Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
The glycerophosphodiester phosphodiesterases are evolutionarily conserved proteins that have been linked to several patho/physiological functions, comprising bacterial pathogenicity and mammalian cell proliferation or differentiation. The bacterial enzymes do not show preferential substrate selectivities among the glycerophosphodiesters, and they are mainly dedicated to glycerophosphodiester hydrolysis, producing glycerophosphate and alcohols as the building blocks that are required for bacterial biosynthetic pathways. In some cases, this enzymatic activity has been demonstrated to contribute to bacterial pathogenicity, such as with Hemophilus influenzae. Mammalian glyerophosphodiesterases have high substrate specificities, even if the number of potential physiological substrates is continuously increasing. Some of these mammalian enzymes have been directly linked to cell differentiation, such as GDE2, which triggers motor neuron differentiation, and GDE3, the enzymatic activity of which is necessary and sufficient to induce osteoblast differentiation. Instead, GDE5 has been shown to inhibit skeletal muscle development independent of its enzymatic activity.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The glycerophosphodiester phosphodiesterases (GP-PDEs) are highly conserved enzymes from bacteria to protozoa, as well as to mammalia (Fig. 1); however, they differ in their enzymatic characteristics and biological functions. The bacterial GP-PDE main function is the hydrolysis of glycerophosphodiesters for the production of glycerol 3-phosphate (Gro3P), which is required for several bacterial biosynthetic pathways. For some bacterial pathogens, these enzymes have an active role in host invasion and bacteria survival, via evasion of the host immune system. Instead, the mammalian glycerophosphodiesterases (GDEs) always have a preferred substrate, which is not necessarily a glycerophosphodiester, and their variety (seven isoforms in human) allows a high degree of tissue and functional specificity. In the present review, we report on the different activities of these enzymes, indicating their physiological functions where these are known.
Initial GP-PDE characterization in Escherichia coli
The E. coli genome contains a regulon, the products of which are responsible for the uptake and metabolism of glycerol and Gro3P, which are sources of the carbon and phosphate that are required for bacteria carbohydrate and phospholipid biosynthesis [1, 2]. In this regulon, the glpT operon includes both the Gro3P transporter gene, glpT, which encodes an ABC transporter that actively transports Gro3P from the periplasm into the cytosol, and also the gene that encodes GlpQ (Fig. 1), a 40-kDa enzyme that catalyzes the hydrolysis of glycerophosphodiesters, with the production of Gro3P and alcohols . GlpQ is localized in the periplasm and has a broad substrate specificity . It hydrolyzes glycerophosphodiester bonds through its recognition of the glycerophospho moiety and, apparently, it does not hydrolyze other types of bonds, such as that of bis(p-nitrophenyl)phosphate. The reported GlpQ substrates are glycerophosphoethanolamine (GroPEtn), glycerophosphocholine (GroPCho), glycerophosphoinositol (GroPIns), glycerophosphoserine (GroPSer) and glycerophosphoglycerol (GroPGro), with essentially equivalent Km and Vmax values. GlpQ also shows a lower affinity for cardiolipin and bis(glycerophospho)glycerol, with no demonstrated activity towards phosphatidyl-dl-glycerol or lysophosphatidyl-dl-glycerol [3, 5]. The native state of GlpQ is as a dimer , and its enzymatic activity is optimal at pH 9 in presence of Ca2+ . GlpQ has been reported to also have transcriptional control over several periplasmic proteins, such as the maltose-binding protein .
The ugp-encoded transport system represents another E. coli transport system for Gro3P . This belongs to the family of binding-protein-dependent transport systems that are found only in Gram-negative bacteria . The ugp operon comprises genes that encode several proteins, among which is UgpQ, comprising a 27-kDa soluble GP-PDE that is a GlpQ homolog (Fig. 1) . UgpQ is a cytosolic enzyme with broad substrate specificity towards glycerophosphodiesters, with maximal activity at pH 7.5 and the requirement for divalent cations for activity [9, 10]. Expression of the UgpQ protein is significantly induced in phosphate-starved wild-type E. coli, where the primary physiological function of UgpQ is the use of glycerophosphodiesters as a source of phosphate, an activity that is performed more efficiently by UgpQ than by GlpQ . Escherichia coli GlpQ and UgpQ possess a significant similarity, suggesting a common evolutionary origin . They showed similar enzymatic activities and mainly differ with respect to cellular localization. In the phylogenetic tree of Fig. 1, they are located in two separate main branches, with the cytosolic UgpQ clustered with the eukaria enzymes and with GP-PDEs of bacteria without a cell wall. Instead, the periplasmic GlpQ appears in a cluster with GP-PDEs of Gram-negative bacteria, for which a role in bacterial pathogenicity has been proposed.
Protein D in Haemophilus influenzae: an antigenically active protein for vaccine formulation
Protein D (also known as hpd or GlpQ) was identified as the E. coli GlpQ homolog in H. influenzae (Fig. 1, in cluster with E. coli GlpQ) [11, 12] and several reports have demonstrated its relevance in the pathogenicity of this Gram-negative bacterium. The encapsulated H. influenzae, so named by the presence of a polysaccharide capsule, causes invasive infections, especially in children, by penetration of the epithelium of the nasopharynx and invasion through the blood capillaries . The nonencapsulated (or nontypeable) strain of H. influenzae is frequently associated with both acute and chronic otitis media and pneumonia, chronic obstructive pulmonary disease and cystic fibrosis . This has resulted in the implementation of vaccination protocols, and protein D has all of the properties necessary for its application as an antigenically active carrier protein for conjugate vaccines, mainly because it is a surface-exposed membrane lipoprotein  that is highly conserved among different H. influenzae strains .
