Crystal structure of a catalytically active GG(D/E)EF diguanylate cyclase domain from Marinobacter aquaeolei with bound c-di-GMP product

Abstract

Recent studies of signal transduction in bacteria have revealed a unique second messenger, bis-(3′-5′)-cyclic dimeric GMP (c-di-GMP), which regulates transitions between motile states and sessile states, such as biofilms. C-di-GMP is synthesized from two GTP molecules by diguanylate cyclases (DGC). The catalytic activity of DGCs depends on a conserved GG(D/E)EF domain, usually part of a larger multi-domain protein organization. The domains other than the GG(D/E)EF domain often control DGC activation. This paper presents the 1.83 Å crystal structure of an isolated catalytically competent GG(D/E)EF domain from the A1U3W3_MARAV protein from Marinobacter aquaeolei. Co-crystallization with GTP resulted in enzymatic synthesis of c-di-GMP. Comparison with previously solved DGC structures shows a similar orientation of c-di-GMP bound to an allosteric regulatory site mediating feedback inhibition of the enzyme. Biosynthesis of c-di-GMP in the crystallization reaction establishes that the enzymatic activity of this DGC domain does not require interaction with regulatory domains.

Introduction

Bis-(3′-5′)-cyclic dimeric guanosine monophosphate (c-di-GMP) is a key signaling molecule in bacteria responsible for regulating cell surface-associated traits and community behavior, such as secretion, cell adhesion, motility, and biofilm formation [1]. Cellular c-di-GMP levels are tightly regulated by the opposing activities of diguanylate cyclases (DGC), which synthesize c-di-GMP from two guanosine triphosphate (GTP) molecules, and phosphodiesterases (PDE), which degrade c-di-GMP [1]. DGC and PDE domains are the most abundant domains encoded in many eubacterial genomes, and they are often encoded together in a single polypeptide with other sensory or regulatory domains. The signature sequence motifs EAL and HD-GYP were identified in the active sites of PDEs that degrade c-di-GMP into pGpG or two GMP molecules, respectively [2, 3]. DGC activity was found to be dependent on a conserved GG(D/E)EF motif in the active site [4, 5]. The consensus sequence RxxD in DGC domains, called the primary inhibitory site (I_p-site), was shown to be important for allosteric regulation of cyclase activity through feedback inhibition [6]. DGC domains lacking the GG(D/E)EF motif are catalytically inactive but function as receptors for c-di-GMP through their I-sites [1]. Based on this observation, catalytically active DGC domains are often referred to as GG(D/E)EF domains.

Crystal structures have been determined for c-di-GMP-bound DGC domains in the PleD protein from Caulobacter crescentus [7, 8], the WspR protein from Pseudomonas aeruginosa [9, 10], and the XCC4471 protein from Xanthomonas campestris [11]. These structures have provided valuable insight into the catalytic mechanism and regulation of DGCs (as reviewed by Kim et al. [12]). Twofold symmetry in the active site of PleD, including in the bound c-di-GMP, suggested that a symmetrical homodimer of GG(D/E)EF domains is required for catalytic formation of 3′-5′ intermolecular phosphodiester bonds [7]. Therefore, it was hypothesized that DGC domains are controlled by allosteric interactions modulating their mutual interaction geometry and that formation of a symmetrical homodimer is required for catalytic activity [12]. Consistent with this hypothesis, Monomers of PleD [4] and WspR [9] are catalytically inactive. The stereochemical model for PleD activation proposed that proper antiparallel dimerization of two GG(D/E)EF domains is controlled by the interaction of its two N-terminal Receiver (REC) domains, one of which is activated by phosphorylation [8]. In contrast, WspR contains a single N-terminal REC domain joined to its C-terminal DGC domain via a long α-helical stalk. Functional dimerization of its DGC domain requires tetramerization of the full-length protein [9].

