Hitting the Target: Emerging Technologies in the Search for Kinase Substrates

See allHide authors and affiliations

Science's STKE  10 Dec 2002:
Vol. 2002, Issue 162, pp. pe49
DOI: 10.1126/stke.2002.162.pe49


Through phosphorylation, protein kinases can alter the activity, localization, protein association, and stability of their targets. Despite the importance to our understanding of all aspects of cell biology, progress toward identifying bona fide substrates of specific protein kinases has been slow. Traditionally used techniques to identify true kinase substrates, such as genetics, yeast two-hybrid screens, and biochemical purification, are often laborious and unreliable. However, several new approaches have recently been developed and used successfully to identify genuine in vivo substrates of certain protein kinases. These methods include screening for phosphorylation of proteins from phage expression libraries, peptide library screens to determine optimal motifs favored by specific kinases, the use of phospho-motif antibodies, and an approach that uses structurally altered kinases and allele-specific adenosine triphosphate analogs and kinase inhibitors. We describe these approaches and discuss their utility and inherent caveats.

If gene expression and protein translation represent the musicians and instruments of a cellular symphony, then posttranslational modification of proteins should be considered the conductor. Aside from containing sequences necessary for three-dimensional folding into domains, many proteins have evolved motifs that allow posttranslational modifications that modulate function. By far the largest group of enzymes that catalyze regulatory posttranslational modifications is the family of protein kinases. These enzymes transfer phosphates from ATP (adenosine triphosphate) to specific hydroxyl groups on substrate proteins. About 2% of the genes encoded by the human genome are predicted to encode protein kinases. The reversible phosphorylation of specific tyrosine, serine, or threonine residues on a target protein can dramatically alter its function in several ways, including activating or inhibiting enzymatic activity, creating or blocking binding sites for other proteins, altering subcellular localization, or controlling protein stability. Therefore, our understanding of the molecular control of cell physiology requires the identification of substrates targeted by specific protein kinases.

Despite the widespread interest in this area, the discovery of physiological substrates for protein kinases has been slow and often unreliable. The catalytic domains of protein kinases share a high degree of sequence and structural similarity, and many subfamilies of kinases exist whose members are likely to have overlapping substrate specificities. The use of genetic or pharmacological disruption of specific kinases also presents inherent problems due to functional compensation by related kinases and lack of specificity in currently available kinase inhibitors. However, several recent in vitro and in vivo technologies have emerged as promising new avenues toward identification of new substrates of protein kinases.

"Classical" Methods for Identifying Kinase Substrates

Approaches to identifying candidate substrates of protein kinases have historically included genetic screens, the isolation of kinase-binding proteins, and biochemical purification. Genetic screens and subsequent epistasis analyses in model organisms such as yeast, worms, and flies have been very successful in placing substrates downstream of specific protein kinases within signaling pathways. For example, in the budding yeast Saccharomyces cerevisiae, screens for mating-defective mutants (sterile or ste mutants) led to the elucidation of a protein kinase cascade leading from cell surface pheromone receptors to the downstream transcription factor Ste12p (1). In Caenorhabditis elegans, epistasis analyses placed the forkhead-family transcription factor DAF-16 downstream of the PI3K (phosphoinositide 3-kinase)-Akt pathway (2), leading to the subsequent demonstration of direct phosphorylation and regulation of mammalian forkhead-family members by the kinase Akt (3). Genetic screens in model organisms have often paralleled and complemented efforts in Xenopus and mammalian cell systems, which have used primarily biochemical approaches to identify kinase substrates.

Many protein kinase substrates have been identified through interaction screens, most commonly yeast two-hybrid screens, with the kinase of interest as "bait." The first demonstration that this was a viable technique for detecting novel kinase substrates came with the identification of Sip1p as a two-hybrid-interacting protein with the yeast Snf1p kinase (4). Sip1p was subsequently shown to be directly phosphorylated by Snf1p. In most cases the interaction between a protein kinase and its substrate is transient, and upon substrate phosphorylation the association is disrupted, allowing a single kinase to catalyze the phosphorylation of multiple substrates. Such interactions are difficult to trap and identify. However, in some cases, kinases bind their substrates by interactions outside of the catalytic pocket, and such associations can often be detected in two-hybrid screens. For instance, filamin was found in a two-hybrid screen for proteins that associate with the NH2-terminal regulatory domain of Pak1 (p21-activated kinase 1) and is a Pak1 substrate (5). However, in cases where a third protein acts as a scaffold to stabilize the interaction between a kinase and its substrate, direct-interacting screens such as two-hybrid screens will likely fail. Although two-hybrid screens to identify kinase substrates have been used with some success, problems with a high rate of false-positives and with distinguishing between kinase-binding partners and bona fide in vivo substrates limit the usefulness of this method. As with all the approaches discussed here, confirmation of candidate substrates by secondary assays is critical.

