- Made available online as an Accepted Preprint 8 January 2009
- Accepted Preprint first posted online on 8 January 2009
G-protein-coupled receptors, cholesterol and palmitoylation: facts about fats
- Cellular and Molecular Pharmacology Section, CNR Institute of Neuroscience, Via Vanvitelli 32, 20129 Milan, Italy1Department of Experimental Medicine, University of Milano-Bicocca, Monza, Italy
- (Correspondence should be addressed to B Chini; Email: b.chini{at}in.cnr.it)
Abstract
G-protein-coupled receptors (GPCRs) are integral membrane proteins, hence it is not surprising that a number of their structural and functional features are modulated by both proteins and lipids. The impact of interacting proteins and lipids on the assembly and signalling of GPCRs has been extensively investigated over the last 20–30 years, and a further impetus has been given by the proposal that GPCRs and/or their immediate signalling partners (G proteins) can partition within plasma membrane domains, termed rafts and caveolae, enriched in glycosphingolipids and cholesterol. The high content of these specific lipids, in particular of cholesterol, in the vicinity of GPCR transmembranes can affect GPCR structure and/or function. In addition, most GPCRs are post-translationally modified with one or more palmitic acid(s), a 16-carbon saturated fatty acid, covalently bound to cysteine(s) localised in the carboxyl-terminal cytoplasmic tail. The insertion of palmitate into the cytoplasmic leaflet of the plasma membrane can create a fourth loop, thus profoundly affecting GPCR structure and hence the interactions with intracellular partner proteins. This review briefly highlights how lipids of the membrane and the receptor themselves can influence GPCR organisation and functioning.
G-protein-coupled receptors–cholesterol interactions
Eighty-ninety per cent of whole-cell cholesterol is contained within the plasma membrane, where it accounts for 20–25% of lipid molecules (Lange 1991, Simons & Ikonen 2000). Cholesterol, which is amphiphilic and has a small −OH head group and a rigid, planar four-ring backbone, influences the functional properties of the plasma membrane by interacting with other resident lipids (Ohvo-Rekila et al. 2002, Ikonen 2008) and proteins (Lee 2004). In particular, it may act on the conformation of membrane proteins: i) by alterations in the physico-chemical properties of the bilayer; ii) through specific and localised molecular interactions; or iii) by a combination of both.
Alterations in the physico-chemical properties of the bilayer
The effects of cholesterol on the fluidity and rigidity of the plasma membrane are well known: a cholesterol molecule can span approximately half a bilayer, preferentially interacting with the saturated fatty chains of phospholipids. By increasing the packing, and hence the lateral assembly of phospholipids, cholesterol reduces membrane fluidity and increases rigidity (Mouritsen & Zuckermann 2004). The ‘rafts/caveolae signalling hypothesis’ suggests that plasma membrane domains enriched in cholesterol and sphingolipids compartmentalise the proteins involved in specific signalling tasks. By regulating lateral mobility and compartmentalisation, cholesterol affects the probability with which receptors interact with their downstream signalling partners, thus optimising their spatial–temporal interactions, hence signalling efficiency and regulation (Simons & Toomre 2000). The trafficking of receptors between regions with different lipid compositions is still unclear, but an interesting hypothesis has recently been advanced for the human δ-opioid receptor based on results obtained with plasmon-waveguide resonance spectroscopy experiments on solid-supported bilayers (Alves et al. 2005). It has been found that the unbound inactive receptor is enriched in phosphatidylcholine bilayers, whereas the activated receptor is concentrated in sphingomyelin bilayers, characterised by a greater thickness. Since it is known that receptor activation involves changes in the orientation of transmembrane helices with an increase in receptor vertical length (i.e. perpendicular to the plane of the plasma membrane; Salamon et al. 2000), it is hypothesised that the driving force for receptor redistribution comes from the increased hydrophobic match of the elongated receptor for the thicker sphingomyelin-enriched bilayer, a sorting mechanism that has previously been observed in other subcellular compartments (Jensen & Mouritsen 2004, Lee 2004).