Feedback inhibition of DGCs by the c-di-GMP product similarly is believed to be mediated by allosterically controlled changes in interdomain interaction geometry. Crystal structures of feedback-inhibited PleD revealed two base-stacked dimers of c-di-GMP bound at two equivalent binding-sites bridging the interface of a dimer of DGC domains with proper 2-fold rotational symmetry [7, 8]. The bound c-di-GMP ligands contact the I_p-site in one DGC domain and a different location on the surface of the symmetry-related domain; this second contact site, centered on residue R313 in PleD from Caulobacter crescentus, was called the secondary inhibitory site or I_s-site [8]. The geometry of the resulting DGC domain dimer prevents interaction of two GG(D/E)EF active sites with one another, thus maintaining them in a catalytically inactive state when c-di-GMP is bound to the interfacial I-sites. The crystal-structures of feedback-inhibited PleD [7, 8] showed that, in addition to the RxxD motif (residues 359–362 in C. crescentus), an additional arginine residues makes a critical contributions to the I_p-site (R359). Crystal structures of feedback-inhibited DGC domains from WspR showed an equivalent dimeric structure with two base-stacked c-di-GMP dimers bound at interfacial I-sites with very similar stereochemical features [9, 10].

The crystal structure of the DGC domain from protein XCC4471 showed product binding in a novel geometry at the active site rather than at the I-sites. While this complex could represent an alternative mode of feedback inhibition of this DGC enzyme, it could also represent an artifact of its crystallization at pH 9.5 [11].

Here we present the crystal structure of the C-terminal DGC domain from protein A1U3W3 from Marinobacter aquaeolei, a Gram-negative, non-spore forming, water-dwelling bacterium. Amino acid sequence analysis of the full-length A1U3W3 protein (UniProt/TrEMBL ID, A1U3W3_MARAV) suggests that it also contains an N-terminal PAS_4 domain (PF08448, residues 17–139). The crystallized DGC domain, which we call G-A1U3W3, spans residues 155–312 in this 316-residue protein [13]. Sequence alignment with the DGC domains of PleD and WspR reveals that the active site GG(D/E)EF motif and inhibitory site RxxD moitf are conserved in the G-A1U3W3 domain (Fig. 1a). As part of a PSI program aimed at providing 3D structures of domain families with incomplete structural coverage, this domain was expressed and purified as part of the Northeast Structural Genomics Consortium (NESG) pipeline (target MqR89A). Surprisingly, co-crystallization of G-A1U3W3 with GTP and subsequent structure solution revealed the presence of a c-di-GMP dimer, the product of DGC catalysis, bound in the I-site at the interface of a domain dimer with very similar geometry to that in the feedback-inhibited structures of PleD and WspR. Therefore, the G-A1U3W3 structure demonstrates for the first time that an isolated DGC domain can possesses full enzymatic activity and synthesize c-di-GMP.

Fig. 1.

Crystal structure of the diguanylate cyclase (DGC) domain from M. aquaeolei protein A1U3W3. a. Sequence alignment of the DGC domains of A1U3W3, PleD from C. crescentus, and WspR from P. aeruginosa by ClustalW [28]. Boxed residues coloured white are invariant in all three sequences, while those colored red are conservatively substituted. The invariant residues involved in GTP/Mg²⁺ or GTP binding in PleD [7] are marked by green or blue triangles, respectively. The conserved RxxD sequence in the I_p-site is marked by red triangles, and the signature GG(D/E)EF motif in the active site is marked by stars. The third arginine in the I_p-site of PleD and A1U3W3 is marked by a magenta triangle. The secondary structure elements found in the crystal structure of G-A1U3W3 are shown above the alignment. Dots above the alignment represent every tenth residue in the G-A1U3W3 domain. This image was generated by ESPript [29]. b. Two views of the crystal structure of the G-A1U3W3 protomer. The β-strands in the protein are colored yellow, the α-helices are colored red, and the loops are colored green. The GG(D/E)EF motif is coloured blue. The bound c-di-GMP dimer is shown in stick conformation colored according to atom type (carbon in cyan, nitrogen in blue, oxygen in red, and phosphorus in orange. c. Least-squares superposition of the DGC domains of A1U3W3 (green and blue), PleD (purple and pink) [7], and WspR (yellow and cyan) [9]. d. Electrostatic surface potential [26] of the G-A1U3W3 protomer oriented as in panel b, with fully saturated red and blue colours representing respectively negative and positive potentials of ±5 kT at an ionic strength of 0.2 M. e. Stereo pair of the intercalated c-di-GMP dimer bound to the I-site of G-A1U3W3. Conserved residues R225, D228, and R256 that ligate c-di-GMP in the I_p-site are shown in green. Residues R182 that ligates c-di-GMP in the I_s-site is shown in magenta. A third G-A1U3W3 protomer that makes crystal-packing interactions at this site is shown in cyan.

