Bioinformatics Advance Access originally published online on July 14, 2006
Bioinformatics 2006 22(18):2192-2195; doi:10.1093/bioinformatics/btl381
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Scoring of predicted GRK2 phosphorylation sites in Nedd4-2
1 Department of Medicine, University of Sydney Sydney, Australia
2 Sydney University Biological Informatics and Technology Centre Sydney, Australia
3 Department of Pathology, University of Sydney Sydney, Australia
4 Department of Physiology, University of Sydney Sydney, Australia
*To whom correspondence should be addressed.
 |
ABSTRACT |
|---|
 TOP ABSTRACT 1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
Motivation: Epithelial Na+ channels (ENaC) mediate the transport of sodium (Na) across epithelia in the kidney, gut and lungs and are required for blood pressure regulation. They are inhibited by ubiquitin protein ligases, such as Nedd4-2. These ligases bind to proline-rich motifs (PY motifs) present in the C-termini of ENaC subunits. Loss of this inhibition leads to hypertension. We have previously reported that ENaC channels are maintained in the active state by the G protein coupled receptor kinase, GRK2. The enzyme has been implicated in the development of essential hypertension [R. D. Feldman (2002) Mol. Pharmacol., 61, 707–709]. Additional findings in our lab pointed towards a possible role for GRK2 in the phosphorylation and inactivation of Nedd4-2.
Results: We have predicted GRK2 phosphorylation sites on Nedd4-2 by combining sequence analysis, homology modeling and surface accessibility calculations. A total of 24 potential phosphorylation sites were predicted by sequence analysis. Of these, 16 could be modeled using homology modeling and 6 of these were found to have sufficient surface exposure to be accessible to the GRK2 enzyme responsible for the phosphorylation of Nedd4-2. The method provides an ordered list of the most probable GRK2 phosphorylation sites on Nedd4-2 providing invaluable guidance to future experimental studies aimed at mutating certain Nedd4-2 residues in order to prevent phosphorylation by GRK2. The method developed could be applied in a wide variety of biological applications involving the binding of one molecule to a protein. The relative effectiveness of the technique is determined mainly by the quality of the homology model built for the protein of interest.
Contact: jarthur{at}med.usyd.edu.au
|
1 INTRODUCTION |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
E3 ubiquitin-protein ligase NEDD4-like protein (or Nedd4-2, UniProt: Q8CFI0) is a member of a family of E3 ubiquitin-protein ligases sharing a similar domain structure. Family members are characterized by the presence of a catalytic HECT (homologous to E6-AP, C-terminus) domain. This domain facilitates ubiquitin attachment to substrate proteins. In addition, they contain multiple WW domains responsible for mediating binding to PY motifs, and may contain an N-terminal C2 (calcium/lipid binding) domain believed to be important for localization (Harvey and Kumar, 1999). Nedd4-2 controls Epithelial Na+ channels (ENaC) surface expression by catalyzing its ubiquitination. This ubiquitination targets ENaC for degradation (Kamynina et al., 2001a, b; Snyder et al., 2004).
It has been suggested (Debonneville et al., 2001; Snyder et al., 2002; Boehmer et al., 2003) that hormones responsible for activating ENaC and voltage-gated Na+ channels may do so by phosphorylating Nedd4-2, thus rendering it unable to interact with the channels. Previous findings in the laboratory of one of us (Dinudom et al., 2004) have suggested a role for GRK2 in controlling the activity of Nedd4-2 in these studies. We used recombinant GRK2 (Pitcher et al., 1999; Lodowski et al., 2003) to show that GRK2 phosphorylates S633 in the C-terminus of ß-ENaC and this renders the channel insensitive to inhibition by Nedd4-2. Studies in the laboratory of one of us have also shown that recombinant GRK2 extensively phosphorylates Nedd4-2 in vitro (Angeles Sanchez-Perez, unpublished data).
Identification of potential phosphorylation sites in Nedd4-2 would enable the construction of mutant Nedd4-2 proteins lacking specific phosphorylation sites. This would allow systematic investigation on the effect of phosphorylation of Nedd4-2 on the regulation of ENaC activity.
