Finally! The structural secrets of a HD-GYP phosphodiesterase revealed



The major sessility-motility lifestyle change and additional fundamental aspects of bacterial physiology, behaviour and morphology are regulated by the secondary messenger cyclic di-GMP (c-di-GMP). Although the c-di-GMP metabolizing enzymes and many receptors have been readily characterized upon discovery, the HD-GYP domain c-di-GMP phosphodiesterase family remained underinvestigated. In this issue of Molecular Microbiology, Bellini et al. provide an important step towards functional and structural characterization of the previously neglected HD-GYP domain family by resolving the crystal structure of PmGH, a catalytically active family member from the thermophilic bacterium Persephonella marina. The crystal structure revealed a novel tri-nuclear catalytic iron centre involved in c-di-GMP binding and catalysis and provides the structural basis to subsequently characterize in detail the catalytic mechanism of hydrolysis of c-di-GMP to GMP by HD-GYP domains.

The second messenger bis-(3′–5′)-cyclic dimeric guanosine monophosphate (c-di-GMP) plays a central role in the control of a variety of fundamental bacterial processes including biofilm formation, motility, virulence and cell differentiation (Mills et al., 2011; Boyd and O'Toole, 2012). Upon stimulation by extra- and intracellular signals, the cellular levels of c-di-GMP are modulated by the opposing activities of c-di-GMP synthesizing GGDEF domain diguanylate cyclases (DGC) and c-di-GMP degrading EAL or HD-GYP domain phosphodiesterases (PDE), each named after a subset of conserved signature amino acids (Fig. 1A; Schirmer and Jenal, 2009). Once thought to be only an exceptional substitute of the EAL domain in certain bacterial genomes, HD-GYP PDEs, which constitute a subset of the larger HD domain superfamily, whose members catalyse phosphohydrolase reactions utilizing a catalytic metal centre (Aravind and Koonin, 1998), make up more than 30% of all predicted c-di-GMP PDEs and have been recognized to be even the predominant c-di-GMP-specific PDEs in some phyla such as Thermotogae and Spirochaetes (Römling et al., 2013).

Figure 1.

EAL and HD-GYP domain-based c-di-GMP phosphodiesterases.

A. Synthesis and degradation of c-di-GMP.

B. Crystal structure of the EAL domain of Blrp1 in complex with two Mn2+ ions (sphere) and c-di-GMP (sticks) (Barends et al., 2009).

C. Close-up view of the active site of Blrp1 with the black arrow indicating the direction of the in-line nucleophilic attack by the hydroxide ion (white sphere).

D. Crystal structure of the HD-GYP domain of PmGH in complex with one Fe2+ ion (sphere) and c-di-GMP (sticks).

E. Close-up view of the active site of PmGH with the black arrows indicating the directions of the in-line nucleophilic attack for generating the intermediate 5′-pGpG. The three spheres represent the three iron ions observed for the c-di-GMP-free structure.

The prototype of a HD-GYP domain protein is RpfG from the plant pathogen Xanthomonas campestris, a multi-phenotype protein involved in virulence, motility and biofilm regulation (Ryan, 2013) and the first protein of this family to be shown to function as a c-di-GMP degrading phosphodiesterase (Ryan et al., 2006). These pioneering biochemical studies revealed that RpfG hydrolyses c-di-GMP into GMP, in contrast to EAL domains, which hydrolyse c-di-GMP to the linear dinucleotide 5′-pGpG (Fig. 1A). Thus, besides serving as an interaction platform alternative to EAL domains for protein–protein interactions (Andrade et al., 2006; Lindenberg et al., 2013), the HD-GYP domain produces the potential secondary messenger 5′-pGpG (Nickels and Dove, 2011) only as an intermediate upon c-di-GMP hydrolysis.

As the neglected small sibling of the EAL PDE domain, HD-GYP domains and their biological significance have not been extensively investigated. Other than the studies in Xanthomonas, the few performed in Pseudomonas aeruginosa, Vibrio cholerae and Bordetella pertussis showed, however, that HD-GYP domain proteins are major determinators of virulence, motility and biofilm formation status and contribute significantly to the alteration of cyclic di-GMP levels in the cell (Hammer and Bassler, 2008; Ryan et al., 2009; Sultan et al., 2011; Zhang et al., 2013).