Historically, the positions of phosphorylation sites on important cellular proteins were often determined prior to identification of the responsible kinase. To identify kinases, proteins in cell lysates were separated by column chromatography and protein fractions were assayed for kinase activity toward a substrate of interest. For example, the 70-kD ribosomal protein S6 kinase (S6K1) was identified with this approach (6, 7). Similar techniques have been used with limited success for the identification of proteins within lysate fractions that can be phosphorylated in vitro by specific protein kinases, as scored by incorporation of 32P. Obvious problems arise from such an approach, including the presence of kinases within the lysates phosphorylating themselves and other proteins, leading to high levels of background phosphorylation and false-positives. Some of these problems have been circumvented in a recent study by the Cohen laboratory searching for novel substrates of SAPK/p38-kinase family members (8). These authors reduced the background phosphorylation problem by making several small adjustments to standard protocols. Together, these modifications allowed the detection of specific bands in lysate fractions that can be phosphorylated by distinct SAPK/p38 family members. However, after the initial detection, this approach still depends on a rather laborious biochemical purification of the candidate substrate. Furthermore, there are intrinsic caveats to all in vitro kinase substrate identification techniques, and these are discussed below.

In Vitro Identification of Kinase Substrate Candidates

In recent years, several novel approaches for the identification of in vitro protein kinase substrates have been developed. Although useful for the identification of candidate substrates for a given kinase, in vitro kinase assays can often be misleading and provide too many candidate substrates to be further validated by more rigorous in vivo analyses. Indeed, protein kinases often display a high degree of promiscuity in vitro. The use of high concentrations of purified kinase in in vitro assays is partially responsible for this loss of specificity. In addition, removal of the kinase from the cell often results in a loss of its physiological regulatory mechanisms. A better understanding of the regulation of a kinase of interest and subsequent adjustments to the kinase purification scheme and assay can circumvent some of these complications. However, all candidate substrates identified with in vitro approaches require validation in vivo. Nevertheless, some of the recently developed in vitro techniques allow a rapid identification of candidate substrates for both known and novel kinases and are, therefore, valuable tools in the search for new kinase targets.

Solid-phase phosphorylation screening of phage expression libraries has been used successfully to identify bona fide substrates of some kinases [for example, extracellular signal-regulated kinase 1 (9) and cyclin-dependent kinases (10)]. This approach uses cDNA libraries cloned into a phage expression vector. Following plaque formation on lawns of Escherichia coli, proteins within the plaques, including those expressed from an individual cDNA clone as well as phage and bacterial proteins, are then immobilized onto nitrocellulose filters. Proteins on these filters are then subjected to phosphorylation by the purified kinase of interest in the presence of [γ-32P]ATP (adenosine triphosphate), and plaques containing positive phosphorylation signals are identified by autoradiography. One of the main benefits of this approach comes from the ease of identifying the candidate substrate by isolating the clone from the phosphorylation-positive plaque. Furthermore, if background phosphorylation of phage and bacterial proteins is a problem with the kinase of interest, it will be detected early on during pilot studies because each plaque will contain the same background proteins. The frequent improper folding of cDNA-encoded proteins within E. coli presents one potential problem. In general, the utility of this approach for identifying genuine substrates of a given kinase will depend greatly on the in vitro specificity of the kinase (11). Kinases that utilize secondary (noncatalytic domain) contacts with their substrates will likely give a higher degree of success in most in vitro kinase screens against full-length native proteins.

Oriented peptide library screens for determining optimal substrate motifs of specific kinases have been useful in distinguishing the substrate specificities of related kinases as well as for identifying candidate substrates through bioinformatics (12-15). This approach uses a degenerate library of peptides (that is, a large mixture of peptides of random sequences) oriented around an invariant serine, threonine, or tyrosine residue to be phosphorylated by a purified kinase in vitro (Fig. 1A). The kinase phosphorylates a small percentage of the billions of peptides within the mixture, and these phosphopeptides are then isolated on a ferric column. This mixture of phosphopeptides is then sequenced by Edman degradation, and the quantity of each amino acid at each position relative to that of the phosphorylated residue is determined. The data obtained from this technique provide a list of residues preferred, tolerated, and selected against by the kinase at each site surrounding the phosphorylation site. These data can then be formatted into a selectivity matrix for protein database searches for candidate kinase substrates with the Web-based program Scansite [ (15)]. Thus far, the amino acid selectivities of kinases analyzed with this peptide library approach match those for known substrates of the tested kinases (11), and Scansite searches with the corresponding selectivity matrices correctly predict previously identified substrates (15, 16). Therefore, this approach will identify those proteins that may be phosphorylated by the given kinase. However, selecting which candidate substrates predicted by Scansite to test in more conventional substrate characterization assays is the real limitation of this approach. The technique is better suited for use in combination with other substrate identification approaches, such as immunoblotting with phospho-motif antibodies (discussed below), and can be very useful in predicting which sites on an independently identified substrate are likely to be phosphorylated.

