The term moonlighting protein is used to describe multifunctional proteins in which several functions can be found in a single strand of amino acids unrelated to splicing, posttranslational changes, and so forth (1–4). It is possible that unstructured domains of proteins may give them moonlighting properties (3). It is of substantial interest that the glycolytic enzyme, glyceraldehyde-3-phosphate dehydrogenase, possesses different functions, which are dependent on the state of oligomerization and its localization (3–7). Recently, the nuclear transport receptor has been shown to go moonlighting (8, 9). Thus, the importin-β-like transport receptor is involved not only in selective protein transport into the nucleus but also in transport of the motor protein Kif17 into primary cilia.
Summing up, moonlighting proteins (MPs) are very special multifunctional proteins, because they perform multiple autonomous, often unrelated, functions without partitioning them into different protein domains. These multiple functions can differ markedly, because it has been shown that they can carry out in some instance a structural function and in some other instance a catalytic function (10).
It has been proposed (11) that the moonlighting phenomenon is a side consequence of the Jacob's principle of the tinkerer's way of performing evolution (12), which means that novel functions can be carried out by adapting the existing devices.
MPs may have several interesting features:
(a)It is possible that a MPs can have more than one domain capable of multiple unrelated functions;
(b)Within the same strand of amino acids (motif), we should distinguish functions involved in the structural assemblage of multimeric complexes from catalytic functions when conformational changes develop;
(c)Within the same motif, the conformational change may in one instance be aimed to produce a structural assemblage of multimeric complexes leading to the formation of allosteric pathways and in another instance not (13, 14).
A final aspect to be considered is the potential modulation by the physicochemical conditions of the microenvironment, where the protein is located to favor moonlighting of one function versus another one.
Our article suggests that the moonlighting concept can be applied to the G protein coupled receptors (GPCRs) since, obviously, they can carry out different functions, for example, as chaperone proteins or as receptors. This is the case of gamma-aminobutyric acid receptor B (GABAB) receptors (15–17).
Thus, some features of the diverse functions of GPCRs can be discussed in the perspective of moonlighting as outlined above. It is even possible GPCRs may have several domains with independent moonlighting functions. In particular,
(a)The same motifs in, for example, third intracellular loop (ICL3), can moonlight by switching between receptor–receptor interactions and interactions with signaling proteins such as G proteins or calmodulin.
(b)Motifs in transmembrane domains can moonlight simply by switching from involvement purely in the structural assemblage of GPCRs into multimeric complexes without allosteric interactions to involvement with the formation of allosteric pathways leading to receptor–receptor interactions.
AAR, alanine-alanine-arginine; RAA, arginine-alanine-alanine; AAE, alanine-alanine-glutamic acid; EAA, glutamic acid-arginine-arginine; BRET, Bioluminescence Resonance Energy Transfer; FRET, Fluorescence Resonance Energy Transfer; GABAB, gamma-aminobutyric acid receptor B; CNS, Central Nervous System; MAPK, Mitogen-activated protein (MAP) kinases; RAS, ras protein; RAF, RAF proto-oncogene serine/threonine-protein kinase; AKT, Protein kinase B; PLC, Phospholipase C; JAK, Janus kinase; STAT, Signal Transducer and Activator of Transcription; IKK, IκB kinase; NF-KB, nuclear factor kappa-light-chain-enhancer of activated B cells; GPCR-RTK, G Protein Coupled Receptor-Receptor Tyrosine Kinase; BDNF, Brain-derived neurotrophic factor; NMDA, N-Methyl-D-aspartic acid; mTOR, mammalian target of rapamycin; AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; ADP, adenosine-diphosphate; ICL3, Intracellular loop 3
During the last decade increasing indications have been obtained that in the central nervous system and other tissues GPCRs form receptor heteromers possessing allosteric receptor–receptor interactions associated with conformational changes in the receptors leading to signal integration (14, 18–30). In agreement, structural plasticity has been demonstrated in GPCRs, for example, through computer-assisted analysis using ad hoc computer programs (29, 31). The N- and C-terminals as well as ICL3 of the GPCRs show a high propensity toward an unstructured conformation. Thus, they are potentially plastic domains, which can undergo interactions with other ligands, particularly with domains of other receptors and receptor interacting proteins. This aspect is of importance not only for the function of single GPCRs, but also for their interactions either with other receptors (receptor–receptor interactions) or, more generally, for the formation of clusters of membrane associated proteins and integration at the plasma membrane level. Based on a mathematical approach, Tarakanov and Fuxe (32) have deduced a set of triplet homologies that may play a major role in protein–protein interactions, including receptor heterodimers and human immunodeficiency virus (HIV) entry. A guide-and-clasp manner of receptor–receptor interactions is proposed where the “adhesive guides” may be the triplet homologies (32). An example is discussed below.
