|Home | About | Journals | Submit | Contact Us | Français|
High cellular concentrations of bis-(3′,5′)-cyclic dimeric guanosine mono-phosphate (c-di-GMP) regulate a diverse range of phenotypes in bacteria including biofilm development. The opportunistic pathogen Pseudomonas aeruginosa produces the PEL polysaccharide to form a biofilm at the air-liquid interface of standing cultures. Among the proteins required for PEL polysaccharide production, PelD has been identified as a membrane-bound c-di-GMP-specific receptor. In this work, we present the x-ray crystal structure of a soluble cytoplasmic region of PelD in its apo and c-di-GMP complexed forms. The structure of PelD reveals an N-terminal GAF domain and a C-terminal degenerate GGDEF domain, the latter of which binds dimeric c-di-GMP at an RXXD motif that normally serves as an allosteric inhibition site for active diguanylate cyclases. Using isothermal titration calorimetry, we demonstrate that PelD binds c-di-GMP with low micromolar affinity and that mutation of residues involved in binding not only decreases the affinity of this interaction but also abrogates PEL-specific phenotypes in vivo. Bioinformatics analysis of the juxtamembrane region of PelD suggests that it contains an α-helical stalk region that connects the soluble region to the transmembrane domains and that similarly to other GAF domain containing proteins, this region likely forms a coiled-coil motif that mediates dimerization. PelD with Alg44 and BcsA of the alginate and cellulose secretion systems, respectively, collectively constitute a group of c-di-GMP receptors that appear to regulate exopolysaccharide assembly at the protein level through activation of their associated glycosyl transferases.
The second messenger bis-(3′,5′)-cyclic dimeric guanosine monophosphate (c-di-GMP)3 is an ubiquitous bacterial signaling molecule that regulates many physiological processes. These processes include but are not limited to exopolysaccharide production, virulence factor expression, cell-cycle progression, and twitching and flagellar motility (1–7). In recent years, a general theme has been identified whereby increased intracellular c-di-GMP concentrations typically regulate the conversion of bacteria from a free-swimming, planktonic form to a sessile, biofilm-embedded community (reviewed in Refs. 8 and 9). The enzymes responsible for the synthesis and degradation of c-di-GMP are referred to as diguanylate cyclases (DGCs) and phosphodiesterases, respectively. These enzymes are typically modular in nature and contain catalytically active domains with highly conserved consensus sequences for which these domains are named. Diguanylate cyclases contain the GGDEF domain, which catalyzes the formation of c-di-GMP through the cyclization of two GTP molecules in a dimerization-dependent mechanism (10–12), whereas phosphodiesterases contain the structurally unrelated EAL or HD-GYP domains that degrade c-di-GMP to either the linear molecule 5′-pGpG or two GMP molecules, respectively (13, 14). Although the mechanistic aspects of c-di-GMP biosynthesis and degradation have been studied in some detail, much less is known about how the molecule exerts its effect on downstream targets. The PilZ domain has been identified as the most widespread c-di-GMP specific receptor (15); however, the exact cellular role of the majority of PilZ-containing proteins studied to date remains undetermined.
One instance where PilZ domain function has been linked to an observable phenotype is in bacterial exopolysaccharide production. The cellulose synthase subunit BcsA from Gluconacetobacter xylinus is essential for cellulose production and contains a C-terminal PilZ domain that binds c-di-GMP (7, 16). In mucoid Pseudomonas aeruginosa strains, production of the exopolysaccharide alginate also requires a PilZ-containing protein, Alg44 (17), and mutations to the c-di-GMP binding residues in this protein abolish alginate production in vivo (1). The identification of these c-di-GMP receptors within polysaccharide secretion systems suggests that other less well characterized polysaccharide loci might also encode c-di-GMP-binding proteins.
