar2xT7 (GenR)Liberati et al. (2006)
 MAR51PAK mucA::Mar2xT7 (Genr)rsmZ ::lacZ, hypermucoidThis study
 MM6175 PA14 ΔPA0012Constructed with pMM6175This study
 MM6223 PA14 ΔPA2960Constructed with pMM6179This study
 MM6225 PA14 ΔPA3542Constructed with pMM6181This study
 MM6316 PA14 Δalg44[36–420]Constructed with pMM6191This study
 MM6427 PA14 Δalg44[43–351]Constructed with pMM6359This study
 MM6314 PA14 Δalg44[31–351]Constructed with pMM6302This study
 MM6438 PA14 Δalg44[475–534]Constructed with pMM6402This study
 MM6432 PA14 Δalg44[712–1089]Constructed with pMM6358This study
 MM6423 PA14 Δalg44[R21A]Constructed with pMM6340This study
 MM6426 PA14 Δalg44[R17AR21A]Constructed with pMM6342This study
 MM6428 PA14 Δalg44[D44A]Constructed with pMM6344This study
 MM6434 PA14 Δalg44[S46A]Constructed with pMM6346This study
 MM6422 PA14 Δalg8Constructed with pMM6384This study
 MM6267PA14::pMM6254ΔmucA cointegrate, hypermucoidThis study
Plasmids
 pBluescriptSK(+)ColE1 αlacZ (Apr)Stratagene
 pMAR2xT7Mariner transposon suicide delivery vector (Gmr, Apr)Liberati et al. (2006)
 pBTK27pMMB67EH derivative with Ptop operator in front of Ptac r (Apr)Kulasekara et al. (2006)
 pRK2013ColE1 mob+traRK2ΔrepRK2 repE (Kmr)Ditta et al. (1980)
 pEXG2ColE1 mob+sacB+ suicide vector (Gmr)Rietsch et al. (2005)
 pPSV35P. aeruginosa oriV, lacIqmob+ PlacUV5 pUC18 MCS expression vector (Gmr)Rietsch et al. (2005)
 pMMB67EHIncQ, RSF1010, lacIq Ptac expression vector (Apr)Furste et al. (1986)
 pQE30ColE1 N-terminal His6 tag expression vector (Apr)Qiagen
 pVL847ColE1 N-terminal MBP-tag expression vector (Apr)This study
 pPA2200pBTK27 with PA2200 PDE from PAO1 with native rbsKulasekara et al. (2006)
 pPA2133pBTK27 with PA2133 PDE from PAO1 with native rbsKulasekara et al. (2006)
 pPA1107pBTK27 with PA1107 DGC from PAO1 with native rbsKulasekara et al. (2006)
 pPA1120pBTK27 with PA1120 DGC from PAO1 with native rbsKulasekara et al. (2006)
 pVL847-PA0012pVL847 with MBP-PA0012 fusionThis study
 pVL847-PA2799pVL847 with MBP-PA2799 fusionThis study
 pVL847-PA2960 aka pVL847-PilZpVL847 with MBP-PilZ fusionThis study
 pVL847-PA2989pVL847 with MBP-PA2799 fusionThis study
 pVL847-PA3353pVL847 with MBP-PA2989 fusionThis study
 pVL847-PA2799pVL847 MBP-PA3353 fusionThis study
 pVL847-ALG44PILZpVL847 with MBP-PA3542[12–140] fusionThis study
 pVL847-PA4324pVL847 with MBP-PA4324 fusionThis study
 pVL847-PA4608pVL847 with MBP-PA4608 fusionThis study
 pMM6081pQE30 with PA3702 (= wspR) ORF from PA14 (Apr)This study
 pMM6389pQE30 with a His6-alg44[1−420] fusionThis study
 pMM6397 aka pMMB-alg44pMMB67EH with PA3542 (= alg44) ORF from strain -alg44[D44A][1−420] derivative of pMM6389This study
 pMM6518pQE30-His6-PA14 with native rbsThis study
