Understanding and exploiting substrate recognition by protein kinases

Introduction

Eukaryotic organisms dedicate some 2% of their genes to encode protein kinases, underscoring the widespread importance of protein phosphorylation in the regulation of myriad cellular actions [1]. Phosphorylation on serine, threonine and tyrosine can modulate protein function in many ways: by controlling subcellular localization, by acting as a tag for protein degradation or stabilization, by triggering the assembly of multiprotein complexes, or by allosterically regulating biochemical activity (e.g. activation or repression of an enzyme or transcription factor). Through regulated protein phosphorylation, processes as diverse and proliferation, migration, differentiation, and cell death are thereby subject to control by protein kinases. Each of these processes is deregulated in cancer, and protein kinases are frequently implicated as culprits in neoplastic initiation and progression. The clinical success of the BCR-Abl tyrosine kinase inhibitor imatinib in treating chronic myeloid leukemia was hailed as a triumph in “bench to bedside” translational medicine, and has led to a flood of pharmaceutical industry activity aimed at therapeutic targeting of kinases [2]. In addition to ten protein kinase inhibitors that have received regulatory approval in the United States, there are nearly 100 others in clinical trials. Aside from oncology, kinases are also being targeted in other disease areas, including diabetes, inflammatory disease, and neurodegenerative disease.

The choice of a kinase as a drug target requires a thorough understanding of its role in disease and in normal physiology. Among the challenges facing researchers in the field are to better understand where and when a kinase is activated, how such regulation occurs, how the kinase impacts physiology at the cellular and organismal level, and perhaps most critically, which proteins serve as substrates of the kinase to carry out its downstream effects. Elucidating such mechanistic details for each of the over 500 human kinases is a formidable task. To address these questions, new “kinomic” methods and approaches have emerged that allow global analysis of kinase activities and targets, including focused RNAi screening, kinome-wide inhibitor profiling, mass spectrometry (MS)-based phosphoproteomics, and analysis of proteome microarrays [3–6]. In this review, I will illustrate ways in which a detailed understanding of protein kinase-substrate interactions can help to elucidate cellular signaling pathways, particularly in the identification of new kinase substrates. I will give special attention to chemical and proteomic tools and approaches described within the last two years that have the capacity for large scale analysis of kinase function.

Reengineering the nucleotide binding pocket

All eukaryotic protein kinases share a common overall fold, comprising a β-sheet rich small N-terminal lobe and a mostly helical large C-terminal lobe (Figure 1A). The ATP binding site is at the interface between the two lobes, and most protein kinase inhibitors typically bind at this site as well in a manner competitive with ATP. The adenine moiety of ATP sits within a hydrophobic pocket in the small lobe, with the N⁶ and sometimes N¹ nitrogens making hydrogen bonds to the kinase. In work carried out over the last decade, Shokat and co-workers have elaborated a strategy for exploiting conserved features of the adenine binding site to engineer kinase mutants specific for modified substrates and inhibitors [7,8]. For most kinases, mutation of the so-called gatekeeper residue (equivalent to Thr338 of human c-Src) to glycine or alanine produces an analog specific (AS) allele that can use non-natural ATP analogs bearing bulky residues at the N⁶ position (e.g. N⁶-benzyl-ATP) as a substrate (Figure 1B). Wild-type (WT) kinases, conversely, cannot use the modified nucleotide due to steric clashes with the gatekeeper. Similarly, AS mutants are uniquely inhibited by analogs of kinase inhibitors with appropriately placed bulky groups. These tools have seen increasing utilization as a means to identify protein substrates of kinases and to pharmacologically target a single kinase. Using WT kinases, labeling of substrates in vitro within a complex mixture such as a cell or tissue extract is problematic due to background from the endogenous kinases present [9]. However, incubation of an AS kinase mutant with radiolabeled N⁶-benzyl-ATP (or a similar analog) in a complex protein mixture results in exclusive labeling of direct substrates of the kinase, since endogenous WT kinases cannot use the analog as a substrate.