In addition to alginate, P. aeruginosa is able to produce two additional exopolysaccharides, PSL and PEL, both of which have been implicated in biofilm formation (18, 19). The PSL polysaccharide, which plays a significant role in P. aeruginosa PAO1 biofilm formation, is comprised of d-mannose, l-rhamnose, and d-glucose (20). The psl operon encodes proteins that are predicted to be functionally similar to those involved in Escherichia coli group 1 capsular and extracellular polysaccharide production (21, 22). Although PSL production is regulated by c-di-GMP at the transcriptional level (23), given that the proteins involved in E. coli group 1 capsular and extracellular polysaccharide production do not require c-di-GMP binding for activation, it is not anticipated that any of the proteins encoded on the psl operon will bind c-di-GMP directly as a means to regulate PSL biosynthesis and export. The PEL polysaccharide is a glucose-rich polymer that is responsible for biofilm formation at the air-liquid interface of standing cultures of P. aeruginosa PA14 (19, 24). The genetic locus responsible for this phenotype has been annotated as the pelABCDEFG operon. Within this operon, the pelD gene product binds c-di-GMP directly and controls PEL polysaccharide production at the protein level (2). PelD is an inner membrane protein with four predicted transmembrane helices and a large cytoplasmic region that contains the c-di-GMP-binding domain. A multiple sequence alignment of PelD orthologs reveals that the protein does not contain a PilZ domain but rather a conserved RXXD motif that resembles the allosteric inhibition site (I-site) found in the GGDEF domains of DGCs. In addition, mutational analysis of this motif has shown that it is required for both c-di-GMP binding in vitro and biofilm production in vivo. These data have led to the proposal that PelD contains a degenerate GGDEF domain that acts as a c-di-GMP receptor with ligand binding occurring at its predicted I-site (2).
In this work, we present the x-ray crystal structure of a cytoplasmic region of PelD, encompassing residues 156–455, in its apo- and c-di-GMP-bound forms. The structure reveals an N-terminal GAF (cGMP-specific phosphodiesterase, adenylyl cyclases, and FhlA) domain and a C-terminal degenerate GGDEF domain that is missing several secondary structure elements normally found in enzymatically active DGCs. In the ligand-bound structure, the GAF domain undergoes a small rotation relative to the GGDEF domain to accommodate c-di-GMP, which is found as a dimer in the I-site of the GGDEF domain. Residues from both the GGDEF domain and the GAF domain interact directly with c-di-GMP; however, only those in the GGDEF domain were found to be essential for ligand binding in vitro. Mutation of these essential c-di-GMP-binding residues in vivo resulted in abolishment of air-liquid pellicle formation that is characteristic of PEL polysaccharide production. Our bioinformatics analysis of the juxtamembrane region of PelD (residues 110–155) suggests that this region of the protein forms a coiled-coil motif and that a dimer may represent the oligomeric state of PelD in vivo. Taken together, our results provide structural data that demonstrates that degenerate GGDEF domains can act as c-di-GMP receptors. Moreover, PelD may act in an analogous manner to proteins in the alginate and cellulose secretion systems, Alg44 and BcsA, respectively, which have been proposed upon c-di-GMP binding, to control the biosynthesis and/or export of their respective polysaccharides.
The cloning, expression, and purification of native PelD156–455 was described previously (25). Because PelD156–455 only contains two methionine residues in 300 total amino acids, an L388M mutant was generated by site-directed mutagenesis (Stratagene) to increase the phasing power obtainable in a selenium single-wavelength anomalous dispersion experiment. Expression of the selenomethionyl-incorporated protein in minimal medium was carried out as per the protocol of Lee et al. (26) using B834 Met− E. coli cells (Novagen) and purified as described for the wild-type protein (25). To probe c-di-GMP binding using ITC, site-directed mutagenesis (Stratagene) of residues Arg161, Arg367, Asp370, and Arg402 was performed. In each case the native residue was substituted with alanine. These single-site mutant constructs were expressed and purified as per the wild-type protein, and their folding was assessed using CD spectroscopy. For c-di-GMP synthesis, the wspR gene was amplified from P. aeruginosa PAO1 genomic DNA using gene-specific primers and cloned into the pET28a expression vector (Novagen). An R242A mutation was introduced into this construct to disable the allosteric inhibition site of the enzyme. WspRR242A expression and purification was carried out as described in De et al. (27).
The protocol for the enzymatic preparation of c-di-GMP was adapted from the protocols of Zähringer et al. (28) and De et al. (27). Briefly, 3.4 μm of purified WspRR242A was incubated overnight at 310 K in a reaction mixture containing 1 mm GTP, 5 mm MgCl2, 50 mm Tris-HCl, pH 7.5, and 50 mm NaCl. The reaction was stopped by heating the sample to 361 K for 5 min. The precipitated protein was removed by syringe filtration prior to adding 5 mm triethylammonium bicarbonate to the reaction mixture. The reaction mixture was loaded onto a 3-ml Resource RPC column (GE Healthcare), and the c-di-GMP was eluted from the column using a linear gradient of 0–50% (v/v) ethanol. Fractions containing c-di-GMP were pooled together, lyophilized, and frozen at 253 K until needed. The identity of the c-di-GMP was confirmed by MALDI-TOF mass spectrometry (Advanced Protein Technology Centre, The Hospital for Sick Children). The solution concentration of c-di-GMP was quantified using the extinction coefficient (ϵ260) of 26,000 m−1 cm−1.