In this paper we describe a computational approach to predict the GRK2 phosphorylation sites of Nedd4-2. Initially, sequence analysis with neural networks is used to identify potential phosphorylation sites. These potential sites are then screened using additional information derived from the analysis of the three-dimensional (3D) structure of Nedd4-2 as predicted by homology modeling. Using this approach we were able to reduce the number of potential phosphorylation sites and suggest target residues on Nedd4-2 for mutation in order to prevent the phosphorylation of Nedd4-2 by GRK2.
|
2 METHODS |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
2.1 Sequence-based identification of phosphorylation sites
In vitro phosphorylation with GRK2 of Nedd4-2 constructs lacking specific domains identified those domains in Nedd4-2 that were phosphorylated. Two programs were then used to identify potential phosphorylation sites within these domains: PREDIKIN, a program that uses the amino acid sequence of a kinase to predict its target sequence (Brinkworth et al., 2003) and NetPhos 2.0, a neural network-based method for predicting potential phosphorylation sites at serine, threonine or tyrosine residues in protein sequences (Blom et al., 1999).
Consideration was then given to the degree of conservation of the potential phosphorylation sites between Nedd4-2 proteins from different species, Nedd4 (a related protein also involved in ENaC regulation) and other proteins known to be phosphorylated by GRK2.
2.2 Homology modeling
The 3D structure of Nedd4-2 was predicted by homology modeling using the MolIDE framework (Canutescu and Dunbrack, 2005). The protein sequence of Nedd4-2 was compared with Version 6.4 of the UniProt database (Bairoch et al., 2005) using PSI-BLAST (Altschul et al., 1997). After four rounds of searching against UniProt, the sequence profiles from each round were used to compare the protein sequence with the Protein Data Bank (PDB) (Berman et al., 2000) to identify proteins of known structure with sequence homology to Nedd4-2. The use of the sequence profile from each round allows the sequence alignment results to be compared as more and more distant homologues are added to the sequence profile. The resulting sequence alignments were examined manually to identify good templates for homology modeling based on low E-value, high sequence identity and a long alignment including the potential phosphorylation sites identified by the sequence analysis.
The selected templates were used to build homology models in MolIDE with the position of backbone atoms and conserved side-chains assigned according to the alignment with the template structures. The side chains for non-conserved residues were built using SCWRL (Canutescu et al., 2003).
2.3 Surface accessibility calculations
The sequence of GRK2 (UniProt: Q99MK8) was used to determine the molecular volume of this protein using the online version of NucProt (Voss and Gerstein, 2005) (http://geometry.molmovdb.org/NucProt/). As a comparison, the structure of the Bos taurus homologue of GRK2 (PDB: 1OMV:A) was used to determine the molecular volume based on the accessible surface area of the molecule using 3V: Voss Volume Voxelator (http://geometry.molmovdb.org/3v/).
The GRK2 enzyme was then modeled as a sphere with radius determined by the molecular volume calculated above. The surface accessibility of each residue in Nedd4-2 was calculated using the method of Lee and Richards (1971) as implemented by Hubbard and Thornton in naccess (http://wolf.bms.umist.ac.uk/naccess/naccess.html) using a probe radius equal to the radius of the spherical model of GRK2.
|
3 RESULTS |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
Nedd4-2 has five distinct domains: a C2 domain (residues 35–137), four WW domains (WW1: residues 221–254, WW2: residues 414–447, WW3: residues 526–559 and WW4: residues 577–610), and a HECT domain (residues 669–1003). Experimental studies in the Cook laboratory (Angeles Sanchez-Perez, unpublished data) have shown phosphorylation specifically occurs on the WW1 and WW2 domains but not on the WW3, WW4 or HECT domains. It is not currently known whether phosphorylation occurs on the C2 domain. Thus we were interested in determining the GRK2 phosphorylation sites within the N-terminal portion of the protein (residues 1–511) including the C2 domain and the first two of the WW domains.