Cyclic di-GMP metabolizing proteins rarely function as stand-alone GGDEF, EAL or HD-GYP domains, but are most commonly associated with regulatory domains providing the basis of post-translational signal integration to modulate the enzymatic activities in response to numerous internal and external stimuli such as light and oxygen (Barends et al., 2009; Kalia et al., 2013). High-resolution structure–function studies on GGDEF and EAL domains have provided detailed insight into their roles in the turnover of c-di-GMP and their regulation by the adjacent signalling domains (Fig. 1 and C; De et al., 2008; Barends et al., 2009; Schirmer and Jenal, 2009, Chen et al., 2012). In contrast, HD-GYP proteins have proven difficult to crystallize and limited information has been available on the characteristics of the active site and substrate/product binding during the catalytic cycle of these proteins. With the exception of the catalytically inactive HD-GYP protein Bd1817 from the bacterial predator Bdellovibrio bacteriovorus (Lovering et al., 2011), HD-GYP proteins have resisted high-resolution crystallography. The structure of Bd1817 provided first insights into c-di-GMP hydrolysis, revealing features such as a binuclear iron centre and clarifying the role of conserved residues in metal binding within the HD-GYP family. However, key questions regarding c-di-GMP conformation, recognition, hydrolysis and regulation of these events remained unanswered.

In this issue of Molecular Microbiology, the Walsh laboratory, in collaboration with the Ryan and Dow laboratory, report the crystal structure of PmGH, an enzymatically active HD-GYP PDE protein from Persephonella marina EX-H1 at 2.03 Å resolution (Bellini et al., 2013). The overall structure revealed that PmGH adopts a head-to-head dimer arrangement in which each monomer consist of an N-terminal regulatory GAF domain connected to a C-terminal HD-GYP domain. Allegorically with the breaking bond phosphodiesterase activity, the structure of the HD-GYP domain resembles an opened two-clawed chela.

The combined information from crystal structures of PmGH determined in complex with metal ions, the substrate c-di-GMP, and the product GMP provided insight into the catalytic mechanism of the family of HD-GYP phosphodiesterases. Unlike the structures of EAL domain phosphodiesterases that usually contain one or two divalent metal ions in the active site (Barends et al., 2009; Schirmer and Jenal, 2009; Tchigvintsev et al., 2010), PmGH was shown to contain three iron ions arranged in a triangular geometry. Site-directed mutagenesis showed that alanine substitutions of the eight metal-ligating residues either abolished or significantly reduced the catalytic activity, indicating that all three metal ions contribute to catalysis, although a more structural role in co-ordination of the active site cannot be ruled out (Bellini et al., 2013). Consistent with the crystal structure of the HD-G_P protein Bd1817 reported earlier (Lovering et al., 2011), the conserved His221 and Asp222 that constitute the HD-GYP motif were shown to be directly involved in metal binding. Thus, the authors further clarify the role for several of the metal co-ordination residues, which have previously been shown to be essential for PDE activity and are highly conserved within the HD-GYP family (Ryan, 2013).

The crystal structure of PmGH in complex with c-di-GMP revealed a ‘U'-shaped (cis) binding conformation for a single c-di-GMP molecule in close proximity to the tri iron centre (Fig. 1; Bellini et al., 2013), distinct from the extended binding conformation in the active site of EAL domains (Fig. 1B) and from the intercalated U-shaped c-di-GMP dimer that binds to the inhibitory I-site in GGDEF domains (Schirmer and Jenal, 2009). The binding conformation of c-di-GMP in the HD-GYP domain of PmGH is similar, though, to the binding mode of c-di-GMP by many PilZ domains (Krasteva et al., 2012). Thus, the structure of PmGH in complex with c-di-GMP further demonstrates the conformational flexibility of c-di-GMP to match a wide variety of recognition motifs. It remains to be investigated whether other cyclic di-nucleotide second messengers possess a similar conformational flexibility (Davies et al., 2012; Corrigan and Gründling, 2013; Nelson et al., 2013), although preliminary chemical data predict that c-di-AMP, a second messenger predominant in Gram-positive bacteria and archaea, has less conformational flexibility (Nakayama et al., 2011).

Interestingly, c-di-GMP binding site and mutational analyses suggest a novel mode of c-di-GMP recognition with three ions forming the c-di-GMP binding site. The phosphate groups positioned at the bend of the U-formed c-di-GMP directly interacts with the central Fe ion in the tri-nuclear iron complex through one of its non-bridging phosphate oxygens. Surprisingly, alanine replacement of conserved amino acids interacting directly with c-di-GMP in the crystal structure had almost no effect on the catalytic activity (Bellini et al., 2013).