Crystallization of the selenomethionyl-incorporated PelD156–455 was performed as previously reported for the native protein (25). Briefly, the purified protein was concentrated to 10 mg ml−1 and crystallized by the hanging drop vapor diffusion technique. Diffraction quality crystals belonging to the orthorhombic space group P21212 were grown in 10% (w/v) PEG 8000, 0.1 m Tris-HCl, pH 7.5, and 0.2 m MgCl2. Crystals were cryoprotected with the crystallization solution supplemented with 20% (v/v) ethylene glycol prior to flash freezing. X-ray diffraction data were collected at the selenium peak on Beamline X29 at the National Synchrotron Light Source (Brookhaven National Laboratory) and processed with the HKL2000 software program (29). A total of six (of six) selenium sites were located using HKL2MAP (30), and density modified phases were calculated using SOLVE/RESOLVE (31). The resulting solvent flattened selenium single-wavelength anomalous dispersion map was of reasonable quality and allowed for manual model building in COOT (32). The model was then refined against the native data using PHENIX.REFINE (33) to final Rwork/Rfree values of 19.6 and 23.8% (Table 1).
Both soaking c-di-GMP into native crystals and co-crystallization with c-di-GMP in the known PelD156–455 crystallization condition were unsuccessful because soaking caused severe fissuring of the apo crystals, whereas co-crystallization failed to yield any crystals. This was not unanticipated, because the c-di-GMP-binding site was located at a crystal contact in the lattice. Therefore, PelD156–455 in the presence of 2.5 mm c-di-GMP (without any preincubation of the protein and ligand) was rescreened for crystallization conditions using commercially available sparse matrix screens. Diffraction quality crystals of PelD156–455 in complex with c-di-GMP were obtained in 22% (w/v) PEG 8000, 0.1 m sodium cacodylate, pH 6.1, and 0.2 m ammonium sulfate. These crystals displayed the symmetry of the orthorhombic space group P212121. As per the apo form of the protein, crystals of the complex were cryoprotected in crystallization solution supplemented with 20% (v/v) ethylene glycol prior to flash freezing, and data were collected at Beamline X29. The data were processed using HKL2000, and the structure was determined using the PHENIX AutoMR wizard using the native structure as a search model. The final model was refined to Rwork/Rfree values of 20.1 and 25.2%.
All of the structural alignments were performed using the secondary structure matching algorithm in COOT. Structural illustrations were generated in Pymol (Schrödinger). Analysis of protein interfaces was performed using the Protein Interfaces, Surfaces and Assemblies (PDBePISA) server provided by the European Bioinformatics Institute (34). Prediction of the coiled-coil region of PelD was carried out using the MARCOIL server provided by the Max Planck Institute for Developmental Biology (35). Interdomain motions were assessed using the DynDom Protein Domain Motion Analysis server (36).
For the ITC experiments, the PelD protein samples and c-di-GMP were prepared in 20 mm Tris-HCl, 150 mm NaCl, 10% (v/v) glycerol, 1 mm DTT, and each solution was degassed before experimentation. ITC measurements were performed with a VP-ITC microcalorimeter (MicroCal Inc., Northampton, MA). Titrations were carried out with 250 μm c-di-GMP in the syringe and 25 μm solution of the indicated PelD sample in the cuvette (except in the experiment where the cell and syringe contents were reversed). Each titration experiment consisted of twenty-five 10-μl injections with 180-s intervals between each injection. The heats of dilution for titrating c-di-GMP into buffer were subtracted from the sample data prior to analysis. The ITC data were analyzed using the Origin v5.0 software (MicroCal Inc.) and fit using a single-site binding model.