Specific and localised molecular interactions
Some membrane proteins directly bind cholesterol, as firstly shown for caveolin-1 (Murata et al. 1995), whose cholesterol-binding domain has not yet been precisely mapped, even if its scaffolding domain, originally considered to be a protein interaction domain, is now thought to play a role in cholesterol binding (Parton et al. 2006). Some membrane proteins involved in the transport of cholesterol possess a conserved five-transmembrane helix domain called the sterol-sensing domain, which bind cholesterol with a measured affinity of around 100 nM (Radhakrishnan et al. 2004). Others bind cholesterol via a conserved hydrophobic binding pocket, such as OSBP (Suchanek et al. 2007) and START proteins (Alpy & Tomasetto 2005), or short cholesterol-binding motifs (Li & Papadopoulos 1998). The fact that cholesterol plays a role in stabilising the conformation of specific G-protein-coupled receptors (GPCRs) has become clear after the first attempts to purify them for pharmacological studies or X-ray diffraction analysis (Schertler & Hargrave 2000, Albert & Boesze-Battaglia 2005). As first noted for rhodopsin, cholesterol greatly improves crystal quality and quantity; furthermore, in the resolved structure of metarhodopsin I, it was found that a cholesterol molecule sits in a pocket formed between adjacent helices of a receptor dimer, thus suggesting that it may help to form or stabilise receptor dimers (Ruprecht et al. 2004). As further discussed below, it has also been observed that cholesterol contributes to dimer formation in the recently resolved structures of the β2-adrenergic receptor, thus further supporting its direct role in regulating GPCR conformation (Cherezov et al. 2007).
GPCR palmitoylation: sites and dynamics
Glycosylation and phosphorylation are certainly the best known post-translational modifications affecting GPCR functions, but lipids are also post-translationally added to most GPCRs in the form of the 16-carbon saturated fatty acid, palmitic acid. One more ‘exotic’ lipidation has been reported for the human prostacyclin receptor, which, in addition to being palmitoylated, has a farnesyl isoprenoid chain attached to the cytoplasmic carboxyl terminal (Miggin et al. 2003).
Cysteines are by far the most frequent acceptor sites of palmitoylation, which occurs on the thiol side chain through a thioester bond. However, it must be kept in mind that, although not yet reported for GPCRs, the addition of palmitic acid to proteins can also occur onto the terminally exposed amides of cysteine (Pepinsky et al. 1998) and glycine (Kleuss & Krause 2003), the hydroxyl group of threonine (Branton et al. 1993) and serine (Jing & Trowbridge 1990, Percherancier et al. 2001), and the ϵ-amino group of lysine via an amide bond (Stanley et al. 1998). In addition, other fatty acids, such as palmitoleate, stearate or oleate, can, much less commonly, bind to cysteines (Liang et al. 2004).
Rhodopsin was the first GPCR found to be palmitoylated at two adjacent cysteines in the carboxyl-terminal cytoplasmic tail (Ovchinnikov et al. 1988) but, since then, many other GPCRs have been experimentally shown to be mono-, bis- or tris-palmitoylated at conserved carboxyl-terminal cysteine residues (see Qanbar & Bouvier 2003 for a review), and yet other GPCRs containing the canonical cysteines are presumably palmitoylated. Interestingly, the μ-opioid receptor is not palmitoylated at the two conserved carboxyl-terminal cysteines, and cysteine-170 in the second intracellular loop (the only other intracellular cysteine) can be the palmitoylation site; however, as its substitution by alanine generates a mutant with very low expression levels, it is has not been possible to measure [3H]palmitate incorporation (Chen et al. 1998).
Some GPCRs are not modified by palmitic acid, including the α-isoform of the thromboxane A2 receptor, which lacks any cysteines in its carboxyl terminal (Reid & Kinsella 2007), and the GnRH receptor, which is unique among class A GPCRs as its cytoplasmic carboxyl terminal tail is only one to two amino acids long and therefore lacks both palmitoylation and phosphorylation sites (Navratil et al. 2006).
The thioester bond linking palmitate to cysteine is labile, thus making the attachment of palmitate to proteins reversible and the turnover rate of the bound fatty acid generally shorter than that of the protein. The addition and removal of palmitate are respectively enzymatically catalysed by palmitoyltransferases (PATs) and thioesterases. Identifying these enzymes proved to be a very challenging task, and it is only recently that a family of proteins with PAT activity has been identified in yeast by means of genetic screening. They all share a common motif referred to as ‘DHHC’, which is enriched in cysteines and contains a conserved aspartate–histidine–histidine–cysteine signature (see Mitchell et al. 2006 for a review). So far, one enzyme (APT1) has been purified from rat liver cytosol that removes palmitate from certain proteins on the cytosolic surface of membranes (Duncan & Gilman 1998). The non-enzymatic palmitoylation of proteins has also been reported, including membrane-bound rhodopsin (Veit et al. 1998).