 pMM6517pQE30-His6-alg44[R17AR21A][1−420] derivative of pMM6389This study
 pMM6519pQE30-His6-alg44[S46A][1−420] derivative of pMM6389This study
 pMM6520pQE30-His6-alg44[R21A][1−420] derivative of pMM6389This study
 pMM6540 aka pMMB-alg44his6PA3542 ORF from strain PA14 with native rbs and tagged at 3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6580 aka pMMB-algUpMMB67EH with algU and native rbs from PA14This study
 pMM6590 aka pPSV-algUpPSV35 with algU and native rbs from PA14This study
 pMM6277pBluescriptSK with an ∼1 kb fragment spanning the 3′ half of alg8 and the 5′ half of alg44This study
 pMM6285pMM6277 with alg44[D44A] alleleThis study
 pMM6286pMM6277 with alg44[R21A] alleleThis study
 pMM6287pMM6277 with alg44[S46A] alleleThis study
 pMM6292pMM6277 with alg44[R17AR21A] alleleThis study
 pMM6254pEXG2 with ΔmucA allele constructed by SOE-PCRThis study
 pMM6175pEXG2 with ΔPA0012 allele constructed by SOE-PCRThis study
 pMM6179pEXG2 with ΔPA2960 allele constructed by SOE-PCRThis study
 pMM6181pEXG2 with ΔPA3542 allele constructed by SOE-PCRThis study
 pMM6302pEXG2 with Δalg44[31–351] allele constructed by SOE-PCRThis study
 pMM6305pEXG2 with Δalg44[36–420] allele constructed by SOE-PCRThis study
 pMM6340pEXG2 with XbaI–XpnI insert from pMM6286 alg44[R21A] alleleThis study
 pMM6342pEXG2 with XbaI–XpnI insert from pMM6292 alg44[R17AR21A] alleleThis study
 pMM6344pEXG2 with XbaI–XpnI insert from pMM6285 alg44[D44A] alleleThis study
 pMM6346pEXG2 with XbaI–XpnI insert from pMM6287 alg44[S46A] alleleThis study
 pMM6358pEXG2 with Δalg44[712–1089] allele constructed by SOE-PCRThis study
 pMM6359pEXG2 with Δalg44[43–351] allele constructed by SOE-PCRThis study
 pMM6402pEXG2 with Δalg44[475–534] allele constructed by SOE-PCRThis study
 pMM6384pEXG2 with Δalg8 allele constructed by SOE-PCRThis study
 pMM6982PA3542 ORF from strain MM6423 alg44[R21A] with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6983PA3542 ORF from strain MM6426 alg44[R17AR21A]with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6984PA3542 ORF from strain MM6428 alg44[D44A] with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6985PA3542 ORF from strain MM6434 alg44[D44A] with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6986PA3542 ORF from strain MM6427 Δalg44[36–420] with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6987PA3542 ORF from strain MM6316 Δalg44[31–351] with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6988PA3542 ORF from strain MM6432 Δalg44[712–1089] with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study
 pMM6989PA3542 ORF from strain MM6438 Δalg44[475–534] with native rbs and tagged at-3′ end with 6xHis codons and cloned in pMMB67EHThis study