DNA manipulations were performed using standard techniques. For the pelD deletion mutant, ΔpelD, allelic replacement strains were constructed by using an unmarked, nonpolar deletion strategy. Flanking regions of pelD were amplified using primers pelD UP and pelD DN primer sets (supplemental Table S1). The resultant PCR product was ligated into the suicide vector, pEX18Gm, via its HindIII restriction site. The plasmid pEX18GmΔpelD was verified by sequencing analysis. Single recombination mutants were selected on LB containing 100 μg ml−1 gentamicin and 25 μg ml−1 irgasan. Double recombination mutants were selected with LB without NaCl containing 10% (w/v) sucrose and confirmed by PCR. The pelD complementation plasmid, pPelD, was constructed by digesting the PCR product generated from the pelD WT primer set with HindIII and SacI. The PCR product was ligated into pUCP18. The plasmid was verified by sequencing. The same strategy was used for the generation of the truncation mutant plasmid, pPelD156–455, except that the pelD 156 F primer was used instead of pelD WT F. The R161A, R367A, D370A, and R402A point mutants were generated from the pPelD plasmid by site-directed mutagenesis (Stratagene). Each plasmid was verified by sequencing. The primers are listed in supplemental Table S1.
The 96-well microtiter dish assay was preformed as described previously with the following modifications (37, 38). 100 μl of mid-log cells (A600 = ~0.5) grown in LB were added to the wells of a 96-well polypropylene plate (Nunc) and incubated statically for 24 h at 298 K. Following incubation, nonattached cells were removed, and the plate was rinsed thoroughly with water. The plates were stained with 150 μl of 0.1% (w/v) crystal violet for 10 min. The plate was rinsed, and adhered crystal violet was solubilized in 200 μl of 95% (v/v) ethanol for 10 min, and then 100 μl was transferred to a new 96-well plate to measure the absorbance at A595.
Mid-log cells (A600 ~ 0.5) grown in LB broth without NaCl were diluted 1/100 in 3 ml of LB broth without NaCl in a glass tube and left undisturbed at 298 K. The pellicles were monitored by visual inspection and photographed after 48 h. Complete coverage at the air-liquid interface of an opaque layer of cells is considered to be indicative of pellicle formation (19).
1 ml of mid-log cells (A600 ~ 0.5) grown in LB were harvested and resuspended in 250 μl of PBS. A 50-μl sample was mixed with 50 μl of 2× Laemmli buffer and boiled for 5 min. Protein concentration was measured using the Pierce 660-nm protein assay and ionic detergent compatibility reagent as described by the manufacturer (Thermo Scientific). Equal total protein was loaded onto a precast 12.5% Tris-HCl polyacrylamide gel and transferred to a PVDF membrane for immunoblotting (Bio-Rad). The membrane was blocked in 5% nonfat milk in TBST for 1 h at 298 K. The membrane was subsequently probed with an absorbed α-PelD antibody at 1:1,000 dilution in 1% nonfat milk TBST for 1 h (see subsequent section for antibody production protocol). The blots were developed with goat α-rabbit HRP-conjugated secondary antibody (Thermo-Scientific) and a Pierce detection kit.
Purified PelD156–455 protein was used to generate antiserum from rabbits using a 70-day standard protocol (Open Biosystems). Antiserum was absorbed using PA14ΔpelD lysates. The lysates were generated from 100 ml of PA14ΔpelD grown to late log (A600 = ~1.0). The cells were centrifuged and resuspended in 3 ml of lysis buffer (50 mm Tris, pH 8.0, 10 mm EDTA, pH 8.0). The cells were lysed by three freeze/thaw cycles followed by sonication and centrifugation. The cell lysate was subsequently used for absorption by mixing 20 μl of α-PelD antisera, 75 μl of PelD lysate in 1 ml of 5% nonfat milk in TBST. The antiserum was absorbed for 4 h at 298 K under constant rotation.
Sequence analysis of PelD suggests that it contains four transmembrane helices with the fourth transmembrane helix extending into the cytoplasm as a long α-helical stalk region (residues 110–155) (Fig. 1A). This helical juxtamembrane region is predicted to be followed by cytoplasmic GAF and GGDEF domains. For structural studies, multiple constructs were made in an attempt to include as much of the cytoplasmic region as possible. However, constructs that included any part of the juxtamembrane region were found to be either insoluble or misfolded when over expressed in E. coli (data not shown). Ultimately, a construct containing only the tandem-arranged GAF and GGDEF domains (residues 156–455) resulted in soluble protein. PelD156–455 was expressed, purified, and crystallized as described previously (25). Crystallization of the selenomethionine-incorporated protein was performed in a similar manner as the native protein (see “Experimental Procedures”), and the 2.6 Å structure was determined using the single-wavelength anomalous dispersion technique and refined against the 2.1 Å native data. PelD156–455 crystallized in the space group P21212 and contains two molecules in the asymmetric unit. Noncrystallographic symmetry restraints were not used during the course of refinement because of subtle deviations between the monomers that resulted in an increase in Rfree when restraints were implemented. The final model was refined to an Rwork of 19.6% and an Rfree of 23.8% (Table 1).
