Search tips
Search criteria 


Logo of plosonePLoS OneView this ArticleSubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)
PLoS ONE. 2007; 2(12): e1317.
Published online 2007 December 19. doi:  10.1371/journal.pone.0001317
PMCID: PMC2267039

Modifying Ligand-Induced and Constitutive Signaling of the Human 5-HT4 Receptor

Hany El-Shemy, Academic Editor


G protein–coupled receptors (GPCRs) signal through a limited number of G-protein pathways and play crucial roles in many biological processes. Studies of their in vivo functions have been hampered by the molecular and functional diversity of GPCRs and the paucity of ligands with specific signaling effects. To better compare the effects of activating different G-protein signaling pathways through ligand-induced or constitutive signaling, we developed a new series of RASSLs (receptors activated solely by synthetic ligands) that activate different G-protein signaling pathways. These RASSLs are based on the human 5-HT4b receptor, a GPCR with high constitutive Gs signaling and strong ligand-induced G-protein activation of the Gs and Gs/q pathways. The first receptor in this series, 5-HT4-D100A or Rs1 (RASSL serotonin 1), is not activated by its endogenous agonist, serotonin, but is selectively activated by the small synthetic molecules GR113808, GR125487, and RO110-0235. All agonists potently induced Gs signaling, but only a few (e.g., zacopride) also induced signaling via the Gq pathway. Zacopride-induced Gq signaling was enhanced by replacing the C-terminus of Rs1 with the C-terminus of the human 5-HT2C receptor. Additional point mutations (D66A and D66N) blocked constitutive Gs signaling and lowered ligand-induced Gq signaling. Replacing the third intracellular loop of Rs1 with that of human 5-HT1A conferred ligand-mediated Gi signaling. This Gi-coupled RASSL, Rs1.3, exhibited no measurable signaling to the Gs or Gq pathway. These findings show that the signaling repertoire of Rs1 can be expanded and controlled by receptor engineering and drug selection.


Heptahelical G protein–coupled receptors (GPCRs) are the largest family of human cell-surface receptors, encompassing more than 340 hormone receptors and 350–460 olfactory receptors [1], [2]. They are activated by peptide hormones, odorants, photons, biogenic amines, phospholipids, and many other extracellular signals. Upon activation, GPCRs undergo conformational changes that allow active and reversible signaling through a limited number of G-protein pathways (Gs, Gi, Gq, G12/13). These signals mediate a wide variety of physiological responses, including heart rate, chemotaxis, cell proliferation, neurotransmission, and hormonal responses. Owing to their physiological importance, GPCRs are of great medical interest. Indeed, they are targets for at least 40% of modern pharmaceuticals [3].

Although many drugs target GPCRs, studies of GPCR signaling in vivo have been hampered by the lack of specific agonists and antagonists for many of the receptors. GPCRs display molecular and functional diversity, such as the type of G-protein signaling pathway associations, different levels of constitutive activity, and different signaling responses due to different ligand-selective conformations [4], [5]. This diversity enables the receptors to transmit unique extracellular signals but hampers efforts to sort out the relative contributions of each signaling pathway or the roles of constitutive signaling for each receptor.

To better study the diversity of GPCRs, we developed receptors activated solely by synthetic ligands (RASSLs) by modifying their structures to render them unresponsive to endogenous hormones. Instead, RASSLs are activated by small-molecule drugs [6], allowing them to be used to activate specific G-protein pathways rapidly and reversibly and to mimic the speed, localization, regulation, and amplification of endogenous GPCR signals [7].

Since it is impractical to convert all GPCRs into RASSLs, we and others have focused on representative Gs-, Gi-, and Gq-coupled GPCRs, which stimulate adenylyl cyclase, inhibit adenylyl cyclase, and stimulate phospholipase-C, respectively. Ro1 (RASSL opioid 1), the prototype RASSL based on a Gi-coupled κ-opioid receptor [8], provided a proof of concept for this strategy. Its Gi response to natural ligands is 0.001% of that of the wildtype receptor, but it is potently activated by the synthetic agonist spiradoline [8][10]. Ro1 decreases heart rate in mice [11] and affects taste sensation in the tongue [9]. In addition to serving as powerful tools to dissect the G-protein signaling in vivo, RASSLs can yield insights into fundamental aspects of receptor diversity [12]. For instance, constitutive signaling of Ro1 led to cardiomyopathy [11], diminished bone formation [13], and induced hydrocephalus [10]. These constitutive signaling phenotypes would have been difficult or impossible to identify by studying endogenous receptors.

Multiple RASSLs have since been made, including a Gs-coupled RASSL based on the melanocortin-4 receptor [14], a Gq-coupled RASSL based on the histamine 1 receptor [15], and a series of RASSLS based on muscarinic receptors [16]. These RASSLs are useful tools; however, it is still advantageous to derive a series of RASSLs with distinct G-protein signaling from the same parental GPCR. It can be difficult to compare the effects of RASSLs based on different parental GPCRs since these RASSLs could have different pharmacokinetics, constitutive activity, desensitization kinetics, and cellular localization.

To better study GPCRs, we built a series of RASSLs based on the human 5-HT4 receptor (Figure 1), which has several advantages over other serotonin receptors. First, its pharmacological properties are well established [17]. Second, its agonists have milder effects (increased gastrokinesis [18], augmented memory acquisition and retention [19], increased chronotropic and inotropic cardiostimulation [20], and enhanced cortisol release [21]) than other serotonergic drugs. Third, it has a large number of synthetic ligands, which allows us to identify differences in their effects on that receptor. Fourth, a single mutation (D100A) in the mouse 5-HT4 receptor dramatically reduced its affinity for serotonin, its endogenous ligand. This mutation also allows synthetic agonists and antagonists for the wildtype receptor to activate the mutant, turning 5-HT4-D100A into a RASSL [22].

Figure 1
Human 5-HT4-based RASSLs.

