Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Filters applied. Clear all
. 2007 Jan 12;2:1.
doi: 10.1186/1745-6150-2-1.

pkaPS: Prediction of Protein Kinase A Phosphorylation Sites With the Simplified Kinase-Substrate Binding Model

Affiliations
Free PMC article

pkaPS: Prediction of Protein Kinase A Phosphorylation Sites With the Simplified Kinase-Substrate Binding Model

Georg Neuberger et al. Biol Direct. .
Free PMC article

Abstract

Background: Protein kinase A (cAMP-dependent kinase, PKA) is a serine/threonine kinase, for which ca. 150 substrate proteins are known. Based on a refinement of the recognition motif using the available experimental data, we wished to apply the simplified substrate protein binding model for accurate prediction of PKA phosphorylation sites, an approach that was previously successful for the prediction of lipid posttranslational modifications and of the PTS1 peroxisomal translocation signal.

Results: Approximately 20 sequence positions flanking the phosphorylated residue on both sides have been found to be restricted in their sequence variability (region -18...+23 with the site at position 0). The conserved physical pattern can be rationalized in terms of a qualitative binding model with the catalytic cleft of the protein kinase A. Positions -6...+4 surrounding the phosphorylation site are influenced by direct interaction with the kinase in a varying degree. This sequence stretch is embedded in an intrinsically disordered region composed preferentially of hydrophilic residues with flexible backbone and small side chain. This knowledge has been incorporated into a simplified analytical model of productive binding of substrate proteins with PKA.

Conclusion: The scoring function of the pkaPS predictor can confidently discriminate PKA phosphorylation sites from serines/threonines with non-permissive sequence environments (sensitivity of appoximately 96% at a specificity of approximately 94%). The tool "pkaPS" has been applied on the whole human proteome. Among new predicted PKA targets, there are entirely uncharacterized protein groups as well as apparently well-known families such as those of the ribosomal proteins L21e, L22 and L6.

Availability: The supplementary data as well as the prediction tool as WWW server are available at http://mendel.imp.univie.ac.at/sat/pkaPS.

Reviewers: Erik van Nimwegen (Biozentrum, University of Basel, Switzerland), Sandor Pongor (International Centre for Genetic Engineering and Biotechnology, Trieste, Italy), Igor Zhulin (University of Tennessee, Oak Ridge National Laboratory, USA).