The overall structure of PelD156–455 shows that the protein, as predicted, contains an N-terminal GAF domain and a C-terminal GGDEF domain (Fig. 1B). A search of the Protein Data Bank using the DALI server (39) indicates that the GAF domain has the highest structural similarity to the GAF domain of a putative two-component response regulator from the cyanobacterium Nostoc sp. PCC 7120 (Protein Data Bank code 3P01, rms deviation of 2.5 Å over 125 equivalent Cα positions) and the GAF-B domain of phosphodiesterase 5A (PDE5A) from Homo sapiens (Protein Data Bank code 3MF0, rms deviation of 2.5 Å over 125 equivalent Cα positions) (40). The ααββαββα topology of the GAF domain of PelD is similar to that of PCC 7120 (supplemental Fig. S1). The difference between the two proteins is a disordered region in PelD156–455 between its α3 helix and β3 strand (residues 247–259) that we were unable to model, which in PCC 7120 protein contains a short β strand followed by a loop region that includes an additional α-helix. Similarly to PCC 7120, the GAF-B domain of PDE5A also contains this additional short β strand, which increases the number of strands in the β sheet core of these two proteins from four to five. Other GAF domains, such as the GAF-A domain of H. sapiens phosphodiesterase 6C (PDE6C), contain another additional short α-helix and β strand after strand β2, giving them a six-stranded β sheet core (41).
Following the GAF domain of PelD is a short linker region (residues 309–317) that connects it to the GGDEF domain. Although this region is disordered in the crystal structure, the two domains still associate with one another through a 653 Å2 interface (assessed using the PISA web server (34)), which is predominantly comprised of the interaction between the α4 helix of the GAF domain and the α1 helix of the GGDEF domain (Fig. 1C). Notably, Asp305 and Arg330, which represent two of the most highly conserved residues among PelD homologs, form a salt bridge with one another at this interface, suggesting that the interaction of the two domains may be critical for biological function and not an artifact of crystallization. Moreover, several hydrophobic residues (Leu165, Ile298, Leu340, Leu369, and Leu390) are solvent-inaccessible as a consequence of this interface. The calculated ΔiG for this interface is −6.3 kcal/mol, suggesting that its formation is energetically favorable (34). The GGDEF domain has a topology of αβαββαβα, which differs from the canonical αβααββαβαβ found in members of this enzymatic fold (supplemental Fig. S2). In particular, PelD lacks both the short α-helix that follows the β1 strand and the final C-terminal β strand that normally comprise GGDEF domains. A search for structural homologs shows that the GGDEF domain of PelD has the closest structural similarity to the degenerate GGDEF domain of P. aeruginosa FimX (Protein Data Bank code 3HVA, rms deviation of 3.5 Å over 118 equivalent Cα positions) (42) and the GGDEF domain of a diguanylate cyclase (Gene locus maqu_2607) from Marinobacter aquaeolei (Protein Data Bank code 3IGN, rms deviation of 3.6 Å over 117 equivalent Cα positions). In functional diguanylate cyclases, the catalytic GGDEF residues required for activity are located between the β2 and β3 strands. However, PelD contains the residues RNDEG at this location, explaining the lack of diguanylate cyclase activity observed for this protein (Fig. 2A) (2). Moreover, when compared with the activated GGDEF domain from Caulobacter crescentus PleD in complex with GTPαS (Protein Data Bank code 2V0N, rms deviation of 4.0 Å over 118 equivalent Cα positions), it becomes clear that the aforementioned secondary structure elements that are absent in PelD contain key residues that are involved in nucleotide binding in active diguanylate cyclases (Fig. 2, B and C). For example, PelD does not contain functionally equivalent residues to Asn335, Asp344, Lys442, and Arg446 of PleD, which are involved in nucleotide binding. The magnesium-coordinating residues in PleD (Glu370 and Glu371 of its GGEEF motif) are conserved in PelD (Asp378 and Glu379); however, in the PelD156–455 structure, they are found ~10 Å away from where they would be needed for substrate binding. A third magnesium-coordinating residue in PleD (Asp327) appears to be both conserved in sequence and location in PelD156–455 (Glu348) and may represent an evolutionary relic. Despite these significant structural differences observed between the catalytically competent and degenerate active sites of PleD and PelD156–455, the conserved RXXD motif of PelD is found between the α2 helix and β2 strand where the allosteric I-site of active diguanylate cyclases is normally located (Fig. 2B, right panel).