As it is a dynamic process, the extent of palmitoylation is subject to regulation. In the case of GPCRs, this often occurs as a result of agonist-induced activation: e.g. isoproterenol stimulation of the β2-adrenergic receptor increases [3H]palmitate incorporation (Mouillac et al. 1992). This increased receptor labelling can be interpreted in two ways: i) agonist binding increases the proportion of receptor being palmitoylated, or ii) agonist binding increases the turnover rate of receptor-linked palmitate, hence it favours the exchange of unlabelled for tritiated palmitate. The latter explanation, which implies an agonist-induced acceleration of depalmitoylation, has been experimentally supported in relation to the β2-adrenergic receptor (Loisel et al. 1999) and other GPCRs, including dopamine D1 (Ng et al. 1994), α1B- (Stevens et al. 2001) and α2A-adrenergic (Kennedy & Limbird 1994), serotonin 4A (Ponimaskin et al. 2002) and δ-opioid receptors (Petaja-Repo et al. 2006). However, it is not true for every GPCR: e.g. the palmitoylation of serotonin 1A receptors is not at all affected by agonist exposure (Papoucheva et al. 2004).
Activation by an agonist is not the only way in which the GPCR palmitoylation rate can be affected. Adam et al. (1999) have shown that high nitric oxide levels reduce both the basal and the isoproterenol-induced incorporation of [3H]palmitate into β2-adrenergic receptors.
How to study cholesterol–GPCR interactions and palmitoylation: a few technical tips
Assaying the effects of cholesterol has proved to be relatively easy since the introduction of cholesterol-depleting agents, such as cyclodextrins (Christian et al. 1997, Zidovetzki & Levitan 2007), and cholesterol synthesis inhibitors, such as statins (Brown & Goldstein 1980). However, such procedures always have both direct and indirect effects, due to the alterations of the physico-chemical properties of the plasma membrane and of membrane domain organisation. On the contrary, measuring direct cholesterol binding to membrane proteins, in general (and GPCRs in particular), is an extremely daunting task. Tritiated steroids can be used to measure cholesterol affinity directly in saturation and/or competition assays usually involving purified proteins, as can be seen from the very elegant work recently done to characterise cholesterol binding to the binding domain of SCAP (Radhakrishnan et al. 2004) and NCP1 (Infante et al. 2008a,b). Photo-activatable cholesterol can also be used to map protein–cholesterol interactions, as originally described in Thiele et al. (2000); in this case, binding can be performed in vivo but, after labelling, an immunoprecipitation or a pull-down step is necessary to detect the incorporation of the reactive steroid in the protein of interest. Finally, the proximity of cholesterol to the membrane nicotinic acetylcholine receptor was deduced from a series of experiments designed to quench the fluorescence of tryptophan residues in the receptor using spin-labelled cholesterol and phospholipid analogues (Jones & McNamee 1988).
The conventional approach to assessing the palmitoylation of proteins (including GPCRs) involves the metabolic labelling of cultured cells expressing the protein of interest with [3H]palmitate, followed by immunoprecipitation, SDS gel electrophoresis and fluorography. The latter is necessary in order to increase the low detection limit of [3H], which takes a very long time (weeks or even months) of gel exposure on films. Maximum sensitivity can be obtained by soaking the gels in 2,5-diphenyloxazole, according to the procedure described by Bonner & Laskey (1974), and the signal can be further enhanced by shrinking the gels in polyethylene glycol 2000 or 4000 (Magee et al. 1995).
Palmitoylation sites can be identified by replacing predicted cysteines with alanines, glycines or serines using site-directed mutagenesis. However, serine is not ideal because its use as an alternative palmitoylation site via an ester linkage has been documented in membrane-associated proteins such as the human transferrin receptor (Jing & Trowbridge 1990) and even a GPCR, the CCR5 receptor (Percherancier et al. 2001).
