Introduction

In normal physiology, the neurotransmitter serotonin (5-hydroxytryptamine, 5-HT) and its receptors regulate behaviours such as aggressiveness, anxiety, sex, sleep, mood, learning, cognition and memory. They are involved in numerous disease states, including depression, anxiety, social phobia, schizophrenia, mania, autism, drug addiction, obesity, obsessive-compulsive, panic and eating disorders. Therefore serotonin receptors are the target of a variety of pharmaceutical drugs. With the exception of the 5-HT3 receptor, a ligand-gated ion channel, serotonin receptors are a group of membrane-bound G-protein-coupled receptors, which, by means of G-proteins, activate intracellular pathways to produce an excitatory or inhibitory response [[1]].

G-proteins are heterotrimers consisting of three subunits: Gα, Gβ and Gγ; they are located on the inner plasma membrane, from which they induce GPCR activation. The Gβ and Gγ subunits form an inseparable complex, the βγ complex [[2]]. In the absence of receptor stimulation the Gα subunit binds guanosine diphosphate (GDP) and the βγ complex, and remains dissociated from the receptor. Binding of the ligand to the GPCR domain outside the cell induces conformational changes of the intracellular GPCR domain, giving rise to GPCR coupling to the G heterotrimer. Consequently, the Gα protein exchanges GDP for guanosine triphosphate (GTP), causing dissociation of the GTP-bound α-subunit from the βγ complex and their separation from the activated receptor. Gα and βγ therefore activate a cascade of further signalling events that finally result in a change in cell function. The process is terminated with GTP hydrolysis to GDP by Gα [[3]].

Various Gα families have been described: they can activate different pathways or even exert opposite effects on the same pathway. In general, the 5-HT1 (1A, 1B, 1D, 1E, 1F) receptor family and 5-HT5 receptors couple with Gαi/o protein family to inhibit adenylate ciclase (AC) activity, reducing the intracellular cyclic adenosine monophosphate (cAMP) level whilst the Gαs family 5-coupled to HT4, 5-HT6 and 5-HT7 receptors triggers a pathway that leads to AC activation and cAMP production. The 5-HT2 (2A, 2B, 2C) receptor family couple with Gαq/11 proteins and stimulate the activity of phospholipase C (PLC) increasing the intracellular inositol trisphosphate (IP3), diacyl-glycerol (DAG) and Ca2+ levels [[1]]. However Gαq can also indirectly alter cAMP production, by decreasing Gαs protein abundance [[4]] or by activating adenylate cyclase 8 (ADCY8) by the PLC/Ca2+/calmodulin pathway [[5]]. Moreover, the Gαi/o family induces a decrease in intracellular cAMP levels through AC inhibition.

The Gβ and Gγ subunits are closely associated forming a βγ complex that can be separated only by denaturation, except in cases when the complex involves β5, whose bond to γ subunits is much weaker. At variance with previous studies, the βγ complex does not remain inert after dissociation from the α subunit, but plays a key role both in the inactive and in the active receptor state [[6]]. The βγ complex has the following functions: i) it is required for optimal receptor-G-protein interaction, because it enhances ligand affinity and receptor-G-protein coupling, hence G-protein activation [[7]]; ii) its subunit composition affects receptor-G-protein coupling specificity [[8],[9]]; iii) it activates specific pathways regardless of the type of Gα subunit involved [[10]]. The role of γ subunits is to transport the βγ complex from the endoplasmic reticulum to the plasma membrane. Although all γ proteins share this property, translocation kinetics differs widely among subunits, ranging from 10 sec of the fastest, γ9, to several minutes of the slowest, γ3 [[11]]. It may be hypothesized that the γ subunits allowing fast translocation are associated with human serotonin receptors with a quicker turnover.

