PEDS Advance Access originally published online on March 24, 2006
Protein Engineering Design and Selection 2006 19(6):277-283; doi:10.1093/protein/gzl010
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Statistical analysis and prediction of functional residues effective for GPCR–G-protein coupling selectivity
1 Graduate School of Information Science, Nara Institute of Science and Technology (NAIST) 8915-5 Takayama, Ikoma, Nara 630-1010, Japan 2 Computational Biology Research Center (CBRC), National Institute of Advanced Industrial Science and Technology (AIST) AIST Waterfront BIO-IT Research Building, 2-42 Aomi, Koto-ku, Tokyo 135-0064, Japan
3To whom correspondence should be addressed. E-mail: taka-mu{at}is.naist.jp
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
One of the important issues in G-protein-coupled receptor (GPCR) functional analysis is the mechanism of GPCR–G-protein coupling selectivity. G-proteins are classified into Gi/o, Gq/11 and Gs families. Although several experimental and computational analyses have been attempted, the mechanism remains unknown to this day. In this study, we have analyzed the multiple sequence alignments of GPCRs of known coupling selectivities by mapping onto the tertiary structure of rhodopsin. We identified several functional residue sites in GPCRs related to coupling selectivity, which are located mainly at the intracellular loops, and found that the occurrence of positively/negatively charged amino acids of the characteristic residues varies depending on the G-protein coupling selectivity. Especially, the occurrence of positively charged amino acids in receptors coupling to Gs family is less than that in receptors coupling to Gi/o and Gq/11 families. It is interesting that some characteristic residues are located near the extracellular terminus of transmembrane helices, which is far from the GPCR/G-protein binding interface. In most of the receptors coupling to Gs family, the occurrence of proline on the position corresponding to the 170th residue on rhodopsin is rare. These findings are vital to improving our understanding of the mechanism of G-protein coupling selectivity.
Keywords: G-protein-coupled receptor/functional residue site/correlation coefficient/rhodopsin/protein-protein interaction
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
G-protein-coupled receptors (GPCRs) form one of the major protein families and perform several functions in signal transduction. When an external ligand, such as a neurotransmitter, a hormone, a lipid, or an odorant molecule, binds to GPCRs, a variety of cellular events are induced, spreading into the cell through the heterotrimeric G-protein (G
, Gß, G
), which is a signal transducing molecule from the GPCR to the effectors such as enzymes and ion channels (Conklin and Bourne, 1993
). GPCRs are known to have important functions in the cell, and hence pharmaceutical companies have been analyzing them with the aim of discovering novel drug targets (Drews, 2000).
The G-proteins are classified into three major families, Gi/o, Gq/11 and Gs, based on the kind of -subunit (Wess, 1998), and each family has a specific influence on the cell. In this article, receptors coupling to Gi/o-, Gq/11- and Gs families are called Gi/o-, Gq/11- and Gs-coupled receptors, respectively. Gi (subfamily of Gi/o) inhibits adenylyl cyclase and activates potassium channels. Go (subfamily of Gi/o) activates potassium channels, inactivates calcium channels and stimulates phospholipase C. Gs activates adenylyl cyclase and calcium channels, and Gq activates phospholipase C. The fact that some GPCRs bind only one particular G-protein family while others bind multiple families has led to the question of what factors determine G-protein coupling selectivity. Several investigations, including chimeric/mutagenesis experiments (Liu et al., 1995; Erlenbach and Wess, 1998; Ilani et al., 2002) and computational analysis of both GPCRs and G-proteins (Horn et al., 2000; Möller et al., 2001; Cao et al., 2003; Sreekumar et al., 2004; Sgourakis et al., 2005), have been carried out to elucidate the mechanisms underlying the coupling selectivity.
Rapid progress in genome studies has led to the discovery of novel GPCRs of unknown functions, often called ‘orphan GPCRs’, and hence it is necessary to determine rapidly their ligands and the associated signal transduction mechanisms. It is crucial as well to predict coupling selectivity, because knowledge of it enables one to choose a suitable assay for ligand identification.