Finally, we reasoned that novel receptors coupling to other signaling pathways could be created by making chimeras of the 5-HT4 receptor with other family members. Altering the G-protein selectivity of GPCRs is often difficult because it is based on receptor conformation determined by multiple regions of the receptor [23]. Changing multiple regions involves large internal mutations that often lead to receptor instability. A better strategy for altering G-protein signaling characteristics is to swap domains between structurally similar receptors within the same family. The 5-HT4 receptor belongs to a family of at least 15 receptors, each with different subfamilies that engage different G-protein signaling pathways. The 5-HT4, 5-HT6, and 5-HT7 subfamilies are Gs coupled, the 5-HT1 subfamily is Gi coupled, and the 5-HT2 subfamily is Gq coupled [24]. These characteristics could expedite our efforts to make purely Gs-, Gi-, and Gq-coupled RASSLs.

Here, we describe a new series of RASSLs developed to modify the ligand-induced and constitutive signaling of the human 5-HT4 receptor. These modified GPCRs will help us better study the effect of constitutive Gs signaling and ligand-induced Gs, Gs/Gq, and Gi signaling in vivo.


Human 5-HT4 D100A is a Gs-coupled RASSL

The D100A mutation in the mouse 5-HT4 receptor converts it into a RASSL [22], but its effects on the human 5-HT4 receptor have not been tested. We now extend these findings to the human 5-HT4 receptor (Figure 1). To determine if antagonists for the wildtype 5-HT4 receptor also activate the human 5-HT4-D100A mutant, we tested a variety of compounds. The mutant receptor was not activated by serotonin (Figures 2, ,3A),3A), but it was activated by agonists (cisapride, zacopride), partial agonists (RS23597, RS39604, RS67333), antagonists (GR113808, RO110-0235), and an inverse agonist (GR125487) for the wildtype 5-HT4 receptor (Figure 2). Interestingly, GR125487 showed a specific response, as demonstrated by the steep concentration-response curve (Figure 3D). In addition, GR113808, GR125487, and RO110-0235 potently activated Gs signaling of 5-HT4-D100A without stimulating the wildtype receptor (Figures 3B–E). The mutant receptor was selectively activated by multiple synthetic ligands (GR113808, GR125487, and RO110-0235) but not serotonin. We named it Rs1 (RASSL serotonin 1) (Figure 1).

Figure 2
Rs1 is a Gs-coupled RASSL.
Figure 3
Rs1 activation occurs in the nanomolar range.

Rs1 has a high level of constitutive signaling

We next examined the constitutive signaling of Rs1 in more detail. Rs1 showed greater constitutive signaling than the wildtype receptor at all levels of transfection (Figure 4A). Constitutive activity was observable when only 25 ng of receptor cDNA per 5×106 HEK293 cells was transfected (Figure 2). The highest level of constitutive activity, achieved with 5.4 µg of receptor cDNA per 5×106 HEK293 cells, was 1.5 times greater than that of the wildtype 5-HT4 receptor (49.6±1.25 nM vs. 32.5±4.04 nM, p<0.005) and >10-fold higher than that of the control receptors (the β2-adrenergic and parathyroid hormone receptors), which have low levels of constitutive signaling (Figure 4A). Despite the high level of constitutive signaling, both the 5-HT4 receptor and Rs1 could still be further activated by zacopride (Figure 4B).

Figure 4
Rs1 has significant constitutive activity.

Ligand-specific Gq signaling in Rs1

Before constructing Rs1-5HT2C chimeras to make a Gq-signaling RASSL, we assayed inositol phosphate 1 (IP1) accumulation by Rs1 via constitutive or ligand-induced signaling. Rs1 showed no measurable difference in constitutive Gq signaling (Figure 5) as compared to mock-transfected cells in the IP1 and calcium mobilization assays. Upon activation by cisapride, zacopride, RS23597, RS39604, or RS67333, Rs1 showed 2–3.5-fold higher Gq signaling than the wildtype 5-HT4 receptor (p<0.005) (Figure 5). Surprisingly, GR113808, GR125487, and RO110-0235 activated predominately Gs (Figure 2) and little Gq signaling (Figure 5) by Rs1. These are the same ligands used to selectively activate the Gs signaling of Rs1 without activating the Gs signaling of the wildtype 5-HT4 receptor (Figure 2). Therefore, we could use drugs with distinct chemical structures (Figure S1) to activate Gs or Gs/Gq signaling of Rs1. The once controversial use of conformation-specific ligands to alter G-protein coupling and other receptor functions has now been demonstrated in several other GPCRs [25][29]. This is the first time that agonist-dependent functional selectivity has been shown in a RASSL.

Figure 5
The wildtype 5-HT4 receptor, Rs1, and the Rs1-C-5-HT2C chimera exhibit different Gq signaling properties.

Replacing the C-terminus of Rs1 with 5-HT2C increases Gq signaling

To make a purely Gq signaling RASSL from Rs1, we exchanged the intracellular loops of Rs1 with those of the Gq-coupled human 5-HT2C receptor. By transferring domains at different junctions of intracellular loops, we made 12 different Rs1-5-HT2C chimeras (Figures S2, S6, S7, S8, S9, S10, S11, S12, S13, S14, S15, S16, S17). To characterize them, we used RS23597 because it activated Gq signaling of Rs1 but not the wildtype 5-HT4 receptor, as measured by calcium mobilization assays (Figure S3C).

Replacing the second (i2) or third intracellular loop (i3) of Rs1 eliminated both Gs and Gq signaling (Figure S3). Only the carboxyl chimera (Rs1-C-5-HT2C) showed enhanced Gq signaling in response to cisapride, zacopride, RS23597, RS39604, and RS67333 (Figure 5), showing that the ligand-induced specificity of signaling was preserved. Constitutive and ligand-induced Gs signaling were also largely preserved (Figures S3A, S3B).