Protein D has an enzymatic activity that is similar to the GlpQ E. coli enzyme , as a GP-PDE without high specificity for the alcohol portion of the substrate , and several studies have correlated this enzymatic activity with bacterial infection [18, 19]. The initial indication came from the observation that 100-fold-higher concentrations of a H. influenzae isogenic mutant lacking this surface-exposed lipoprotein were required to cause otitis in rats challenged with these bacteria . Moreover, protein D appears to be involved in the pathogenesis of upper respiratory tract infections of nontypeable H. influenzae, probably through enhancement of the functional damage of ciliated epithelial cells, with a decrease in the frequency of the ciliary beat and an increase in the loss of cilia . This kind of damage is not seen with a protein-D-deficient bacterial strain . Even if nontypeable H. influenzae is usually regarded as a non-invasive pathogen, Ahren et al.  demonstrated that protein D is necessary for the adhesion and internalization of these bacteria into human monocytes. This explains the persistence of these bacteria in certain infections because, in this way, they are shielded from the highly phagocytic polymorphonuclear leukocytes, as well as from antibody-mediated bactericidal activity and from antibiotics [21, 22]. Determination of the molecular mechanism in host-cell adhesion came from subsequent studies performed by Fan et al. , who demonstrated that protein D provides an alternative mechanism for H. influenzae to acquire choline for the synthesis of lipopolysaccharide-phosphorylcholine from glycerophospholipids, the major source of choline in the host cells. This decoration of the bacteria cell wall, which is known to be a particular feature of many bacterial species that reside predominantly in the respiratory tract [24-26], allows the bacteria to escape the host immune system because this mimics the characteristics of the host cells, where phosphorylcholine is a component as part of phosphatidylcholine [27-29]. In addition, the expression of lipopolysaccharide-phosphorylcholine allows bacteria to interact with the platelet-activating factor receptor of host cells , which leads to the invasion of epithelial cells and the sequestration from host clearance mechanisms .
Based on this evidence, in animal models, protein D has been used as an active vaccine against homologous and heterologous strains . The pneumococcal conjugate vaccine (Synflorix™; GlaxoSmithKline, Rixensart, Belgium) that has been licensed in Europe and in more than 110 countries around the world since 2008 is a 10-valent vaccine that uses a nonlipidated form of nontypeable H. influenzae protein D as the protein carrier component [33, 34]. The antibodies raised by the protein D vaccine inhibit the enzymatic activity of recombinant protein D in in vitro assays, and the measurement of these antibody responses from serum samples is a useful tool for assessing the level of vaccine-induced protective immunity . Moreover, when the passive immunization induced by antibodies raised against protein D was verified, there was protection against the development of nontypeable H. influenzae-induced acute otitis media in chinchillas [36, 37].
The search for antigenically active proteins in other pathogens
On a similar basis, the GP-PDEs of two pathogen spirochetes were cloned in a search of outer membrane proteins that might serve as protective immunogens. Unfortunately, the data were less striking compared to H. influenzae, mainly as a result of the cellular localization of these spirochete GP-PDEs.
Gpd was cloned in Borrelia hermsii (Fig. 1, in cluster with E. coli GlpQ), one of the etiological agents of the tick-borne relapsing fever . The Gpd DNA sequence predicted that Gpd contains a 20-amino acid signal peptide typical of lipoproteins ; indeed, subsequent studies with tritiated palmitate have demonstrated that Gpd is an outer membrane lipoprotein . In the Borreliae species that express Gpd, this is required for Gro3P production starting from deacylated phospholipids . This metabolic pathway appears to contribute to cell proliferation during host infection, which leads to an increased cell density of B. hermsii in the host blood compared to Lyme disease spirochetes, which do not express Gpd at all . Even if the enzyme activity monitored by GroPCho hydrolysis is associated with the membrane fraction of the spirochete, immunogold labeling and electron microscopy analysis have not detected Gpd associated with the outer surface of intact spirochetes; this indicates that, in B. hermsii, Gpd is most probably bound to the periplasmic side of the inner or outer membrane .
Borrelia hermsii Gpd is now used only for the serological diagnosis of patients with tick-borne relapsing fever because the presence of antibodies against this spirochete has been detected upon infection [41, 42]. Moreover, the cloning of Gpd homologs in several Borreliae species, and their conservation, has allowed the setting-up of molecular and serologic techniques for the diagnosis of relapsing fever borreliosis [43-45].
Intriguingly, the search for the extracellular antigens of the syphilis pathogen Treponema pallidum led to controversial results. Upon infection, this spirochete can produce rapid humoral and cellular responses, and it is killed by macrophages after antibody-mediated opsonization and subsequent phagocytosis . Unfortunately, this first immune response does not eradicate the few organisms that are present during the latency, which leads to the establishment of a lifelong chronic infection in the absence of appropriate antibiotic treatment . For this reason, the identification of cell surface antigens is a prerequisite for vaccine design. Potential targets of the opsonic antibodies produced during infection were obtained via the differential screening of a genomic expression library with a T. pallidum-specific rabbit immune serum. This immunological approach resulted in the cloning and sequencing of the GP-PDE (GlpQ) of T. pallidum (Fig. 1, in cluster with E. coli GlpQ) . In this scenario, the GlpQ residing on the outer leaflet of the T. pallidum outer membrane was proposed to bind the Fc region of immunoglobulins, thus limiting the antibody cytotoxic and opsonic capacities, and contributing to escape from the host immune system. In further investigations, immunization with the recombinant enzyme significantly protected rabbits from subsequent T. pallidum challenge, with attenuation of lesion appearance and development at the sites of infection . For this reason, GlpQ was proposed as part of a vaccine cocktail to achieve complete immunity against T. pallidum challenge . This was substantiated by the conservation of the GlpQ sequence in several T. pallidum species  and, within a few years, this led to the improvement of the GlpQ-based vaccine formulation. Recently, a DNA-based vaccine constructed by fusing T. pallidum GlpQ with interleukin-2, using chitosan nanoparticles as the vector, effectively attenuated the development of syphilitic lesions .