One of the distinct features of HD-GYP domain proteins is that they can break down the intermediate 5′-pGpG to GMP rather efficiently (Stelitano et al., 2013). The hydrolysis of the P-O bond in 5′-pGpG may be promoted by the unique substrate-binding conformation as well as the tri-nuclear cluster. Based on the structural data, the authors suggest that the catalytic mechanism of PmGH involves a metal-ion-activated hydroxide ion for nucleophilic attack on the scissile P–O bond. Considering that the intermediate 5′-pGpG can only be generated by an in-line nucleophilic attack from position (a) or (b) (Fig. 1E), it is not clear from the current structures where the bridging nucleophilic hydroxide is located. To fully establish the trimetal mechanism, further biochemical and structural studies are also needed to identify the proton donor required for the protonation of the O3′ leaving group and establish the role of the third metal ion, which may not be needed for water activation, but may play a role stabilizing the transition state or the developing charge on the leaving group. These high-resolution structure–function studies ideally include active HD-GYP domains containing bi- or trinuclear metal centres, HD-GYP domains in complex with a non-hydrolysable substrate and catalytically inactive mutants with the iron centre intact.

Although the crystal structures of PmGH did not reveal how the HD-GYP phosphodiesterase domain is regulated by the putative regulatory GAF domain nor has the GAF ligand been identified (Bellini et al., 2013), one interesting observation made by the authors is that although the Tyr285 from the conserved ‘GYP’ motif was seen to make direct contact with c-di-GMP in the crystal structure, alanine substitutions of Gly284, Tyr285 or Pro286 did not seem to impact catalysis significantly. Because a saturating concentration (100 μM) of c-di-GMP was used for the enzymatic assay, the observation does not rule out a perturbation of Km and substrate binding that may only be observed at lower substrate concentration. It was reported, however, by Ryan et al. (2010; 2012) that the GYP motif of the HD-GYP protein RpfG is crucial for the formation of a protein–protein complex between RpfG and a GGDEF domain protein, but not required for the basic catalytic activity. It is conceivable that the surface-exposed GYP motif may play a regulatory role in controlling the enzymatic activity and c-di-GMP degradation through protein–protein interaction. In addition, Miner et al. recently found that only the reduced (ferrous) form of a HD-GYP protein from V. cholerae is active in breaking down c-di-GMP, which raises the intriguing possibility that some HD-GYP proteins could be under the control of redox regulation (Miner et al., 2013).

According to the crystal structure, the direct contact of the Tyr285 from the conserved ‘GYP’ motif with c-di-GMP and its concomitant requirement in protein–protein interaction between the RpfG phosphodiesterase and GGDEF domain proteins (Ryan et al., 2012), in order to mediate type IV pilus-dependent motility, seem to be mutually exclusive. However, the possibility remains that conformational changes associated with c-di-GMP binding and degradation determine protein–protein interaction and the subsequent physiological events similar to trigger enzymes that regulate gene expression in response to substrate availability (Commichau and Stulke, 2008). Indeed, the EAL domain protein YciR in Escherichia coli was recently suggested to be a trigger enzyme that controls expression of the major biofilm regulator CsgD through c-di-GMP sensing rather than its hydrolytic activity (Lindenberg et al., 2013).

By analysing HD-GYP proteins phylogenetically, the authors identified two subfamilies of HD-GYP proteins, one of which may not contain a tri-nuclear iron cluster due to the lack of one or two metal-ligating residues (Bellini et al., 2013). Whether this subfamily of HD-GYP proteins is catalytically competent or not remains to be determined. However, at least for one member of this subfamily, experimental evidence suggests a role in c-di-GMP turnover (Ryan et al., 2009).

Phylogenetic and sequence analysis in the current study also implies that the HD-GYP family contains several catalytically inactive members (Bellini et al., 2013). Experimental evidence indicates that one of three HD-GYP domain proteins from P. aeruginosa, although affecting biofilm formation and virulence, does not seem to possess c-di-GMP hydrolysing activity (Ryan et al., 2009). Sequence diversity in combination with functional divergence of EAL domain proteins has been well documented. In addition to the enzymatically active EAL domains, catalytically incompetent EAL domain proteins were discovered to function as c-di-GMP receptors. Examples are FimX involved in type IV pili twitching motility (Jain et al., 2012; Römling et al., 2013) and LapD which regulates adhesin expression through regulation of periplasmic protease activity (Newell et al., 2011). Other c-di-GMP non-binding EAL domains function as protein–protein interaction domains antagonizing the activity of transcription factors. Examples include the light sensor BluF and the stand alone EAL-like domains STM1344 and STM1697 (Tschowri et al., 2009; Wada et al., 2011; Ahmad et al., 2013).

In addition, the HDOD domain that is known to be associated with c-di-GMP signalling is structurally related to the HD-GYP protein (Galperin, 2006). The enzymatically inactive HDOD domain, such as the one in the YuxH protein of Bacillus subtilis, lacks metal-binding residues and may represent a non-enzymatic variant of the HD-GYP domain (Z.-X. Liang, unpubl. data).

In conclusion, we expect that the arrival of the first structure of a catalytically active HD-GYP domain will spur further studies to fully unveil the functional spectrum and biological impact of this important family of c-di-GMP signalling proteins.