These data indicated that i2 and i3 are both necessary for Gs-coupling of the 5-HT4 receptor. They also suggested that the C-terminus of 5-HT2C contains a G-protein coupling domain for Gq signaling or that the C-terminus of 5-HT4 receptor promotes Gq signaling. However, we were unable to completely alter the G-protein preference of this receptor for Gq signaling. We amplified the Gq signaling of Rs1 by domain swapping the C-terminus with 5-HT2C. None of the other 12 chimeras showed enhanced Gq signaling, even when multiple internal segments were combined. We did not proceed further with these experiments because substitutions of multiple internal domains also decreased cell-surface expression of the receptors (data not shown). Since Gq signaling of Rs1-C-5-HT2C was activated by cisapride, zacopride, and RS23597 but not by serotonin, we named it Rs1.1 (Table 1).

Table 1
Controlling the G-protein signaling of Rs1

Moreover, when Rs1-C-5-HT2C was activated by cisapride, zacopride, RS23597, RS39604, or RS67333, signaling was 4.5–6.8-fold higher than constitutive signaling (p<0.005). Signaling increased 1.3-fold in response to GR125487 (p<0.05), 1.7-fold in response to GR113808 (p<0.05), and 2.2-fold in response to RO110-0235 (p<0.005). This suggests that the carboxylic tail may play a role in functional selectivity in 5-HT4 or 5-HT2C receptors (Figure S3).

A purely Gs signaling RASSL with low levels of constitutive signaling

Since the high constitutive activity of the 5-HT4-D100A mutant causes significant phenotypes in transgenic mice [30] and could not be controlled by inverse agonists [31], we attempted to lower the Rs1 constitutive activity by making additional point mutations. We focused on the D66N and W272A mutations, which reduce constitutive signaling of the mouse 5-HT4 receptor [31], [32]. Rs1-D66A, Rs1-D66N, and Rs1-W272A significantly reduced constitutive signaling (Figure 6A). The cell-surface expression of Rs1-D66A and Rs1-D66N was similar to that of Rs1 (Figure 6B), so the reduction in constitutive signaling was probably not linked to lower cell-surface expression. Surprisingly, the D66A and D66N mutations also abolished zacopride-induced Gq signaling (Figure 6C). Thus, we created two RASSLs exhibiting pure Gs signaling and low constitutive signaling. Unfortunately, the efficacy of the ligand-induced Gs response was also significantly compromised, diminishing the utility of these receptors.

Figure 6
Abolishing the constitutive activity of Rs1 eliminates ligand-induced Gs and Gq signaling.

Engineering Rs1 for Gi Signaling

To engineer a Gi-signaling RASSL based on Rs1, we replaced its intracellular loops with those of 5-HT1A, a Gi signaling receptor [33]. Of four Rs1-5HT1A chimeras (Figures 7, S4, S18, S19, S20, S21), only the two containing i2 and i3 from Rs1 were expressed at a level similar to that of Rs1 (Figure S5D). Replacing those loops abolished constitutive and ligand-induced Gs signaling at both low and high levels of receptor cDNA (25 ng and 4.8 µg per 5×106 cells) (Figures S5A, S5B). Interestingly, this RASSL showed no evidence of constitutive signaling via the Gi or Gs pathway. These findings strongly imply that both i2 and i3 are required for Gs signaling of Rs1. In addition, activation of the Rs1-i3-5-HT1A chimera with zacopride significantly inhibited cAMP accumulation induced by 10 µM apomorphine (agonist for dopamine 1 receptors) in HEK293 cells co-transfected with 1.5 µg of Rs1 receptor and 0.5 µg of dopamine 1 receptor (per 5×106 cells; Figure S5C and Figure S8A). This inhibition was smaller than that of µ-opioid receptor stimulated by [D-Ala2, D-Leu5]-enkephalin (DADLE). Both responses were abolished by 50 nM pertussis toxin, indicating the involvement of Gi signaling (Figure 7B). Unfortunately, the potency (amount of drug needed to reach an effect) of ligand-induced Gi signaling was significantly reduced (Figure 7C). While these results are encouraging, future experiments are needed to determine if low potency may be due to nonspecific effects, or will reproduce in other cell types. Since Rs1-i3-5HT1A exhibited Gi but not Gs signaling, we named it Rs1.3 (Figure 1, Table 1).

Figure 7
Replacing i3 of Rs1 with that of 5-HT1A results in a receptor, Rs1-i3-5-HT1A, with weak Gi signaling.


A new series of RASSLs

We report here a new series of RASSLs to study multiple G-protein signaling pathways. Many GPCRs activate multiple G-protein signaling pathways and exhibit a wide range of constitutive signaling activities. Our new RASSLs will help us better study the effect of stimulating canonical signaling pathways (Gs/Gq, Gs, Gi), with different constitutive signaling using a single receptor backbone and different synthetic agonists. These new RASSLs will also allow a systematic examination of the different functional domains of the 5-HT4B receptor.

These new RASSLs could help us better dissect the physiological significance of constitutive signaling in vivo. Constitutive signaling is crucial for many physiological processes and many diseases. Up to 40% of all GPCRs [34], including the 5-HT4 receptor [35], show significant constitutive activity, and Rs1 could be a good model for these receptors. Like the mouse 5-HT4-D100A receptor [22], Rs1 had a higher level of constitutive signaling than the wildtype 5-HT4 receptor. Using the tetracycline transactivator (Tet) system, we have already made an Rs1 transgenic mouse in which Rs1 expression is driven by the osteoblast-specific Col1α-1 2.3-kb promoter fragment. These mice exhibited dramatically increased bone formation [30]. These and other findings strongly suggest that constitutive signaling can drive potent phenotypic changes in vivo. We found only constitutive Gs signaling in Rs1 but did not observe any constitutive Gq signaling. It would be interesting to generate new or modify existing RASSLs with increased Gi or Gq constitutive signaling in the future to examine the effect of both agonist-mediated and constitutive signaling in various physiological processes. Since signaling would only be dependent on expression rather than on circulating hormones, these RASSLs have an intrinsic advantage for studying constitutive signaling when combined with a Tet system.