It is well known that proteins having a common function are related through the evolutionary process. On the basis of this principle, many computational methods, such as conventional homology search (Smith and Waterman, 1981; Altschul et al., 1990), protein family profile search (Eddy, 1998) and evolutionary trace (Lichtarge et al., 1996; Madabushi et al., 2004), were developed. However, GPCRs that couple to the same G-protein family do not always show an evolutionary relationship. For example, the muscarinic acetylcholine receptors having very high sequence similarities bind different G-protein families. On the other hand, no common patterns in sequences of GPCRs with the same coupling selectivity were found (Wess, 1998; Möller et al., 2001). These features make it difficult to predict coupling selectivity. In order to resolve this issue, it is essential to analyze the sequences from the perspective of the GPCR structure, because the G-protein coupling process should be restricted by the tertiary structure at the binding site in which the sequence similarity is not always conserved although the structure of this region is similar.
In this work, we conducted position-specific profile analysis by using multiple sequence alignments of the Class A GPCR family [this classification is used in GPCRDB (Horn et al., 2003)] to reveal G-protein coupling selectivity, since the Class A GPCR family is the largest and has a 3D structural representative (rhodopsin). Functional residue sites that are involved in G-protein selection were determined by comparing their alignment profiles with G-protein coupling selectivity information. We mapped the functional residue sites onto the tertiary structure of bovine rhodopsin (Palczewski et al., 2000) and investigated their steric locations and the amino acid frequencies at these positions. We found that the functional residue sites are located mainly at the intracellular side of helices, although some of them are located at the extracellular side of helices and the extracellular loop. We also found that charged residues in the intracellular domain and proline at the 170th position mapped onto the tertiary structure of rhodopsin are quite effective for discriminating the Gs-coupled receptors.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
GPCR dataset for evaluation of functional residue sites
GPCR sequences whose binding G-protein is well characterized were gathered from the Class A GPCR family because this family includes rhodopsin, whose 3D structure is already known, thereby enabling alignment of the other Class A members with this structure. ‘Class A’ is used in the GPCRDB (Horn et al., 2003). The 111 human GPCR sequences (Gi/o-coupled receptors: 55; Gq/11-coupled receptors: 34; Gs-coupled receptors: 22) were obtained from Swiss-Prot (Boeckmann et al., 2003), TrEMBL and partly from Entrez Protein (Maglott et al., 2005). Recently, the concept of multiplicity of G-protein coupling was proposed (Hermans, 2003) in which phosphorylation switching, receptor overexpression, etc. induce GPCR binding of multiple types of G-proteins. Because the promiscuous G-protein coupling mechanisms are not well understood, GPCRs that bind multiple types of G-proteins as reported in the literature (Alexander et al., 2001) were excluded from the dataset.
Multiple sequence alignment
Several points must be considered for the multiple sequence alignment of GPCR sequences. Firstly, membrane protein sequences have both hydrophobic transmembrane (TM) domains and hydrophilic domains; therefore, it is difficult to generate multiple sequence alignment that is applicable to both domains. Most multiple sequence alignment tools have given undesirable alignment results, such as the inclusion of unexpected gaps in TM domains, etc. Secondly, Class A GPCRs have several highly conserved residues as indicated previously (Van Rhee and Jacobson, 1996), and such residues that are dominant in each TM domain provide significant information for the accurate sequence alignment of large members of the Class A GPCR family. Finally, these alignments can be mapped onto the tertiary structure of rhodopsin, a member of the Class A GPCR family. Taking these points into consideration, we applied the profile hidden Markov model (HMM) (Eddy, 1998), which was mapped onto the tertiary structure of rhodopsin. The profile HMM enables multiple sequence alignment of common protein domains. To construct the HMM profile, we obtained 3482 Class A GPCR sequences from GPCRDB 7.0 (Horn et al., 2003), excluding sequences with amino acid residues Asx, Glx and Xaa and constructed a Class A non-redundant sequence set, including the bovine rhodopsin sequence and having 30% sequence identity or less. Multiple sequence alignment of this non-redundant sequence set was generated using CLUSTAL W 1.82 (Thompson et al., 1994). Next, we used HMMER (Eddy, 1998) to construct the HMM profile from the alignment. We used the hand model architecture construction option in order to assign the column that was aligned with rhodopsin residue to the match state in the model. The parameters of the model were modified as the model had match states in the TM helical regions and the insertion states could not be repeated in the TM non-helical regions. The TM helical regions were determined from the tertiary structure of bovine rhodopsin (1L9H) at 2.6 Å resolution (Okada et al., 2002) using DSSP program (Kabsch and Sander, 1983). HMMER aligned GPCR sequences including the bovine rhodopsin sequence to the HMM profile. We analyzed the columns that were aligned with rhodopsin residue, and had more than 90% non-gaps from the generated alignment, and called them ‘valid columns’.