Some studies designed to clarify the functional consequences of GPCR palmitoylation have used 2-bromopalmitate to inhibit the reaction (see for example Petaja-Repo et al. 2006), as initially suggested by Webb et al. (2000). However, caution is necessary because inhibition by 2-bromopalmitate extends to all palmitoylated proteins (including the immediate G-protein α-subunit partners of GPCRs), and thus it may well affect their functions as well.
Cholesterol regulation of GPCR function
Cholesterol depletion experiments have shown that it is very difficult to distinguish the direct and indirect effects of cholesterol on receptor function. Although the effects of cholesterol depletion on several functions have been reported for many GPCRs (Chini & Parenti 2004, Pucadyil & Chattopadhyay 2006, Patel et al. 2008), here we describe only a few GPCRs for which direct and specific effects of cholesterol have been clearly documented.
In the case of the human oxytocin receptor (OTR), it has been shown that high-affinity agonist (but not antagonist) binding is highly dependent on the cholesterol content of the plasma membrane (Klein et al. 1995). The specificity of cholesterol was determined by means of structure–activity analyses of a large number of steroids replacing cholesterol and, as only some were capable of restoring the affinity of OTR in cholesterol-depleted membranes, the authors concluded that indirect effects on the physico-chemical properties of the membrane could be excluded (Gimpl et al. 1997). Modulation of ligand binding affinity has also been shown in the case of the galanin receptor, the specificity of which was again supported by experiments in which only a limited number of steroids could restore agonist affinity to cholesterol-depleted membranes (Pang et al. 1999). Interestingly, biochemical and fluorescence recovery after photobleaching experiments have shown that only a small fraction of OTRs seem to be located in lipid rafts (Gimpl & Fahrenholz 2000, Reversi et al. 2006); as this receptor fraction is capable of activating specific signalling events (the stimulation of cell growth), cholesterol may be involved in modulating very selective receptor-induced signalling events (Guzzi et al. 2002, Rimoldi et al. 2003). However, molecular modelling and mutagenesis studies have thus far failed to map any directed binding sites in the transmembrane regions (Politowska et al. 2001), and the cholesterol-binding region of receptor has not yet been identified.
Functional consequences of palmitoylation
It has been shown that all of the properties of GPCRs are affected by the absence of palmitoylation, but the effects depend on the type of receptor.
Receptor maturation and membrane delivery
It has been reported that the lack of palmitoylation reduces the number of receptors reaching the cell surface in various GPCRs, including dopamine D1 (Ng et al. 1994), vasopressin V2 (Schulein et al. 1996, Sadeghi et al. 1997), TSH (Tanaka et al. 1998), chemokine CCR5 (Percherancier et al. 2001), histamine H2 (Fukushima et al. 2001) and δ-opioid receptors (Petaja-Repo et al. 2006). This suggests that early palmitate binding, presumably at the level of the endoplasmic reticulum, is required for the correct processing of a number of immature and mature GPCRs, as well as for receptor progression along the biosynthetic route from synthesis to plasma membrane localisation.
Compartmentalisation of the serotonin 1A receptor within sphingolipid/cholesterol-rich membrane domains (lipid rafts and caveolae; see Chini & Parenti 2004 for a review), measured as the association with Triton X100-insoluble floating fractions (TIFFs), also depends on palmitoylation (Renner et al. 2007). However, this is not a general rule because the GnRH receptor, which lacks palmitoylation sites (see above), is constitutively associated with TIFFs (Navratil et al. 2006). Interestingly, fusion of the palmitoylated carboxyl terminal of the LH receptor (normally excluded from lipid rafts) to the GnRH receptor removes the chimeric GnRH/LH receptor from TIFFs (Navratil et al. 2006), whereas fusion of the palmitoylated carboxyl terminal from non-mammalian GnRH receptors preserves the TIFF localisation, thus suggesting that neither the extension of the carboxyl terminal extension nor palmitoylation are signatures for raft targeting.