In addition, the 5-HT4 RASSLs have a large set of agonists with different activities that could be useful in future studies. Many of the Rs1 agonists used in this study activated Rs1 with an EC50 in the nanomolar range, allowing to us activate Rs1 effectively. Having an abundant selection of ligands for Rs1 helped us find strong evidence of ligand-induced functional selectivity on Rs1. Having a greater variety of drugs to choose from could be valuable for in vivo studies. For instance, we could use GR125487 to activate Rs1 while suppressing the basal activity of wildtype 5-HT4. Alternatively, we could potentially study Rs1 constitutive signaling by using drugs such as RO116-0086 or RO116-1148 [31] to decrease constitutive signaling of the wildtype 5-HT4 receptor but not of Rs1.

Finally, this new series of RASSLs may make it possible to perform in vivo studies in which Rs1 is activated with minimal side effects. Since knockout of the 5-HT4 receptor does not cause overt side effects [36], treatment with GR113808, GR125487, and R0110-0235 (antagonists and inverse agonists for the wildtype receptor) may have minimal side effects as well. This eliminates the need to knock out the endogenous 5-HT4 receptors for most in vivo experiments [10]. Since the same agonists can be used to activate all of the RASSLs within this series and therefore+ engage different G-protein signaling pathways (Table 1), we can more easily compare the effect of activating the Gs, Gs/q, or Gi pathway.

While these new RASSLs may be useful for studying G protein signaling in vivo, they have several limitations. Despite significant efforts to screen all available antagonists, we could not identify inverse agonists that would lower constitutive signaling by Rs1. Fortunately, we can use conditional expression systems such as the tetracycline transactivator system to control Rs1 expression and constitutive activation. Although we were successful in finding mutations that reduced Rs1 constitutive signaling these same mutations adversely affected the agonist mediated signaling (Rs1.2), suggesting that the D66N and W272A mutations are critical for all aspects of Gs signaling. Other mutations could be explored that may provide a more optimal reduction in Gs basal activity without affecting ligand-induced receptor activation. In addition, the relatively small Gi signaling of Rs1.3 can only be activated by a relatively high concentration of zacopride, thus limiting its usefulness in vivo.

Insights into G-protein signaling by Rs1

Our study yielded insights into the G-protein selectivity and functional selectivity (differential effects of ligands on the same receptor) of Rs1. Of the 12 Rs1-5-HT2C chimeras that are expressed on the cell surface, none of the i2 or i3 chimeras showed any Gs or Gq signaling. Evidently, these intracellular loops of Rs1 are crucial for signaling via those pathways. The importance of i2 and i3 for Gs signaling of Rs1 is further supported by the lack of Gs signaling by the Rs1-i2-5-HT1A and Rs1-i3-5-HT1A chimeras. This is the first study showing the importance of i2 and i3 in both Gs and Gq signaling of the human 5-HT4-D100A receptor.

We also found that i3 domain swapping abolished all Gs signaling and enabled Rs1 to stimulate Gi signaling of 5-HT1A. The role of i2 and i3 in the Gi signaling of 5-HT1A receptor has been extensively reported. The entire N-terminus [37] and C-terminus of i2 [38] of 5-HT1A are thought to be sufficient to support G-protein coupling, but not signaling. On the other hand, the N-terminus [39] and C-terminus of i3 of 5-HT1A [40], [41] seem to be essential for the Gi signaling of 5-HT1A. In fact, replacing the N-terminus of the i3 of the α2-adrenergic receptor with that of 5-HT1A resulted in a chimera that signals like a 5-HT1A receptor when stimulated by a α2-adrenergic receptor agonist [42]. Since Rs1-5HT1A chimeras with multiple internal domains replaced are not significantly expressed on the cell surface (data not shown), it may be difficult to further improve the potency of the Rs1.3 using our current approach. We hypothesize that replacing the N- and C-terminal portions of i2 and i3 instead of the whole i2 and i3 loops may increase the potency of Rs1.

We also showed that various drugs can differentially activate G-protein signaling of Rs1. Functional selectivity has been reported for many receptors. It led to divergent fates of internalization for the dopamine D1 receptor [25], various binding specificities for gonadotropin-releasing hormone receptors, and different levels of activation of G proteins for the β2-adrenergic [26], mu-opioid [27], dopamine D2 [28], and human 5-HT2A [29] receptors.

The indoleamine derivatives GR113808, GR125487, and RO110-0235 did not fully activate Gq signaling of Rs1, Rs1.1, or the 5-HT4 receptor. On the other hand, the benzamide derivatives cisapride, zacopride, RS23597, RS39604, and RS67333 activated the Gq signaling of Rs1. These findings may reflect distinct conformational changes caused by indoleamine and benzamide derivatives.

The possibility of functional selectivity is further supported by the results obtained with Rs1-C-5HT2c and Rs1 point mutants (Rs1-D66A and Rs1-D66N). The D100A mutation and replacement of the C-terminus amplified Gq signaling by the 5-HT4 receptor. The addition of D66A and D66N abolished Gq signaling. Since D100A is located in the binding pocket of the 5-HT4 receptor, this mutation in Rs1 may have changed the configuration of the binding pocket, making the receptor more susceptible to Gq activation by zacopride and RS23597. This response was even more pronounced when the D100A mutation was combined with domain swapping of the C-terminus with that of 5-HT2C. Thus, it is reasonable to hypothesize that these changes modified the ligand-selective receptor conformation [5], changing the receptor susceptibility to functional selectivity.