The lengths of several loops of GPCRs, such as IL3 and C-terminal loop, vary markedly. Few valid columns are included in the middle regions of the loops, because the columns on the loops include many gaps and the columns on loops longer than those of rhodopsin are not aligned with rhodopsin residue. The evaluation of the residues in the middle regions of the loops is difficult with our method. However, as the N- or C-terminus of long loops, rather than the middle regions, is intimately involved in G-protein recognition (Wess, 1998), most of the columns in the N- or C-terminus of long loops were included in the valid columns.
Scoring for detection of functional residue sites
In this work, Pearson's correlation coefficient score (CCS) is defined as the amino acid profile distance between different types of receptor groups that are classified according to the type of coupled G-protein family. Although several scoring schemes were proposed, it was suggested that CCS shows the best performance for evaluating alignment accuracy (Marti-Renom et al., 2004).
Pearson's CCS for evaluating amino acid profile distance is expressed as
 | (1) |
is the probability of type i amino acid occurrence at position c in rhodopsin (1L9H) for a query GPCR that binds G-protein type g (Gi/o, Gq/11 and Gs) and 
is that for the GPCR that does not bind G-protein type g at position c. is estimated using the weighted frequency method (Henikoff and Henikoff, 1994) and the nine-component Dirichlet mixture density method (Sjölander et al., 1996). 
is the mean value of . CCS represents the distance between two probability and distributions to measure the significance of amino acid frequency in relation to the coupling selectivity.
We used seven classes of amino acid residues modified from Mirny's category (Mirny and Shakhnovich, 1999) as the amino acid types. The seven classes are aliphatic {AVLIMC}, aromatic {FWYH}, polar {STNQ}, positively charged {KR}, negatively charged {DE}, proline {P} and glycine {G}. Special {PG} in Mirny's category is divided into proline and glycine classes because these residues have different conformational properties in TM proteins.
Generation of random CCS distribution
To evaluate the significance of CCS, random CCS distributions were constructed for each G-protein binding type by the random sampling method. From 3482 Class A GPCR sequences, equal numbers of sequences for 55 Gi/o-coupled receptors, 34 Gq/11-coupled receptors and 22 Gs-coupled receptors were randomly chosen (see GPCR dataset section), and one valid column was randomly selected from their multiple alignments. Random CCS distribution was calculated from the random frequency estimated from these residues in the selected column with the same method described above. Random CCS distributions were generated by 1 000 000 trials of this procedure using ‘Mersenne Twister’ (Matsumoto and Nishimura, 1998) as the random generator. We defined the P-value as the probability of a CCS occurring in the random CCS distribution.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Significance of CCS
In order to identify the residue positions that are responsible for binding a particular G-protein type, Pearson's CCS was calculated for this G-protein type and the remaining G-protein type for each residue position that is mapped onto the tertiary structure of rhodopsin by multiple alignment. If a residue position shows a significantly low CCS for binding a particular G-protein type, that position is considered to be effective for binding the G-protein or disrupting the remaining G-protein type.
To test the statistical significance of this analysis, we generated random CCS distributions by using random parameters. The random CCS distributions differed among the Gi/o, Gq/11- and Gs-protein-coupled receptors (Figure 1) and suggested the dependence of the CCS distribution on the data size of the GPCR sequences. This finding indicates that the raw CSSs of different G-protein types should not be compared in the absence of any compensation. Therefore, we used the P-value, which is the probability of a CCS occurring in the random CCS distribution. In this study, we chose residue positions with P-values less than 0.01 (significance level) as listed in Table I and called them functional residue sites because their significantly small P-values indicated correlation with a specific G-protein type.