The Triplets AAR (or RAA) and AAE (or EAA) Homologies in GPCR Heteromers to Guide-and-Clasp Binding not only Between Protomers but also to Calmodulin and Gi Protein
Based on a bioinformatics approach, Tarakanov and Fuxe (32) have deduced a set of triplet homologies that may be responsible for receptor–receptor interactions. This set consists of two nonintersecting subsets: “protriplets” and “contra-triplets.” Any protriplet appears as a homology in at least one heterodimer but does not appear as a homology in any nonheterodimer. Just the reverse, any contra-triplet appears in at least one nonheterodimer but does not appear in any heterodimer. For example (Table 1), the triplet of amino acid residues alanine-alanine-arginine (AAR) (Ala-Ala-Arg) appears as a homology in seven receptor heterodimers: A2AR-D2R, A2AR-mGluR5R, D2R-mGluR5R, mGluR5R-GABAB1, D2R-CCKBR, 5-HT1BR-5HT1DR, and MHC1-CD8, whereas its symmetric triplet arginine-alanine-alanine (RAA) appears as a homology in four heterodimers: 5-HT1BR-5HT1DR, A1R-D1R, TLR1-TLR2, and TLR2-TLR6. Thus, both triplets AAR and RAA appear together as the homologies in the heterodimer 5-HT1BR-5HT1DR. Moreover, the triplet AAR may play a role in the HIV entry (Table 1) and the repeat proteins as well as in the Tax-CREB interactions (31). In the triplet AAR (and/or RAA), positively charged R (Arg) may interact with negatively charged E (Glu) from another triplet alanine-alanine-glutamic acid (AAE) (and/or glutamic acid-arginine-arginine (EAA)). Such electrostatic interaction between triplets AAR-AAE (and/or their symmetric variants) may guide-and-clasp protein–protein interactions. Taking together with the bioluminescence resonance energy transfer (BRET)/fluorescence resonance energy transfer (FRET) experimental results, these bioinformatics predictions suggest the existence of a basic set of common triplets in the two participating proteins that may participate in the protein–protein interaction interfaces (Fig. 1).
Table 1. Example of triplet homologies in receptor heteromers and HIV entry
A2A-D2, A2A-mGluR5, D2-mGluR5,
5-HT1B-5-HT1D, A1-D1, TLR1-TLR2, TLR2-TLR6,
D2-mGluR5, D1-D2, mGluR5-GABAB1,
Moonlighting via Allosteric Receptor–Receptor Interactions in Receptor Heteromers
In this review, we will discuss indications how changes in the allosteric receptor–receptor interactions among receptors through the formation of different types of receptor heterodimers and/or higher order heteromers (receptor mosaics) and/or receptor/protein complexes can change the function of the individual receptor present as a homomer (Fig. 2). Agonists and allosteric modulators can also produce changes in the allosteric receptor–receptor interactions. In this way, through conformational changes in these receptor complexes moonlighting may develop in a single GPCR by, for example, switching its coupling to other types of G proteins, to β-arrestins, to GPCR-receptor tyrosine kinases (RTKs) and to ion channel receptors and through changes in its orthosteric site altering its transmitter specificity (24, 33–38).
Alternatively, a single GPCR through conformational changes due to its participation in an altered receptor complex may undergo changes in cotrafficking with increased internalization and can become possible transcription factors through their altered conformation and possible receptor fragments formed. As possible transcription factors the GPCRs would have a novel function by directly modulating gene expression and likely trophism, possible actions of relevance for neurodegeneration. Allosteric receptor–receptor interactions in G protein coupled receptor-receptor tyrosine kinase (GPCR-RTK) heteromers likely will lead to moonlighting of the participating GPCR and RTK protomers with high relevance for neurodegeneration (Fig. 2) (27, 38, 39).