Construction of plasmids

Detailed procedures to construct the plasmids used in this study for complementation, protein expression and allelic exchange experiments are available upon request. Isolation and manipulation of recombinant DNA molecules used standard procedures (Ausubel et al., 1987) or the various product manufacturers' instructions, and all constructs were sequenced to rule out PCR errors. Primers for PCR and DNA sequencing are described in Table S1 available online as supplemental information. In brief, for expression and purification of the diguanylate cyclase WspR as His6-tagged protein, gene PA3702 from strain PA14 was amplified and cloned into pQE30 at the BamHI site. For construction of plasmids expressing MBP fusions to P. aeruginosa PilZ proteins, we amplified the individual open reading frames (ORFs) by PCR and cloned them into pVL847 for expression from Ptac. For complementation or ectopic induction of the mucoid phenotype alg44 and algU ORFs including their ribosome binding sites were amplified by PCR from PA14 genomic DNA as the template, gel purified, digested with EcoRI and HindIII and ligated into the IncQ plasmid pMMB67EH for expression from the tac promoter to generate plasmid pMM6397 and pMM6580 (Table 1). The algU ORF was also cloned in the expression vector pPSV35 to create plasmid pMM6590. The alg44 ORF was tagged in the 3′ end with six His codons by PCR and cloned into pMMB67EH to created plasmid pMM6540, used in the cell fractionation experiments. The 5′-end portion of alg44 containing the PilZ domain (amino acids 1–140) was amplified by PCR, digested with BamHI and HindIII and ligated into plasmid pQE30 to generate the expression plasmid pMM6389 for the purification of His6-Alg44 (amino acids 1–140) proteins. PCR fragments around 1 kb in size (∼500 bp each side of the deletion) were constructed by the method of gene splicing by overlap extension (Horton et al., 1989) in order to engineer deletions in the mucA PA0012, PA2960, alg44 and alg8 genes. The suicide plasmid pEXG2 was used as the vector (Rietsch et al., 2005). Plasmids pMM6254, pMM6175, pMM6179, pMM6384, pMM6181, pMM6302, pMM6359, pMM6305, pMM6402, pMM6358 carried a SOE-PCR fragment with a full deletion in mucA, PA0012, PA2960, alg8, alg44 and internal deletions nucleotide 31–351 (PilZ domain deletion), nucleotide 43–351 (PilZ domain deletion), nucleotide 36–420 (PilZ domain deletion), nucleotide 474–535 (Alg44 putative transmembrane domain), nucleotide 712–1089 (deletion of the NolF domain) respectively. For allelic exchange of genes with single codon substitutions an ∼1 kb insert spanning part of gene upstream of alg44 (alg8) and about half of alg44 was amplified by PCR and ligated first in pBluescript for site-directed mutagenesis to produce pMM6277. pMMB67EH plasmids carrying alg44 mutant alleles expressed from Ptac as C-terminal His6-tagged proteins were constructed by amplifying the corresponding chromosomal genes created by allele exchange and clone them as HindIII/EcoRI fragments (pMM6982 to pMM6989 plasmids).

Site-directed mutagenesis

Oligonucleotide-directed mutagenesis of plasmid pMM6277and plasmid pMM6389 was performed using the Quick-change protocol (Clontech) and the primer pairs described in Table S1. Mutant clones were screened by PCR followed by digestion at restriction sites engineered with silent mutations next to the substituted codon. Inserts from the pMM6277 derivatives mutagenized by PCR (pMM6292, pMM6286, pMM6285, pMM6287) were subcloned into pEXG2 to produce pMM6342 (R17R21 A), pMM6340 (R21A), pMM6344 (D44A), pMM6346 (S46A).

Construction of mutants by allelic exchange

Mutations engineered in the alg44 and alg8 alleles and carried by the suicide vector pEXG2 were conjugated into P. aeruginosa PA14 by triparental mating (E. coli HB101/pRK2013 was used as the helper strain) selecting for gentamicin resistant recombinants followed by positive selection for the loss of vector sequences following growth on sucrose containing media (Rietsch et al., 2005). Chromosomal mutants were verified by PCR with the outer primers used for the construction of the suicide plasmid inserts followed by direct DNA sequencing of the amplicons.

Insertional mutagenesis

The P. aeruginosa PAK mutant MAR51 with the mucA::MAR2xT7 insertion was isolated following MAR2xT7 mutagenesis as previously described (Liberati et al., 2006) using PAK rsmZ::lacZ as the recipient (Goodman et al., 2004). Transposon insertional mutants were analysed by semi-random PCR followed by DNA sequencing as described previously (Liberati et al., 2006). Construction of the PA14 MucA mutant MM6267 and its sibling MM6268 was carried out by conjugating the suicide vector pEXG2 into P. aeruginosa PA14 by triparental mating (E. coli HB101/pRK2013 was used as the helper strain) selecting for gentamicin resistance and growth on sucrose-containing plates. The E. coli ZK57 was mutagenized at the adrA locus using a lambda-Red one-step gene inactivation protocol using Kanr-tagged PCR products as described by Datsenko and Wanner (Datsenko and Wanner, 2000).

