Kinase-catalyzed Crosslinking: A Comparison of ATP-crosslinker Analogs

Abstract

Protein phosphorylation is catalyzed by kinases to regulate cellular events and disease states. Identifying kinase-substrate relationships represents a powerful strategy to understand cell biology and disease yet remains challenging due to the rapid dynamics of phosphorylation. Over the last decade, several γ-phosphoryl modified ATP analogs containing crosslinkers were developed to covalently conjugate kinases, their substrates, and their associated proteins for subsequent characterization. Here, kinetics and crosslinking experiments demonstrated that the UV-activated analogs, ATP-aryl azide and ATP-benzophenone, offered the most robust crosslinking, whereas electrophilic ATP-aryl fluorosulfate promoted the most effective proximity-enabled crosslinking. The data will guide future applications of kinase-catalyzed crosslinking to study normal and disease biology.

Graphical Abstract

Phosphorylation is a reversible post-translational modification that influences the function of proteins and controls important biological events. Kinases catalyze phosphorylation,¹ and kinase activity must be highly regulated for healthy and normal biological functions, including proliferation, cell signaling, and differentiation. Uncontrolled phosphorylation due to aberrant kinase activity can alter cell biology and lead to diseases, such as cancers and neurodegeneration. Due to their role in disease-linked abnormal phosphorylation, kinases are attractive drug targets, with a plethora of FDA-approved kinase inhibitors for the treatment of cancer and autoimmune disease.² To facilitate new drug development, phosphorylation events must be thoroughly characterized to understand complex cellular processes and identify disease-promoting kinases.

A challenge with studying kinase-catalyzed phosphorylation is fast product release,³ which makes the isolation of kinase-substrate interactions difficult. In addition, kinase substrate specificity can be influenced by associated scaffold proteins,⁴ which make characterization of proteins in proximity of a phosphorylation event helpful. Methods to monitor kinases, their substrates, and phosphorylation-regulating associated proteins are needed to comprehensively investigate diseased or healthy states.

To provide enabling tools for characterizing kinase-mediated biology, analogs of the universal phosphoryl-donating cosubstrates of kinases, adenosine 5’-triphosphate (ATP, Figure 1A–B),¹ have been developed. Specifically, ATP analogs containing a modification at the terminal γ-phosphoryl are accepted by kinases to transfer the labeled γ-phosphoryl onto the hydroxyl of a serine (Ser), threonine (Thr), and tyrosine (Tyr) of a protein substrate (Figure 1A).^{5, 6} ATP analogs modified at the γ-phosphoryl are broadly accepted by Ser/Thr or Tyr kinases.⁷ Among the various ATP analogs, ATP-crosslinkers (Figure 1B) contain a reactive group at the γ-phosphoryl that acts as molecular glue to covalently conjugate phosphoproteins to their kinases or associated proteins in a phosphorylation-dependent manner (Figure 1C).⁶ By covalently linking a phosphoprotein substrate to its active kinase, the normally dynamic and transient kinase-substrate pair is stabilized. In addition, crosslinking can also occur between nearby scaffolding proteins and the substrate, allowing monitoring of both kinase-substrate pairs and phosphorylation-regulating associated proteins in a single experiment.

Figure 1. ATP analogs and kinase-catalyzed crosslinking.

A. Kinase-catalyzed phosphorylation by kinases with either ATP or an ATP-crosslinker. B. Structures of ATP-crosslinker analogs ATP-aryl azide (ATP-ArN₃, 1a), ATP-benzophenone (ATP-BP, 1b), ATP-methylacrylamide (ATP-MAc, 1c), ATP-hexanoyl bromide (ATP-HexBr, 1d), and ATP-aryl fluorosulfate (ATP-AFS, 1e). C. ATP-crosslinkers can covalently conjugate substrate to kinase or an associated protein. D. Kinase-catalyzed crosslinking with UV-activatable analogs ATP-ArN₃ (1a) or ATP-BP (1b) that contain an aryl azide (ArN₃) or benzophenone (BP) photocrosslinker (PCL). E. Kinase-catalyzed crosslinking with electrophilic analog ATP-MAc (1c), where the electrophile (E) methylacrylamide (MAc) reacts with nucleophilic (Nu) cysteine (Cys) residues. F. Kinase-catalyzed crosslinking with the proximity-reactive aryl fluorosulfate (AFS) or hexanoyl bromide (HexBr) groups in ATP-HexBr (1d) and ATP-AFS (1e), where nucleophilic (Nu) amino acids (Lys, Tyr, His, for ATP-AFS; Cys, Lys, and His for ATP-HexBr) react with the Br and F leaving groups (LG) on the analogs.