Because access to the allosteric I-site is obstructed by crystal contacts in the apo form of PelD156–455, we rescreened for new crystallization conditions for PelD156–455 after mixing the protein with c-di-GMP (for details, see “Experimental Procedures”). This new crystal form, which diffracted to 2.3 Å, exhibited the space group symmetry of P212121 and contained one molecule in the asymmetric unit. After structure determination by molecular replacement, electron density that resembled two mutually intercalated c-di-GMP molecules was clearly observable in the I-site of PelD and allowed for the straightforward placement of two c-di-GMP molecules into the structure (Fig. 3A). The final model of the PelD156–455·c-di-GMP complex was refined to an Rwork of 20.1% and an Rfree of 25.2% (Table 1).
Overall, no gross structural rearrangements were observed between the apo and holo forms of PelD156–455 (overall Cα rms deviation of 1.6 Å). A Cα alignment of the apo and c-di-GMP bound GGDEF domains shows that the GAF domain undergoes a 14° rigid body rotation to facilitate ligand binding (Fig. 3B), which slightly decreases the buried interface between the two domains from 653 to 609 Å2 (because of disruption of the Ser294/His338 interaction). Ser294 and His338 are not conserved among PelD homologs, indicating that this interaction may not be important for function. Moreover, the overall conformational change is relatively small compared with what has been observed for other c-di-GMP receptors such as Vibrio cholerae PlzD (43), which undergoes a 123° interdomain rotation or Pseudomonas fluorescens LapD in which autoinhibition by its signaling helix is relieved by c-di-GMP binding resulting in homodimerization (44). Whether or not full-length PelD utilizes either of these mechanisms remains to be seen. Within the I-site, two c-di-GMP molecules interact with one another through hydrogen bonds between the N1 and N2 atoms of the guanine bases and the oxygen atoms of the phosphate groups in the same manner observed previously for the ligand in the c-di-GMP-bound structures of the GGDEF-containing proteins PleD from C. crescentus (11) and WspR from P. aeruginosa (27) as well as in the small molecule crystal structure of c-di-GMP itself (45). In addition, the interaction between the conserved RXXD motif and c-di-GMP is similar to that seen in the PleD and WspR structures with hydrogen bonds existing between Arg367 and the guanine O6 atom and phosphate group and between Asp370 and the N1 and N2 atoms of the guanine base (Fig. 3C). Arg402, located on the α3 helix of the GGDEF domain, also interacts with c-di-GMP through hydrogen bonds between its Nϵ and Nη1 atoms and the O6 and N7 atoms of guanine. This side-on interaction between the guanidium group and the guanine base differs from the end-on interaction observed for the corresponding arginine residue in the PleD structure (11). WspR on the other hand, contains a lysine residue at this position that does not interact with c-di-GMP (supplemental Fig. S3). Instead, c-di-GMP inhibits WspR by making additional contacts with another WspR molecule resulting in a product-inhibited tetramer. Arg161 is the sole residue from the GAF domain that interacts with c-di-GMP through interactions with the phosphate group and the N7 atom of guanine. This interdomain cross-linking by c-di-GMP observed between the GAF and GGDEF domains of the PelD156–455 structure is similar to what is seen in the PleD structure where c-di-GMP interacts with both its GGDEF and Rec-like adapter domain. Although domain immobilization through I-site binding serves to prevent the association of the GGDEF active half-sites in PleD, the significance of the interdomain cross-linking observed in PelD156–455 is not immediately obvious given the degeneracy of its GGDEF domain.
To determine the relative contribution of each of the c-di-GMP binding residues to ligand binding, isothermal titration calorimetry was performed on wild-type PelD156–455, as well as on R161A, D367A, R370A, and R402A site-directed mutants (Fig. 4). In agreement with previous characterization, the wild-type protein bound c-di-GMP with moderate affinity (Kd of 1.9 μm) (2). In comparison, the R161A mutant showed a relatively modest but significant reduction in binding affinity (Kd of 4.9 μm). In both of these experiments, a consistently anomalous ligand:protein stoichiometry of ~1.4:1 was observed. However, based on the crystallographic data presented in this work and previous solution studies that have shown, c-di-GMP exists as a dimer in solution at the concentrations used in the ITC experiments (46), the concentration of c-di-GMP was adjusted to give a stoichiometry of 2:1. After making this concentration adjustment, the reverse experiment was performed (i.e., titrating PelD into c-di-GMP) to ensure that only a single binding event was occurring. In this scenario, a protein-ligand binding stoichiometry of 0.5:1 was obtained, suggesting that a single binding event is occurring. It is likely that the anomalous stoichiometry observed in these experiments arises from a subpopulation of misfolded/inactive PelD (given the consistency of this anomalous stoichiometry), although errors in protein or ligand concentration may also be a factor. In agreement with previous surface plasmon resonance data (2), the R367A, D370A, and R402A mutants reduced the ligand binding affinity to below the detection limit of the calorimeter, demonstrating that most of the binding specificity for c-di-GMP resides in the GGDEF domain of PelD.