Protein purification

For His-tagged protein expression and purification, E. coli BL21/pACT3 pMM6389 carrying the gene for His6-alg44[1−420] or plasmids expressing the various point mutants (pMM6517, pMM6518, pMM6519, pMM6520), were grown in 1 l of LB until the culture reached an optical density (at 600 nm) between 0.8 and 1.0. Isopropyl-1-thio-beta-d-galactopyranoside (IPTG) was added to a final concentration of 1 mM and the culture was grown at 15°C for 18 h. Cells were harvested by centrifugation and resuspended in 20 ml of ice-cold lysis buffer (2 mM imidazole, 500 mM NaCl, 1 mM PMSF, 50 mM NaHPO4, pH 8.0 at 4°C), containing 500 μl of protease inhibitor cocktail (Sigma). The cell suspension was sonicated while kept on ice with a Brandson Sonic disruptors at 40% duty, 10 s on/30 s off for 15 cycles. The lysate was first centrifuged at 4000 g for 10 min at 4°C in a Sorval SS34 rotor, and then at 45 000 g for 1 h in a Type 70 Ti rotor (Beckman). Column chromatography was performed at 4°C. The supernatant was incubated for 2 h with 1.5 with a prewashed Ni2 ± NTA beads (Qiagen). The column was washed with 10 column-volumes of lysis buffer and 10 column-volumes of washing buffer (50 mM imidazole, 500 mM NaCl, 1 mM PMSF, 50 mM NaHPO4, pH 8.0 at 4°C). The His6-Alg44PilZ proteins were eluted with 250 mM imidazole buffer (250 mM imidazole, 500 mM NaCl, 1 mM PMSF, 50 mM NaHPO4, pH 8.0 at 4°C. Fractions containing higher concentrations of the 15.5-kDa recombinant protein, as determined by SDS-PAGE, were pooled and dialysed overnight against buffer A (5 mM β-mercaptoethanol, βME, 0.5 mM PMSF, 20 mM Tris-HCl, pH 7.5) at 4°C. The preparation was further purified by anion exchange chromatography using a mono-Q sepharose column (Amersham-Pharmacia). His6-Alg44PilZ was collected in the flow-through while the contaminating bands were retained by the column. Salt concentration was increased to 200 mM NaCl and the preparation concentrated by centrifugation in a Vivaspin 15R ultrafiltration tube (Viva Science) with a molecular weight cut-off of 5 kDa. The proteins were stored at −20°C after addition of one volume of storage buffer (20 mM Tris-HCl, pH 7.5, 200 mM NaCl, 5 mM βME, 80% glycerol). Protein concentration was determined by the BCA method (Pierce). The degree of purification was estimated by Coomassie brilliant blue staining after SDS-PAGE (Laemmli, 1970). For purification of MBP-tagged proteins expressed from the pVL847 derivatives, a similar procedure was followed but high flow amylase resin (New England Biolabs) was used following the manufacturer's instructions. The MBP fusion proteins were dialysed and concentrated by ultrafiltration.

Cell fractionation and Western blot analysis

Bacteria were grown overnight at 37°C, diluted 1:100 in 100 ml of LB and grown to OD600−0.2 at 37°C (1 h 30 min). Cultures were induced with 0.25 mM IPTG for 0–1.5 h or as indicated in the legends and centrifuged at 4000 g for 20 min at 25°C. The spheroplasts (cytoplasm + membranes) and the periplasmic fractions were obtained using the Wood's procedure (Wood, 1978) modified as described (Robles-Price et al., 2004). In brief, cells were suspended in 5 ml of a solution containing 0.5 mg ml of lysozyme/, 40 mM Tris-HCl (pH 8.0), 0.5 M sucrose incubated at 30°C water bath. Na+2 EDTA was added at 4 mM for 2 min to increase outer membrane permeability, and MgCl2 was added to the final concentration of 20 mM to complex the excess EDTA and avoid extensive cell lysis. The cell suspension was incubated for 30 min to allow spheroplast formation. The solution was centrifuged at 8000 g for 15 min to pellet the spheroplasts. The supernatant containing the periplasmic proteins was collected, while the pellet was resuspended in 20 mM Tris-HCl, 200 mM NaCl, 0.5 mM EDTA, 1 mM PMSF, sonicated for 15 cycles 40% duty, 10 s on/30 s off cycles. Unbroken cells were removed by low speed centrifugation (4000 g for 15 min), and the membranes were collected by centrifuging the lysate at 100 000 g for 1 h in the Beckman Type 70 Ti rotor. Supernatant (cytoplasmic fraction) and pellet (total membranes) were mixed with 1× SDS-sample buffer. The cell fractions were run on an SDS-PAGE gel (12% monomer), and transferred to a PVDF membrane. Typically, 0.15 OD600 × ml bacteria-equivalent amounts of proteins were loaded per lane. PVDF membranes were blocked overnight with 20 mM Tris-HCl pH 7.5, 150 mM NaCl (TBS) with 3% Non-Fat Dry Milk at 4°C, rinsed in TBST (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.05% Tween-20) and probed with monoclonal antibodies diluted in TBS-1% Non-Fat Dry Milk. Anti-His antibody (H3; Santa Cruz Biotech) was diluted 1:1500, anti-β-lactamase (QED Biosciences) was diluted 1:4000, while monoclonal anti-RNA polymerase σ70 antibodies (Neoclone) was diluted 1:20000. Membranes were washed and probed with goat anti-mouse IgG(H + l) conjugated with HRP (KPL). Supersignal West Pico chemiluminescent substrate (Pierce) was used for detection. For reprobing, blots were stripped with 62.5 mM Tris-HCl pH 6.8, 2% SDS, 100 mM β-mercaptoethanol for 30 min at 55°C followed by several washes in TBS.