Fig. 1.

Schematic of two approaches for identifying substrates of protein kinases. (A) The use of an oriented peptide library to determine substrate specificity of a kinase and subsequent identification of candidate substrates. A degenerate mixture of peptides (x, any amino acid) oriented around a Ser, Thr, or Tyr to be phosphorylated is subjected to an in vitro kinase assay with the kinase of interest. The small fraction, typically 1%, of resulting phosphopeptides is then isolated on a ferric column. After elution from the column, the phosphopeptides are quantitatively sequenced by Edman degradation to determine the peptide-substrate specificity of the kinase. This information can then be used both to search protein databases with the Scansite program and to generate phospho-motif-specific antibodies, the combination of which identifies candidate substrates. (B) A method for the in vivo study of kinase function and for kinase substrate identification developed by Shokat and colleagues (22) (see text for details). This approach uses ATP analogs (bumped ATP) that can distinguish between wild-type and mutant (holed) kinases. The in vitro or in vivo use of a bumped-ATP/holed-kinase pair allows selective labeling of substrates specific for the kinase of interest. The same technology has been modified for use of allele-specific inhibitors (bumped inhibitors) to study kinase function in vivo. WT, wild-type.

In Vivo Identification of Bona Fide Kinase Substrates

Aside from genetic approaches, in vivo screens for kinase substrates have been slow to develop. In vivo [γ-32P]ATP labeling of cells followed by separation of phosphoproteins by chromatography or two-dimensional gel electrophoresis has limited value when used alone. In general, in vivo detection of substrate candidates for individual protein kinases requires an effective and highly specific means of inactivating that kinase. This can be done by pharmacological inhibition [although inhibitors that can strictly distinguish between specific kinases are rare (17)], antibody injection, mouse knockout technology, or, more recently, by small-interfering RNAs (siRNAs). Knowledge of a specific stimulus or condition that maximally activates the kinase of interest is also crucial. Although less prevalent than during in vitro assays, overexpression of kinases within cells will often lead to nonspecific phosphorylation of some proteins. In addition, unlike in vitro methods, it is difficult to determine if a phosphorylation event detected in vivo is the direct action of the kinase of interest or occurs through a downstream kinase. With these caveats in mind, some new technologies have recently been developed that seem quite promising for the identification of in vivo targets of protein kinases.

Knowledge of the minimal requirements for a kinase to phosphorylate a given protein sequence, determined by previously identified phosphorylation sites or peptide library methods, can be used to raise phospho-motif-specific antibodies (18). These antibodies are raised against a library of degenerate phosphopeptides with residues required by the specific kinase and a phosphoserine, phosphothreonine (pT), or phosphotyrosine locked in at the appropriate positions. In theory, these antibodies should recognize substrates of the given kinase only once they are phosphorylated. Phospho-motif antibodies can be used to identify specific kinase substrates. This was demonstrated with an antibody to the minimal Akt substrate motif (16, 19), which was raised against an xxxRxRxx(pT)xxxx peptide mixture (where x represents any amino acid) (18). The activity of PI3K, an upstream activator of Akt, was required for immunoblotting of the majority of the proteins recognized by this antibody. Our laboratory used the apparent molecular size of protein bands on phospho-Akt substrate immunoblots to narrow down searches for candidate Akt substrates with Scansite (16). This approach successfully identified known and novel Akt substrates that were confirmed by using conventional methods of substrate validation. One problem with this method is that in any given mass range Scansite will predict several strong substrate candidates. However, the search can be narrowed down even further by determining the isoelectric point (pI) of a given protein recognized by the antibody, because Scansite can search for proteins in restricted ranges of both mass and pI. The existence of kinases related to Akt that phosphorylate the same motif and that are also regulated by PI3K (for example, serum- and glucocorticoid-regulated kinase and S6K1) presents another potential problem with this approach. For example, the phospho-Akt substrate antibody also recognizes the phosphorylated form of S6 (a substrate of S6K1) (18). Use of Scansite with a matrix based on the selectivity of the phospho-Akt substrate antibody, rather than that of the Akt protein kinase, and with imposed restrictions of isoelectric point and molecular size range allowed prediction of S6 as a major spot visualized by the antibody on two-dimensional immunoblots (18). Novel phosphoproteins identified in this manner can be further analyzed by Scansite to predict which related kinase is likely to be responsible.