To examine the effect of c-di-GMP binding in vivo, an unmarked, nonpolar pelD deletion mutant was generated in P. aeruginosa PA14 and used for complementation studies. As expected, this mutant was deficient in both biofilm formation in a microtiter dish, as well as pellicle formation in standing cultures (Fig. 5). When this mutant was complemented with pelD on a plasmid, wild-type levels of biofilm and pellicle were restored. However, when the pelD deletion mutant was complemented with the pelD gene containing the R367A, D370A, or R402A point mutants, biofilm and pellicle formation were reduced to that of the noncomplemented strain. Western blot analysis clearly shows that the lack of complementation is not due to a loss of protein expression because comparable amounts of PelD are expressed in all cases (Fig. 5B, bottom panel). These results correlate well with the ITC data and suggest that c-di-GMP binding is essential for PEL polysaccharide formation. Interestingly, complementation of the pelD deletion mutant with the pelD gene containing the R161A point mutant resulted in a partial recovery of biofilm formation. Moreover, pellicle formation was clearly observed in standing cultures, although some cells remained in a planktonic state. This intermediate phenotype, when taken together with the ITC data, suggests that Arg161 may act as a fine-tuning mechanism in that it is required for both the highest affinity binding of c-di-GMP to PelD in vitro as well as maximum biofilm production in vivo; however, when mutated, it does not severely hinder binding or completely abolish biofilm/pellicle formation. The complete conservation of Arg161 across all PelD homologs lends further support to the functional importance of this residue. Lastly, complementation was performed with a construct containing the same boundaries as the crystallized fragment of PelD, PelD156–455. Perhaps not unsurprisingly, this plasmid was unable to restore biofilm or pellicle formation, confirming that the GAF-GGDEF-containing fragment of PelD is insufficient for biological activity, and the juxtramembrane region and transmembrane domains of PelD are essential for its function.
Recent structural and functional characterization of c-di-GMP-binding proteins has expanded the repertoire of protein domains known to bind the cyclic dinucleotide. Examples of these novel c-di-GMP-binding domains include a noncanonical receiver domain from V. cholerae VpsT (47), the AAA σ54 interaction domain from P. aeruginosa FleQ (48), and degenerate EAL domains from P. aeruginosa FimX and P. fluorescens LapD (42, 44). Composite GGDEF-EAL-containing proteins have been shown to contain degenerate domains that regulate the activity of their associated catalytically active domain. For example, CC3396 from C. crescentus contains a catalytically competent EAL domain whose activity is stimulated when its associated degenerate GGDEF domain binds GTP at its nonfunctional active site (49). In the present work, we provide the first structural data confirming that degenerate GGDEF domains can also function as c-di-GMP receptors. However, unlike the degenerate EAL domain-containing receptors FimX and LapD, which bind the monomeric dinucleotide at their degenerate active sites, PelD utilizes a conserved allosteric I-site commonly found in active DGCs to bind dimeric self-intercalated c-di-GMP. This mode of binding has also been observed in the product-inhibited structures of the DGCs PleD and WspR. It has been proposed that c-di-GMP dimerization and I-site binding sets the upper concentration limit of c-di-GMP in the cell by preventing further biosynthesis (11, 27). Lending further support to this theory is the observation that the inhibition constants (Ki) for c-di-GMP binding to the I-site of both PleD and another active DGC, C. crescentus DgcA, are both in the range of ~1 μm (50), which is very similar to the measured c-di-GMP dissociation constant for PelD (1–2 μm). Given that c-di-GMP binding to PelD is essential for PEL polysaccharide production, it appears that dimeric c-di-GMP contains the necessary molecular determinants to elicit production of a key matrix component of P. aeruginosa PA14 biofilms (24).