Controversial results have been obtained in parallel studies focusing more on the GlpQ structure, which revealed that T. pallidum GlpQ is a hydrophilic lipoprotein that is assumed to be anchored by N-terminal lipids to the periplasmic leaflet(s) of the peptidoglycan-cytoplasmic membrane, and not to be exposed on the outer membrane of the pathogen [50, 51]. In addition to the different localization data, the main contrast with previous observations was highlighted because rabbits hyper-immunized with GlpQ were not protected against intradermal challenge with virulent treponemes .
The search for genes required for Pasteurella multocida virulence and survival in the host has been similarly unsuccessful, which has led to the study of another analog of GlpQ: P. multocida lipoprotein D (Fig. 1, in cluster with E. coli GlpQ). Pasteurella multocida is a ubiquitous animal pathogen that is associated with a range of diseases, which include hemorrhagic septicemia, atrophic rhinitis and fowl cholera . Pasteurella multocida lipoprotein D has GP-PDE activity similar to other bacterial GlpQ, which is modulated by, but not dependent on, its N-terminal lipidation . Surface immunoprecipitation assays using intact P. multocida demonstrated that lipoprotein D was not surface-exposed and cannot be considered a virulence factor useful for vaccine design . Indeed, vaccine trials using both the purified nonlipidated protein and the lipidated recombinant protein did not stimulate protective immunity, although the mice had high antibody titers to lipoprotein D .
Mycoplasma GP-PDEs: evidence for their role in pathogenicity
Mycoplasma pneumoniae is the causative agent of atypical pneumonia, and comprises a mollicute that obtains the building blocks for its cellular macromolecules from the host tissue . Exogenous lipid metabolism is particularly relevant for Mycoplasma because the phospholipids that represent the majority of pulmonary surfactant have a major role in its nutrition. The M. pneumoniae genome encodes two potential GP-PDEs (MPN420 or GlpQ, and MPN566), although only GlpQ is functional (Fig. 1). MPN566 has no GP-PDE enzymatic activity, and inactivation of its gene does not result in any detectable phenotype. Inactivation of glpQ results in reduced bacteria growth, loss of hydrogen peroxide production and a complete loss of M. pneumoniae cytotoxicity towards HeLa cells. Mycoplasma pneumoniae GlpQ not only possesses GP-PDE activity, but also controls the expression of a set of genes that encode lipoproteins, the glycerol facilitator and a metal ion ABC transporter .
GP-PDE activity of the Mycoplasma hyorhinis GPD protein (Fig. 1) has also been reported . Mycoplasma hyorhinis is the major contaminant of tissue cultures  and is a pathogen that has been implicated in a variety of diseases in swine [56, 57]. The M. hyorhinis GPD sequence predicts a polypeptide of 241 amino acids that has no signal sequences or hydrophobic motifs common to membrane proteins. Nevertheless, M. hyorhinis GPD activity was detected almost exclusively in the membrane fraction . Interest in M. hyorhinis has increased since its detection in human gastric cancer tissues [58, 59]. By contrast to other human and animal mycoplasmas, M. hyorhinis is not only present on the surface of host cells, but also can invade and survive within nonphagocytic melanoma cells . Several enzymatic activities of M. hyorhinis contribute to the damage to the host cell membranes, which include nonspecific phospholipase A activity and potent GDP activity .
Other bacterial GP-PDEs
In the Gram-positive Bacillus subtilis, the cytoplasmic GP-PDE known as YqiK (Fig. 1) was proposed to have a role in osmoprotection . YqiK is regulated by the same operon (yqiHIK) as other hydrolytic enzymes, and high-salinity growth conditions induce the up-regulation of the transcription of this operon . Microorganisms actively adjust their turgor to the prevailing external osmolarity by dynamic modulation of the osmotic potential of their cytoplasm [62, 63]. Interestingly, deletion of the yqiHIK operon impairs the growth of B. subtilis at high salinity .
Other bacterial GP-PDEs show activity towards more complex substrates, even if their final product is always Gro3P. A GP-PDE described in cell-free extracts from the Gram-positive Bacillus pumilus, DSM27, hydrolyzes phosphodiester bonds between adjacent glycerol units. The reported substrates are polyglycerophosphates, such as purified cell-wall teichoic acid, as well as deacylated, unsubstituted lipoteichoic acid, di(glycerophospho)glycerol (deacylated cardiolipin) and mono(glycerophospho)glycerol .
A GP-PDE was identified in the Gram-positive bacterium Staphylococcus aureus (Fig. 1)  during the sequencing of the methicillin resistance determinant that is responsible for the intrinsic resistance of staphylococci to β-lactam antibiotics . The expression of this GP-PDE in S. aureus is constitutive, although its function has not been defined; however, it appears to be catalytically active because its expression can complement UgpQ in a UgpQ-deleted strain of E. coli.
Streptomyces have been extensively studied because they are the most proficient producers of naturally occurring therapeutic molecules, such as antibiotics, immunosuppressants and antitumor agents . The Streptomyces coelicolor genome encodes seven genes that are putative GP-PDEs, three of which are secreted (GlpQ1–3) and the other four cytoplasmic (UgpQ1–4) . To date, no information is available on their enzymatic activities .