Review

Many G-protein isoforms, huge number of possible heterotrimers

We have explored UniProt (http://www.uniprot.org) and Entrez Gene (http://www.ncbi.nlm.nih.gov/gene) databases to establish how many protein isoforms are currently known for each G-protein subtype (Table 1). For example, we have found 10 isoforms of the Gαi2 subtype. Generally, the isoforms do not derive from new gene loci but from different gene expression regulation of the main transcript. In particular, they are due to alternative splicing, use of alternative transcription start sites (TSS) and alternative start codons. All these detailed data are reported in Additional file 1 along with some annotations. In brief, we found many more protein isoforms than expected: Gα, Gβ, and Gγ proteins may actually be as many as 53, 38, and 20, respectively, raising the potential heterotrimers to about 40,000 combinations. Most of the isoforms we reported are shorter and lack one or more functional domains compared to the reference isoforms, so they could have a reduced functional activity. In particular, the shorter Gα subtype isoforms lack GTP domains, the Gβs lack WD domains whereas the Gγs have no alterations lying in their functional domains. Surprisingly, up till now, many isoforms have not yet been confirmed at protein level according to UniProt, so we looked for this information in the literature. Unfortunately, the papers merely report the name of the investigated G-protein and do not specify its particular isoform. We also found nomenclature inaccuracies, which are probably due to the former names of G-protein: for example when the authors write “the Gαq protein” it is unclear whether they mean GNAQ (Gαq) or the entire family including also GNA11 (Gαq11), GNA14 (Gαq14), or GNA15 (Gαq15 also Gαq16); similarly, “Gαi protein” may indicate GNAI1 (Gαi1), GNAI2 (Gαi2), or GNAI3 (Gαi3). The same applies to “Gβγ” [[12]]. The nomenclature is not univocal even in some databases. For example, in UniProt GNG8 is also called GNG9, and GNGT2 is also called GNG8 or GNG9; therefore a paper examining GNG9 could refer either to GNG8 or to GNGT2. To overcome this ambiguity, the authors should indicate the protein or gene identifier or protein sequence.

Table 1 G-protein isoforms
Full size table

How heterotrimer composition affects 5-HT receptor behaviour

The mechanisms underpinning formation of one heterotrimer rather than another are poorly understood, but post-translational modifications of G-proteins and of membrane environments are likely to be involved [[14]]. Nonetheless heterotrimer composition is critical, because it determines what pathway is activated. For example, activation of 5-HT1A receptor inhibits basal phosphoinositide hydrolysis in the dorsal raphe nucleus but not in the hippocampus, most likely due to different heterotrimer compositions in the two tissues [[15]]. Moreover a single heterotrimer can activate multiple pathways simultaneously, because some Gα proteins have multiple effects. For example, some Gαi/o family proteins inhibit AC leading to intracellular cAMP reduction, whereas others can also inhibit Ca2+ or activate K+ channels [[16]]. To complicate matters further, the same serotonin receptor can couple to different heterotrimers [[7],[17],[18]]; the same ligand may therefore simultaneously activate multiple pathways but be unable to regulate a specific one. The mechanism appears to be irrational, since a single switch (receptor) is unlikely to be able to control a large number of lights (pathways).

Ligand-receptor binding affinity affects G-protein-receptor affinity and vice versa, as described in Spodoptera frugiperda Sf9 cells [[7]]. Such affinity also depends on heterotrimeric composition; for example, coupling of Gαi3 to 5-HT1A or 5-HT1B receptor was more effective than that of Gαi2 and Gαo in enhancing agonist [3H]-5-HT affinity [[14]].

Since a variety of psychiatric disorders and/or drug responses are held to be related to altered ligand-receptor affinity, association studies have mainly explored receptor and downstream effector polymorphisms to explain the genetic basis of such different phenotypes [[19]–[26]]. However, given that G-proteins can affect ligand affinity, their variations should also be considered in association studies. For these reasons it is important to gain insights into the role, the interactors and the expression of G-proteins in determining cell responses as a consequence of receptor activation.

The observations that in some GPCRs the G-protein complex can modulate receptor activity state and that the transition from active to inactive state depends on the Gα subunit associated with the receptor make the study of G-proteins even more intriguing [[14]]. In other words, GPCRs can switch from inactive to active even in the absence of binding to an agonist. This mechanism is still poorly understood and may have important pathological as well as physiological implications. In particular significant activation even without serotonin has been described with coupling of Gαz to 5-HT1A receptor, but not to the other 5-HT1 receptors [[27]].

Finally, a scenario is emerging where different G-protein combinations can bind the same receptor type, conferring a different ligand affinity and activating several pathways. This warrants investigation of the heterotrimeric combinations that may form in humans, their distribution in different tissues, and the differences in ligand binding affinity among the heterotrimers resulting from binding of one receptor type and various G-proteins.

How many couplings between 5-HT receptors and G-proteins are known

In order to find out what is known about heterotrimer associations with serotonin GPCRs, we have performed a literature search. The papers specifically addressing receptor-G-protein complexes were scanty, therefore data were available for quite a small number of complexes out of the possible thousands. Table 2 presents an exhaustive list of all known combinations of the three types of G-proteins and their associations with serotonin receptors in human neural tissues or in similar models. In particular, for each receptor, we have reported the experimentally assessed complexes formed with the G-proteins, the tissues or contexts where the complexes were determined and their references. We have also annotated the couplings assessed as not present along with the particular experimental context. The 5-HT1p and 5-HT3 receptors were excluded, because the former is expressed in the nervous enteric system (not the central nervous system), the latter because it is a serotonin-gated ion channel not coupled to G-proteins, whereas 5-HT5B is a pseudogene in humans according to EntrezGene and the related protein is absent in UniProt.

