An affinity-based, cysteine-specific ATP analog for kinase-catalyzed crosslinking

Abstract

Kinases mediate cell signaling pathways by catalyzing protein phosphorylation. Irregularities in kinase activity are directly associated with disease conditions. Therefore, methods to identify substrates of a particular kinase are needed to understand signaling cascades in normal and diseased states. Photocrosslinking ATP analogs provide powerful tools to study kinases by covalently linking kinases with substrates. However, the involvement of UV light and nonspecific reactivity of current ATP-photocrosslinkers challenge kinase-substrate identification. We report here an affinity-based crosslinking ATP analog, ATP-methylacrylamide (ATP-MAc), that contains a cysteine-reactive acrylamide crosslinking group, which avoids the UV irradiation and non-specific reactivity of prior analogs. Using in vitro kinase assays, ATP-MAc acts as a kinase cosubstrate and covalently crosslinks only kinases containing cysteines in the active site. ATP-MAc was also able to crosslink cellular proteins in lysates, documenting compatibility with cell-based studies. Phosphorylation-dependent affinity-based crosslinking with ATP-MAc will have applications in kinase substrate identification to understand kinase-mediated cell signaling.

Graphical Abstract

Characterization of kinase-substrate pairs is an enabling strategy to understand kinase-related diseases and identify new drug targets. An affinity-based kinase-dependent crosslinker, ATP-MAc, is introduced to capture kinase-substrate interactions under physiological conditions. Unlike prior ATP-photocrosslinking analogs, ATP-MAc covalently links only cysteine-containing kinases with their substrates without the need for UV light.

Protein phosphorylation is a ubiquitous post-translational modification that regulates cellular signal transduction.^[1] Protein kinases catalyze phosphorylation of hydroxyl-containing serine, threonine, and tyrosine in eukaryotic protein substrates.^[2] Kinases are effective drug targets since their aberrant activity is often associated with diseases, including cancer.^[3] Characterization of kinase substrates is important to decipher the molecular details of kinase signaling. Unfortunately, the transient nature of kinase-substrate interactions^[4] has challenged characterization of kinase substrates. New methods are needed to fully map cell signaling in normal and diseased states.

Chemical approaches that identify kinase-substrate pairs typically convert the transient kinase-substrate interactions into stable complexes that can be purified and characterized.^[5] As one approach, we have developed crosslinking probes by modifying adenosine 5’-triphosphate (ATP), which is the universal cosubstrate of kinases, at the ɤ-phosphoryl with photocrosslinkers, such as arylazide (ATP-ArN₃) or benzophenone (ATP-BP) (Figure 1A).^[6] Upon UV irradiation, ATP-ArN₃ and ATP-BP covalently link kinases to their substrates in a phosphorylation-dependent manner to facilitate enrichment and identification of kinase-substrate pairs (Figure 1B).^[7] While ATP-ArN₃ and ATP-BP have the advantages of broad reactivity and inducible activation, the disadvantages include nonspecific crosslinking and undesired biological effects due to UV irradiation.^[8] For an example, in our prior work, kinase-catalyzed crosslinking with p53 and ATP-ArN₃ identified the UV-induced p53 kinase DNAPK, which suggests that UV light influenced the outcome of the study.^[7a]

Figure 1.

(A) Chemical structures of ATP and the ɤ-phosphate-modified ATP analogues ATP-arylazide (ATP-ArN₃), ATP-benzophenone (ATP-BP), and ATP-methylacrylamide (ATP-MAc). (B) Kinase-mediated phosphorylation of hydroxyl-containing amino acids (serine, threonine, or tyrosine) with ATP-ArN₃ or ATP-BP (PCL, photocrosslinker), followed by UV irradiation produces a covalent crosslink between the kinase and substrate. (C) Kinase-mediated phosphorylation followed by Michael addition with an active site cysteine by ATP-MAc produces a covalent linkage between the kinase and substrate.

To develop new crosslinking ATP analogs that avoid UV irradiation, we exploited affinity-based crosslinking groups that will form a covalent bond with specific amino acids when in close proximity.^[9] Roughly 200 kinases have reactive cysteine residues located in four regions near the ATP binding pocket.^[10] Considering that some kinases contain cysteines, we developed an ATP-analog with a cysteine-reactive methylacrylamide crosslinking group (ATP-MAc, Figure 1A). While the modified ɤ-phosphoryl group of ATP-MAc is transferred to the kinase substrate, Michael addition between the cysteine in the kinase active site and the methylacrylamide group will crosslink the kinase to its substrate (Figure 1C). The methylacrylamide group was selected due to the reduced electrophilicity imparted by the α-methyl to avoid reactivity with other nucleophilic groups.^[11] The two crucial features of ATP-MAc are compatibility with physiological conditions without UV light and specific targeting of cysteines to eliminate nonspecific reactivity.

To design an ATP-MAc analog appropriate for kinase-catalyzed crosslinking, the structure of the cysteine-containing kinase EGFR was analyzed to reveal that the distance between the ɤ-phosphorus of ATP and the closest cysteine residue in the active site is 8.4 Å (Figure 2).^[12] In addition, the structures of kinases bearing cysteines in the four active site regions were used to dock ATP-MAc (Figure 1A), and the distances were 7.3 to 16.1 Å (Figure S1). Based on the structural analysis, ATP-MAc 1 and 2 were designed with distances between the ɤ-phosphorus and acrylamide CH₂ groups of 22.0 Å and 12.4 Å, respectively, to explore the optimal crosslinking distance. ATP-MAc 1 and 2 were synthesized by incubating methacrylic anhydride with diamines 4 or 5 to give amines 6 or 7 (Scheme 1). Final compounds 1 and 2 were obtained by coupling amines 6 and 7 with ATP.^{[6b, 13]}

Figure 2.

(A) Crystal structure of the EGFR kinase domain (cyan) with the ATP analog, thiophosphoric acid o-((adenosyl-phospho)phospho)-S-acetamidyl-diester, in active conformation (pdb 2GS6). (B) The distance between the ɤ-phosphorus of the ATP analog and sulfur of the EGFR active site cysteine was measured as 8.4 Å using PyMOL software.^[14] Atoms color-coding are C-green, H-gray, N-blue, O-red, P-orange, S-yellow.

Scheme 1.

Synthesis of ATP-MAc 1 and 2.

To test the reactivity of ATP-MAc 1 and 2 in kinase-catalyzed labeling, a fluorescence labeling reaction (Figure 3A) was conducted where diFITC-cystamine 10 (Scheme S1) was incubated with ATP-MAc 1 or 2, PKA kinase, and myelin basic protein (MBP) substrate under reducing conditions (Scheme S2). We hypothesized that PKA will catalyze transfer of the modified ɤ-phosphoryl group of ATP-MAc to MBP, with subsequent reaction of the methylacrylamide with the reduced FITC-thiol to fluorescently label MBP (Figure 3A). After separation by SDS-PAGE and visualization of florescence, MBP was fluorophore labeled by PKA and diFITC-cystamine 10 in the presence of both ATP-MAc 1 and 2 (Figure 3B, lanes 4 and 6). In contrast, reactions without PKA (Figure 3B, lanes 3 and 5) or with staurosporine, a pan kinase inhibitor, did not show fluorescence labeling (Figure 3B, lanes 7 and 8), which confirmed kinase dependency. These results indicated that ATP-MAc 1 and 2 are kinase cosubstrates with reactive methylacrylamide Michael acceptors.

Figure 3.