In Enterobacter aerogenes, an operon similar to the E. coli ugp was identified in which every gene has an E. coli homolog other than the phosphodiesterase (GpdQ; Fig. 1) that is unrelated to the E. coli UgpQ [69, 70]. The crystal structure of the E. aerogenes GpdQ emphasizes its difference compared to all other bacterial GP-PDEs with respect to the absence of the conserved triosephosphate isomerase barrel fold in the catalytic site [71, 72]. Indeed, GpdQ is clustered separately in the phylogenetic tree of Fig. 1. GpdQ appears as a structurally novel GP-PDE and its secondary-structure prediction suggests that it more properly belongs to the α/β-sandwich metallo-dependent phosphoesterase family [71, 72].
GpdQ has been shown to have very broad substrate specificity because it can catalyze the hydrolysis not only of GroPEtn, but also of phosphomonoesters, diesters and triesters, in addition to phosphothiolates [71, 73]. Because it can hydrolyze several organophosphates, this GpdQ has been studied for the bioremediation of soil, through the detoxification of organophosphate pesticides and products of the degradation of nerve agents [69, 70]. Accordingly, through directed evolution, Yip et al.  enhanced the activity of GpdQ towards larger and nonphysiological substrates.
Crystal structures of bacterial GP-PDEs
There are several bacterial GP-PDEs with uncharacterized biochemical activities where their crystal structures have been partially solved. A common feature of this enzyme family is the presence of the classical triosephosphate isomerase barrel fold .
The first GP-PDE structure determined using X-rays was the TM1621 protein of the bacteria Thermotoga maritima . The crystallographic packing in the TM1621 structure suggests that the biologically relevant form is a monomer composed of 11 β-strands, 10 α-helices and four 310-helices. The structure of the T2047 GP-PDE of Agrobacterium tumefaciens reveals instead that it forms a hexamer and, in particular, a ‘trimer of dimers’, with a channel passing through the center of the assembly; this has never been observed in other members of the GP-PDE family .
The crystal structure of the Thermoanaerobacters tengcongensis GP-PDE, ttGDD, was obtained with a calcium atom chelated by three conserved residues and a glycerol molecule bound in the catalytic groove. This led to the proposal of a mechanism of catalysis through two reaction steps, with the glycerol and the phosphate moieties forming a cyclic phosphate intermediate that is stabilized by the calcium ion .
Saccharomyces cerevisiae GP-PDEs
Based on sequence similarities to E. coli GlpQ and on the identification of a specific GP-PDE motif (Pfam accession number PF3009) , YPL110c and YPL206c were identified as genes that encode putative GP-PDEs in S. cerevisiae . YPL110c encodes a cytoplasmic protein of 1223 amino acids with a molecular mass of 138 kDa (Gde1p; Fig. 1) . In a S. cerevisiae strain carrying the deletion of the YPL110c gene, the massive accumulation of GroPCho provided the first evidence that Gde1p is responsible for GroPCho hydrolysis, which is used as a phosphate source . Indeed, increased Gde1p expression was reported in microarray studies performed under low-phosphate conditions . In addition to the GP-PDE domain localized at the C terminus, Gde1p has also ankyrin repeats  and an SPX (from the SYG1, Pho81, XPR1 proteins) domain . When taken together with the observation that a number of proteins involved in phosphate homeostasis also contain N-terminal SPX domains , this observation would appear to be in agreement with the hypothesis that Gde1p could also act by binding potential partners involved in phosphate metabolism [79, 84].
Subsequent studies have better clarified the role of the product of the YPL206c gene. This encodes Pgc1p, which is a 321 amino acid protein with a predicted molecular mass of 37 kDa that belongs to the superfamily of phospholipase-C-like enzymes. Pgc1p controls the phosphatidylglycerol content of the cell membranes by cleavage of phosphatidylglycerol to diacylglycerol and glycerophosphate .
Glycerophosphodiester catabolism in yeast is strongly affected by nutrient availability. Both GroPIns and GroPCho catabolism are induced by phosphate limitation and the corresponding inositol and choline produced are shuttled into phospholipid biosynthesis. Thus, the combined actions of phospholipases, glycerophosphodiester transport and GP-PDE activities should provide the needed supply of inositol, choline, phosphate and glycerol for their recycling and use in different biochemical routes [86, 87].
No alteration in GroPIns metabolism is observed in either the YPL110c or YPL206c mutant strains, which suggests another gene that remains to be identified should determine the metabolic fate of GroPIns . However, the evidence that GroPIns can be used as a phosphate and inositol source suggests that there should be an enzyme with GroPIns phosphodiesterase activity [84, 86, 88, 89].