Our studies with Rs1 provide a proof-of-concept for making a series of RASSLs with different signaling properties. Recently, Armbruster et al. made a series of RASSLs based on the muscarinic M3 and M4 receptors, which have low constitutive activity. These RASSLs each couple different G-protein signaling pathways and can be activated by clozapine-N-oxide, an inert ligand with high bioavailability [16]. These RASSLs nicely complement our Rs1 RASSLs with varying constitutive activity. In addition, we predict that some RASSLs with the same canonical G-protein signaling (Gs, Gi, or Gq) will have different in vivo phenotypes due to noncanonical signaling. This growing collection of RASSLs will greatly facilitate our efforts to understand the physiological significance of the inherent signaling diversity of GPCRs.

An ideal series of RASSLs would have receptors with different combinations of low and high basal signaling, with robust ligand-induced effects for each major pathway, and potent inverse agonists. Although we have not achieved this goal with the Rs1 series, we are hopeful that it can be achieved with other receptors in the future. Indeed, RASSLs based on the muscarinic receptors [16] show great promise, as there are naturally occurring, or published mutants of the muscarinic reports that activate each of the major G protein signaling pathways.

Materials and Methods

Constructing human 5-HT4 mutant cDNA and Rs1-5-HT1A, and Rs1-5-HT2C chimeras

The human 5-HT4 receptor cDNA (a gift from Dr. Bryan Roth, University of North Carolina) was used in all experiments. To improve expression and allow detection of the receptor, we added a signal peptide from influenza hemagglutinin [43] and a FLAG epitope (DYKDDDDA) at the N-terminus. 5-HT4 was then subcloned by PCR; the primers, ATCGATCGgcggccgcGTGAGCAAGGGCGAGGAGCTGTTC and ATCGATCG gcggccgcCTAAGTGTCACTGGGCTGAGCAGCC, were inserted into the Not1 restriction site (gcggccgc) of the pUNIV-5-HT2C-INI plasmid to replace the 5-HT2C-INI (a gift from Dr. Bryan Roth) in frame with the signal peptide and the FLAG epitope. The receptor was then mutated (D100A) with a Quick-Change site-directed mutagenesis kit (Stratagene, La Jolla, CA) with primer GTCTTGTTCGGACATCTCTGgccGT CCTGCTCACAACGGCAT-CG (Figures 1, S2). The following mutant sense primers were used: 5-HT4-D66A, TTCATTGTATCTCTTGCTTTTGCGgcaCTGCT GGTTTCGGTGCTGGT-GATG; 5-HT4-D66N, TTCATTGTATCTCTTGCTTTTGCGaacC TGCTGGTTTCGGTGCT-GGTGATG; and 5-HT4-W272A, GTTGCTTCTGCCTCTGCT GGGCGCCAgccTTTGTCA-CCAATATTGTGG. The sense primer used to replace the carboxyl chimera for Rs1.1 (Table 1) was AGTTACTCTTCCgcgGCCGCGAATTCAG TGGATCCACTAGTAAC. The Rs1-5-HT1A and Rs1-5-HT2C chimeras were made by PCR fusion. The mutations are indicated by boldface, lower-case letters.

HEK293 maintenance and electroporation

Early-passage (≤20) HEK293 cells were maintained in high-glucose DMEM (Invitrogen, Carlsbad, CA) supplemented with sodium pyruvate (Invitrogen) and 10% Fetalplex (Gemini Bio-Products, West Sacramento, CA). Receptors were electroporated into HEK293 cells as described [44]. The electroporated cells were reconstituted into a suspension using DMEM with 10% heat-inactivated, dialyzed fetal bovine serum (Thermo-Fisher Scientific, Logan, UT). The transfection efficiency was monitored by flow cytometry, and the cell-surface expression of the receptor was determined by FLAG ELISA (enzyme-linked immunosorbent assay) the next day.


5-HT, isoproterenol, and 3-isobutyl-1-methylxanthine (IBMX) were purchased from Sigma-Aldrich (St. Louis, MO). Cisapride, zacopride, GR113808, GR1254875, RS23597-190 HCl, RS39604 HCl, and RS67333 HCl were from Tocris (Bristol, UK). Human parathyroid hormone peptide (amino acids 1–34) was from Bachem Biosciences (King of Prussia, PA). RO110-0235 was generously donated by Renee Martin (Roche, Palo Alto, CA).

Measuring cell-surface expression by FLAG ELISA

Cell-surface receptor expression was measured with a FLAG ELISA as described [45]. Cells seeded in poly-D-lysine-coated 96-well plates were fixed with 100 µl of 4% paraformaldehyde for 10 min at room temperature, washed, and stained with 100 µl of staining buffer (DMEM, 10% FBS, and 1 mM CaCl2) containing anti-FLAG M1 antibody (1[ratio]1000; Sigma-Aldrich) for 1 h at 25°C. The samples were washed three times with wash buffer (PBS and 1 mM CaCl2) and stained with 100 µl of staining buffer with rat anti-mouse IgG antibody conjugated with horseradish peroxidase (1[ratio]1000; Bio-Rad Laboratories, Hercules, CA). After 30 min, the samples were washed with wash buffer, placed on a rocker for 10 min, and washed again. This process was repeated two more times. Then, 2,2-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) liquid substrate (200 µl; Sigma-Aldrich) was added to the samples. After rocking for 30 min, 200 µl of the substrate was transferred to new 96-well plates, and optical density was measured with Victor 3 (PerkinElmer, Waltham, MA) at 405 nm. All samples contain three replicates, and all experiments were repeated at least three times.

Measurement of cAMP production in intact cells

To improve assay consistency and minimize pipetting error in the 384-well plates, we modified the high-range HTRF assay (CisBio International, Bagnols-sur-Cèze, France) by seeding, stimulating, and lysing the cells in 96-well plates and using the lysate instead of live cells to determine cAMP production. The remainder of the analysis was performed according to the manufacturer's instructions

Gi Assay

Gi signaling was examined in cells transfected with 1.5 µg of receptor cDNA, 0.6 µg of human dopamine 1 receptor cDNA, and pcDNA3 (up to 6 µg). The co-transfectants were stimulated first with 100 µl of KRBG buffer containing IBMX for 10 min at room temperature and then with 50 µl of PBS containing 10 µM apomorphine (agonist for the dopamine 1 receptor) and 10 µM zacopride for 10 min at 37°C. The cells were lysed in 50 µl of lysis buffer, and 5 µl of lysate was used in the HiRange HTRF assay.

Fluorometric imaging plate reader assay to measure calcium mobilization

To measure calcium mobilization, 4.8 µg of receptor cDNA, 0.6 µg of DsRed plasmid, and 0.6 µg of human bombesin receptor cDNA were electroporated into 5×106 HEK293 cells as above [44]. Hank's balanced salt solution (10 ml) with 20 mM HEPES, 0.25 mM probenecid acid (Sigma-Aldrich), and 2% pluronic acid (Sigma-Aldrich) was added to each bottle of Calcium 4 (Molecular Devices, Sunnyvale, CA), and 100 µl of the resulting solution was added to each well for 1 h at 37°C before measurement. Assays were performed with a FLEX Station (Molecular Devices), with excitation of 485 nm, emission of 525 nm, and cut-off of 515 nm, as recommended by the manufacturer.

Determination of IP1 production in intact cells

A modified version of the IP1 protocol was used (CisBio International). HEK293 cells were washed once with calcium-free PBS and dissociated from flasks with cell dissociation buffer (Invitrogen). Cells (5×106) were electroporated as described above. Then, 105 cells were placed in DMEM supplemented with 10% decomplemented, dialyzed against FBS, and seeded onto 96-well plates coated with poly-D-lysine. The next day, the cells were stimulated with agonists in 50 µl of 1× stimulation buffer for 30 min at 37°C and lysed for 10 min with 9 µl of lysis/detection buffer. Then, 14 µl of lysate was added to 384-well plates and subjected to High-Range HTRF assay as described above, except that 3 µl of cAMP-d2 and anti-cAMP-cryptate solution were added to each well.

Data analysis

cAMP and IP1 values were analyzed with GraphPad Prism 4 (GraphPad Software, San Diego, CA). Calcium mobilization results were analyzed with SoftMax Pro v5 (Molecular Devices). Statistical significance was determined with paired Student's t tests.

Supporting Information

Figure S1

Chemical structures of the compounds used in the study. All the chemicals used in the experimental are shown.

(1.92 MB EPS)

Figure S2

Rs1-5-HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(12.17 MB EPS)

Figure S3

The second and third intracellular loops (i2 and i3) of Rs1 are crucial for Gs and Gq signaling. (A, B) Rs1-5-HT2C chimeras with swaps of i2 and i3 could no longer process Gs signals, at either 25 ng or 4.8 µg of receptor cDNA per 5×106 HEK293 cells. (C) Gq signaling of Rs1 was abolished when i2 and i3 of Rs1 were replaced with those of 5-HT2C. The Gq signaling was measured by calcium mobilization assay. (D) Only chimeras with a single domain swap were expressed on the cell surface. The results represent three independent experiments. All figures were representative of three independent experiments.

(2.03 MB EPS)

Figure S4

Rs1-5-HT1A chimeras. All modifications were made on Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.82 MB EPS)

Figure S5

Replacing the second or third intracellular loop (i2 or i3) of Rs1 with 5-HT1A alters G protein signaling. (A, B) Rs1-5-HT1A chimeras with swaps of the second (i2) and third intracellular (i3) loops no longer signal via the Gs pathway, regardless of whether cells were transfected with 25 ng or 4.8 µg of receptor cDNA per 5×106 HEK293 cells. This suggests that the second and third intracellular loops are crucial for acute Gs signaling of Rs1. ***p<0.001 vs. mock transfected (t test). (C) Replacing i3 of Rs1 resulted in a Gi signaling receptor. All HEK293 transfectants were electroporated with 0.6 µg of the human dopamine 1 receptor and 1.5 µg of Rs1, Rs1-5HT1A chimeras, or the mu-opioid receptor. Rs1 and Rs1-5HT1a chimeras were treated with 10 µM apomorphine (an agonist for dopamine 1 receptor) or with 10 µM apomorphine and 10 µM zacopride. Transfectants with 0.6 µg of the human dopamine 1 receptor and 1.5 µg of the mu-opioid receptor served as positive controls. 10 µM DADLE was used in place of zacopride to stimulate mu-opioid receptors. ***p<0.005, *p<0.05 vs. apomorphine (t test). (D) Rs1, (i1, i2 and i30 were expressed on the cell surface. Cell-surface expression and calcium mobilization of the chimeras were examined at 4.8 µg of receptor DNA per 5×106 HEK293 cells. The results are representative of three independent experiments. Values are mean±SD. ***p<0.001 vs. mock transfected (t test).

(1.67 MB EPS)

Figure S6

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.98 MB EPS)

Figure S7

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.96 MB EPS)

Figure S8

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.00 MB EPS)

Figure S9

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.99 MB EPS)

Figure S10

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.99 MB EPS)

Figure S11

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.99 MB EPS)

Figure S12

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.00 MB EPS)

Figure S13

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.99 MB EPS)

Figure S14

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.00 MB EPS)

Figure S15

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.99 MB EPS)

Figure S16

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.00 MB EPS)

Figure S17

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.96 MB EPS)

Figure S18

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.00 MB EPS)

Figure S19

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(2.96 MB EPS)

Figure S20

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.40 MB EPS)

Figure S21

High-resolution representation of Rs1-5HT1A and Rs1-5HT2C chimeras. All modifications made to Rs1 (Figure 1). Red lines indicate the junctions of chimeras. The amino acids exchanged are shown by the amino acid alignment.

(3.03 MB EPS)


We thank Kimberly Scearce-Levie and Renee Martin for insightful review of the manuscript, Bryan Roth for the human 5-HT4B and 5-HT2C receptors cDNA, Robert J. Lefkowitz for the human 5-HT1A receptor cDNA, Mark von Zastrow for the human β2-adrenergic receptor cDNA, human µ-opioid receptor cDNA, and DADLE, Renee Martin for RO110-0235 (Roche Pharmaceuticals Division, Palo Alto, CA), Mary Weglarz for manuscript preparation and Gary Howard and Stephen Ordway for editorial review.


Competing Interests: The authors have declared that no competing interests exist.

Funding: This study was supported by NIH grant HL-60664 and American Heart Association Predoctoral Fellowship 0415005Y (to WCC), NIH Fellowship Training Grant 2T32DK07418-26 (to ECH), J David Gladstone CIRM Fellowship (to ECH), and Kirschstein NRSA Fellowship F32A624044-02 (to JKN).


1. Karchin R, Karplus K, Haussler D. Classifying G-protein coupled receptors with support vector machines. Bioinformatics. 2002;18:147–159. [PubMed]
2. Fredriksson R, Lagerstrom MC, Lundin LG, Schioth HB. The G-protein-coupled receptors in the human genome form five main families. Phylogenetic analysis, paralogon groups, and fingerprints. Mol Pharmacol. 2003;63:1256–1272. [PubMed]
3. Brink CB, Harvey BH, Bodenstein J, Venter DP, Oliver DW. Recent advances in drug action and therapeutics: relevance of novel concepts in G-protein-coupled receptor and signal transduction pharmacology. Br J Clin Pharmacol. 2004;57:373–387. [PMC free article] [PubMed]
4. Maudsley S, Martin B, Luttrell LM. The origins of diversity and specificity in g protein-coupled receptor signaling. J Pharmacol Exp Ther. 2005;314:485–494. [PMC free article] [PubMed]
5. Kenakin T. Ligand-selective receptor conformations revisited: the promise and the problem. Trends Pharmacol Sci. 2003;24:346–354. [PubMed]
6. Scearce-Levie K, Coward P, Redfern CH, Conklin BR. Engineering receptors activated solely by synthetic ligands (RASSLs). Trends Pharmacol Sci. 2001;22:414–420. [PubMed]
7. Srinivasan S, Santiago P, Lubrano C, Vaisse C, Conklin BR. Engineering the melancortin-4 receptor to control constitutive and ligand-mediated Gs signaling in vivo. PLoS One. In press [PMC free article] [PubMed]
8. Coward P, Wada HG, Falk MS, Chan SD, Meng F, et al. Controlling signaling with a specifically designed Gi-coupled receptor. Proc Natl Acad Sci USA. 1998;95:352–357. [PubMed]
9. Zhao GQ, Zhang Y, Hoon MA, Chandrashekar J, Erlenbach I, et al. The receptors for mammalian sweet and umami taste. Cell. 2003;115:255–266. [PubMed]
10. Sweger EJ, Casper KB, Scearce-Levie K, Conklin BR, McCarthy KD. Development of hydrocephalus in mice expressing the Gi-coupled GPCR Ro1 RASSL receptor in astrocytes. J Neurosci. 2007;27:2309–2317. [PubMed]
11. Redfern CH, Coward P, Degtyarev MY, Lee EK, Kwa AT, et al. Conditional expression and signaling of a specifically designed Gi-coupled receptor in transgenic mice. Nat Biotechnol. 1999;17:165–169. [PubMed]
12. Pauwels PJ. Unravelling multiple ligand-activation binding sites using RASSL receptors. Trends Pharm Sci. 2003;24:504–507. [PubMed]
13. Peng J, Bencsik M, Lu W, Storer A, Burghardt A, et al. Osteopenia produced by expression of a Gi-activating receptor in osteoblasts. J Bone Miner Res. 2006;21:s1–s530.
14. Srinivasan S, Vaisse C, Conklin BR. Engineering the melanocortin-4 receptor to control G(s) signaling in vivo. Ann N Y Acad Sci. 2003;994:225–232. [PubMed]
15. Bruysters M, Jongejan A, Akdemir A, Bakker RA, Leurs R. A G(q/11)-coupled mutant histamine H(1) receptor F435A activated solely by synthetic ligands (RASSL). J Biol Chem. 2005;280:34741–34746. [PubMed]
16. Armbruster BN, Li X, Pausch MH, Herlitze S, Roth BL. Evolving the lock to fit the key to create a family of GPCRs potentially activated by an inert ligand. Proc Natl Acad Sci USA. 2007 [PubMed]
17. Claeysen S, Sebben M, Becamel C, Eglen RM, Clark RD, et al. Pharmacological properties of 5-Hydroxytryptamine(4) receptor antagonists on constitutively active wild-type and mutated receptors. Mol Pharmacol. 2000;58:136–144. [PubMed]
18. Crowell MD, Mathis C, Schettler VA, Yunus T, Lacy BE. The effects of tegaserod, a 5-HT receptor agonist, on gastric emptying in a murine model of diabetes mellitus. Neurogastroenterol Motil. 2005;17:738–743. [PubMed]
19. Orsetti M, Dellarole A, Ferri S, Ghi P. Acquisition, retention, and recall of memory after injection of RS67333, a 5-HT4 receptor agonist, into the nucleus basalis magnocellularis of the rat. Learn Mem. 2003;10:420–426. [PubMed]
20. Kaumann AJ, Levy FO. 5-Hydroxytryptamine receptors in the human cardiovascular system. Pharmacol Ther. 2006;111:674–706. [PubMed]
21. Lefebvre H, Cartier D, Duparc C, Lihrmann I, Contesse V, et al. Characterization of serotonin(4) receptors in adrenocortical aldosterone-producing adenomas: in vivo and in vitro studies. J Clin Endocrinol Metab. 2002;87:1211–1216. [PubMed]
22. Claeysen S, Joubert L, Sebben M, Bockaert J, Dumuis A. A single mutation in the 5-HT4 receptor (5-HT4-R D100(3.32)A) generates a Gs-coupled receptor activated exclusively by synthetic ligands (RASSL). J Biol Chem. 2003;278:699–702. [PubMed]
23. Wong SK. G protein selectivity is regulated by multiple intracellular regions of GPCRs. Neurosignals. 2003;12:1–12. [PubMed]