It is expected that any pathological change in the moonlighting properties of a receptor would have major consequences for the signaling, especially of the GPCRs with multiple interactions with other receptors and proteins and thus having a multitasking role (so-called hub receptors (14)). A moonlighting dysfunction in a hub receptor of key networks may substantially contribute to the spread of pathology in neurodegenerative diseases (40).
Moonlighting via D1R-D2R Heteromers
Altered G Protein Selection
The D1R-D2R heterodimer/receptor mosaic is an example of how the formation of a heteromer can switch the G protein coupling of participating receptors (D1R: Gs; D2R: Gi) to other G proteins (Gq) (35, 41). Thus, on coactivation of D1R and D2R in this heterodimer/receptor mosaic, a selective Gq/11 activation occurs, producing increases in phospholipase C (PLC) activity and a rapid rise in intracellular calcium levels without influencing adenylate cyclase activity regulated by Gs and Gi proteins (Fig. 3A). In this way, the PLC can become activated, leading to intracellular calcium release and increased levels of calcium/calmodulin dependent protein kinase II alpha contributing to synaptic plasticity (42). This represents an interesting example of moonlighting as the allosteric receptor–receptor interactions between the D1R and D2R leads to coupling to Gq with activation of PLC on their coactivation. Thus, a new function has developed in the D1R and D2R through their formation of a D1R-D2R heteromer the dysfunction of which may contribute to neuropsychiatric diseases.
Altered D1R and D2R Recognition
The D1R-D2R heteromer has also been discovered in brain to have a unique pharmacology by the George group (35, 42). The D1R agonist SKF83959 was a selective agonist for this receptor heteromer by being a full agonist at the D1R partner and a partial agonist at the D2R existing in a pertussis toxin-insensitive state leading to rapid activation of Gq/11 and PLC with a robust intracellular release of calcium (Fig. 3B). On the other hand, the D1R-like agonist SKF83959 does not activate adenylyl cyclase (AC) linked D1R and D2R and may, therefore, be a unique agonist for the D1R-D2R heteromer. Thus this work represents an interesting example of how moonlighting changes in the receptor binding domains of the two protomers in the D1R-D2R heteromers generate new recognition specificity and new drugs in neuropsychopharmacology (43).
The results indicate that the antagonistic A2AR-D2R allosteric receptor–receptor interaction in A2AR-D2R heteromer favors β-arrestin2 recruitment to the D2R protomer with subsequent cointernalization associated with a reduced time onset of Akt phosphorylation followed by a rapid dephosphorylation (Fig. 3C). Thus, β-arrestin2 action becomes more rapid and short lasting and, in this way, mimics G protein-mediated signaling (44). Thus, the moonlighting achieved via the antagonistic allosteric A2AR-D2R interactions is to favor a rapid onset of β-arrestin2-mediated D2R signaling through a D2R/Gi uncoupling favoring the formation of a possible A2AR-D2R-β-arrestin2 complex (44).
A beautiful moonlighting phenomenon in the A2AR-D2R heteromer is that the positively charged D2R N-Terminal ICL3 epitope (VLRRRRKRVN) may switch between binding to the negatively charged A2AR epitope (SAQEpSQGNT), localized in the medium segment of the C-terminus of the A2AR (22) to several negative epitopes of calmodulin (45).This moonlighting phenomenon is modulated by calcium ions since they disrupt the binding of calmodulin to the D2R but not to the A2AR (45) favoring the binding of the D2R N-terminal ICL3 epitope to the A2AR epitope or to epitopes of Gi/o. As demonstrated in Fig. 1, it is possible that the conformational change in the ICL3 of D2R may favour moonlighting to Gi/o via inter alia electrostatic interaction between triplets AAR(in D2R ICL3) and AAE (Gi/alpha1) (and/or their symmetric variants) contributing to guide-and-clasp D2R-Gi interactions (Fig. 1, see also Fig. 3D).