Table 2 Assessed couplings between G-proteins and serotonin receptors
Full size table

Many papers have addressed G-protein combinations with 5-HT1A, 5-HT1B, 5-HT2A and 5-HT2C while only one or two papers refer to 5-HT1E, 5-HT1F or 5-HT6. The experimental models usually involve transfection of human genes into Spodoptera frugiperda Sf9 cells, since they express low levels of mammalian G-proteins, thus avoiding competition with endogenously expressed G-proteins in [35S]GTPγS binding assay [[16]].

However, to extend these findings to humans is not necessarily correct. For example, the poor coupling of Gαq to 5-HT1A and 5-HT2C could be due to a large portion of the expressed but inactive Gαq [[84]], maybe because Gαq is not post-translationally modified by palmitoylation in Sf9 as in humans [[85]]. Other authors have used hamster, mouse and human cells transfected with rat, mouse and bovine constructs. The main methods used to assess the compositions of the heterotrimers are immunoprecipitation and western blot analysis, the binding with radio-ligands and FRET (Förster Resonance Energy Transfer) by using fluorescent ligands.

For the majority of serotonergic receptors, coupling data are available only in relation to Gα family proteins without specifying which Gβ and Gγ were coupled. However, the few data on Gβ and Gγ only concerned Gβ1 with Gγ2. Gα proteins coupled to receptors are the most commonly studied G-proteins, because they are held to indicate the pathway activated by receptor stimulation. In contrast, Gβ and Gγ proteins are believed merely to play a structural role, that is to stabilize the receptor complex, but they actively participate in signal transduction by activating specific pathways.

We also annotated the second messengers activated downstream G-proteins, when these data were available in the related paper, since they allow to take into account the converging effect of various Gα proteins and the antagonistic/additive effects of Gβγ.

Investigation methods for the assessment of G-protein activation

We have shown that G-protein heterotrimers recruited by serotonin receptors have been evaluated experimentally. This dearth of data is mainly due to the cumbersome methods used to identify the heterotrimers involved in the effects of ligands and to some technical limitations. In fact, to assess the receptor-mediated G-protein activation, both indirect and direct assays are available [[16]]. Indirect methods, in spite of their good sensitivity, are focused on measuring concentrations of second messengers but the evaluation of these data can be complicated since most receptors can activate different G-proteins.

In particular, since Gαs and Gαi/o proteins activate or inhibit AC respectively, their activation can be indirectly detected determining intracellular adenosine triphosphate (ATP) conversion into cAMP. It is measured using [α-32P]ATP as the enzyme substrate or using cAMP antibodies. These methods cannot follow quick fluctuations as they are based on static measurements after cell lysis. In the case of the AC inhibiting Gαi/o proteins, another problem regards the too low dynamic ranges of inhibition detection. To deal with this specific problem, chimeric Gαi/o proteins were developed, but they do not exactly mimic the natural G-proteins.

To test ligand efficacy on Gαq/11-coupled receptors, [3H]IP3 concentration as product of PLC activity can be measured using [3H]PIP2 (phosphatidyl inositol 4,5-bisphosphate) substrate. Alternatively, antibodies can be used, but, since IP3 has a short half life, it is preferred to detect its stable metabolite inositol-1-phosphate (IP1), although this is a more downstream product. Also Ca2+ concentration, by dyes generating fluorescence upon binding of free Ca2+, can be determined to assess Gαq/11 activation, although these probes can influence calcium levels and kinetics. Moreover, Ca2+-sensitive photoproteins, as aequorin, can detect calcium in specific cell compartments by fusion with targeting sequences. This approach is not so sensitive and consists of laborious procedures, such as fusion protein production, transfection and assay calibration.

Gα12/13 activation can be assessed by determining Rho guanine nucleotide exchange factors (RhoGEFs) by immunoblotting, a not highly sensitive technique. Moreover, since RhoGEFs are activated also by Gαq/11, there are crosstalk problems that can be partially overcome by small interfering RNA (siRNA) knockdown.

To directly and quantitatively assess the Gα protein activation, [35S]GTPγS binding assay is employed. Upon Gα subunit activation, it binds the mimic substrate, so remaining blocked in the active form as it cannot hydrolyze this substrate. The blocked Gα can be measured after isolation and it can be immunoprecipitated to identify the specific Gα subunit. However, this approach is mainly suitable to evaluate Gαi/o-coupled receptor activation. This assay can be effectively combined with the use of drugs stimulating or inhibiting specific G-proteins, for example, Pertussis toxin (PTX), Mastoparan, Mastoparan-S, Cholera toxin, Suramin, Pasteurella multocida toxin (PMT). Alternatively, it is possible to use G-protein-deficient mice or gene silencing by siRNA, although studies have to take into account the cellular compensatory mechanisms that alter the expression level of other G-proteins. An exhaustive review of these and other techniques was made by Denis et al. [[16]].