A comparison of the apo and holo forms of PelD156–455 reveals only minor structural rearrangements upon ligand binding. Moreover, PelD156–455 elutes as a monomer both in the absence and presence of c-di-GMP on an analytical size exclusion column (supplemental Fig. S4). These observations prompted us to utilize a bioinformatics approach to examine regions of the protein where overexpression and purification had proven intractable as a means to propose how c-di-GMP binding to PelD stimulates PEL polysaccharide production. Analysis of the juxtamembrane region using the coiled-coil prediction program MARCOIL (35) suggests that residues 129–153 (≥98% confidence level) form a coiled-coil motif, implying that the oligomeric state of PelD in vivo may be dimeric (supplemental Fig. S5A). The boundaries of the coiled-coil prediction lie just upstream of the α1 helix from the GAF domain, indicating that the coiled-coil may simply be an extension of this N-terminal helix (Fig. 6). Perhaps the most convincing line of evidence for this hypothesis comes from the examination of several other crystal structures of GAF domains that show they are often associated with α-helical stalk domains (including coiled-coil motifs). The structure of a P. aeruginosa GAF domain of unknown function (PA5279; Protein Data Bank code 3E98) solved by the Joint Center for Structural Genomics contains an N-terminal coiled-coil that mediates dimerization of the protein. In addition, phosphodiesterase 2A from H. sapiens contains tandem GAF domains, both of which contain α-helical stalk regions that form the homodimerization interface (51). It is also of interest to note that the putative coiled-coil motif of PelD is predicted to form a single continuous α-helix with the fourth transmembrane domain, suggesting that perhaps the homodimerization interface may extend into the inner membrane (supplemental Fig. S5B). In addition to functioning as protein-protein interaction modules, it is pertinent to note that GAF domains are also known to function as cyclic mononucleotide-binding domains (typically cGMP or cAMP). However, previously published biochemical experiments suggest that PelD does not bind cyclic mononucleotides (2).
To date, there has been very little characterization of any of the proteins encoded on the pel operon, and thus it is difficult to speculate how c-di-GMP binding to the cytoplasmic GGDEF domain of PelD stimulates the polymerization and export of PEL polysaccharide. A recent bioinformatics analysis of the seven Pel proteins predicts that the only protein that localizes exclusively to the cytoplasm is the putative family 4 glycosyl transferase, PelF (21). Because it is the only glycosyl transferase in the pel operon, it has been suggested that PelF is the PEL polymerase protein that assembles the polysaccharide from an as yet unidentified sugar nucleotide precursor. Therefore, it may be that c-di-GMP binding to PelD regulates the PEL polymerase activity of PelF through a protein-protein interaction in an analogous manner as to what has been proposed for the alginate and cellulose secretion systems (Fig. 7) (1, 52, 53). Alginate polymerization is carried out by the inner membrane protein Alg8, which contains four transmembrane domains and a cytoplasmic family 2 glycosyl transferase domain (52, 54). The ability of Alg8 to synthesize alginate is dependent on the c-di-GMP binding activity of the PilZ-containing protein Alg44, and thus it has been proposed that Alg44 acts in concert as a co-polymerase with Alg8 to not only to synthesize alginate but also facilitate its transport across the inner membrane. In the bacterial cellulose export apparatus, BcsA is a large multitopic inner membrane protein that contains both a family 2 glycosyl transferase domain and a c-di-GMP binding PilZ domain in the cytoplasm (7). Although PelD contains a degenerate GGDEF domain instead of a PilZ domain, it may play a functionally equivalent role in PEL polymerization and export. Characterization of the other Pel proteins as well as examination of potential protein-protein interactions will provide further insight into the mechanism of PEL synthesis in P. aeruginosa and will determine whether a unifying theme of synthase driven polysaccharide secretion can be established among these systems. Moreover, c-di-GMP signaling and the protein receptors thereof are an attractive target for antibiotic development given their exclusivity among bacteria.
We thank Maria Amaya, Marko Erak, and Patrick Yip for technical assistance.
*This work was supported in part by Canadian Institutes of Health Research Grant MT43998 (to P. L. H.). This work was also supported by National Institutes of Health Grant R01 AI077628-01A1 (to M. R. P.) and National Science Foundation Grant MCB0822405 (to M. R. P.). Beamline X29 at the National Synchrotron Light Source was supported by the U.S. Department of Energy and the National Institutes of Health National Center for Research Resources.
This article contains supplemental references, Table S1, and Figs. S1–S5.
The atomic coordinates and structure factors (codes 4DMZ and 4DN0) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).
3The abbreviations used are: