C-di-GMP: the dawning of a novel bacterial signalling system

Authors


*E-mail ute.romling@mtc.ki.se; Tel. (+46) 8524 87319; Fax (+46) 833 0744.

Summary

Bis-(3′-5′)-cyclic dimeric guanosine monophosphate (c-di-GMP) has come to the limelight as a result of the recent advances in microbial genomics and increased interest in multicellular microbial behaviour. Known for more than 15 years as an activator of cellulose synthase in Gluconacetobacter xylinus, c-di-GMP is emerging as a novel global second messenger in bacteria. The GGDEF and EAL domain proteins involved in c-di-GMP synthesis and degradation, respectively, are (almost) ubiquitous in bacterial genomes. These proteins affect cell differentiation and multicellular behaviour as well as interactions between the microorganisms and their eukaryotic hosts and other phenotypes. While the role of GGDEF and EAL domain proteins in bacterial physiology and behaviour has gained appreciation, and significant progress has been achieved in understanding the enzymology of c-di-GMP turnover, many questions regarding c-di-GMP-dependent signalling remain unanswered. Among these, the key questions are the identity of targets of c-di-GMP action and mechanisms of c-di-GMP-dependent regulation. This review discusses phylogenetic distribution of the c-di-GMP signalling pathway in bacteria, recent developments in biochemical and structural characterization of proteins involved in its metabolism, and biological processes affected by c-di-GMP. The accumulated data clearly indicate that a novel ubiquitous signalling system in bacteria has been discovered.

Embarrassment of the riches: GGDEF and EAL domains in bacterial genomes

The principal reason for studying GGDEF and EAL domains – and, ultimately, for this review – is quite simple: there are just too many of them. The protein domains designated GGDEF and EAL are encoded in the genomes from diverse branches of the phylogenetic tree of bacteria. The domain names originate from the conserved sequence motifs, Gly–Gly–Asp–Glu–Phe and Glu–Ala–Leu; however, domain sizes are much larger, i.e. approximately 180 and 240 residues – for GGDEF and EAL respectively. Public protein databases currently contain over 2200 proteins with one domain or another, which makes GGDEF and EAL the most numerous domains, whose functions until recently have not been understood. Some databases still list GGDEF and EAL as ‘domains of unknown function’, DUF1 and DUF2 (e.g. SMART, http://smart.embl.de), so that annotation of these domains as uncharacterized often shows up in databases. In this review we describe the significant progress achieved in biochemical and physiological characterization of these protein domains. It has now been shown that GGDEF and EAL are involved in synthesis and hydrolysis, respectively, of bis-(3′-5′)-cyclic dimeric guanosine monophosphate (c-di-GMP) (Fig. 1). This unusual cyclic nucleotide is emerging as an important bacterial second messenger that has been overlooked despite decades of intensive microbiology research (see also D’Argenio and Miller, 2004; Galperin, 2004; Jenal, 2004).

Figure 1.

Known input signals and output of c-di-GMP metabolism. GGDEF and EAL domains conduct the turnover of c-di-GMP. pGpG is degraded to two GMP by an unknown phosphodiesterase. Various domains N-terminal of GGDEF or EAL receive and transmit the input signals (left): phosphorylation [Rec, CheY-homologous receiver domain; HisKA, histidine kinase A (phosphoacceptor) domain; HATPase, histidine kinase-like ATPases; Hpt, histidine phosphotransfer domain], ion binding (haemerythrin, HHE cation-binding domain), protein/peptide binding (TRP, tetratrico peptide repeats; CBS, domain in cystathionine beta-synthase and other proteins; CHASE, cyclase/histidine kinases-associated sensory domain), binding of phosphothreonine/phosphotyrosine (FHA, fork-head-associated domain), binding of amino acids (PBPb, bacterial periplasmic substrate-binding proteins), binding of gaseous molecules [PAS/PAC, Per (periodic clock protein), Arnt (Ah receptor nuclear translocator protein), Sim (single-minded protein); haemerythrin], binding of carbohydrates (7TMR-DISMED2, 7TM receptors with diverse intracellular signalling modules), DNA binding (HTH–LUXR, helix–turn–helix, Lux regulon), binding of nucleotides/nucleosides (GAF, domain present in phytochromes and cGMP-specific phosphodiesterases; cNMP, cyclic nucleotide-monophosphate binding domain; CBS), light (BLUF, sensor of blue-light using FAD). The output behaviour by variation of c-di-GMP concentration is shown on the right.

Phylogenetic distribution of the GGDEF and EAL domains

GGDEF and EAL domains are found in the deeply branching phyla of bacteria, such as Thermotogales and Aquificales, but are not present in proteins encoded by the genomes of any Archaea or Eukarya (not counting highly diverged and presumably misassigned sequences; Galperin, 2004). This suggests that c-di-GMP-mediated signalling is an exclusively bacterial trait. However, c-di-GMP-mediated signalling pathways are not equally abundant in diverse bacterial species (Fig. 2). The genome of Escherichia coli K12, for example, encodes 19 proteins with the GGDEF domain and 17 with the EAL domain, whereas Bacillus subtilis has four and three, respectively, and even the tiny genome of Rickettsia prowazekii encodes one of each (Galperin et al., 1999; 2001; Galperin, 2004). The current champion, Vibrio vulnificus, encodes 66 proteins with the GGDEF domain and 33 with the EAL domain, with the human pathogen Vibrio cholerae coming not far behind (Galperin, 2004). Among bacteria, GGDEF and EAL domains are absent in proteins encoded by the sequenced genomes of Bacteroidetes, Chlamydiales and Fusobacteria (Fig. 2).

Figure 2.

Phylogenetic distribution of the GGDEF domains in sequenced prokaryotic genomes.

Flexibility through modularity

Existing experimental evidence suggests that GGDEF and EAL are soluble cytoplasmic domains. Although few single-domain proteins exist, the GGDEF and EAL domains are usually found in multidomain proteins. Often a GGDEF and an EAL domain occur in the same protein, whereby the GGDEF domain is located N-terminally from the EAL domain (few exceptions exist). Extending the complexity,  the  GGDEF  and  EAL  domains are  located  C-terminally from, often multiple, sensory and signal transduction domains (Fig. 1). The different N-terminal domains are capable of acquiring a wide range of signals, i.e. phosphorylation, protein binding, binding of gaseous molecules, light. A significant fraction of the GGDEF and/or EAL domains is linked to cytoplasmic sensory domains such as PAS and GAF involved in binding of small molecular ligands or protein–protein interactions (Taylor and Zhulin, 1999; Hurley, 2003). Another sizable fraction is linked to N-terminal periplasmic or integral membrane sensory domains whose ligand binding specificity is unknown (Nikolskaya et al., 2003; Zhulin et al., 2003). In some cases, periplasmic domains belong to well-characterized families of solute-binding proteins, for example, those involved in amino acid binding, suggesting that the presence of an amino acid can modulate the turnover of c-di-GMP in the same fashion as they modulate activities of histidine kinases. Thus, the overall architecture of GGDEF and/or EAL domain proteins (sensor + output domain) is similar to that of sensor histidine kinases and methyl-carrier chemotaxis proteins (Galperin, 2004).

Biochemical activity of the GGDEF domains

A key paper in 1998 from Moshe Benziman's group (Tal et al., 1998) provided the first connection between enzymatic activities of proteins containing GGDEF and EAL domains and turnover of c-di-GMP (Fig. 1), which has been identified as an allosteric activator of cellulose synthase in the bacterium Gluconacetobacter xylinus (formerly Acetobacter xylinum) more than 15 years ago (Ross et al., 1987). Tal et al. (1998) cloned and characterized six proteins of almost identical domain composition, namely PAS-GGDEF-EAL and PAS-GAF-GGDEF-EAL, where PAS and GAF are structurally related ligand-binding domains, the first of which typically binds haem and flavins (Taylor and Zhulin, 1999), and the second one cAMP and cGMP (Hurley, 2003). Of these six paralogous proteins, three had c-di-GMP synthetase (diguanylate cyclase, DGC) activity and three had c-di-GMP phosphodiesterase (PDE-A) activity, hydrolysing c-di-GMP to linear diguanylate pGpG (Tal et al., 1998). Further hydrolysis of pGpG to two 5′-GMP molecules, referred to as PDE-B activity by Tal et al. (1998), was apparently catalysed by unrelated enzymes.

This combination of activities led to the proposal that GGDEF domain, which is distantly related to mammalian adenylate cyclases (Pei and Grishin, 2001), might be responsible for c-di-GMP production (Galperin et al., 1999; 2001; Ausmees et al., 2001; Galperin, 2004). These suggestions have been supported by a variety of genetic data (Ausmees et al., 2001; Aldridge et al., 2003; Simm et al., 2004; Tischler and Camilli, 2004) and, recently, by direct biochemical experiments (Paul et al., 2004; Ryjenkov et al., 2005). Specifically, Jenal and colleagues demonstrated the DGC activity of the purified Caulobacter crescentus PleD response regulator protein (CheY-CheY-GGDEF; Paul et al., 2004), while Gomelsky and co-workers showed the same GTP-dependent DGC activity for another six randomly chosen GGDEF domain proteins from representatives of diverse branches of the bacterial phylogenetic tree, namely, Thermotogae, Deinococcus-Thermus, Cyanobacteria, Spirochaetes, α- and γ-divisions of the Proteobacteria(Fig. 3; Ryjenkov et al., 2005). None of the GGDEF domain proteins showed activity with any other nucleotides, indicating that they are dedicated to c-di-GMP synthesis (Ryjenkov et al., 2005).

Figure 3.

Functionally verified GGDEF and EAL domains in different bacterial phyla. Blue star indicates species for which DGC and/or PDE activities have been demonstrated in vivo. Red star indicates species for which DGC and/or PDE activities have been demonstrated in vitro.

Mechanistic aspects of c-di-GMP formation

The crystal structure of the PleD protein (Hecht and Newton, 1995; Aldridge et al., 2003; Paul et al., 2004) in the presence of its reaction product, c-di-GMP, was recently solved (Protein Data Bank access code 1w25; Chan et al., 2004). As predicted by Pei and Grishin (2001), the structure of its GGDEF domain is very similar to that of eukaryotic adenylyl cyclases and includes a putative GTP binding catalytic site formed by the residues of the GGDEF motif as well as other conserved residues of the GGDEF domain (Fig. S1A in Supplementary material). The three-dimensional (3D) structure of the GGDEF domain readily explained why most amino acid substitutions, such as changing either of the Gly residues in the GGDEF motif to Ala (Kirillina et al., 2004), abrogate the activity of this domain. The GGDEF domain appeared to contain half of the catalytic site where a single GTP molecule may be bound, suggesting that the enzymatically active form of DGC is a dimer of two GGDEF domains, as seen in the crystal structure (Chan et al., 2004). The phosphorylation of one of the CheY-like receiver domains of PleD is envisioned to trigger formation of such an active dimer with a catalytic site formed between the two subunits. The proposed mechanism is in agreement with biochemical observations that individually expressed and purified GGDEF domains from various proteins have propensity to form dimers. However, such dimers possess only low-level DGC activity, while the presence of activated sensor domains (GAF or phosphorylated CheY-like receiver domain) results in significantly higher DGC activity (Paul et al., 2004; Ryjenkov et al., 2005). Noteworthy, sensor histidine kinases and chemotaxis methyl-carrier proteins, like GGDEF domains, function as dimers, suggesting possible common themes in the mechanisms of regulation of these otherwise different signalling systems.

Biochemical activity of the EAL domains

While these findings left the EAL domain (Fig. S1B) as the most plausible candidate for the PDE-A activity seen by Tal et al. (1998) and Chang et al. (2001) in G. xylinus proteins, there still remained a possibility that both GGDEF and EAL domains are required for the PDE-A activity. Although in vivo experiments showed that proteins containing EAL but not GGDEF domain were capable of c-di-GMP degradation (Simm, 2004; Tischler and Camilli, 2004), they could not fully resolve the issue of whether EAL is sufficient for PDE-A activity or just reversing the activity of GGDEF domains present in the same cell. Schmidt et al. (2005) showed that purified EAL domains from E. coli, YahA and Dos (Delgado-Nixon et al., 2000) proteins possessed PDE-A activity, i.e. hydrolysed c-di-GMP into linear dimeric GMP, pGpG. The earlier report that E. coli Dos protein had cAMP phosphodiesterase activity (Sasakura et al., 2002) could not be confirmed. PDE-A activity was dependent on Mg2+ or Mn2+, strongly inhibited by Ca2+ and did not require protein oligomerization. EAL domains and full-length EAL domain proteins also catalysed the subsequent hydrolysis of the reaction product, pGpG, into two 5′-GMP molecules. However, the latter PDE-B activity was several orders of magnitude slower and probably not physiologically relevant. This is consistent with the observation by the Benziman group (Ross, 1987; Tal et al., 1998) and suggests that hydrolysis of pGpG is catalysed by a different enzyme(s).

One can envisage that pGpG is rapidly degraded by cellular nucleases. However, as production and hydrolysis of pGpG are apparently uncoupled, a regulatory role of pGpG as a player in the c-di-GMP-mediated response or as a signalling molecule on its own is also possible, as has been found for other nucleotides and dinucleotides (Ismail et al., 2003; Gralla, 2005).

Another domain with PDE-A activity?

Several bacterial genomes encode proteins with the GGDEF but not proteins with EAL domains. The most conspicuous case is Thermotoga maritima, which encodes nine proteins with the GGDEF domain but not a single protein with the EAL domain. This led to the suggestion that this bacterium could have proteins with PDE-A activity but with another domain, i.e. HD-GYP (Galperin et al., 1999; 2001; Galperin, 2004). Whether HD-GYP has c-di-GMP hydrolase activity remains to be tested but other bacteria, including Treponema pallidum, Campylobacter jejuni and several other important human pathogens, have a protein with a GGDEF domain but do not have proteins with either EAL or HD-GYP domains. If these GGDEF domain proteins are active, how do these pathogens reduce their c-di-GMP levels? As a possibility, c-di-GMP is degraded by phosphodiesterases with relaxed substrate specificity in the respective bacteria.

So many domains, so few phenotypes

The first phenotype resulting from a mutation inactivating a gene encoding a GGDEF domain protein was detected when cell differentiation was analysed in C. cresentus (Hecht and Newton, 1995). However, it was not until the systematic genetic analysis of multicellular bacterial behaviour was initiated that mutants in genes encoding GGDEF and EAL domain proteins started to pop up more frequently (Ausmees et al., 1999; Jones et al., 1999; Römling et al., 2000; Boles and McCarter, 2002; D’Argenio et al., 2002). The realization that multicellular behaviour can be assayed by simply testing for adherence to and growth on polystyrene or glass surfaces, by pellicle formation at the liquid–air interface or by expression of distinct colony morphologies, greatly facilitated high throughput screening for mutants and hence lead to the functional identification of GGDEF and/or EAL domain proteins.

The peculiar colony morphology, which shows a surprising convergence among bacteria, has been termed wrinkled, rugose or rdar (Fig. 4; Römling et al., 1998; Boles and McCarter, 2002; D’Argenio et al., 2002; Spiers et al., 2002; Rashid et al., 2003; Kearns et al., 2005) and results from the presence of extracellular matrix components, which allow bacterial cells to form highly structured entities. The extracellular matrix consists of exopolysaccharides and other extracellular matrix components, often called ‘slime’, if the components identity is uncertain. Cellulose and an acetylated cellulose derivative were the first identified exopolysaccharides, the production of which was shown to be activated by GGDEF domain-containing proteins in such diverse bacteria as G. xylinus, E. coli, Salmonella enterica serovar Typhimurium (S. Typhimurium), Rhizobium leguminosarum and Pseudomonas fluorescence (Tal et al., 1998; Ausmees et al., 1999; Zogaj et al., 2001; Spiers et al., 2002). Adhesive fimbriae represent yet another class of extracellular matrix components whose synthesis depends on the GGDEF domain proteins (D’Argenio et al., 2002; Römling, 2005). Apparently, production of GGDEF domain proteins and the resulting elevated c-di-GMP concentrations favour the production of adhesive matrix components, which leads to elevated multicellular behaviour.

Figure 4.

Rdar colony morphologies of S. Typhimurium and E. coli. Colonies were grown on LB without salt agar at 37°C for 24 h (S. Typhimurium ATCC14028) and at 28°C for 48 h (E. coli DSM6601). The agar medium contained Congo Red (40 µg ml−1) and Coomassie brilliant blue (20 µg ml−1).

Consequently, decrease of c-di-GMP concentration leads to the opposite behaviour. Consistent with the function of EAL domain proteins as PDE-A, biofilm formation and autoaggregation are all suppressed upon overproduction of EAL domain proteins, but various types of motility are activated (Hecht and Newton, 1995; Drenkard and Ausubel, 2002; Kirillina et al., 2004; Simm et al., 2004; Tischler and Camilli, 2004). Swimming motility defect resulting from loss of flagella motion in an hns mutant of E. coli was overcome by producing an EAL domain protein (Ko and Park, 2000). This finding showed that downregulation of c-di-GMP concentrations leads to functional activation of structural components that can be uncoupled from the synthesis of the respective structures.

Interestingly, overproduction of or mutations affecting the GGDEF and EAL domain proteins do not lead to an ‘all or nothing’ phenotype. Individual GGDEF/EAL domain proteins modulate biofilm architecture, indicating differential activity in the individual cells within the biofilm (Huber et al., 2002; Bomchil et al., 2003). Such spatial heterogeneity is not surprising, as biofilms, with their water channels and complex architecture, contain various microenvironmental niches. Moreover, biofilms are not static, but consist of both moving and sessile cells (Klausen et al., 2003). Thus, it is reasonable to assume that the decision to stay in the biofilm or to move away is made on a nanoscale by individual cells based on the assessment of their local microenvironment.

The role of GGDEF and EAL domain proteins goes beyond production of extracellular matrix components and stimulation of motility. In C. crescentus, the GGDEF-domain protein PleD controls not only flagellum ejection but also cell morphology (stalk formation; Aldridge and Jenal 1999). Further, cell-to-cell communication in Myxococcus xanthus and Burkholderia cepacia required to co-ordinate multicellular behaviour is affected by GGDEF domain proteins (Gronewold and Kaiser, 2001; Huber et al., 2002), while the EAL domain protein YhjH has been shown to affect growth competition between different strains of S. enterica (Rychlik et al., 2002).

Functions beyond multicellular behaviour

GGDEF/EAL domain proteins appear to alter the expression of other GGDEF and EAL domain proteins (Kirillina et al., 2004) as well as protein composition in membranes. Early studies of uncharacterized E. coli proteins now known to contain the EAL domain suggested that they might play a role in resistance to bacteriophages and heavy metal ions (Brown et al., 1986; Chae and Yoo, 1986). In the cyanobacterium Synechococcus elongatus, the production of photopigments and the photosynthetic reaction centre protein is modulated by the response regulator PsfR that contains a GGDEF domain (Thomas et al., 2004). It is not entirely clear whether the above described phenotypes are related to multicellular behaviour.

In addition, GGDEF and EAL domain proteins are apparently involved in controlling interactions of pathogens with their various hosts ranging from plants to animals (Merkel and Stibitz, 1995; Milton et al., 1995; Ausmees et al., 1999). An EAL domain-containing protein is required to suppress virulence-inhibiting genes in Bordetella pertussis (Merkel et al., 1998a). Furthermore, some GGDEF/EAL domain-containing proteins were shown to be produced by the bacterial pathogens V. cholerae and Vibrio vulnificus exclusively during infection of mice or humans, but not in laboratory media (Camilli and Mekalanos, 1995; Lee et al., 2001; Kim et al., 2003). The list of processes regulated by GGDEF/EAL domain proteins will undoubtedly continue to grow.

Functions of GGDEF-EAL domain fusions

As the GGDEF domain is sufficient for DGC activity and the EAL domain for PDE-A activity, what is the impact of proteins containing an N-terminal GGDEF and C-terminal EAL domain? Is one of the domains non-functional or can these proteins switch between the DGC and PDE-A states depending on conditions?

GGDEF-EAL domain proteins whose activities were tested in vitro (Tal et al., 1998; M. Tarutina, D.A. Ryjenkov and M. Gomelsky, unpublished) show only one activity, suggesting that one of the two domains is inactive. Sequence analysis of the EAL domains from the three DGCs from G. xylinus revealed differences in several highly conserved amino acid motifs preserved in the EAL domains from active PDE-As (Schmidt et al., 2005). Similarly, some GGDEF domains in GGDEF-EAL domain proteins with PDE-A activity deviate from the consensus sequence and different subclasses of GGDEF domains could be distinguished (Pei and Grishin, 2001). However, the molecular basis for inactivity is not entirely clear and the assignment of amino acids definitely required for enzymatic functionality requires saturating analyses. The hypothesis of inactive domains, if proven, would explain the biochemical paradox of the GGDEF-EAL domain proteins. However, it would pose new questions, i.e. what roles do enzymatically inactive domains play? What is the function of proteins that contain one or even two seemingly enzymatically inactive domains?

However, some in vivo observations suggest that a GGDEF-EAL domain protein may possess both DGC and PDE-A activities. For example, the knockout of diguanylate cyclase 1 (dgc-1), which is responsible for 80% of the c-di-GMP production in G. xylinus (Tal et al., 1998; Bae et al., 2004), showed decreased cellulose production under some growth conditions but, surprisingly, elevated cellulose production under others (Bae et al., 2004). Furthermore, ScrC, a GGDEF-EAL domain protein from Vibrio parahaemolyticus, has the opposite function on two transcriptional fusions monitoring swarming and biofilm formation when expressed alone or in its operon context scrABC (Boles and McCarter, 2002). Inactivation of homologous GGDEF-EAL domain-containing proteins in different bacteria sometimes causes opposite phenotypes. For example, the knockout of yciR (STM1703) of S. Typhimurium caused cellulose overproduction, indicating that YciR has PDE-A activity (Garcia et al., 2004), whereas, a mutation in yciR diminished biofilm formation in B. cepacia, indicating that the encoded protein has DGC activity (Huber et al., 2002). Clearly, additional work is needed to determine the full enzymatic capacity of proteins containing both GGDEF and EAL domains.