|
|
Domains including predicted functional residue sites
The residue position on each GPCR is described as the position mapped onto the tertiary structure of bovine rhodopsin (1L9H) that is used as the template for the multiple alignment of GPCR sequences (Figures 2 and 3; Table I). It was shown that the sequences of the Class A GPCR family [according to the classification used in GPCRDB (Horn et al., 2003)] share several conserved residues at each position. The functional residue sites are represented by blue spheres in the rhodopsin structure, as viewed from the intracellular side (left figure) and from the axis parallel to the membrane (right side) in Figure 3. Figures 2 and 3 and Table I indicate that the number of predicted functional residue sites in the intracellular loops is larger than those in the other domains. This is reasonable because GPCRs interact with G-proteins on the cytoplasmic side. We also found other notable functional residue sites on the extracellular side of TM helices, especially in TM3.
|
|
Figures 2 and 3a show the locations of the predicted functional residue sites in Gi/o-coupled receptors, as mapped onto the tertiary structure of rhodopsin. They are found at positions 33, 36, 107 and 279 on the extracellular side of TM helices, position 115 in the TM domain (TM3) and positions 65, 148, 233, 244, 311 and 319 on the intracellular side of protein. In the last case, positions 65 and 319 are adjacent to each other whereas the other positions are distributed sparsely (Figure 3a).
As shown in Figures 2 and 3c, the predicted functional residue sites in Gs-coupled receptors are classified into three clusters that are located at the third intracellular loop (IL3), both IL1 and the C-terminal loop, and the TM3-4 domain. In the IL3 cluster, position 248 is located at the intracellular side of TM4 and position 233 is located at the middle of IL3. In the IL1-C-terminal cluster, three positions 67, 69 and 312 are located closely at the intracellular side of GPCRs mapped onto the tertiary structure of rhodopsin (Figure 3c).
It is noted that there are few residue positions reaching the significance level in Gq/11-coupled receptors (Figures 2 and 3b; Table I). They are position 312 at the membrane-bound helix VIII (Figure 2) of the C-terminal loop, position 231 in IL3 and position 107 on the extracellular side of TM3. This analysis suggested that the sequences of Gq/11-coupled receptors are less capable of selecting G-protein types than those of Gi/o- and Gs-coupled receptors. This may be one of the reasons for the failure of several prediction methods based on intracellular domain analysis to distinguish Gq/11-coupled receptors from other G-coupled receptors (Cao et al., 2003).
Table II shows the occurrence of positively or negatively charged amino acids on the intracellular side of the protein. The distribution of positively/negatively charged amino acids at each position shows characteristic features related to G-protein coupling selectivity. In particular, position 248 in IL3 is distinct in that positively charged amino acids are found in many (>20) of the Gi/o- or Gq/11-coupled receptors but not in the Gs-coupled receptors except for one case. In the case of negatively charged residues, the fraction at position 311 shows a large difference (>0.3) between Gi/o-coupled receptors and Gq/11- or Gs-coupled receptors. By contrast, the negatively charged residue fraction at position 312 shows a large difference (>0.2) between Gs-coupled receptors and Gi/o- or Gq/11-coupled receptors.
|
It is suggested that the functional residue sites are located mainly in the intracellular domains of GPCRs. However, Table I and Figure 2 indicate the existence of these sites also on the extracellular side. For example, positions 33, 36, 106, 107, 170, 171 and 279 are located at or near the extracellular terminus of helices. Among these positions, the 170th position shows the characteristic influence of a specific amino acid on the extracellular side of the helix. As shown in Table III, the occurrence of proline at the position corresponding to the 170th residue on bovine rhodopsin (170P) is observed predominantly in the Gi/o- and Gq/11-coupled receptors but not in the Gs-coupled receptors, except for the rare cases of 5-HT-7 receptor, adrenomedullin receptor and vasopressin V2 receptor. On the other hand, the occurrence of proline at the 171st position in the Gs-coupled receptors is similar to that in the Gi/o-coupled receptors.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Analysis of the predicted functional residue sites mapped onto the tertiary structure of rhodopsin reveals potentially important residues that are related to G-protein coupling selectivity. These residues are classified into two groups: (i) those in the intracellular domain that interact directly with G-proteins and (ii) those in other domains, such as TM and extracellular domains, which are distant from the GPCR/G-protein binding interface. In what follows, we described the relationship between potential functional residue sites and G-protein coupling selectivity in each domain.
Intracellular domains
Since residues in the intracellular domain interact directly with G-protein, the residues in this domain play an important role in determining G-protein coupling selectivity. Several mutational analyses have suggested the importance of the intracellular domains, especially IL2 and the N/C-terminus of IL3 (Wess, 1998), for determining coupling selectivity. Table I shows single or multiple substitution mutation data of G-protein coupling or receptor activation at positions of the functional residue sites. As shown in Table I, half of the functional residue sites in the intracellular domain are involved in G-protein coupling or receptor activation. The result indicated the importance of residues in the intracellular domain.
Previous mutational analyses have indicated that charged residues in IL play a key role in the recognition and activation of G-proteins (Wang, 1997). We therefore investigated the relationship between the predicted functional residue sites and the occurrence of charged residues. Table II shows the occurrence of positively or negatively charged amino acids on the intracellular domain. The occurrence of positively charged residues at position 248 in IL3 is more frequent in the Gi/o- or Gq/11-coupled receptors than in the Gs-coupled receptors. Several point mutation experiments indicated that positively charged amino acids at position 248 are required for coupling to Gi/o or Gq/11 family (Cotecchia et al., 1990; Lee et al., 1996; Wang, 1997; Wade et al., 1999). They indicated that the lack of positively charged amino acids at position 248 might disrupt coupling to the Gi/o or Gq/11 family. These point mutation experiments suggested the validity of our evaluation of position 248. Although several residues at the C-terminus of IL3, the CCSs of which have no statistical significance, show similar propensity, the difference in the positively charged amino acid frequency at positions 247 and 252 between Gs-coupled receptors and the other receptors is large (>0.25). Mutational experiments indicated that the positively charged residues at positions 247 and 252 play a role in the coupling to Gi/o or Gq/11 family (Wang, 1997, 1999). These results suggest that the existence of charged residues at several positions of the IL3 C-terminus affects coupling selectivity.
Extracellular domain and TM domain
Although it is well known that the intracellular domain is one of the most important sites for G-protein coupling selectivity, it has been reported that residues in the other domains, such as EL3 of LH/CG receptors (Gilchrist et al., 1996), play a significant role in the G-protein coupling selectivity. These residues are definitely remote from the intracellular domain that is directly involved in G-protein coupling. As shown in Table I, several functional residue sites are located at the extracellular side of protein. It is considered that the extracellular domain and the extracellular side of TM are involved in agonist binding (Strader et al., 1994; Wess, 1996). These residues might affect or limit the conformational changes for GPCR activation and disrupt coupling with particular G-proteins. We found that position 170 shows the characteristic occurrence of a specific amino acid (proline) on the extracellular side of the helix, which is observed predominantly in Gi/o- and Gq/11-coupled receptors, but not in Gs-coupled receptors. It is known that the proline residue in the helix structure generally plays the role of a helix breaker, which is essential for structure folding. Previous work has suggested that 170P at the extracellular side of the fourth TM helix (TM4) is important for the activation of some GPCRs (Lu et al., 2001; Leung et al., 2004) or agonist binding (Javitch et al., 2000). However, 170P may have roles other than agonist binding because 170P is not always conserved in a particular GPCR family, such as serotonin receptor family, histamine receptor family and dopamine receptor family, although each family binds the same ligand. Therefore, 170P also has roles other than agonist binding. These analyses suggest that 170P may contribute directly or indirectly to determining G-protein coupling selectivity, although there are no mutational experiments to support this suggestion.
What defines G-protein coupling selectivity?
It is assumed that the residues surrounding GPCR/G-protein interaction sites determine G-protein coupling selectivity. Several sites in different domains of the same receptor affect G-protein coupling selectivity, and point mutations effective for changing G-protein coupling selectivity are applied at IL2, IL3 and the C-terminal loop (Wess, 1996; Wong, 2003). In those loops, we also found several significant residue positions by means of statistical analysis. Those residues are dispersed widely at different positions, and we cannot find a local common sequence region, such as a motif pattern, related to coupling selectivity. These findings suggest that not only one or a few residues located close to the region in contact with G-protein define coupling selectivity, but other distant residues collaborate as well. It may be that although these residues are isolated along a one-dimensional sequence, they can come near a region in the 3D structure (Figure 3) and the residues that do not interact directly with G-proteins also influence G-protein activation through conformational change of the GPCR.