Identifying the protein substrates detected using AS kinases generally requires biochemical purification and can pose a challenge, particularly if the targets are present at low levels (see for example [10]). Some recent innovations that have facilitated identification of substrates tagged using AS kinases include the use of antibodies to phosphotyrosine (pTyr) to affinity purify tyrosine kinase substrates [11] and an S. cerevisiae strain collection expressing each protein with an affinity tag from its endogenous locus to pick out substrates of yeast kinases [12]. In addition, the Shokat laboratory has found that most AS kinases tested can use ATP-γ-S analogs as the substrate, which yields thiophosphorylated rather than phosphorylated protein products. Subsequent alkylation of the incorporated thiophosphate group with a thiol-reactive electrophilic tag marks direct substrates of the kinase with a unique epitope (Figure 1C) [13,14••]. Though the electrophilic tag used also reacts with cysteine residues present in most proteins, by using high affinity antibodies raised against the full epitope (including the thiophosphate moiety), direct substrates can be specifically immunopurified and identified by MS. The group has validated this approach by isolating a known substrate of the Erk2 mitogen-activated protein kinase (MAPK) from permeabilized embryonic fibroblasts expressing an Erk2 AS allele from the endogenous locus. Interestingly, Zhou and co-workers have reported that at low pH (<4.0), electrophiles such as iodoacetamido-biotin react selectively with thiophosphate over cysteine thiols, allowing site-selective tagging of thiophosphorylated kinase substrates with biotin [15], which could allow for highly efficient purification on immobilized streptavidin. Further, Green and Pflum have shown that several kinases can use ATP biotinylated at the γ position as a substrate, thus transferring phosphobiotin directly to substrates [16]. Though the range of kinases that can accept biotin-ATP as a substrate and precise catalytic parameters for its use have not been determined, this approach holds promise as a generally accessible way to identify new kinase substrates through direct labeling.

Interactions with the protein substrate

The protein substrate binding site has been structurally characterized for a number of kinases through X-ray crystallography in complex with pseudosubstrate inhibitors and short peptide substrates [17–21]. The repertoire of kinase-peptide complex structures is rapidly expanding through the use of high affinity bisubstrate analogs, in which ATP is tethered to the phosphorylation site of a peptide substrate [22–25] (see [26] for a discussion of general features of substrate binding). Interactions with residues surrounding the phosphorylation site vary considerably between kinases, reflecting differences in sequence specificity (Figure 2). This selectivity at the active site of the kinase is reflected in specific sequence motifs found at phosphorylation sites in its protein substrates, and provides one mechanism by which different kinases phosphorylate distinct protein substrates. Protein kinase phosphorylation motifs can be experimentally determined by screening of either immobilized [27–29] or solution phase [30,31] peptide arrays. Peptide libraries immobilized on beads have recently been described for profiling tyrosine kinase specificity [32•]. A pool of beads, each bearing an individual peptide, is treated with the kinase. Phosphorylated beads are then detected using alkaline phosphatase-conjugated anti-pTyr antibodies and then manually segregated from the larger pool of unphosphorylated beads, followed by MS peptide sequencing. The method could theoretically be modified to incorporate fluorescence detection, allowing automated bead selection and higher throughput [33]. Unfortunately, the lack of suitably general antibodies for detecting pSer and pThr [34] restricts the approach to tyrosine kinases.

Conclusions

Though a great deal of new information regarding protein kinase signaling networks has been uncovered in recent years, clearly the current state of knowledge is only the tip of the iceberg. That the majority of phosphorylation sites uncovered by each successive quantitative phosphoproteomics study are novel suggests that thousands, perhaps tens of thousands more sites exist in humans in excess of our current tally. For most of the ones catalogued thus far, the responsible kinases and the functional significance for cellular regulation remain obscure, and this fraction is only likely to increase as phosphoproteomes become more fully defined. The methods described here will hopefully help narrow this growing gap in our knowledge.