Phospho-motif and general phospho-specific antibodies have also been used to immunoprecipitate phosphorylated proteins in order to identify candidate kinase substrates by mass spectrometry (MS) methods. A recent study used the phospho-Akt substrate antibody to immunoprecipitate and identify specific PI3K-dependent phosphoproteins from insulin-stimulated adipocytes (19). One problem with this method is that phospho-motif antibodies are generated against peptides and, hence, will often not be able to bind and immunoprecipitate proteins in their native folded state. In fact, efficient immunoprecipitation with the phospho-Akt substrate antibody requires prior denaturation of proteins within cell lysates (19). General phospho-tyrosine-specific antibodies and MS have also been used to isolate and identify protein substrates of receptor-tyrosine kinases (20, 21). As with other in vivo techniques, this approach requires specific means of activating and inhibiting the kinase whose substrates are being analyzed.

A promising new chemical genetic technology for identifying kinase substrates has been developed (Fig. 1B) (22). This approach takes advantage of the structurally conserved ATP-binding pocket within all kinases to generate mutant alleles that can utilize specific ATP analogs, in addition to ATP. The mutations create a "hole" in the nucleotide-binding domain in a region where the N6 amine of ATP usually sits. This mutation allows the use of an ATP analog with a large hydrophobic moiety, or "bump," positioned off of the N6 amine. Importantly, wild-type kinases cannot utilize these ATP analogs, and these mutations appear to be silent with respect to kinase activity and substrate specificity (22). Assays with a "holed"-kinase mutant in the presence of a "bumped"-[γ-32P]ATP derivative allow the specific phosphorylation, and thus labeling, of substrates targeted by this kinase. In principle, this approach should be applicable to both in vivo labeling of substrates when the mutant kinase is expressed, as well as in vitro biochemical approaches discussed above. Unlike other in vivo approaches, this method eliminates any doubts as to whether the phosphorylation event is direct, because only the kinase being assayed can transfer phosphate from the labeled-ATP analog. However, delivering sufficient quantities of radiolabeled "bumped"-ATP analog into the cell to allow visualization and purification of substrate proteins is likely to present a problem. Although issues with in vitro specificity of kinases still arise, the use of a "bumped"-ATP/"holed"-kinase pair eliminates all background problems when searching for substrates by in vitro kinase assays on fractions from cell lysates. This latter approach has recently been used with a "holed"-JNK allele and has successfully identified a novel JNK substrate (23). The "bump and hole" technology has also been adapted to create kinase-allele-specific inhibitors that are powerful tools for further defining substrate specificities and the overall cellular function of specific kinases (24). This method is particularly useful in model organisms such as yeast, where replacing the wild-type allele with one encoding a "holed" kinase that is sensitized to the "bumped" inhibitor is trivial (25, 26). Overall, the use of "bumped"-ATP analogs and inhibitors for specific "holed" kinases should be a valuable tool for kinase substrate identification. However, the applicability of this approach to all families of protein kinases has yet to be demonstrated.

The Future

The future progress of kinase substrate searches will involve refinement of the technologies discussed here, as well as the development of novel in vitro and in vivo approaches. With the availability of complete genome sequences, proteomic approaches are rapidly emerging to study protein function and modification. Arrays of an organism's entire complement of kinases or its entire proteome ordered onto protein chips represent exciting new tools for the high-throughput identification of kinase substrates in vitro (27, 28). Technologies for the analysis of the "phosphoproteome" are also coming to fruition. Methods for isolating phosphoproteins from in vivo sources and rapidly identifying them by MS are under development (29). This underscores the need for specific inhibitors of kinases to identify candidate kinase substrates through these new approaches. The use of siRNA to uncover the in vivo repertoire of substrates for a specific kinase will also prove invaluable when studying whole-proteome phosphorylation. Finally, the best methods to identify genuine substrates of specific kinases are likely to be combinatorial approaches. For instance, protease cleavage of whole-cell lysates and immunoprecipitation of specific phosphopeptides with phospho-motif antibodies followed by tandem MS (MS/MS) to rapidly identify a specific subset of a cell's phosphoproteins could be a powerful technique in the near future. Regardless of the preferred approach, the search for kinase substrates will go on. Our ability to read the cellular sheet music depends on it.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
View Abstract

Navigate This Article