Figures

Figure 1
Figure 1
Variation of hydrophobicity and of flexibility over the motif region. The graph depicts the mean value deviations of the hydrophobicity-related property EISD840101 [29] and the flexibility scale VINM940104 [30] over the 81 positions that encompass the learning set sites. The mean values are presented as deviations from the UNIREF average (baseline) in percent of UNIREF standard deviations. The plots were smoothed by applying sliding windows (running averages) over 5 residues. Mean values were calculated using two different sequence sets: (i) one that contains all entries from the learning set, and (ii) one that consists of all proteins that are phosphorylated only once in the learning set. The difference between these two curves is not dramatic although, as a trend, the property values appear to fall back more sharply to the database values if only proteins with single PKA phosphorylation sites are taken into account.
Figure 2
Figure 2
Cumulative distribution of distances between successive sites in learning set proteins with multiple phosphorylated serine/threonine residues. The figure demonstrates that about two thirds of all distances are within the extended motif length of approximately 50 positions. The maximum distance, which exceeds the displayed x-axis, is 1759 amino acids.
Figure 3
Figure 3
Structure of the inhibitor peptide PKI bound to the PKA enzyme: N-terminal region of the substrate. Key arginines from the substrate peptide (RCSB Protein Data Bank entry 1JLU [92]) are highlighted. The left part of the figure shows the surface of PKA in ochre, the backbone of the substrate peptide in silver and the arginines -6, -3 and -2 of the substrate in blue. Arginines -3 and -2 interact with the binding cleft and thereby make major contributions to substrate specificity. A set of acidic enzyme residues interacts with these arginines (zoomed detail-view to the right): Glu170 and Glu230 for Arg-2, Glu127 for Arg-3 and Glu203 for Arg-6 [3]. The pictures were generated using VMD [93].
Figure 4
Figure 4
Preference for positive charge at positions located N-terminally with regard to the phosphorylated site. The upper graph depicts the increased occurrence of positively charged residues (His, Lys, Arg) compared to the expected database occurrence of 13.6% (deduced from UNIREF). The lower part of the figure shows the correlation coefficients R between amino acid frequencies and ZIMJ680104 (isoelectric point) [87] property values. Both plots demonstrate that the preference for basic residues is highest at positions -3 and -2, but encompasses at least the entire region between amino acids -6 and -2.
Figure 5
Figure 5
Structure of the inhibitor peptide PKI bound to the PKA enzyme: C-terminal region of the substrate. Overall (left) and detail views (right) of the substrate region that lies on the C-terminal side of the phosphorylated serine in complex with the kinase PKA (RCSB Protein Data Bank entry 1JLU [92]) are shown. Ile+1, His+2 and Asp+3 of the PKI substrate as well as the surface of the PKA enzyme to the left are colored according to residue types: white/gray for apolar, green for polar, blue for basic, and red for acidic amino acids. Compared with Figure 3, the orientation of the complex roughly corresponds to a counterclockwise rotation of 90 degrees around the vertical axis. The detail view to the right shows the hydrophobic patch at the surface of PKA which interacts with the substrate residue that lies C-terminally adjacent to the phosphorylated site. The pictures were generated using VMD [93].
Figure 6
Figure 6
Multiple alignment of the binding site regions across PKA orthologue sequences. Starting with the mouse sequence (accession NP_032880) of the protein in the crystal structure 1JLU [92], we searched for orthologues of the catalytic subunit of PKA with the ANNOTATOR suite [45]. In the alignment (generated with T-COFFEE [94]), we present 40 variants thereof ranging as far as from yeast to human (sequence position numbering is without leading methionines according to the 1.29 Å rule [56,57]). The figure focuses on the protein polypeptide stretch that encompasses the residues forming the surface of the binding site at substrate position from -3 to +1. Red triangles (at Glu127, Glu170 and Glu230 in the numbering of 1JLU without the leading methionine in NP_032880) mark positions that form the pocket for substrate residues -3 and -2. Blue triangles (at Leu198, Pro202 and Leu205) mark the hydrophobic pocket-forming positions that accept substrate residue +1 [4–7].
Figure 7
Figure 7
Mean distances between pairs of neighboring predicted sites depending on the total number of predicted sites in the query proteins. The red line displays the linear regression (y = 59.3 - 0.271x; R = -0.66) calculated using these data points.
Figure 8
Figure 8
Approximation of the empirical score distribution of non-phosphorylated sites. The empirical score distribution was approximated using equations 16 and 17. With a correlation coefficient of 0.9988, the applied polynomial fit of 3rd order provides a sufficiently accurate approximation of the expected false-positive rate. The parameters with respect to equation 17 are: u = -1.76847, λ1 = -0.766775, λ2 = 0.166677 and λ3 = -0.0298602. The polynomial fit was calculated using the XMGRACE tool [81].

Similar articles

See all similar articles

Cited by 53 articles

See all "Cited by" articles

References

    1. Madhani H. Functional analysis of protein kinase networks in living cells: beyond "knock-outs" and "knock-downs". Methods. 2006 - PubMed
    1. Knighton DR, Zheng JH, Teneyck LF, Ashford VA, Xuong NH, Taylor SS, Sowadski JM. Crystal-Structure of the Catalytic Subunit of Cyclic Adenosine-Monophosphate Dependent Protein-Kinase. Science. 1991;253:407–414. doi: 10.1126/science.1862342. - DOI - PubMed
    1. Knighton DR, Zheng JH, Teneyck LF, Xuong NH, Taylor SS, Sowadski JM. Structure of A Peptide Inhibitor Bound to the Catalytic Subunit of Cyclic Adenosine-Monophosphate Dependent Protein-Kinase. Science. 1991;253:414–420. doi: 10.1126/science.1862343. - DOI - PubMed
    1. Bramson HN, Kaiser ET, Mildvan AS. Mechanistic Studies of Camp-Dependent Protein-Kinase Action. Crc Critical Reviews in Biochemistry. 1984;15:93–124. - PubMed
    1. Kemp BE, Graves DJ, Benjamini E, Krebs EG. Role of Multiple Basic Residues in Determining Substrate-Specificity of Cyclic Amp-Dependent Protein-Kinase. Journal of Biological Chemistry. 1977;252:4888–4894. - PubMed

LinkOut - more resources

Feedback