Moonlighting via D2-5-HT2A Heteromers
Modulation of the Balance of PLC Activation versus AC Inhibition
The potential existence of D2LR-5-HT2AR heteromers in living cells and the functional consequences of this interaction have been demonstrated (46, 47). Thus, by means of a proximity-based BRET approach, it has been shown that the D2LR and the 5-HT2AR form stable and specific heteromers when expressed in HEK293T mammalian cells. Furthermore, when the D2LR-5-HT2AR heteromeric signaling was analyzed, we found that the 5-HT2AR-mediated PLC activation was synergistically enhanced by the concomitant activation of the D2LR (Fig. 3C). Thus, when both receptors were simultaneously challenged a specific and significant rise of the intracellular calcium levels was observed, as determined by means of a NFAT-luciferase reporter gene assay. Conversely, when the D2LR-mediated AC inhibition was assayed costimulation of D2LR and 5-HT2AR within the heteromer led to inhibition of the D2LR functioning (Fig. 3C), thus suggesting the existence of a 5-HT2AR-mediated D2LR trans-inhibition phenomenon (46).
Moonlighting of the 5-HT2AR versus the D2R signaling may thus take place via allosteric receptor–receptor interactions induced by coactivation of the orthosteric binding sites in the D2R-5-HT2AR heteromer favoring 5-HT2AR and reducing D2R signaling. Possible novel therapeutic strategies for treatment of schizophrenia may be explored by targeting this heteromer in view of the role of D2R receptor antagonists and 5-HT2AR antagonists in the treatment of schizophrenia (46).
Moonlighting via A1-P2Y1 Receptor Heteromers
Altered A1R Recognition
Changes in A1R pharmacology have been discovered by Nakata and his team in A1R-P2Y1 receptor heteromers (34, 48). This heteromerization results in a conformational change in the A1R binding pocket leading to the appearance of an A1R receptor with P2Y1-like agonistic pharmacology (Fig. 3B). In fact, a P2Y1 agonist binds to the A1R receptor and produces an inhibition of AC that is blocked by an A1R antagonist. Therefore, this A1R-mediated ATP response can be one of the mechanisms that might account for the ATP-induced inhibition of transmitter release, as it is coreleased from nerve terminal networks. The moonlighting achieved via the allosteric A1R-P2Y1 receptor interaction has led to an altered conformation in the A1R orthosteric site with the formation of a P2Y1-like A1R.
Moonlighting via GPCR-RTK Heteromers and Its Relevance for Neurodegeneration and Depression-induced Atrophy of Neurons
It has been discovered that the dopamine D2R stimulation of mitogen-activated protein kinases is mediated by a cell type-dependent transactivation of receptor tyrosine kinases (39). There may also exist RTK ligand-independent mechanisms for RTK transactivation with participation of both GPCRs and the RTK in a multireceptor signaling complex (49, 50). In fact, D2R agonist increases the coimmunoprecipitation of D2R and epidermal growth factor receptors in neuroblastoma cells, thereby suggesting that D2R activation induces the formation of a macromolecular signaling complex (39) as a D2R/RTK heteromer.
Furthermore, a physical interaction between the A2AR and the fibroblast growth factor receptor (FGFR1) has been demonstrated (38). Coactivation of these two classes of receptors caused a substantially increased activation of the mitogen-activated protein (MAP) kinases (MAPK) pathway associated with increases in synaptic plasticity with spine morphogenesis. More recently, it has been demonstrated that the dopamine D2R agonist quinpirole is able to block the synergistic actions of FGFR1 and adenosine A2A receptors on corticostriatal plasticity (51).
It has been suggested that the DA hypothesis of schizophrenia is based on the D2R-like antagonist properties of antipsychotic medication and the potential of D2R agonists to induce schizophrenia-like symptoms (52). FGF-2 expression is also altered after antipsychotic treatment as shown in postmortem brains of schizophrenic patients. In addition, antagonistic allosteric A2AR-D2R receptor–receptor interactions have been demonstrated in the central nervous system opening up the possibility to explain the findings from Flajolet et al. (38) on the basis of a FGFR1-A2AR-D2R receptor complex (51). Thus, again moonlighting properties may develop, leading to an enhancement of the MAPK pathway provided the D2 receptor is not activated.
In 2007, receptor tyrosine kinase transactivation in response to antidepressant drug treatment has been postulated to take place via a new receptor–receptor interaction mechanism between 5-HT1AR and FGFR1 receptors in the dorsal raphe-cortical neurons and their target regions (27).