Coupling to G protein and signalling
Palmitoylation can markedly influence receptor coupling to G proteins, and hence the intracellular strength of signalling, or have no effect. The agonist-induced stimulation of adenylyl cyclase by dopamine D1 (Jin et al. 1997) and serotonin 4A receptors (Ponimaskin et al. 2002) is not affected by mutated palmitoylation sites, whereas the isoproterenol stimulation of adenylyl cyclase mediated by the palmitoylation-negative C341G mutant of the β2-adrenergic receptor is greatly reduced. This and the decreased ability of C341G to form a guanyl nucleotide-sensitive, high-affinity state for agonists indicate that palmitate attachment is necessary for efficient coupling between receptor and Gs protein (O'Dowd et al. 1989). Later studies offered insights into the mechanism underlying such uncoupling by showing that C341 is constitutively highly phosphorylated, and thus behaves as a wild-type receptor desensitised by the agonist (Moffett et al. 1993). The increased phosphorylation is due to the greater accessibility of C341 to protein kinase A in comparison with the wild-type receptor (Moffett et al. 1996). These findings, together with the observation that the palmitoylation of cysteine-341 is dynamically regulated by the agonist (Loisel et al. 1999), highlight the key role played by the addition/removal of palmitate in regulating the phosphorylation of β2-adrenergic receptors and, hence, their signalling efficiency.
Palmitoylation can also ‘direct’ pleiotropic GPCRs to signal preferentially through a certain pathway. This has been shown in the case of endothelin receptor A, in which mutation of the palmitoylation sites does not affect the agonist stimulation of adenylyl cyclase but impairs phospholipase C activation, thus suggesting that the presence of palmitate regulates the coupling to Gq, but not to Gs (Horstmeyer et al. 1996). Likewise, the V2 receptor-linked MAPK (but not the adenylyl cyclase) pathway is affected by the palmitoylation state of the receptor (Charest & Bouvier 2003).
Desensitisation and internalisation
As agonist stimulation usually promotes GPCR desensitisation and internalisation (as well as signalling), it is not surprising that palmitoylation also regulates these processes as mentioned above in the case of the β2-adrenergic receptor. However, different mechanisms apply to the various receptors. It has been reported that the palmitoylationless V2 receptor undergoes agonist-promoted internalisation less efficiently than the wild-type receptor (Charest & Bouvier 2003), and so the palmitoylation state modulates both signalling (MAPK) and endocytosis. The reduced internalisation positively correlates with less recruitment of β-arrestin2 to the non-palmitoylated receptor. Interestingly, the phosphorylation state of the mutant is not affected, thus suggesting that palmitoylation per se regulates the receptor's affinity for β-arrestin.
The picture is completely different in the case of the serotonin 4A receptor (Ponimaskin et al. 2005), which has a complex palmitoylation pattern at two adjacent cysteines (328 and 329) and a more distal cysteine-386. The combined substitution of cysteines 328 and 329 with serines generates a constitutively active mutant that is hyperphosphorylated and more effectively desensitised and internalised than the wild-type receptor through a β-arrestin2-dependent mechanism. On the contrary, the C386S and C328/329/386S mutants do not show any constitutive activity despite being as highly phosphorylated as the C328/329S mutant. In addition, the agonist-induced desensitisation and internalisation of single and triple mutants are less efficient than those of the double mutant. There is thus a complex interplay between site-dependent palmitoylation and phosphorylation and their related functional consequences, while not forgetting that the palmitoylation state of serotonin 4A receptor is dynamically regulated by agonist activation (Ponimaskin et al. 2001).
In essence, palmitoylation can do everything to GPCRs, and no predictions can be made concerning its functional relevance for a particular GPCR until experimental proof is obtained. The logical question to ask is how the attachment of palmitate(s) to the same domain of GPCRs can subserve so many different roles. Is the impact of palmitoylation influenced in any way by the length of the fourth intracellular loop formed upon the membrane insertion of palmitate and/or by the presence of other post-translational modification(s) (i.e. phosphorylation) within the loop? Does the presence of one or more palmitate(s) make any difference, by forming one or more loop(s) or making the loop(s) more or less flexible? These questions need to be addressed in a systematic way, although the task is extremely challenging given the number of different variables to take into account.
What is certain is the fact that, as was suggested by Miggin et al. (2003), ‘looping’ of the carboxyl terminal (Fig. 1) is important in the case of the prostacyclin receptor. Its tail can form two loops, one proximal to the seventh α-helix and reversibly anchored to the membrane by palmitates bound to cysteines 308 and 311, and a more distal one formed by the stable membrane insertion of a farnesyl isoprenoid bound to cysteine-383. Depending on which loop is present, the receptor signals through different effectors, i.e. adenylyl cyclase and/or phospholipase C.