Glycerophosphodiesters in plants
GP-PDE activity with broad glycerophosphodiester substrate specificity was initially described in the cell wall and vacuole of plant cells; whereas compounds such as bis-p-nitrophenyl phosphate, ADP-glucose and ADP-ribose were not hydrolyzed [90, 91]. The initial enzyme identification came from Daucus carota , Acer pseudoplatanus , Nicotiana tabaccum  and Lupinus albus , whereas the more comprehensive characterization was performed in Arabidopsis thaliana . By genome analysis, 13 potential homologs were identified and subdivided into two groups: the first comprising proteins with only one GP-PDE domain (or canonical type A enzymes, AtGDPD1-6) and the second including proteins with two putative GP-PDE domains (type B, AtGDPDL1-7) . Expression studies indicated that some of these proteins have overlapping expression patterns suggesting functional redundancy, whereas others show distinguishable expression patterns with a potential specificity of action. The type A enzymes are more closely related to the bacterial GP-PDEs and their catalytic domain has the conserved crucial amionoacids for E. coli GlpQ function. Indeed, purified recombinant A. thaliana AtGDPD1 displays a Mg2+-dependent, GP-PDE catalytic activity toward GroPGro, GroPCho and GroPEtn . The type B enzymes are noncanonical GP-PDEs unique to plants. Recombinant A. thaliana AtGPDPL1 shows limited enzymatic activity toward glycerophosphodiesters, even if it is less efficient than AtGDPD1; in this case, noncanonical substrates cannot be excluded . In addition, other studies could not demonstrate any enzymatic activity for AtGDPDL3 (also named SHV3), further supporting the existence of possible nonglycerophosphodiester substrates for this class of enzymes . AtGDPDL3 was reported to play a role in pectin network formation at the plasma membrane and cell wall stability .
AtGDPDs genes are up-regulated by inorganic phosphate deprivation, whereas salt and osmotic stress induce up-regulation of AtGDPDL genes . In agreement with these findings, loss-of-function of the plastid-localized AtGDPD1 induced a decrease of GP-PDE activity, Gro3P and inorganic phosphate content, and seedling growth rate compared to the wild-type plant, indicating that this enzyme is devoted to the glycerophosphodiester degradation pathway as a source of inorganic phosphate .
In L. albus two GP-PDEs have been described; both were reported to be induced by phosphate deprivation, to possess a GP-PDE activity and to regulate root hair development and density, suggesting their role in plant acclimation to phosphate deprivation .
GP-PDE from Plasmodium falciparum
The relevance of the lipid scavenging and metabolic pathways that occur during the different phases of infection of host erythrocytes by the malaria parasite led Denloye et al.  to identify and characterize a GP-PDE expressed in this organism, PfGDPD, in their search for promising targets for anti-malaria drug development. PfGDPD is a 475-amino acid protein with a clear homology with bacterial GP-PDEs (Fig. 1). There are homologs of PfGDPD in other protozoans, such as Toxoplasma gondii and Cryptosporidium parvum . In P. falciparum-infected erythrocytes, PfGDPD is distributed in different host-cell compartments, which suggests that it has more than one function during malaria progression . To date, the only information available on the PfGDPD enzyme is that it is a Mg2+-dependent GP-PDE with similar kinetic parameters to E. coli UgpQ, and that different isoforms can exist . Indeed, PfGDPD appears to be clustered in the same group of the E. coli UgpQ in the phylogenetic tree shown in Fig. 1.
GP-PDE activity in Musca domestica
Hildenbrandt and Bieber  characterized the GP-PDE hydrolytic activity in the mature larval stage of insects, including in M. domestica . When the larvae enter the pupal stage, they undergo extensive breakdown of phospholipids and changes in their phospholipid composition. GP-PDE activity is considered to be essential during this phase of metamorphosis, when the organism extensively hydrolyzes its cellular constituents and reassembles the components into the tissues of the adult organism. This enzymatic activity has a pH optimum of 7.2, with GroPCho as the preferred substrate, although GroPEtn, GroPIns, GroPSer and GroPGly are also hydrolyzed to Gro3P and the free alcohols .
Mammalian GDE activity was initially characterized in tissue extracts from rat kidney [98-101], brain [102-104], liver [105, 106] and in uterine secretion . Subsequent to these studies, two different GDE species have been identified: one associated with the membrane fraction (mainly plasma membrane) and one in the cytosol [106, 108]. In rat extracts from different brain regions at different ages, GDE activity was shown to be specific for GroPCho, and to be regionally and developmentally regulated . In human temporal cortex extracts, two different GDE activities have been identified with different mechanisms of hydrolysis of the phosphodiester bond of glycerophosphodiesters: one produces choline-phosphate from GroPCho, and the other produces choline/ethanolamine from GroPCho/GroPEth . Subsequently, seven different mammalian GDEs were cloned (Fig. 1) . As indicated, all the mammalian GDEs are in the same group of E. coli UgpQ (Fig. 1). In this group, a first branch clusters together the cytosolic S. cerevisiae Gde1p and the unique mammalian GDE cytosolically-located GDE5. The second branch is further subdivided into a group comprising two mammalian GDEs located in internal membranes and with specific activity against glycerolysolipids instead of glycerophosphodiesters (GDE4 and GDE7), a second group comprising non-mammalian GP-PDEs, and a third group including GDE1 and the three serpentine GDEs (GDE2, GDE3 and GDE6), so-called for the presence of multiple membrane spanning domains in their structure (Fig. 1).
GDE1 enzymatic activity and regulation by G-protein-coupled receptors (GPCRs)
The first human GDE to be cloned was originally named Membrane Interacting protein of RGS16 (MIR16; then renamed GDE1) because it was found in two-hybrid screening using, as a bait, a regulator of G-protein signaling (RGS16) , which is a GTPase-activating protein for members of the Gαi family [112, 113]. This interaction suggested that GDE1 is regulated in some way by G-proteins or GPCRs. The connection with GPCRs was further confirmed by a second yeast two-hybrid screening performed by Bachmann et al. , who showed that GDE1 can bind to the PRAF2 protein, which is an interactor of the G-protein-coupled chemokine receptor CCR5 .
Sequence analysis and three-dimensional modeling indicated that human GDE1 shares significant homology and strong structural similarities with bacterial GP-PDEs , suggesting that the structure and molecular architecture were well preserved through evolution .