Material and Methods

Cloning, expression, and purification of G-A1U3W3

The DGC domain of A1U3W3 (G-A1U3W3) from M. aquaeolei was cloned, expressed and purified following standard NESG protocols [14]. In brief, the GGDEF domain (residues 149–316) of the Maqu_2607 gene (UniProt/TrEMBL ID, A1U3W3_MARAV; NESG ID, MqR89A) was amplified from genomic DNA by PCR and cloned between the NdeI and XhoI sites of the pET21_NESG vector (Novagen) in-frame between a C-terminal affinity tag (LEHHHHHH) and an initiating methionine. The resulting plasmid (MqR89A-21.2) coding for a 177-residue recombinant protein was transformed into codon-enhanced Escherichia coli BL21(DE3) pMGK cells. A 1 L culture was grown at 37 °C in MJ9 minimal media [15] supplemented with selenomethionine, lysine, phenylalanine, threonine, isoleucine, leucine, and valine for the production of selenomethionine-labeled proteins [16]. When the OD₆₀₀ reached 0.6, protein expression was induced with 1.0 mM isopropyl-1-thio-β-D-galactopyranoside (IPTG), and the temperature was lowered to 17 °C. Following overnight incubation, the cells were harvested by centrifugation and stored at −80 °C.

Frozen, induced cells were harvested and resuspended in 500 mM NaCl, 40 mM imidazole, 1 mM TCEP, 0.02% (w/v) NaN₃, 50 mM Tris pH 7.5. Following lysis by sonication, the supernatant was loaded onto an ÄKTAxpress system (GE Healthcare), and the protein was purified using a two-step protocol consisting of IMAC (HisTRAP HP) chromatography followed by gel-filtration (HiLoad 26/60 Superdex 75) chromatography in 100 mM NaCl, 5 mM DTT, 0.02% (w/v) NaN3, 10 mM Tris-HCl pH 7.5. Protein-containing fractions were pooled and concentrated to 7.5 mg/mL. Protein purity and molecular mass were evaluated using SDS-PAGE and MALDI-TOF mass spectrometry (20448.6 Da observed vs. 20256.5 Da expected for the selenomethionine-labeled protein). The final yield was 46.4 mg per liter of culture. The pET expression vector (NESG MqR89A-21.2), has been deposited in the PSI Materials Repository (http://psimr.asu.edu/).

Crystallization and data collection

Initial crystallization screening of G-A1U3W3 was performed in 1:1 sitting-drop vapour-diffusion reactions in the presence of 3 mM GTP at 4 and 18 °C. Small clusters of diamond-shaped crystals appeared after 2 months in a single condition at 4 °C. The precipitant contained 3 mM GTP, 100 mM magnesium acetate, 30% (w/v) MPD, 100 mM Na-cacodylate, pH 7.0. The crystals used for data collection grew in hanging-drop reactions and reached a maximum length of 100 μM after 5 weeks. The crystals were cryoprotected with paratone oil prior to flash-freezing in liquid propane for data collection. Single-wavelength anomalous diffraction (SAD) data were collected at the peak x-ray absorption wavelength of selenium on beamline X4C at the National Synchrotron Light Source at Brookhaven National Laboratory (Table 1). Data from a single crystal of G-A1U3W3 maintained at 100 K were processed with HKL2000 [17].