Moreover, cause of GTPase-accelerating proteins (GAPs) that accelerate GTPase activity of Gα-protein subunits, the measuring of GTPase activity in vivo and in vitro differ. In addition, for in vivo studies, methods having a subsecond time resolution for GTP hydrolysis must be adopted [[86],[87]].

Heterotrimers activation effects

It would be important to consider the synergic effects of the entire activated heterotrimer in order to evaluate ligand effects, drug efficacy and side effects such as hallucination onset. For this reason, we annotated also the second messengers during literature revision. Unfortunately, as can be seen in Table 2, few studies assessed the coupling of all the heterotrimer subunits and few of them assessed the second messengers. Regarding these cases, only Gβ1γ2 were present, so it was not possible to verify if and how different Gβγ combinations can affect Gα induced pathways. In general, the physiological significance of the different Gβγ pairs is unclear, since they participate in complex interactions with receptors, Gα subunits and effectors [[6]].

A more detailed description of the pathway downstream G-proteins was performed by Millan et al. [[88]], however it should be taken into account that the signalling downstream a receptor is ligand-dependent. For example, some agonists of 5-HT2A can induce hallucinations but other structurally related ones do not [[89]].

Expression analysis in human brain tissues

We believe that our collected data can be used for guiding experiments which seek new couplings. However, we verified if it was possible to reduce the number of combinations by filtering out those not allowed in a particular neural tissue due to one or more components that are not expressed. In Additional file 2, we report the expression profiles of Gα, Gβ, Gγ and the serotonin receptor by using three proteomic databases according to a previous work [[90]]: Human Protein Reference Database (http://www.hprd.org); Human Proteinpedia (http://www.humanproteinpedia.org) and Human Protein Atlas (http://www.proteinatlas.org). We also used expression data obtained from three transcriptomic databases: Human Transcriptome Map (http://bioinfo.amc.uva.nl/HTMseq), Cancer Gene Anatomy Project database (http://cgap.nci.nih.gov) and Allen Brain Atlas database (http://www.brain-map.org). An issue that arose in the course of this investigation was the partial conflict between microarray and RNA-Seq data retrieved from Allen Brain Atlas and, to a lesser extent, protein and transcript expression data. Generally, these discrepancies could be resolved by relying on proteomic data, which, if present, are usually more dependable than transcriptomic data. For example, regarding GNG1, it seems to be absent in all tissues and their sub-tissues assessed by Allen Brain Atlas RNA-seq. Instead, according to Allen Brain Atlas microarray data, GNG1 expression results as being very variable among sub-tissues of each tissue. Since in the cerebellar cortex also Protein Atlas data are available, they solve this contradiction claiming GNG1 absence. According to our expression data, most G-proteins are expressed in the majority of brain tissues, thus confirming the possible existence of a big number of heterotrimer combinations in nearly all neural tissues. Of course, co-expression of a receptor and G-proteins in a brain tissue does not imply that they are functionally coupled to each other. However, the available databases, being manually annotated, do not contain the all expression data reported in the literature, so the Additional file 2 may be incomplete.

Conclusion

The large number of human G-proteins that our searches found demonstrates that a very large amount of possible heterotrimers can be formed but unfortunately only a few have been assessed. Naturally, a limitation of the studies carried out in vitro is that the reported couplings do not always match to the couplings found in vivo [[91]]. However, knowledge of all the G-proteins that bind to each receptor would allow linking each receptor to all the possible activated pathways. Association of a receptor with multiple G-proteins would also highlight activation of different pathways in different tissues. This is important, because G-protein gene mutations or polymorphisms could alter transduction efficacy, thus explaining the non-activation of a pathway despite the presence of the right ligand and the absence of nucleotide variation in the receptor. Therefore precise knowledge of the role and distribution of G-proteins would greatly contribute to the evaluation of G-protein gene polymorphisms and to the development of drugs targeting specific G-proteins.

Finally, since receptor binding to a G-protein considerably modifies receptor behaviour, it could be that the G-proteins define many receptor subtypes. For this reason it is more appropriate to consider a receptor not individually but in association to each permitted heterotrimer. This also suggests that a number of experiments should be performed again, like the biochemical studies exploring the affinity constants between ligands and a receptor not considering if, and which, G-proteins were associated.

Additional files