(A) Kinase-catalyzed fluorophore labeling of MBP (blue circle) with PKA kinase, diFITC-cystamine 10, and ATP-MAc 1 or 2. (B) MBP was incubated with or without PKA, ATP-MAc 1 or 2, and diFITC-cystamine 10, separated by SDS-PAGE, and visualized by in-gel fluorescence (pMBP) or Sypro Ruby total protein stain (MBP). As a control, staurosporine (STSP) kinase inhibitor was included. Reproducible trials are in Figure S19. (C) MPB and PKA were incubated with ATP or ATP-MAc 1 or 2, and then treated with TFA to cleave the phosphoramidate bond, releasing the methylacrylamide group. Reactions were separated by SDS-PAGE, followed by staining with ProQ diamond phosphoprotein stain (pMBP) or SYPRO® Ruby total protein stain (MBP). (D) Phosphorylated MBP (pMBP) in part C was quantified from four independent trials by ImageQuant software, with mean and standard error shown. Reproducible trials are in Figure S20.

The efficiency of ATP-MAc 1 and 2 as cosubstrates was assessed next by quantifying the extent of phosphorylation. ATP, ATP-MAc 1, or ATP-MAc 2 were incubated with PKA and MPB, followed by cleavage of the methylacrylamide modification on the ɤ-phosphoryl group with acid to generate the same phosphorylated product in all reactions. The reaction mixtures were separated by SDS-PAGE and stained with ProQ diamond phosphoprotein stain to monitor the extent of MBP phosphorylation. Phospho-MBP was observed in all reactions (Figure 3C, lanes 2, 3, and 4), consistent with the cosubstrate activities of ATP-MAc 1 and 2. Quantification of four independent trials showed 68 ± 9% conversion with ATP-MAc 1 and 60 ± 9% conversion with ATP-MAc 2, compared to ATP (100%, Figure 3D). The data suggest that both ATP-MAc analogs have acceptable efficiency. ATP-MAc 1 was used in further studies because its longer 22 Å linker provides the appropriate span (Figure S1) needed for crosslinking with cysteine-containing kinases.

ATP-MAc 1 was next tested with two cysteine-containing kinases, epidermal growth factor receptor (EGFR) and fibroblast growth factor receptor 4 (FGFR4) to assess crosslinking compatibility. EGFR and FGFR4 control critical signaling pathways linked to cell proliferation, differentiation, migration, and transcription.^[15] Importantly, EGFR and FGFR4 undergo autophosphorylation and have reactive cysteines in the hinge binding region and the glycine rich loop regions of the ATP binding sites,^[10] making them susceptible to kinase-catalyzed crosslinking with ATP-MAc. To confirm that crosslinking requires a reactive cysteine, PKA kinase, which does not contain a reactive cysteine in its active site, but undergoes autophosphorylation, was also tested.^[10] EGFR, FGFR4, and PKA were separately incubated with ATP-MAc 1, and then crosslinking was observed after SDS-PAGE separation and visualization with EGFR, FGFR4, and PKA antibodies. In the cases of EGFR (89 kDa) and FGFR4 (65 kDa), high molecular weight crosslinked complexes were observed corresponding to dimers (EGFR, 178 kDa; FGFR4, 130 kDa) and trimers (EGFR4, 267 kDa; FGFR4, 195 kDa) only in the presence of ATP-MAc (Figure 4A and 4B, compare lanes 5 and 1). Crosslinked complexes were lost when the reactions were treated with EGFR-selective inhibitor PD153035^[16] (Figure 4A, lane 2) or pan kinase inhibitor staurosporine^[17] (Figure 4B, lane 2), consistent with kinase dependence. Crosslinked complexes were also absent when the cysteine residues of the kinases were alkylated with iodoacetamide prior to crosslinking (Figure 4A and 4B, lane 3) or when ATP-MAc was incubated with the thiol, dithiothreitol, before reaction (Figure 4A and 4B, lane 4), indicating that the thiol and acrylamide functional groups are necessary for crosslinking. In contrast, higher molecular weight crosslinking was not observed with PKA (Figure 4C, lane 5), suggesting that ATP-MAc 1 crosslinks only cysteine containing kinases.

Figure 4.