Table 1.

X-ray data and refinement statistics^a

	A1U3W3 CTD
Crystal Parameters
Space group	P 43 21 2
Cell dimensions:
a, b, c (Å)	59.2, 59.2, 118.9
α, β, γ (°)	90, 90, 90
Matthews coefficient (Å3/Da)	2.6
Solvent Content (%)	52.3
Data Collectionb
Wavelength (Å)	0.97900
Resolution (Å)	33-1.83 (1.86-1.83)
Rsym (%)	3.3 (43.9)
No. of unique reflections	35,667
No. of reflections in Rfree set	1,810
Mean redundancy	15.4 (14.2)
Overall completeness (%)	99.8 (100)
Mean I/σ	78.6 (6.0)
Refinement Residualsc
Rfree (%)	24.1 (25.3)
Rwork (%)	19.5 (19.2)
Completeness (%)	99.6 (96.5)
Model Qualityd
RMSD bond lengths (Å)	0.006
RMSD bond angles (°)	1.2
MolProbity Ramachandran distribution
Most favored (%)	98.8
Allowed (%)	1.2
Disallowed (%)	0.0
Mean main chain B-factor (Å2)	31.7
Mean overall B-factor (Å2)	35.5
Mean solvent B-factor (Å2)	48.5
Model Contents
Protomers in ASU	1
Protein residues	A152-317 e
Ligands	2 c-di-GMP
No. of protein atoms	1302
No. of ligand atoms	92
No. of water molecules	168
PDB accession code	3IGN

^aStandard definitions were used for all parameters [27]. Entries in parentheses report data from the limiting resolution shell. Data collection and refinement statistics come from SCALEPACK [17] and PHENIX [21], respectively. The abbreviations RMSD and ASU stand for root-mean-square deviation and asymmetric unit, respectively.

^bAll observations with I ≥ −3σ_I were included in calculating data-quality statistics.

^cReflections with f ≥ 1.34 σ_f were included in calculating R-factors.

^dCalculated using PSVS 1.4 program [23].

^eResidue 317 is a leucine from the C-terminal affinity tag, while three N-terminal residues from the expressed domain are missing.

Structure Solution and Refinement

SnB [18] was used to locate 5 of the 6 selenium sites in the asymmetric unit (ASU) and to calculate phases to 2.1 Å resolution. Solvent-flattening and partial model building by RESOLVE [19] led to a model in which 72% of the residues had properly placed side chains. The model was completed using iterative cycles of manual rebuilding in COOT [20] and computational refinement at 1.83 Å in PHENIX [21] (Table 1). The geometrical parameters for c-di-GMP were generated using the eLBOW module from the PHENIX suite [21] using the c-di-GMP model from PDB id 2RDE [22].

Structure validation and deposit

The quality of the final structure was assessed using PSVS [23] and PROCHECK [24]. The atomic coordinates and structure factors are available in the Protein Data Bank under accession code 3IGN.

Results and Discussion

The crystal structure of G-A1U3W3 (Table 1; Fig. 1) was determined at 1.83 Å resolution using the selenomethionyl SAD method [25]. Electron density was interpretable for 166 of 177 residues in the protomer, corresponding to residues 152–316 from full-length A1U3W3 plus a leucine residue from the C-terminal affinity tag. The G-A1U3W3 domain has a canonical DGC fold. A central, mostly antiparallel five-stranded β-sheet with topology β2-β3-β1-β4-β5 is surrounded by five α-helices (Fig. 1b). Least-squares superpositions of the backbone Cα atoms of G-A1U3W3 with the DGC domains of PleD [7] or WspR [9] show root-mean-square-deviations (RMSD) of 1.6 or 1.7 3, respectively (Fig. 1c). As in PleD and WspR, the characteristic GG(D/E)EF motif in G-A1U3W3 is located on a tight turn connecting strands β2 and β3 (Fig. 1c).