The hypothesis that the tertiary structure at the intracellular domain determines G-protein coupling selectivity can account for the fact that the change of G-protein coupling selectivity is due to either a few intracellular residues or the residues remote from the intracellular domain (Wong, 2003). Our study shows several functional residue sites in the TM and extracellular domains that contribute to G-protein activation, as shown in Table I and Figures 2 and 3. It may be that these functional residue sites affect directly or indirectly the conformational change at the intracellular domain. Mutational analyses have shown that some of them are important for G-protein activation (Gilchrist et al., 1996; Lu et al., 2001).
Knowledge of GPCR conformation is not sufficient to understand the mechanism of G-protein coupling selectivity due to the absence of a high-resolution structure of the GPCR/G-protein complex. Although mutational data provide information about G-protein coupling selectivity, the experimental difficulty prevents us from obtaining experimental data for all of the GPCRs to study this selectivity. In silico analysis can provide information of the relationship between GPCRs and G-proteins, based on the many known sequences of GPCRs. This study complements mutational analyses and provides a new view of coupling selectivity.
|
Acknowledgements |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
We thank Dr T. Hirokawa and Mr Y. Yabuki for helpful discussion. This work was supported by a grant-in aid for special projects in genome science. This work was carried out using the computational facility of CBRC (AIST). We would like to thank Dr M. M. Gromiha and Dr X. Suresh (CBRC, AIST) for helpful discussion and for grammatically checking the previous version of this manuscript.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAcknowledgementsReferences |
|---|
Alexander S., Mathie A., Peters J., MacKenzie G., Smith A. (2001) Trends Pharmacol. Sci. 22:5–103.[Medline]
Altschul S.F., Gish W., Miller W., Myers E.W., Lipman D.J. (1990) J. Mol. Biol. 215:403–410.[CrossRef][Web of Science][Medline]
Arora K.K., Krsmanovic L.Z., Mores N., O'Farrell H., Catt K.J. (1998) J. Biol. Chem. 273:25581–25586.
Boeckmann B., et al. (2003) Nucleic Acids Res. 32:D23–26.
Cao J., Panetta R., Yue S., Steyaert A., Young-Bellido M., Ahmad S. (2003) Bioinformatics 19:234–240.
Carlson S.A., Chatterjee T.K., Murphy K.P., Fisher R.A. (1998) Mol. Pharmacol. 53:451–458.
Conklin B.R. and Bourne H.R. (1993) Cell 73:631–641.[CrossRef][Web of Science][Medline]
Cotecchia S., Exum S., Caron M.G., Lefkowitz R.J. (1990) Proc. Natl Acad. Sci. USA 87:2896–2900.
Drews J. (2000) Science 287:1960–1964.
Eddy S.R. (1998) Bioinformatics 14:755–763.
Erlenbach I. and Wess J. (1998) J. Biol. Chem. 273:26549–26558.
Geng L., Wu J., So S.P., Huang G., Ruan K.H. (2004) Arch. Biochem. Biophys. 423:253–265.[CrossRef][Medline]
Gilchrist R.L., Ryu K.S., Ji I., Ji T.H. (1996) J. Biol. Chem. 271:19283–19287.
Henikoff S. and Henikoff J.G. (1994) J. Mol. Biol. 243:574–578.[CrossRef][Web of Science][Medline]
Hermans E. (2003) Pharmacol. Ther. 99:25–44.[CrossRef][Web of Science][Medline]
Horn F., van der Wenden E.M., Oliveira L., Ijzerman A.P., Vriend G. (2000) Proteins 41:448–459.[CrossRef][Medline]
Horn F., Bettler E., Oliveira L., Campagne F., Cohen F.E., Vriend G. (2003) Nucleic Acids Res. 31:294–297.