The challenge of fine-tuned networking

The inactivation of genes encoding GGDEF and EAL domain proteins frequently modulates the amplitude of a phenotype or regain of function is achieved under different environmental conditions, but seldom causes a complete phenotype change (Aldridge and Jenal, 1999; Gronewold and Kaiser, 2001; Garcia et al., 2004; Thomas et al., 2004). For example, cellulose biosynthesis in S. enterica on LB agar plates without salt is activated by the GGDEF domain protein AdrA (Römling et al., 2000), while in a carbon source-rich, trace element-poor medium, STM1987 fulfils this function (Garcia et al., 2004). Thus, the functionality of the two GGDEF domain proteins to mediate the phenotype ‘cellulose biosynthesis’ was restricted to certain environmental conditions. This finding brings up the question of how to manage a branched network of DGCs and PDE-As and avoid unwanted cross-talk between different c-di-GMP targets in a single bacterial cell. We speculate that several mechanisms might work in parallel to accomplish specificity of c-di-GMP action. One involves temporal and environmental control of GGDEF and EAL domain protein expression on a transcriptional and post-transcriptional level. Cell density-dependent expression of GGDEF and EAL domain proteins was observed by DNA array analysis (Johnson et al., 2005). Temperature-regulated expression of the GGDEF domain protein HmsT by proteolysis controls the haemin storage phenotype of Yersinia pestis (Perry et al., 2004). The second possible mechanism relies on the dependence of enzymatic activities of DGCs and PDE-As on environmental or intracellular stimuli. In line with this suggestion, the activities of PleD and of the Rrp1 protein from Borrelia burgdorferi are strongly dependent on the phosphorylation status of the CheY domains (Chan et al., 2004; Ryjenkov et al., 2005). Hence, a given enzyme produces or degrades c-di-GMP only when a specific signal has been received. A third mechanism probably involves colocalization of enzymes of c-di-GMP turnover with their targets. Two pieces of experimental evidence are in line with this scenario. (i) G. xylinus DgcA and PdeA proteins were found to co-purify with cellulose synthase (Ross et al., 1987). (ii) C. crescentus PleD is localized near the base of the flagellum, where its DGC activity is required for flagellum ejection. A fourth potential mechanism is a feedback inhibition of the enzymatic activity by the product, c-di-GMP. Such inhibition by c-di-GMP has been observed for PleD (Paul et al., 2004), in which an allosteric binding site for c-di-GMP is formed jointly by the signalling and GGDEF domain (Chan et al., 2004). Whether this mechanism is unique to PleD or common among DGCs remains to be tested. However, significantly different in vivo activities of c-di-GMP formation have been observed among proteins (R. Simm and U. Römling, unpublished results), suggesting that a broad range of enzymatic activities is present.

The intracellular levels of c-di-GMP are in the micromolar range or lower, as judged by measurement in several proteobacterial species (Weinhouse et al., 1997; Simm et al., 2004; Tischler and Camilli, 2004). Even if a reliable procedure for quantification is set up, the significance of a measurement of the average cytoplasmic c-di-GMP concentration is not entirely obvious. C-di-GMP concentrations probably fluctuate locally, as discussed above. The measured and real concentrations of c-di-GMP in the cytoplasm might differ. For example, the majority of the c-di-GMP in G. xylinus is apparently bound by a yet undefined membrane protein (Weinhouse et al., 1997) and is probably released in response to certain signals, which might lead to very short cytoplasmic c-di-GMP spikes. Benziman and colleagues have shown that only a fraction of the cellular c-di-GMP concentration is required to achieve almost optimal activation of cellulose biosynthesis (Tal et al., 1998), suggesting that various biological processes might require different thresholds of c-di-GMP concentrations.

Networking – yes! – but where?

Where do GGDEF/EAL domain proteins stand in the established regulatory network of signal transduction in bacteria. A subfraction of GGDEF/EAL domain proteins is intimately linked to two-component signalling systems. In some bacteria, genes encoding EAL domain proteins are coexpressed with sensor kinase and response regulator genes (Merkel et al., 1998b; Tischler and Camilli, 2004; Kulasekara et al., 2005). Many functionally characterized GGDEF/EAL domain proteins have an N-terminal receiver domain (Gronewold and Kaiser, 2001; Drenkard and Ausubel, 2002) that is phosphorylated by cognate sensor kinases (Aldridge et al., 2003; Kulasekara et al., 2005). In S. enterica, production of the GGDEF domain protein AdrA is transcriptionally activated by the response regulator CsgD (Römling et al., 2000). These examples show that the phosphotransfer network established by two-component systems integrates components of the c-di-GMP signalling pathway on several levels.

In other instances, EAL domain proteins might be directly involved in gene regulation, although there are no experimental data that support this idea. However, several proteobacterial proteins contain EAL domains and predicted DNA-binding domains, and two of them have experimentally verified PDE-A activity (Tischler and Camilli, 2004; Schmidt et al., 2005). What, if any, is the relation of their PDE-A activity to DNA binding? One working hypothesis that arises is that binding of the small molecule c-di-GMP to the EAL domain affects DNA binding or transcriptional activation.