Novel research in this field has increased our understanding of the mechanisms of antidepressant drug action and led to the development of new strategies for treatment of depression. For the first time, it has been possible to identify in the mesencephalic 5-HT neurons and their telencephalic targets a 5-HT1AR-FGFR1 heteromer the activation of which may contribute to a relief from depression (Borroto-Escuela et al., unpublished data). Ongoing work based on the present hypothesis seems to identify that the activation of the 5-HT1AR-FGFR1 heteromer can mediate improvement in 5-HT neuronal communication and neurotrophism. Moonlighting in the FGFR1 of the GPCR-FGFR1 heteromers may develop as a result of the allosteric receptor–receptor interaction in the heteromer achieved inter alia by coactivation of its protomers. This allosteric change in the catalytic region of the FGFR1 may change its recruitment of adaptor proteins, which may alter the pattern of activation of its five major signaling pathways: the ras protein (RAS)/RAF proto-oncogene serine/threonine-protein kinase (RAF)/MAPK pathway, the PI3K/protein kinase B (AKT)/mammalian target of rapamycin (mTOR) pathway, the PLC gamma pathway, the Janus kinase (JAK)/signal transducer and activator of transcription (STAT) pathway and the IκB kinase/nuclear factor kappa-light-chain-enhancer of activated B cells pathway. It may be that the receptor–receptor interaction may favor the selection of some of its pathways, for instance the RAS/RAF/MAPK pathway, according to the results obtained (Borroto-Escuela et al., unpublished data). The same may also be true for the demonstrated receptor–receptor interaction in the FGFR1-A2AR heteromer (38). The results open up the exciting possibility that moonlighting in the GPCR-RTK heteromers may inter alia lead to altered functions of the RTK protomer with relevance both for trophism and neurodegenerative processes including depression-induced atrophy of neurons.
There exists substantial evidence that brain neurotrophic factors and especially brain-derived neurotrophic factor (BDNF) play a significant role in depression. It is reduced in depressed patients and treatment with antidepressants increases hippocampal BDNF levels (53, 54). Peripheral BDNF has even been shown to exert antidepressant-like actions in behavioral and cellular models (55). Furthermore, a highly interesting article has recently appeared demonstrating that the rapid antidepressant actions of N-methyl-D-aspartic acid (NMDA) antagonists (56) may be produced via mTOR-dependent synapse formation (57). These effects were found to be dependent on α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor signaling. According to our hypothesis, the mechanism for these events may be moonlighting in the function of RTK-NMDA-AMPA receptor complexes, where blockade of NMDA receptors may lead to a dominance of AMPA receptor modulation of the RTK receptor signaling panorama moving the RTK signaling predominantly into the PI3K/AKT/mTOR pathway. Based on above, the RTK receptors involved may mainly be TrkB (receptor for BDNF) and FGFR1. Thus, allosteric receptor–receptor interactions in higher order RTK-ion channel receptor complexes may also play an important role in counteracting depression and stress induced atrophy in neurons through moonlighting of the signaling functions of the RTK receptors involved.
Moonlighting via Receptor Interacting Proteins
Increasing evidence indicates that signaling efficiency/specificity for muscarinic acetylcholine receptors family (mAChRs) is determined in part by accessory proteins that physically interact or are found in the microenvironment of the receptor (58). Several proteins have been shown to interact with mAChRs, including other GPCRs, kinases, and scaffold proteins such as b-arrestin (59, 60). These proteins, along with classical core signaling entities (receptor, G protein and effectors), contribute to form a signalosome complex at the cytoplasmic face of the receptor (61, 62). Previous studies using receptor-derived peptides from specific regions of the M1R and M2R receptors have shown the C-terminal tail of the ICL3 is critical for receptor-G protein interaction and the resulting signal transduction mediated by G proteins (63). More recently, specific motifs in the ICL3 of M1R and M3R receptors have been shown to bind some accessory proteins with high affinity (calmodulin, oncogenic SET protein, and small GTPase Rho) (60). This experimental evidence points to a specific role of ICL3 in receptor-G protein coupling signal transduction and multiprotein complex formation.
Coexpression of the M3R ICL3 minigene dramatically reduces both carbachol-mediated G protein activation and inositol phosphate accumulation. Minigene expression also abolishes activation of M3R and M5R receptor mitogen-activated protein kinases pathway (64). These data, together with results of coimmunoprecipitation of different scaffold and kinase proteins, provide experimental evidence for the role for the third cytoplasmic loop of the human M3 muscarinic receptor in G-protein activation and multiprotein complex formation (59). It is also of substantial interest that the third intracellular loop of the M5R like of the D2R (22, 26) may play a regulatory role both in receptor function and heteromerization (M3R/M5R heteromer) (64).