Ilani T., Fishburn C.S., Levavi-Sivan B., Carmon S., Raveh L., Fuchs S. (2002) Cell Mol. Neurobiol. 22:47–56.[CrossRef][Medline]
Javitch J.A., Shi L., Simpson M.M., Chen J., Chiappa V., Visiers I., Weinstein H., Ballesteros J.A. (2000) Biochemistry 39:12190–12199.[CrossRef][Medline]
Kabsch W. and Sander C. (1983) Biopolymers 22:2577–2637.[CrossRef][Web of Science][Medline]
Koradi R., Billeter M., Wuthrich K. (1996) J. Mol. Graph. 14:51–55.[CrossRef][Web of Science][Medline]
Lee N.H., Geoghagen N.S., Cheng E., Cline R.T., Fraser C.M. (1996) Mol. Pharmacol. 50:140–148.[Abstract]
Leung M.Y., Al-Muslim O., Wu S.M., Aziz A., Inam S., Awadh M., Rennert O.M., Chan W.Y. (2004) Am. J. Med. Genet. A 130:146–53.[Medline]
Lichtarge O., Bourne H.R., Cohen F.E. (1996) J. Mol. Biol. 257:342–358.[CrossRef][Web of Science][Medline]
Liu J., Conklin B.R., Blin N., Yun J., Wess J. (1995) Proc. Natl Acad. Sci. 92:11642–11646.
Lu Z.L., Saldanha J.W., Hulme E.C. (2001) J. Biol. Chem. 276:34098–34104.
Madabushi S., Gross A.K., Philippi A., Meng E.C., Wensel T.G., Lichtarge O. (2004) J. Biol. Chem. 279:8126–8132.
Maglott D., Ostell J., Pruitt K.D., Tatusova T. (2005) Nucleic Acids Res. 33:D54–8.
Marti-Renom M.A., Madhusudhan M.S., Sali A. (2004) Protein Sci. 13:1071–1087.[CrossRef][Web of Science][Medline]
Mason D.A., Moore J.D., Green S.A., Liggett S.B. (1999) J. Biol. Chem. 274:12670–12674.
Matsumoto M. and Nishimura T. (1998) ACM Trans. Model. Comput. Simul. 8:3–30.
Mirny L.A. and Shakhnovich E.I. (1999) J. Mol. Biol. 291:177–196.[CrossRef][Web of Science][Medline]
Möller S., Vilo J., Croning M.D. (2001) Bioinformatics 17:S174–181.[Abstract]
Okada T., Fujiyoshi Y., Silow M., Navarro J., Landau E.M., Shichida Y. (2002) Proc. Natl Acad. Sci. USA 99:5982–5987.
Palczewski K., et al. (2000) Science 289:739–745.
Sgourakis N.G., Bagos P.G., Papasaikas P.K., Hamodrakas S.J. (2005) BMC Bioinformatics 6:104.[CrossRef][Medline]
Sjölander K., Karplus K., Brown M., Hughey R., Krogh A., Mian I.S., Haussler D. (1996) Comput. Appl. Biosci. 12:327–345.
Smith T.F. and Waterman M.S. (1981) J. Mol. Biol. 147:195–197.[CrossRef][Web of Science][Medline]
Strader C.D., Fong T.M., Tota M.R., Underwood D., Dixon R.A.F. (1994) Annu. Rev. Biochem. 63:101–132.[CrossRef][Web of Science][Medline]
Sreekumar K.R., Huang Y., Pausch M.H., Gulukota K. (2004) Bioinformatics 20:3490–3499.
Stitham J., Martin K.A., Hwa J. (2002) Mol. Pharmacol. 61:1202–1210.
Thompson J.D., Higgins D.G., Gibson T.J. (1994) Nucleic Acids Res. 22:4673–4680.
Van Rhee A.M. and Jacobson K.A. (1996) Drug Dev. Res. 37:1–38.
Wade S.M., Lim W.K., Lan K.L., Chung D.A., Nanamori M., Neubig R.R. (1999) Mol. Pharmacol. 56:1005–1013.
Wang H.L. (1997) J. Neurochem. 68:1728–1735.[Web of Science][Medline]
Wang H.L. (1999) J. Neurochem 72:1307–1314.[CrossRef][Medline]
Wess J. (1996) Crit. Rev. Neurobiol. 10:69–99.[Web of Science][Medline]
Wess J. (1998) Pharmacol. Ther. 80:231–264.[CrossRef][Web of Science][Medline]
Wong S.K. (2003) Neurosignals 12:1–12.[CrossRef][Web of Science][Medline]
Wu S.V., Yang M., Avedian D., Birnbaumer M., Walsh J.H. (1999) Mol. Pharmacol. 55:795–803.
Zhang L., Huang G., Wu J., Ruan K.H. (2005) Biochemistry 44:11389–11401.[Medline]
Received September 6, 2005; revised January 17, 2006; accepted February 20, 2006.
Edited by Mark Sansom
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||