In vitro enzymatic characterization showed that GDE1 has a GroPIns phosphodiesterase activity that produces Gro3P and inositol in the presence of post-nuclear fractions of HEK293 cells overexpressing GDE1 (Table 1). This GDE1 activity was enhanced by the addition of 10 mm Mg2+ but not Ca2+ and was optimal at pH 7.5 . In competition experiments, a 10-fold excess of GroPCho had no significant effects on this GroPIns phosphodiesterase activity. However, GroPEtn and GroPSer blocked GDE1 activity by only approximately 30% and 80%, respectively, which suggested that GroPIns45P2 and GroPSer might also be GDE1 substrates (Table 1) . The GDE1 substrate specificity was verified by a global analysis of GDE1−/− mice using a MS-based metabolomics platform . These studies showed an absolute elevation in GroPSer levels that was more-or-less ‘mass-balanced’ by the reduction in the corresponding serine levels in the GDE1−/− mouse brain, which indicated that GDE1 catalyzes the hydrolysis of GroPSer (Table 1). GroPIns was also confirmed as a physiological substrate of GDE1 because it was profoundly elevated in the central nervous system of these GDE1−/− mice . GroPEtn and GroPCho showed similar levels in GDE1+/+ and GDE1−/− brains . Other studies have demonstrated that GDE1 can also hydrolyze substrates different from the glycerophosphodiesters, as in the case of the biosynthesis of anandamide (arachidonoyl ethanolamine), one of the endogenous ligands of cannabinoid receptors (Table 1) . This biosynthetic pathway does not appear to contribute to the in vivo production of this relevant neurotransmitter, anandamide, as verified in GDE1−/− mice, where anandamide production (in live neurons) and the levels of anandamide (measured by MS in mouse brain) were unaltered compared to wild-type mice . The cellular localization of GDE1 in rat kidney and liver indicated that it is an integral membrane glycoprotein that is localized at the plasma membrane (Fig. 2), and also at intracellular membranes . Membrane topology studies have suggested a model in which the catalytic GDE domain faces the lumen/extracellular space, and the C-terminus faces the cytoplasm . Thus, GDE1 can be classified as an ecto-, endoluminal enzyme that is active in these specific cell compartments.
Table 1. The mammalian glycerophosphodiesterases.
The original hypothesis that GDE1 can be regulated by GPCRs was confirmed in in vitro functional assays using GroPIns as substrate because its activity was increased by isoproterenol, an agonist for β-adrenergic receptors, and decreased by phenylephrine and lysophosphatidic acid, the ligands of the α-adrenergic receptor and the lysophospholipid receptors, respectively .
Multiple GDE2 functions: focus on the kidney and the nervous system
Human GDE2 (also named GDPD5) is widely expressed, with relative low levels in the kidney and prostate, and its gene was precisely mapped to 11q13.4-13.5, and contains 17 exons and 16 introns .
The intracellular localization of GDE2 was investigated via the overexpression of a tagged version of GDE2 in COS-7 , HEK293  and HeLa cells . In HeLa cells, GDE2 was shown to be located at the endoplasmic reticulum, as well as at the plasma membrane, depending on cell confluence, with a predominant plasma-membrane localization in confluent cells (Fig. 2) . Periz et al.  showed that GDE2 contains a 43-amino acid intracellular N-terminal region, six transmembrane domains, an intracellular C-terminal domain of 82-amino acid residues and two 13-amino acid intracellular connecting loops between the transmembrane domains.
GDE2 enzymatic activity has been shown to hydrolyse GroPCho, with the release of free choline and Gro3P (Table 1), as shown both with cell lysates and in mIMCD3 (intact mouse inner medullary collecting duct) cells . In these mIMCD3 cells, osmotic stress conditions resulting from high salt concentrations cause a decrease in the GDE2 mRNA half-life, with the consequent lowering of GDE2 protein levels and a decrease in GroPCho hydrolysis .
GroPCho acts as an osmosprotector by stabilizing intracellular macromolecules and by protecting inner-medullary cells against the perturbing effects of high NaCl and urea; therefore, by controlling the level of GroPCho in vivo, GDE2 acts as an osmoregulated enzyme . In addition, high extracellular salt and urea levels have no direct effects on GDE2 enzymatic activity but can post-translationally regulate and inhibit the GDE2 .
In addition to this osmoprotective GDE activity, there have been several reports correlating GDE2 function with neuronal differentiation. Indeed, GDE2 is necessary and sufficient to induce motor-neuron differentiation and its catalytic activity has an important role in this process . In vivo, GDE2 is primarily expressed by mature motor neurons and not in undifferentiated progenitors; ablating GDE2 expression in the spinal cord using small-interfering (si)RNAs results in the loss of post-mitotic motor neurons and an increase in cell death . In line with these observations, data have been obtained using Neuro2A mouse spinal cord neuroblastoma cells, a well-established model of neurite growth induced by retinoic acid treatment . Here, GDE2 up-regulation upon retinoic-acid treatment is sufficient to induce neurite formation that is blocked upon GDE2 down-regulation by siRNAs .
The different steps that involve GDE2 in the regulation of neuronal transcriptional programs have been defined in more recent studies. First, the antioxidant enzyme peroxiredoxin1 was identified as the enzyme involved in GDE2 activation . Yan et al.  demostrated that peroxiredoxin1 causes a reduction of the intramolecular disulfide bond formed between cysteine 25 and cysteine 576 of chicken GDE2, which is a link between the N-terminal and C-terminal domains of GDE2 that normally inhibits its function.