Kinase-catalyzed crosslinking with ATP-MAc 1. Recombinant EGFR (A), FGFR4 (B), PKA (C), and a mixture of ERK2 and p53 (D), were incubated with ATP-MAc 1 before separation by SDS-PAGE and visualization using SyproRuby total protein stain or antibodies (α) to EGFR, FGFR4, PKA, ERK2, or p53. Arrows show EGFR (A), FGFR4 (B), PKA (C), or ERK2 and p53 (D), whereas brackets indicate higher molecular weight complexes due to crosslinking. As controls, kinases were treated with PD153035 (PD), staurosporine (ST), or magnolin (MG) kinase inhibitors, or iodoacetamide (IA) to alkylate kinase cysteines. As another control, dithiothreitol (DT) was preincubated with ATP-MAc 1 to inactivate the methylacrylamide group. Repetitive trials are shown in Figures S21-S24. (E) Untreated or EGF-treated HeLa cell lysates were incubated with ATP-MAc 1 without or with PD153035 inhibitor (PD), before separation by SDS-PAGE and visualization with antibodies (α) to EGFR. Repetitive trials are shown in Figure S25.

As a second test of ATP-MAc 1, ERK2 crosslinking to its known substrate, p53,^[18] was monitored. ERK2 is a downstream kinase in the MAP kinase pathway, which is linked to cell differentiation, proliferation, and survival.^[19] In contrast to EGFR and FGFR4, the active site cysteine in ERK2 is located in the DFG region,^[20] which will test the influence of active site cysteine location on crosslinking. Incubation of ATP-MAc 1 with recombinant ERK2 (68kDa) resulted in the formation of high molecular weight complexes corresponding to ERK2 dimer (136 kDa) and trimer (204 kDa) (Figure 4D, lanes 6), similar to EGFR and FGFR4. When p53 (80 kDa) was included, p53 was also observed in the high molecular weight (~284 kDa) crosslinked complexes (Figure 4D, lane 7), which suggested that p53 crosslinked to ERK2 after oligomerization. Crosslinking was reduced in the presence of magnolin, an ERK kinase inhibitor,^[21] iodoacetamide, and DTT (Figure 4D, lanes 3–5), which documented the kinase, cysteine, and acrylamide dependence of crosslinking. These results confirm that ATP-MAc 1 can be applied to a variety of cysteine-containing kinases to crosslink substrates.

Finally, to confirm that ATP-MAc 1 can crosslink kinase-substrate pairs in complex lysate mixtures, HeLa cells were stimulated with EGF to activate the EGFR kinase, lysed, and then incubated with ATP-MAc 1. Crosslinking of endogenous EGFR in the lysates was visualized after SDS-PAGE separation, transfer to a PVDF membrane, and probing with EGFR antibody. In addition to EGFR (175 KDa), high molecular weight crosslinked bands were observed (Figure 4E, lane 3), which are likely a combination of EGFR-EGFR, EGFR-substrate and EGFR-kinase complexes. Crosslinked complexes were reduced in reactions without EGF stimulation or ATP-MAc (Figure 4E, lanes 1 and 2) or with EGFR-selective inhibitor PD153035 (Figure 4E, lane 4). Crosslinking was also observed in the presence of excess glutathione (Figure S26). These results indicate that ATP-MAc crosslinks kinase-substrate pairs in a kinase-dependent manner in complex cellular mixtures containing glutathione.

In conclusion, we have developed an affinity-based crosslinking ATP analog, ATP-MAc, that acts as a kinase cosubstrate, crosslinks only kinases with active site cysteines, and performs kinase-dependent crosslinking in a lysate mixture. This cysteine-specific crosslinker avoids the UV light requirement and non-specific reactivity of prior ATP-photocrosslinkers. As future work, ATP-MAc will be utilized in kinase-catalyzed crosslinking and immunoprecipitation (K-CLIP)^[7a] to discover substrate and interacting proteins of cysteine-containing kinases from cellular mixtures and live cells.^[22] Phosphorylation-dependent affinity-based crosslinking provides a powerful tool to identify kinases, their substrates, and phosphorylation-dependent interactions in cell signaling pathways.