Analytical gel filtration chromatography, monitored by in-line static light-scattering, refractive index, and UV detectors showed that the G-A1U3W3 domain is a monomer in the crystallization stock solution (Suppl. Fig. S1). The crystal structure of G-A1U3W3 contains one protein molecule in the asymmetric unit (ASU) of the lattice (Fig. 2). The GG(D/E)EF catalytic motif at residues 234–238 (GGEEF in A1U3W3) does not have a bound ligand and is located in an asymmetrical environment in the lattice. However, the RxxD I_p-site motif at residues 225–228 interacts with a dimer of c-di-GMP that bridges a 2-fold packing interface with an adjacent protomer (Fig. 2). The resulting dimer with proper 2-fold symmetry seems likely to represent the physiological c-di-GMP-inhibited form of the domain, as discussed below.

Fig. 2.

Crystallographic dimer of the DGC domain from M. aquaeolei protein A1U3W3. a. The packing is shown in one unit cell in the crystal structure of the G-A1U3W3 domain (PDB id 3IGN). Three protomers interacting with the same c-di-GMP dimer are colored green, magenta, and cyan; this c-di-GMP dimer is colored according to atom type (with carbon in grey, nitrogen in blue, oxygen in red, and phosphorus in orange). It interacts with I_p-site in the green protomer, the I_s-site in the magenta protomer, and residue K304 in the cyan protomer. The remaining symmetry-related protomers and c-di-GMP dimers are coloured yellow and orange, respectively. A close-up view of the boxed area is shown in Fig. 1e. b. Crystallographic dimer bridged by the bound c-di-GMP molecules. The green and magenta protomers from panel a are shown with their signature GG(D/E)EF motifs coloured blue and purple, respectively. Based on the arguments presented in the text, this dimer probably represents the feedback-inhibited conformation the G-A1U3W3 domain. c,d. Least-squares superposition of the c-di-GMP-bound crystallographic dimer of G-A1U3W3 (magenta and green) on crystallographically observed dimers of DGC domains from PleD (purple and slate in panel c, from PDB id 2V0N [7, 8]) or WspR (yellow and cyan in panel d, from PDB id 3I5C [9, 10]).

The GTP-interacting and putative catalytic residues observed in previous DGC domain structures [7] are conserved in A1U3W3 (Fig. 1a), supporting its function as an enzymatically active DGC domain. Analysis of the electrostatic surface potential [26] of the G-A1U3W3 protomer (Fig. 1b) reveals a negatively charged pocket at the conserved GG(D/E)EF active site (Fig. 1d). Many of the conserved residues in a sequence alignment of the DGCs of known structure (Fig. 1a) map to the surface of the protein. These residues could mediate functionally important interprotein interactions, either with a second DGC domain to form a catalytically active dimer or with some common regulatory domain.

While G-A1U3W3 was crystallized in the presence of GTP, two molecules of c-di-GMP are observed in the crystal structure (Fig. 1b, e). Their four guanyl bases intercalate to form a c-di-GMP dimer that is bound between the primary inhibitory site (I_p-site) in one protomer and the secondary inhibitory site (I_s-site) in a second protomer (Fig. 2). This interface between protein molecules related by proper 2-fold rotational symmetry buries 803 Ų of solvent-accessible surface area per protomer. The conformation of the c-di-GMP dimer in G-A1U3W3 and the location of its I_p-site relative to the active-site GG(D/E)EF motif are similar to those observed in the c-di-GMP-bound structures of PleD and WspR. A similar constellation of residues surrounding the c-di-GMP dimer is observed for the I_p-site (R225, D228, and R256) and I_s-site (R182) of a second monomer (Fig. 1e). However, in G-A1U3W3 the c-di-GMP dimer also interacts with residue K304 in a third neighbouring domain (Fig. 1e), what was not observed in the c-di-GMP-bound structures of PleD and WspR.