GDE2 regulation of motor neuron differentiation was shown to be partially mediated by its interaction with Gαi2 . In more detail, this functional interaction, which has a role in spinal motor-neuron development, occurs between Gαi2 in the inactive (GDP-bound) form and the C-terminal domain of GDE2 . This observation raised the possibility that proteins belonging to the GDE family can use similar components of the GPCR signal-transduction machinery to perform their functions.
GDE2 promotion of neurogenesis has revealed a different molecular mechanism compared to that postulated for GDE2 osmoregulation of kidney cells. For osmoregulation, GDE2 activity induces changes in intracellular GroPCho levels . The induction of motor-neuron differentiation is triggered by GDE2 through noncell-autonomous cleavage of the GPI anchor from the ‘reversion-inducing cysteine-rich protein with Kazal motifs’ (RECK) that is located in the neighboring cells (Table 1) . RECK normally activates Notch signaling in cortical progenitors by directly inhibiting metalloprotease-dependent processing of the Notch ligand Delta-like 1 [129, 130]. GDE2 activity promotes Delta-like 1 shedding and inactivation by cleavage of the GPI-anchor of RECK and, in this way, GDE2 regulates the timing of cortical progenitor differentiation . GDE2-mediated hydrolysis of the RECK GPI anchor is not a phospholipase D-like hydrolysis, which also suggests a different attack of the phosphodiester bond compared to that reported for the other GDE2 substrate, GroPCho (Table 1) . GDE1 was almost inactive under similar conditions, whereas some hydrolysis was observed for the GPI anchor of the heparan sulfate proteoglycans, glypicans 4 and 2, by GDE2 and the other serpentine GDEs: GDE3 and GDE6 (Table 1) .
Altogether, these data suggest a model in which GDE2 expressed in newly-formed cortical neurons provides a feedback signal to down-regulate the levels of Notch signaling in adjacent cortical progenitors, through the cleavage of the RECK GPI anchor, thereby inducing neuronal differentiation . Through the analysis of GDE2−/− mice, the specific neuronal area that is influenced by GDE2 activity was also characterized .
GDE3 induction of osteoblast differentiation
GDE3 (also named as GDPD2) was originally cloned using a RNA differential display procedure in MC3T3-E1 cells (clonal osteoblast cells derived from newborn mice calvaria) undergoing osteocyte differentiation . GDE3 appeared to be present at high levels only in differentiated cells, and kinetics studies indicated that GDE3 mRNA was specifically expressed in a premature stage of MC3T3-E1 development . Tissue GDE3 expression is concentrated not only mainly in the femur and calvaria, but also in the spleen .
Subsequent studies demonstrated a GroPIns-specific phosphodiesterase activity for GDE3 (Table 1), whereas no hydrolysis was observed with GroPIns4P, GroPCho, GroPEtn and GroPSer . As for other GP-PDEs, GDE3 is also a metal-dependent enzyme, although this differs from GDE1 because its activity is dependent on Ca2+, with an optimum at 5 mm. Compared to GDE1 and the bacterial GP-PDEs, GDE3 has a peculiar phospholipase C-like activity because it hydrolyzes GroPIns to glycerol and Ins1P (Table 1) .
GDE3 protein alignment with GDE1 and bacterial GP-PDEs showed conserved amino acids in the catalytic region, such as arginine 231, which is considered to be relevant for GP-PDE activity. Indeed, mutation of arginine 231 (i.e. GDE3R231A) leads to impaired GroPIns hydrolysis . Hydropathy and motif analysis of the GDE3 protein sequence showed a seven-transmembrane region of 161 amino acids, and the predicted N-linked glycosylation sites suggested the extracellular localization of the protein portion that contains the catalytic domain. This extracellular localization of the GDE domain was also confirmed by fluorescence-activated cell sorting analysis with HEK293 cells over-expressing GDE3, using a specific antibody that recognizes this domain as its epitope . In addition, heterologous GDE3 expression in HEK293 cells, or induction of MC3T3-E1 cells differentiation, led to increased phosphodiesterase activity measured in the extracellular medium, with no effects on the intracellular GroPIns pool, again supporting the extracellular activity of GDE3 (Fig. 2) [133, 134].
Under these GDE3 over-expression or induction-of-expression conditions, a major modification in cell morphology was observed. The expression of GDE3 showed its prevalent localization at the cell periphery and in lamellipodial extensions. Although wild-type HEK293 cells are generally spread, the expression of GDE3 induced cell rounding and disappearance of actin filaments, as visualized using rhodamine phalloidin [133, 134]. The GDE3 enzymatic activity appears to be necessary for these effects because the catalytically inactive GDE3R231A mutant did not induce these morphological changes [133, 134]. A similar cellular localization and actin cytoskeleton association was seen also in MC3T3-E1 cells , where GDE3 expression induced actin cytoskeleton disorganization, resulting in a clear disassembly of the stress fibers . The molecular mechanisms of the GDE3-dependent modulation of the actin cytoskeleton remain to be determined.
Although initially considered as a differentiation-regulated gene, GDE3 was later reported as an inducer of osteoblast differentiation . Stable expression of GDE3 in MC3T3-E1 cells appeared to accelerate the program of osteoblast differentiation and mineralization (such as increased alkaline phosphatase activity and increased calcium content), with the appearance of osteocyte markers and a decrease in stress fibers, as well as consequent morphological changes . GDE3 expression in MC3T3-E1 cells also induced a decrease in growth rate compared to wild-type osteoblasts, a process that appears to be mediated by increased GroPIns hydrolysis, which is in agreement with data showing GroPIns as an inducer of thyroid cell proliferation . These data indicate that increased GDE3 levels accelerate the program of osteoblast differentiation.