Synthesis of [32P]-c-di-GMP

Synthesis of [32P]-c-di-GMP was performed using α-labelled [32P]-GTP (3000 Ci mmol−1, 10 μCi μl−1; Amersham Bioscience) and purified His6-WspR (Kulasekara et al., 2006). Reactions were carried out in a total volume of 100 μl containing 200 μCi α-[32P]-GTP (66 pmol),dephosphorylated with 250 mM NaCl, 25 mM KCl, 10 mM MgCl2, 75 mM Tris-HCl pH 8.0 and 12 μM of enzyme for 1 h. Residual [32P]-GTP was 5 units of calf intestinal phosphatase (New England Biolabs) for 10 min. The proteins were denatured at 95°C for 5 min and precipitated by centrifugation for 5 min at 16 000 g. The supernatant was spun in an Ultrafree-MC column (Millipore) (cut-off of 5000 Da) for 30 min at 5000 g. The [32P]-c-di-GMP was quantified by storage phosphor imaging after thin-layer chromatography on PEI-cellulose plates (Sigma) with 1.5 M KH2PO4 pH 3.65 as the mobile phase.

Organic synthesis of c-di-GMP

Total synthesis of c-di-GMP was carried out at Nagoya University using the procedure described by (Kawai et al., 2003). Purity and stability of lyophilized c-di-GMP were determined by first resuspending the molecule in 0.9% NaCl, creating a 2 mM solution; the results were then confirmed by high performance liquid chromatography (HPLC) analysis and electrospray ionization (ESI)-time of flight (mass spectrometry). A 100 mM stock solution of each nucleotide in 0.9% NaCl was prepared and stored at 4°C until use.

Photolabelling of proteins by UV cross-linking of nucleotides

Proteins (5 μM) were incubated for 15 min at room temperature in binding buffer (200 mM NaCl, 0.5% acetone, 5 mM β-mercaptoethanol as scavenger, 20 mM Tris-HCl pH 7.5) usually containing 100 μM nucleotide (c-di-GMP, ATP, GTP) with 2.5 μCi [32P]-c-di-GMP or [32P]-ATP or [32P]-GTP (3000 Ci mmol−1), respectively (to final specific activity ∼0.2–2 Ci mmol−1), unless otherwise indicated. Samples were irradiated at 254 nm for 5 (or 20) min at room temperature (or in ice) with a UVG-54 Mineral Light lamp (UVP) at a distance of 3 cm (Intensity ∼14 000 μwatt cm−2). After irradiation, samples were treated with 2× SDS-PAGE sample buffer (40% glycerol, 8% SDS, 2% β-mercaptoethanol, 40 mM EDTA, 0.05% Bromophenol Blue, 250 mM Tris-HCl, pH 6.8) and boiled for 5 min. Proteins were separated on by SDS-PAGE (20% gel) and detected by storage phosphor imaging with a Typhoon scanner (Amersham Bioscience) and Coomassie Blue staining.