Some examples of lipidation patterns affecting the carboxyl terminals of GPCRs. The insertion of lipid anchors (palmitoyl and/or farnesyl) into the plasma membrane can cause the formation of one or more intracellular loops that change the topology of carboxyl terminals, thus affecting interactions with receptor partner proteins. Palmitate-dependent loops can be disrupted by agonist-promoted depalmitoylation, whereas an isoprenyl loop (IP receptor) is stable. TPβ, the β-isoform of the thromboxane A2 receptor; V2, vasopressin V2 receptor; IP, prostacyclin receptor.
Palmitoylation and cholesterol: filling a structural gap?
Moench et al. (1994) used fluorescent derivatives of palmitate incorporated into rhodopsin to show that the two fatty acyl chains reside within the membrane and are not held in the protein core (Moench et al. 1994). This finding induced them to hypothesise that the insertion of palmitate into the lipid bilayer creates a fourth intracellular loop between the cytoplasmic end of α-helix 7 and fatty acid anchors. The X-ray crystal structure of rhodopsin has indeed confirmed that there is a fourth intracellular loop consisting of an α-helix (α-helix 8) that lies parallel to the plane of the membrane and is kept in position by the palmitate anchor which thus plays an important structural role.
In addition to a structural role inside monomeric receptors, palmitate anchors may play an important role in forming/stabilising receptor oligomers; as in the recently resolved X-ray structures of β2-adrenergic receptors, it has been found that mono-palmitoylated receptor moieties are assembled in supramolecular complexes by interactions occurring via their lipidic anchors (Cherezov et al. 2007). Notably, two cholesterol molecules were intercalated within the palmitic residues, and contribute to filling the adjacent protomer interfaces (Fig. 2). On the basis of these observations, it seems that GPCR dimerisation essentially takes place as a result of lipid-mediated interactions. Although these findings may not correspond to the in vivo situation (cholesterol is added during the purification procedure to help the crystallisation process), it is clear that the role of palmitoylation and cholesterol in oligomerisation is an intriguing question that needs to be carefully examined.
Schematic of palmitate- and cholesterol-mediated interfaces deduced from β2-adrenergic receptor crystals. In the recently resolved X-ray structures of β2-adrenergic receptors (Cherezov et al. 2007), receptor dimers appear to be organised in complexes whose major interactions are mediated by palmitate residues, bridged by cholesterol molecules; here, depicted in yellow.
Thus far, the effect of palmitoylation on oligomerisation has only been investigated in serotonin 1A receptors by Kobe et al. (2008), who used the fluorescence resonance energy transfer (FRET) technique in murine N1E-115 neuroblastoma cells transiently co-transfected with receptor chimeras fused to cyan and yellow fluorescent proteins. They observed that the agonist decreased FRET intensity in wild-type receptors, but increased it in palmitoylation-negative receptors. These shifts from negative to positive FRET have been interpreted as changes in the way in which palmitoylated or non-palmitoylated receptors are spatially oriented within oligomers. As palmitoylation positively regulates the number of serotonin 1A receptors residing within rafts (Renner et al. 2007), it can also be suggested that raft oligomers differ from non-raft oligomers, but the functional implications of these findings have still to be ascertained.
Final remarks
Despite years of effort, a number of key aspects of GPCRs are still a matter of investigation. As the active and inactive states of heterotrimeric complex structures have now been solved, the G-protein cycle is fairly well documented and understood; however, there are still no conclusive data concerning the dynamics of the association between GPCRs and the G-protein α subunits, and their physical dissociation during the activation process is still a matter of debate. Similarly, the specificity of receptor coupling to different G proteins has not been definitely mapped. There is also controversy concerning the monomeric/oligomeric GPCR state that may best accomplish G-protein docking and subsequent activation. Palmitoylation and/or cholesterol binding may play a more significant role than previously thought in all of these events, and more exhaustive studies of GPCR–lipid interactions may well contribute to exciting insights and advances in these fields.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.
Funding
This study was supported by grants from the Italian Association for Cancer Research (AIRC 2006) and Fondazione Cariplo (grant 2006.0882) to B Chini, and a FIRB grant (RBIN04CKYN) from the Ministry for Universities and Research (MIUR) to M Parenti.
- Revision received 17 November 2008
- Accepted 8 January 2009
- © 2009 Society for Endocrinology