3.1 Sequence-based prediction
Using the sequence analysis techniques described above, we predicted 24 potential phosphorylation sites described below:
- C2 domain sites (S41, S56, S76, T93, T128)
- Sites between the C2 domain and WW1 domain (S141, T155, S165, S168, S192)
- WW1 domain sites (S253)
- Sites between the WW1 and WW2 domains (S259, S261, S284, T304, S306, S337, S371, T396)
- WW2 domain sites (S418)
- Sites between the WW2 and WW3 domains (S475, S477, S478, S493)
3.2 3D models of Nedd4-2
The sequence of Nedd4-2 was compared with all sequences in the PDB as described above in order to find homologous proteins of known structure. No single protein in the PDB had homology to Nedd4-2 across the entire N-terminal region of the protein covering the C2, WW1 and WW2 domains. Thus it was necessary to model each domain independently.
The C2 domain was modeled on the crystal structure of C2A/C2B fragment of Synaptotagmin III from Rattus rattus (PDB: 1DQV [PDB] :A, E-value = 7 x 10–34, sequence identity = 35%, resolution = 3.2 Å). The alignment covered 139 residues over the C2 domain including the potential phosphorylation sites S56, S76, T93, T128, S141, T155, S165 and S168. The model is shown in Figure 1.
|
The WW1 and WW2 domains were modeled on the NMR structure of WW3-4 domains of Suppressor of Deltex from Drosophilia melanogaster (PDB: 1TK7 [PDB] :A, E-value = 8 x 10–12, sequence identity = 38%). In the case of the WW1 domain, the alignment stretched over 68 residues including four potential phosphorylation sites: S253, S259, S261 and S284. In the WW2 domain, the alignment stretched over 67 residues and also included four potential phosphorylation sites: S418, S475, S477 and S478. The models are shown in Figure 1.
It was thus possible to create homology models for regions of the protein containing 16 of the 24 potential phosphorylation sites. These models provided the basis for the surface accessibility calculations.
3.3 Using surface accessibility to clarify potential phosphorylation sites
The phosphorylation sites predicted using sequence analysis are determined solely on the basis of sequence information, i.e. the presence of a sequence motif associated with phosphorylation. However, when the protein is folded into its 3D structure, some of these potential phosphorylation sites will be located within the interior of the protein or in parts of the surface screened by other residues. Thus GRK2 will be unable to access these residues in order to phosphorylate them.
The surface accessibility of the polar side-chain atoms of each potential phosphorylation site was calculated using naccess with a probe radius designed to model the GRK2 enzyme as described above. The polar side-chain atoms being, in this case, the O atom involved in the phosphorylation. Six residues were found to be accessible to the GRK2 enzyme with S261 being the most exposed. The relative surface accessibility of all the other potential phosphorylation sites was then calculated by normalizing the surface accessibility to the surface accessibility of S261. The full list of exposed residues and their exposure normalized to that of S261 is shown in Table 1.
|
3.4 Verification of the method
The method was verified by application to proteins with experimentally determined GRK2 phosphorylation sites. Six proteins with experimentally confirmed GRK2 phosphorylation sites were identified as potential test cases. Sequence-based prediction of GRK2 phosphorylation sites was undertaken as described above and then homology modeling was attempted. It was only possible to build a homology model for one of the proteins. This was owing to the inability to find a target alignment spanning a long enough region to contain the experimentally determined phosphorylation site and produce a reasonably, globular protein model.
The ß5 tubulin protein from Sus scrofa is known to be phosphorylated by GRK2 at residues T409 and S420 and only at these residues (Yoshida et al., 2003). We were able to use our method to predict GRK2 phosphorylation binding sites on this protein. Sequence-based prediction showed 11 potential phosphorylation sites (S40, S48, S75, S115, S126, S172, T274, T285, S322, T409 and S420). We built a homology model of ß5 tubulin based on the crystal structure of tubulin from Bos taurus (PDB: 1FFX:B, E-value = 0, sequence identity = 97%, resolution = 3.95 Å). Surface accessibility calculations then predicted a single phosphorylation site at S420. The second experimentally determined phosphorylation site, T409, was confirmed by visual inspection of the model, secondary structure predictions undertaken as part of the modeling process, and reference to the template PDB structure, to be located in a lengthy random coil region between two helices. As a result, this region of the protein may display greater flexibility in solution and thus may at times have greater surface accessibility than that shown in the crystal structure, leading to our inability to predict this phosphorylation site with our method. However, this could not be confirmed using the alternative ‘hot loops’ definition of disorder in DisEMBL (Linding et al., 2003). In any case, we were able to correctly identify one of the two experimentally determined phosphorylation sites and correctly eliminate the other nine potential phosphorylation sites predicted by sequence analysis alone.
|
4 DISCUSSION |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
Sequence-based analysis of Nedd4-2 identified 24 potential phosphorylation sites. Of these, 16 could be modeled using homology modeling. Using the extra structural information provided by the homology modeling, only six of these were found to be significantly accessible to a probe modeling the large GRK2 enzyme. In addition, we have normalized the surface accessibility scores of the six residues exposed to GRK2. Thus, in addition to determining which residues are accessible to GRK2, we are also able to rank these residues from the most accessible to the least accessible and thus order them according to their relative likelihood of being phosphorylation sites of GRK2.
The outcome of this analysis provides much needed guidance to further experimental studies aimed at mutating residues in Nedd4-2 in order to prevent the phosphorylation of Nedd4-2 by GRK2. The most likely phosphorylation sites can be mutated first, followed by the less probable phosphorylation sites in a series of studies as far as the time and money available for the experimental work will allow.
One limitation in this particular application of the method developed above was the inability to find a single protein sequence in the PDB displaying homology to Nedd4-2 across the entire N-terminal region. This has two main consequences.
First, one third of the potential phosphorylation sites predicted through sequence analysis were located in the gaps between the three sequence alignments used to create the homology models. As a result, no structural information about these residues was available and no prediction about their relative accessibility to the GRK2 enzyme can be made. Hence there may be additional phosphorylation sites among these eight residues, with the consequence that they must also be considered in planning future experimental work.
Second, because three separate models were used in the surface accessibility calculations, the final list of residues accessible to GRK2 is likely to contain a number of false positives. In the complete protein structure, the three separate domains are likely to be closely associated in such a way that some of the residues accessible to GRK2 in the individual domains will become inaccessible owing to shielding from one or more of the other domains in the complete structure.
In theory, more sophisticated modeling techniques, such as ab initio protein folding, could be used. However, given the size of Nedd4-2, it is unclear whether such a determination of the structure would be accurate enough to improve the prediction of phosphorylation sites.
The method is relatively straightforward. It makes use of well-known and freely available bioinformatics software and therefore could be easily applied to other proteins. Nonetheless, an integration of the various parts into a single workflow in order to automate the process would make the technique more easily available to a wider audience. One strategy for such an integration of the various applications would be to use PERL, or a similar scripting language, to create a single ‘pipeline’ of the various applications, parsing and collating the results as necessary to prepare the input into applications further down the process. The success of this strategy would depend mainly on the selection of the structural target for homology modeling. This part of the process would require the development of a heuristic algorithm to balance the need to model long, contiguous regions of the target sequence, have a high quality alignment and include as many as possible of the potential phosphorylation sites. The strategy would also, of course, require permission to use or licensing of the various bioinformatic software applications required.
In this particular case a ranked list of six potential phosphorylation sites is predicted providing some guidance for experimental mutation studies on Nedd4-2. The technique developed here could also be applied in a wide range of similar applications involving the binding of one molecule to a protein. All that is required is the ability to
- define a binding sequence motif to identify the residues on the protein where the molecule of interest may bind to or interact with the protein,
- calculate a molecular volume for the binding molecule and effectively model the molecule with a sphere of appropriate radius,
- create an accurate homology model of the protein of interest.
|
Acknowledgments |
|---|
J.W.A. would like to thank Dr Simon Hubbard for his advice on modifying naccess to use larger probe sizes. A.S.P. and D.I.C. would like to thank Robert J. Lefkowitz for the gift of purified GRK2 protein.Conflict of Interest: none declared.
|
FOOTNOTES |
|---|
Associate Editor: Anna Tramontano
Received on February 9, 2006; revised on June 6, 2006; accepted on July 6, 2006
|
REFERENCES |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
Altschul, S.F., et al. (1997) Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res, . 25, 3389–3402
Bairoch, A., et al. (2005) The Universal Protein Resource (UniProt). Nucleic Acids Res, . 33, D154–D159
Berman, H.M., et al. (2000) The Protein Data Bank. Nucleic Acids Res, . 28, 235–242
Blom, N., et al. (1999) Sequence and structure-based prediction of eukaryotic protein phosphorylation sites. J. Mol. Biol, . 294, 1351–1362[CrossRef][Web of Science][Medline].