Enzymatic-activity-independent regulation of skeletal muscle development by GDE5
GDE5 (also named GPCPD1 or GDPD6) is highly expressed in the human brain, and a genome-wide association study suggested that GDE5 expression can contribute to variations in cortical surface area . Northern blotting showed GDE5 mRNA mainly in skeletal muscle, with down-regulation of its expression in several types of skeletal muscle atrophies induced by aging and denervation, as well as in diabetic mouse models . This decreased GDE5 expression might represent an adaptation response to counteract the pathology .
By contrast to all of the other mammalian GDEs, GDE5 localizes to the cell cytoplasm, as also confirmed under overexpression conditions in HEK293 cells (Fig. 2) . Mouse GDE5 from postnuclear fractions of transfected HEK293 cells, or purified from virus-infected SF9 cells, is a Mg2+-dependent GDE that can hydrolyze GroPCho and GroPEtn, whereas no activity was detected with GroPGro, GroPIns and GroPSer as substrates (Table 1) .
The functional activity of GDE5 has been investigated in models of myogenic differentiation, such as C2C12 mouse myoblasts. In this model system, siRNA-mediated GDE5 knockdown resulted in the induction of differentiation, with myotube formation and the appearance of myogenic differentiation markers . In parallel, stable expression of GDE5 in the same cellular model suppressed myogenic differentiation. The unexpected finding was that a catalytically inactive mutant of GDE5, obtained by truncating its sequence to form a deletion mutant (amino acids 1–470; GDE5ΔC471), has behavior similar to the wild-type protein, emphasizing that the GDE5 effect on myogenic differentiation is not related to its GDE enzymatic activity . The molecular mechanisms behind this phenomena are still under investigation. Finally, transgenic mice have been produced that specifically overexpress GDE5ΔC471 in their skeletal muscle . These GDE5ΔC471 transgenic mice showed smaller-sized skeletal muscles, with decreased expression levels of genes related to structural proteins of type II muscle fibers compared to control mice . All of these data indicate that GDE5 negatively regulates skeletal muscle development in an enzymatic-activity-independent manner.
Recently characterized GDEs
Three other mammalian GDEs have been recently cloned (Fig. 1) but, to date, little information is available on their cellular distribution and enzymatic activity.
The human GDE4 (also named GDPD1) gene maps to 17q22 and contains 10 exons and nine introns. The cDNA sequence codes for a protein of 314 amino acids that contains a GDE domain, two putative transmembrane regions and an N-terminal signal peptide .
Information on GDE4 intracellular localization has come from overexpression studies, in which GFP-GDE4 was transfected into HEK293 and neuroblastoma N2a cells. GDE4 was mainly localized in the perinuclear region and the cell periphery, and appeared to be membrane bound (Fig. 2). Its expression did not induce any evident phenotype or changes in cell morphology .
GDE4 does not show any activity on classical glycerophosphodiester substrates, such as GroPIns or GroPCho, although it has lysophospholipase D activity toward lysophospholipids (such as lysophosphatidylcholine), which suggests a role in the modulation of the levels of lysophosphatidic acid, an important mediator in a variety of cellular events, such as cell proliferation and platelet activation (Table 1; N. Yanaka, N. Ohshima and T. Kudo, unpublished data).
GDE6 (also named GDPD4) is another less-studied GDE, its ORF of 1899 nucleotides encodes a protein of 633 amino acids, which contains seven transmembrane regions, as determined by hydropathy analysis .
Tissue localization studies have indicated that GDE6 mRNA is predominantly or solely observed in testis, and in situ hybridization analysis has revealed that GDE6 mRNA is detected in testicular germ cells at the spermatocyte stages. Intracellular GDE6 localization was monitored under overexpression conditions in COS-7 cells transfected with GFP-GDE6. The fluorescent signal appeared to be punctate and localized in the perinuclear region, with no significant change in COS-7 cell morphology under these conditions (Fig. 2) .
Recent data on mouse GDE7 have been obtained showing that GDE7 (also named GDPD3) has 53.4% amino acidic identity with GDE4, and a similar transmembrane organization and intracellular localization (Fig. 2; N. Yanaka, N. Ohshima and T Kudo, unpublished data). Similar to GDE4, GDE7 also has lysophospholipase D activity toward lysophospholipids (Table 1; N. Yanaka, N. Ohshima and T Kudo, unpublished data).
By contrast to the bacterial GP-PDEs that have been studied for decades, the mammalian GDEs are relatively recently identified proteins. For this reason, questions remain concerning the functional role of some of them and, in particular, for GDE4, GDE6 and GDE7, the most recently identified members of the GDE family. An intriguing aspect of these studies is the GDE substrate specificity. Originally, only glycerophosphodiesters were considered to be hydrolyzed by mammalian GDEs, whereas it is now apparent that GDEs also have lysolipase-like activities, and that new substrates are continuously being identified. Further investigations will be required to solve the topological localization of the GDE enzymatic activities because some of these GDEs have both intracellular and extracellular activity towards vicinal cells. These studies will be motivated by the great number of patho/physiological functions that have been reported to be dependent or regulated by mammalian GDEs, and these investigations will certainly result in the possibility of pharmacological exploitation.
We thank Dr A. Fontana and Dr A. Cutignano (Istitute of Biomolecular Chemistry, National Research Council, Naples) for discussion and help with the MS analysis, as well as Dr C. P. Berrie for editorial assistance. The cited work in the authors laboratories was supported by PON projects: 01_00862/1 and 01_00117, the Italian Association for Cancer Research Grant IG10341 (Milan, Italy), grant FIT DM 24/09/2009, Legge 46/82 and FaReBio di Qualità Ministry of Economy and Finance.