Very recently tandem-affinity purification and mass spectrometry have been used as a systematic approach to characterize multiprotein complexes in the acetylcholine muscarinic receptor subfamily (60). Specific interacting proteins from each receptor subtype were identified systematically, representing a major methodological advance in the identification of GPCR-interacting protein complexes involving a cytoplasmic minigene construct, encoding the third intracellular loop and the C-terminal tail tagged to the tandem-affinity-cassette of each receptor subtype. In addition, collected data allowed for the first time the construction of an acetylcholine muscarinic receptor interactome network (e.g., receptorsome), reflecting receptor complex properties and tendencies. Through the “interactome” concept, it was possible to propose signaling roles for interacting partners, which had not been previously described. In this article, four Gα proteins for M1R, M3R, and M5R receptor subtypes (Gαq, Gα11, Gα12, and Gα13) and three Galpha-i isoforms (Galpha-i1, Galpha-i2, and Gapha-i3) for M2 and M4 receptor subtypes, as well as two different Gβ isoforms (Gbeta1 and Gbeta4) and three Ggamma isoforms (Ggamma2, Ggamma7 and Ggamma11) were identified (60).
Muscarinic receptors were also found to selectively bind to different structural components in the cell that may have a role in receptor internalization and the regulation of cell spreading and migration. In addition, the identification was made of several muscarinic receptor subtype-specific signaling proteins such as adenosine-diphosphate-ribosylation factors, elongation factor 1-A (eEF-1A), oncogenic SET protein, Rac1, and different isoforms of phospholipase C and the protein kinase C (60). Thus, there exist both G protein dependent and independent signaling.
On the basis of inter alia the above findings, it seems likely that moonlighting may develop in the intracellular loops and C terminal of the GPCRs especially in the long ICL3 as a result of dynamic allosteric interactions between different types of G proteins and other receptor interacting proteins in these domains of the receptor. Because of these dynamic conformational changes, the same amino acid strand in these domains may in one state bind to a certain G protein and in a different state to another signaling protein as a result of changes in guide-clasp interactions of this motif to the interaction motives of these two signaling proteins. In this way, the receptor may change its signaling function and dynamic multitasking actions may develop.
In this case, with receptor interacting proteins, the moonlighting will mainly function to have intracellular signals change the function of the GPCR as a result of intracellular demands. In the case of moonlighting in receptor heteromers, it is instead the result of a demand from the environment to integrate two or multiple signals into novel functions of a single amino acid strand, which takes place through allosteric receptor–receptor interactions in the extracellular, transmembrane, and intracellular domains. The integration of signals in higher order receptor oligomers (receptor mosaics) may result in a more complex moonlighting phenomenon.
The field of MPs continues to expand in this case into the field of receptor proteins, especially in relation to their formation of GPCR heteromers and GPCR-RTK heteromers and postulated ion channel receptor-RTK receptor heteromers. The allosteric receptor–receptor interactions lead to changes in the guide-clasp interactions of key receptor motives involved in agonist/antagonist binding and binding of signaling proteins so that moonlighting develops. Thus, another type of pharmacology and another signaling protein spectrum becomes selected. The moonlighting in the GPCR-RTK and postulated AMPA-NMDA-RTK heteromers may have special relevance for counteracting neurodegenerative processes such as depression induced atrophy of neurons in view of the impact of RTK signaling on nerve cell survival and differentiation. Moonlighting in GPCR homomers may likely also develop due to the binding of multiple signaling proteins including the G proteins to the intracellular domains of the GPCRs altering via allosteric mechanisms the guide-clasp interactions of the receptor strands of amino acids with the selection of different signaling proteins.
This work was supported by grants from the Swedish Research Council (04X-715), Hjärnfonden, Torsten and Ragnar Söderberg, Telethon TV3's La Marató Foundation 2008 and M.M. Wallenberg Foundation to KF, Karolinska Institutets Forskningsstiftelser 2010 to D.O.B-E. and also by grants SAF2008-01462 and Consolider-Ingenio CSD2008-00005 from Ministerio de Ciencia e Innovación and ICREA Academia-2010 from the Catalan Institution for Research and Advanced Studies to FC. A.O.T. has not received any support for this work.