The appearance of the c-di-GMP product in our structure was unexpected. A recent study of WspR showed that the isolated DGC domain had no detectable enzymatic activity, neither in a cell-based assay nor in vitro [9]. However, in our experiments, co-crystallization of the isolated DGC domain of A1U3W3 with GTP resulted in formation of the c-di-GMP product. Therefore, the G-A1U3W3 domain slowly catalyzed the synthesis of c-di-GMP during several weeks of crystal growth.

The crystal structures of PleD and WspR both support the hypothesis that that cyclase activity requires homodimerization of DGC domains, to form an active site with two-fold symmetry matching that within the c-di-GMP product [7, 9]. Formation of the catalytically active DGC-domain dimer is believed to be controlled by steric interactions with the N-terminal REC domain in each of these proteins, when the REC domain is in its activated conformational state [12]. The synthesis of c-di-GMP by the isolated DGC domain of protein A1U3W3 suggests that the combination of low water activity and high enzyme concentration during crystallization of this domain drove transient formation of the catalytically active dimer structure. The enzyme concentration in the crystallization reaction was 3.75 mg/ml (190 μM) immediately after 1:1 dilution with precipitant, already much higher than the concentration used in a typical in vitro enzyme assay, and this concentration probably approximately doubled during the crystal-growth period (i.e., due to vapor diffusion between the crystallization reaction and the precipitant solution in the well). Notably, non-activated PleD was observed to have such a tendency to dimerize at elevated protein concentrations, providing a biophysical precedent for our hypothesized catalytic mechanism during crystallization of G-A1U3W3 [8].

As indicated above, the c-di-GMP synthesized by the G-A1U3W3 domain is bound between the I_p site in one domain and the I_s site in another, at the interface of a domain dimer with proper two-fold symmetry (Fig. 2b). Two base-stacked c-di-GMP dimers were also bound at two such symmetry-related I-sites in PleD and WspR. Feedback inhibition in these DGCs was hypothesized to occur by c-di-GMP locking two DGC domains into an inactive dimer via their I-sites, thus separating their GG(D/E)EF active sites and preventing catalysis [7–10]. The c-di-GMP-bridged dimer of G-A1U3W3 observed in its crystal structure has equivalent geometry to the feedback-inhibited dimers of the DGC domains from PleD (Fig. 2c) and WspR (Fig. 2d). This dimeric organization of G-A1U3W3 similarly prevents interaction of its GG(D/E)EF active sites (Fig. 2), thereby blocking catalytic activity. Therefore, our structural observations suggest that feedback inhibition of the A1U3W3 protein probably employs the same allosteric mechanism used by PleD and WspR.

Summary

We have determined the crystal structure of the DGC domain from protein A1U3W3 from M. aquaeolei. The structure shows the c-di-GMP product bound in the active site even though only the GTP substrate was added to the crystallization reaction. This observation indicates that the product was synthesized in situ in the crystallization reaction containing the isolated DGG domain. Therefore, our crystal structure demonstrates that the isolated DGC domain by itself possesses diguanylate cyclase activity.

Abbreviations

ASU

Asymmetric unit

c-di-GMP

Bis-(3′–5′)-cyclic dimeric guanosine monophosphate

DGC

Diguanylate cyclase

GG(D/E)F

Diguanylate cyclase domain domain

GTP

Guanosine triphosphate

I-site

Inhibitory site

IPTG

Isopropyl-1-thio-β-D-galactopyranoside

NESG

Northeast Structural Genomics Consortium

PDB

Protein Data Bank

PDE

Phosphodiesterase

PSI

NIH Protein Structure Initiative

RMSD

Root-mean-square-deviation

SAD

Single-wavelength anomalous diffraction