Mechanism(s) of c-di-GMP action – an entirely unresolved question

Studies of the c-di-GMP-mediated signalling in bacteria are still in their infancy and there are more questions than answers. The most intriguing question is the nature of the target(s) of c-di-GMP action. In their original studies of the regulation of G. xylinum cellulose synthase, Benziman and colleagues detected c-di-GMP binding to a 67 kDa soluble protein whose N-terminal region is very similar to an internal sequence in the cellulose synthase β-subunit, BcsB (Mayer et al., 1991). This led to the suggestion that c-di-GMP binding is a property of the cellulose synthase itself. However, subsequent studies revealed that c-di-GMP could have been interacting with a 200 kDa membrane-bound protein complex (Weinhouse et al., 1997) that has not been further characterized.

Could there be a single c-di-GMP-binding adaptor protein, akin to CRP for cAMP, or are there multiple targets for c-di-GMP action? A single adaptor would be expected to have the same phylogenetic distribution as the GGDEF and EAL domains but no such protein could be identified (M.Y. Galperin, unpubl. obs.). Therefore, c-di-GMP is likely to operate through several different targets.

Practical importance of c-di-GMP

There is an urgent need to overcome the intrinsically high resistance of biofilm-forming pathogens to traditional antimicrobial agents. Here, one strategy is to develop therapies that interfere with the formation of the multicellular communities. Biofilm formation of Pseudomonas aeruginosa, S. enterica, E. coli, V. cholerae and Y. pestis has already been demonstrated to be manipulable by varying c-di-GMP concentrations (Kirillina et al., 2004; Simm et al., 2004; Tischler and Camilli, 2004). As biofilm formation of those important human pathogens is associated with chronic disease, the carrier state or transmission, c-di-GMP metabolism is emerging as an attractive target for clinical intervention. On the other hand, c-di-GMP added extracellularly to Staphylococcus aureus inhibited cell biofilm formation and attachment to human epithelial cells suggesting that c-di-GMP itself could be used as a therapeutic agent to interfere with multicellular behaviour and adherence mechanisms (Karaolis et al., 2005a). These findings once again demonstrated the link between basic microbiological research and practical aspects of antibacterial defence as prominently stated in a recent open letter by US microbiologists (Altman et al., 2005). However, effects of c-di-GMP treatment with a potential for therapeutical interference in cancer were also observed in eukaryotic cells (Amikam et al., 1995; Steinberger et al., 1999; Karaolis et al., 2005b). The observation of effectively applied extracellular c-di-GMP brings forth the intriguing possibility that specific receptors or uptake systems for c-di-GMP exist in pro- and eukaryotes.

Concluding remarks

It is evident that there are more questions than answers regarding c-di-GMP-dependent signalling. However, we are in the midst of an explosion of interest in c-di-GMP. The next few years will likely prove crucial for establishing the scope of targets of this novel second messenger and determining its mechanism(s) of action. C-di-GMP has risen to prominence due to the massive microbial genome sequencing and an increased interest in the natural bacterial lifestyle, the biofilm. We find ourselves at the dawning of a novel signalling system in bacteria. C-di-GMP is rapidly earning its place alongside its better-understood cousins, cAMP and cGMP, as a ubiquitous second messenger.

Acknowledgements

Work in the laboratory of U.R. is supported by the Karolinska Institutet (Elitforskartjänst), the Mukoviszidose e.V. and Vetenskapsrådet Grant 621-2004-3979. Work on c-di-GMP in the laboratory of M.G. is supported by the NSF Grant  MCB-0316270.

Note added in proof

After this review was completed, phosphodiesterase activity of the EAL domain proteins has been reported in two more papers by Hisert et al. (2005) A glutamate-alonine-leucine (EAL) domain protein of Salmonella controls bacterial survival in mice, antioxidant defence and killing of macrophages: role of cyclic oliGMP. Mol Microbiol56: 1234–1245. And Bobrov et al. (2005) The phosphodiesterase activity of the HmsP EAL domain is required for negative regulation of biofilm formation in Yersinia pestis. FEMS Microbiol Lett, in press.

Ancillary