24. Hoyer D, Hannon JP, Martin GR. Molecular, pharmacological and functional diversity of 5-HT receptors. Pharmacol Biochem Behav. 2002;71:533–554. [PubMed]
25. Ryman-Rasmussen JP, Griffith A, Oloff S, Vaidehi N, Brown JT, et al. Functional selectivity of dopamine D1 receptor agonists in regulating the fate of internalized receptors. Neuropharmacology. 2007;52:562–575. [PMC free article] [PubMed]
26. Ghanouni P, Gryczynski Z, Steenhuis JJ, Lee TW, Farrens DL, et al. Functionally different agonists induce distinct conformations in the G protein coupling domain of the beta 2 adrenergic receptor. J Biol Chem. 2001;276:24433–24436. [PubMed]
27. Saidak Z, Blake-Palmer K, Hay DL, Northup JK, Glass M. Differential activation of G-proteins by mu-opioid receptor agonists. Br J Pharmacol. 2006;147:671–680. [PMC free article] [PubMed]
28. Gay EA, Urban JD, Nichols DE, Oxford GS, Mailman RB. Functional selectivity of D2 receptor ligands in a Chinese hamster ovary hD2L cell line: evidence for induction of ligand-specific receptor states. Mol Pharmacol. 2004;66:97–105. [PubMed]
29. Urban JD, Clarke WP, Zastrow M von, Nichols DE, Kobilka B, et al. Functional selectivity and classical concepts of quantitative pharmacology. J Pharmacol Exp Ther. 2007;320:1–13. [PubMed]
30. Hsiao EC, Boudignon B, Chang PW, Bencsik M, Peng J, Manalac C, Halloran BP, Conklin BR, Nissenson RA. A stone mouse: transgenic expression of an engineered Gs-coupled receptor in osteoblasts produces increased bone mass. J Bone Miner Res. 2007;22(Suppl 1):S6.
31. Joubert L, Claeysen S, Sebben M, Bessis AS, Clark RD, et al. A 5-HT4 receptor transmembrane network implicated in the activity of inverse agonists but not agonists. J Biol Chem. 2002;277:25502–25511. [PubMed]
32. Rivail L, Giner M, Gastineau M, Berthouze M, Soulier JL, et al. New insights into the human 5-HT4 receptor binding site: exploration of a hydrophobic pocket. Br J Pharmacol. 2004;143:361–370. [PMC free article] [PubMed]
33. Liu YF, Ghaharemani MH, Raesnik MM, Jakobs KH, Albert PR. Stimulation of cAMP synthesis by Gi-coupled receptors upon ablation of distinct Gαi protein expression. Gi subtype specificity of the 5-HT1A receptor. J Biol Chem. 1999;274:16444–16450. [PubMed]
34. Seifert R, Wenzel-Seifert K. Constitutive activity of G-protein-coupled receptors: cause of disease and common property of wild-type receptors. Naunyn Schmiedebergs Arch Pharmacol. 2002;366:381–416. [PubMed]
35. Claeysen S, Sebben M, Becamel C, Parmentier ML, Dumuis A, et al. Constitutively active mutants of 5-HT4 receptors are they in unique active states? EMBO Rep. 2001;2:61–67. [PubMed]
36. Compan V, Charnay Y, Dusticier N, Daszuta A, Hen R, et al. Feeding disorders in 5-HT4 receptor knockout mice. J Soc Biol. 2004;198:37–49. [PubMed]
37. Thiagaraj HV, Ortiz TC, Devereaux MC, Jr, Seaver B, Hall B, et al. Regulation of G proteins by human 5-HT1a receptor TM3/i2 and TM5/i3 loop peptides. Neurochem Int. 2007;50:109–118. [PubMed]
38. Kushwaha N, Harwood SC, Wilson AM, Berger M, Tecott LH, et al. Molecular determinants in the second intracellular loop of the 5-hydroxytryptamine-1A receptor for G-protein coupling. Mol Pharmacol. 2006;69:1518–1526. [PubMed]
39. Ortiz TC, Devereaux MC, Parker KK. Structural variants of a human 5-HT1A receptor intracellular loop 3 peptide. Pharmacology. 2000;60:195–202. [PubMed]
40. Hayataka K, O'Connor MF, Kinzler N, Weber JT, Parker KK. A bioactive peptide from the transmembrane 5-intracellular loop 3 region of the human 5HT1a receptor. Biochem Cell Biol. 1998;76:657–660. [PubMed]
41. Malmberg A, Strange PG. Site-directed mutations in the third intracellular loop of the serotonin 5-HT1A receptor alter G protein coupling from Gi to Gs in a ligand-dependent manner. J Neurochem. 2000;75:1283–1293. [PubMed]
42. Eason MG, Liggett SB. Identification of a Gs coupling domain in the amino terminus of the third intracellular loop of the alpha 2A-adrenergic receptor. Evidence for distinct structural determinants that confer Gs versus Gi coupling. J Biol Chem. 1995;270:24753–24760. [PubMed]
43. Guan XM, Kobilka TS, Kobilka BK. Enhancement of membrane insertion and function in a type IIIb membrane protein following introduction of a cleavable signal peptide. J Biol Chem. 1992;267:21995–21998. [PubMed]
44. Dumuis A, Bouhelal R, Sebben M, Cory R, Bockaert J. A nonclassical 5-hydroxytryptamine receptor positively coupled with adenylate cyclase in the central nervous system. Mol Pharmacol. 1988;34:880–887. [PubMed]
45. Scearce-Levie K, Lieberman M, Elliott HH, Conklin BR. Engineered G protein coupled receptors reveal independent regulation of internalization, desensitization and acute signaling. BMC Biol. 2005;3:3. [PMC free article] [PubMed]

Articles from PLoS ONE are provided here courtesy of Public Library of Science