Boehmer, C., et al. (2003) Serum and glucocorticoid inducible kinases in the regulation of the cardiac sodium channel SCN5A. Cardiovasc. Res, . 57, 1079–1084
Brinkworth, R.I., et al. (2003) Structural basis and prediction of substrate specificity in protein serine/threonine kinases. Proc. Natl Acad. Sci. USA, 100, 74–79
Canutescu, A.A. and Dunbrack, R.L., Jr. (2005) MolIDE: a homology modeling framework you can click with. Bioinformatics, 21, 2914–2916
Canutescu, A.A., et al. (2003) A graph-theory algorithm for rapid protein side-chain prediction. Protein Sci, . 12, 2001–2014[CrossRef][Web of Science][Medline].
Debonneville, C., et al. (2001) Phosphorylation of Nedd4-2 by Sgk1 regulates epithelial Na+ channel surface expression. EMBO J, . 20, 7052–7059[CrossRef][Web of Science][Medline].
Dinudom, A., et al. (2004) The kinase Grk2 regulates Nedd4/Nedd4-2-dependent control of epithelial Na+ channels. Proc. Natl Acad. Sci. USA, 101, 11886–11890
Feldman, R.D. (2002) Deactivation of vasodilator responses by GRK2 overexpression: a mechanism or the mechanism for hypertension? Mol. Pharmacol, . 61, 707–709
Harvey, K.F. and Kumar, S. (1999) Nedd4-like proteins: an emerging family of ubiquitin-protein ligases implicated in diverse cellular functions. Trends Cell Biol, . 9, 166–169[CrossRef][Web of Science][Medline].
Kamynina, E., et al. (2001a) A novel mouse Nedd4 protein suppresses the activity of the epithelial Na+ channel. FASEB J, . 15, 195–203
Kamynina, E., et al. (2001b) Distinct characteristics of two human Nedd4 proteins with respect to epithelial Na+ channel regulation. Am. J. Physiol. Renal Physiol, . 281, 469–477.
Lee, B. and Richards, F.M. (1971) The interpretation of protein structures: estimation of static accessibility. J. Mol. Biol, . 55, 379–380[CrossRef][Web of Science][Medline].
Linding, R., et al. (2003) Protein disorder prediction: implications for structural proteomics. Structure, 11, 1453–1459[Medline].
Lodowski, D.T., et al. (2003) Keeping G proteins at bay: a complex between G-protein-coupled receptor kinase 2 and Gbetagamma. Science, 300, 1256–1262
Pitcher, J.A., et al. (1999) Feedback inhibition of G protein-coupled receptor kinase 2 (GRK2) activity by extracellular signal-regulated kinases. J. Biol. Chem, . 274, 34531–34534
Snyder, P.M., et al. (2002) Serum and glucocorticoid-regulated kinase modulates Nedd4-2-mediated inhibition of the epithelial Na+ channel. J. Biol. Chem, . 277, 5–8
Snyder, P.M., et al. (2004) Relative contribution of Nedd4 and Nedd4-2 to ENaC regulation in epithelia determined by RNA interference. J. Biol. Chem, . 279, 5042–5046
Voss, N.R. and Gerstein, M. (2005) Calculation of standard atomic volumes for RNA and comparison with proteins: RNA is packed more tightly. J. Mol. Biol, . 346, 477–492[CrossRef][Web of Science][Medline].
Yoshida, N., et al. (2003) Identification of sites of phosphorylation by G-protein-coupled receptor kinase 2 in beta-tubulin. Eur. J. Biochem, . 270, 1154–1163[Web of Science][Medline].
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  N. H. Lee, B. J. Haas, N. E. Letwin, B. C. Frank, T. V. Luu, Q. Sun, C. D. House, S. Yerga-Woolwine, P. Farms, E. Manickavasagam, et al. Cross-Talk of Expression Quantitative Trait Loci Within 2 Interacting Blood Pressure Quantitative Trait Loci Hypertension, December 1, 2007; 50(6): 1126 - 1133. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||