Protein-[32P]-c-di-GMP filter binding assay

Filter binding assays were used to measure the stability of the protein-ligand complexes, binding specificity and the relative affinity of various Alg44PilZ protein variants for c-di-GMP. For competition experiments, Alg44PilZ : c-di-GMP complexes were allowed to form in binding buffer (200 mM NaCl, 5 mM β-mercaptoethanol, 20 mM Tris-HCl pH 7.5) containing 2.5 μM protein, 25 μM [32P]-c-di-GMP (0.2 Ci mmol−1) and increasing concentrations of unlabelled c-di-GMP, GTP or GMP in the 5 μM to 25 mM range (total volume 20 μl) for 10 min at room temperature. Reactions were filtered through an Immobilon-NC nitrocellulose membrane (Millipore) using a dot-blot apparatus (Bio-Rad). Wells were washed with 1.0 ml of binding buffer, air dried and exposed to a storage phosphor cassette for quantification with a Typhoon Imager (General Electric). Liquid scintillation counting was also performed in preliminary experiments and the two quantification methods produced comparable results. For Koff determinations, complexes formed in presence of 300 μM [32P]-c-di-GMP were separated from unbound nucleotide using a G-25 spin column equilibrated with binding buffer and centrifuged at 1000 g for 1 min. The flow-through was loaded onto the dot-blot apparatus at various time points and the bound radioactivity was measured as above. For the analysis of the relative binding affinity of wild-type Alg44PilZ and its variants, proteins (2.5 μM) were incubated in binding buffer with increasing concentration of c-di-GMP at constant specific activity (0.2 Ci mmol−1) for 10 min then analysed as above. Data were reported as percentage binding compared with the wild-type protein control (%Bwt). Data were fitted with non-linear regression models using GraphPad Prism 3.02.

Isothermal calorimetry

Isothermal titration calorimetry experiments were performed using a VP-ITC microcalorimeter (MicroCal). His6-Alg44[1–140] or MBP-Alg44[12−140] were dialysed against binding buffer (20 mM Tris-HCl, pH 7.5., 200 mM NaCl, 0.5 mM EDTA) and concentration determined by absorbance at 280 nm using a molar extinction coefficient of 5240 (for His-tagged protein; molecular mass 16.7 kDa) or 69 960 (for MBP-tagged protein; molecular mass 58 kDa) M−1 cm−1. The concentration of ligand was established gravimetrically. A solution containing 150 or 300 μM c-di-GMP in binding buffer was used as titrant. The heat of reaction per injection (μcal s−1) was determined by integration of the peak areas using the Origin Version 5.0 software (OriginLab). Heat of binding (H°), the stoichiometry of binding (n), and the dissociation constant (Kd) were calculated from plots of the heat evolved per mol of ligand injected versus the molar ratio of ligand to receptor (Ladbury and Chowdhry, 1996; Ladbury, 2004).

Uronic acid assay

P. aeruginosa was grown in LB plates inoculated with 50 μl of overnight cultures and grown at 37°C for 20 h. Cells were collected, washed twice in 0.9% NaCl or PBS and the alginate in the supernatants was precipitated with one volume of cold isopropanol at −20°C. The pellet was resuspended in 1 ml of water. The solution was treated with DNase I, RNase A for 2 h followed by proteinase K digestion for one h. The uronic acid concentrations were determined by the uronic acid assay (Knutson and Jeanes, 1968). In Brief, 350 μl of sample or its dilutions were added to 3 ml of borate-sulphuric acid reagent (0.1 M K3BO3 in concentrated sulphuric acid), and 0.1 ml of carbazole reagent (0.1% carbazole in absolute ethanol). The mixture was incubated 55°C for 30 min and absorbance at 530 nm was measured. Uronic acid concentrations were determined from a standard curve with alginic acid from brown algae (Sigma).

Acknowledgements

This work was supported by the NIH Grant R37 AI021451 to S.L. and the Cystic Fibrosis Foundation postdoctoral fellowship to V.T.L. We thank Dr Anja Brencic for critical reading of the manuscript.

Ancillary

Advertisement