Because γ-phosphoryl modified ATP analogs are accepted by all kinases,⁷ which results in crosslinking of all possible kinase-substrate-associated proteins in a complex mixture, the conjugated complexes of a single protein of interest must be selectively isolated prior to analysis. Specifically, enrichment methods have been coupled with kinase-catalyzed crosslinking to isolate and study only the crosslinked complexes of a selected phosphoprotein substrate. In the Kinase-catalyzed CrossLinking And Streptavidin Purification (K-CLASP) method,⁸ crosslinking occurs between a biotinylated peptide substrate to allow streptavidin enrichment of peptide-kinase and peptide-associated protein complexes. K-CLASP is useful when the kinases and associated proteins of a specific phosphorylated amino acid are needed. When all possible phosphorylated sites on a full length protein substrate are of interest, the Kinase-catalyzed CrossLinking and ImmunoPrecipitation (K-CLIP) method relies on immunoprecipitation of the phosphoprotein substrate to enrich crosslinked complexes containing kinase and associated proteins.⁹ After enrichment using K-CLASP or K-CLIP methods, crosslinked complexes can then be identified via gel or liquid chromatography-tandem mass spectrometry (LCMS/MS) analysis. Coupled with LC-MS/MS, unanticipated kinase(s) and associated scaffolding proteins of a phosphoprotein substrate can be discovered from any appropriate cellular mixture. Using the K-CLASP and K-CLIP methods, kinase-catalyzed crosslinking is a versatile tool to both monitor and discover phosphorylation-mediated cell biology.^{10, 11}

A variety of ATP-crosslinker analogs have been developed for kinase-catalyzed crosslinking.^{6, 12–14} These ATP-crosslinkers utilize two mechanisms for covalent conjugation: photoactivation-mediated (Figure 1D) or proximity-enabled (Figure 1E and 1F) reactions. ATP-aryl azide (ATP-ArN₃, 1a) and ATP-benzophenone (ATP-BP, 1b) are two photocrosslinking analogs that rely on UV irradiation to activate the aryl azide (ArN₃) or benzophenone (BP) crosslinking group and conjugate substrate to kinases and associated proteins (Figure 1D).^{6, 12} While ATP-BP and ATP-ArN₃ maintain broad reactivity, a weakness is their reliance on UV light, which can influence the biological activity of certain cellular proteins, potentially biasing the identified complexes. For example, UV-activated DNA-dependent protein kinase (DNA-PK) was enriched in crosslinked samples of the p53 phosphoprotein in K-CLIP studies using UV and ATP-ArN₃,⁹ which suggested that UV irradiation affected the cellular conditions.

To avoid UV activation, three electrophilic ATP-crosslinker analogs were developed to react with nearby nucleophilic amino acids of the kinase or associated protein. ATP-methylacrylamide (ATP-MAc, 1c) uses an acrylamide electrophile as a Michael acceptor to react with nucleophilic thiol of cysteine (Figure 1E).¹³ However, only half of the 535 known human kinases contain a reactive cysteine near the active site,¹⁵ which limits applications of ATP-MAc. To expand affinity-based crosslinking to all kinases, ATP-hexanoyl bromide (ATP-HexBr, 1d) and ATP-aryl fluorosulfonate (ATP-AFS, 1e) were developed to crosslink with multiple amino acid side chains, in addition to Cys (Figure 1F),¹⁶ allowing crosslinking with any kinase.¹⁴ Specifically, the bromine or fluorine leaving groups of ATP-HexBr or ATP-AFS, respectively, are displaced by nucleophilic side chains of Lys, His, and Cys (ATP-HexBr) or Lys, His, and Tyr (ATP-AFS).¹⁶ The toolbox of five ATP-crosslinkers (Figure 1B) expands the possible applications of kinase-catalyzed crosslinking to study kinase biology.

With a toolbox of ATP-crosslinkers available, here we evaluated the strengths and weaknesses of each analog to promote successful application to any cellular mixture and protein of interest. The five crosslinking ATP analogs (Figure 1B) were tested with two representative kinases: the Ser/Thr-specific cyclic AMP-dependent protein kinase (PKA) and the Tyr-specific epidermal growth factor receptor (EGFR). In addition to their different amino acid preference, both PKA and EGFR regulate normal cell function, including cell signaling, and are linked to a variety of diseases states, including cancer.^{17, 18} PKA and EGFR also are well characterized biochemically,¹ making them ideal test kinases.

As a first step to compare the ATP-crosslinkers, the cosubstrate compatibilities of the analogs were assessed using an NADH-coupled kinase kinetics assay, a previously described method.^{19, 20} Briefly, production of the ADP byproduct of the kinase reaction with ATP or an ATP analog was monitored by coupling with the activities of pyruvate kinase and lactate dehydrogenase, which consume light-absorbing NADH. Kinase activity was indirectly measured by loss of NADH absorption. ATP or the five ATP-crosslinkers were separately incubated with recombinant PKA or EGFR kinase, peptide substrate, and the enzyme-coupled assay components. Absorbance measurements at 340 nm were taken every 30 seconds to monitor the kinase-coupled oxidation of NADH. The rate of NADH consumption was plotted against time to determine initial rates, and a Michaelis-Menten plot was generated from the initial rates to calculate kinetics values (Figure S1).

The kinetics analysis confirmed that each analog was a viable cosubstrate of both PKA and EGFR, with catalytic turnover observed with all analogs (Table 1). As expected, ATP exhibited the best turnover with both PKA and EGFR among all cosubstrates tested, as illustrated by the highest catalytic efficiencies (k_cat/K_M, Table 1, 430 ± 80 s ⁻¹mM⁻¹ with PKA and 13 ± 2 s⁻¹mM⁻¹ with EGFR). Evident in the catalytic efficiency with ATP, catalytic turnover with EGFR in general was slower than with PKA. Among the five ATP-crosslinkers, ATP-HexBr and ATP-AFS displayed the best catalytic efficiency (Table 1, 52–85 s⁻¹mM⁻¹ with PKA and 1.5–6.9 s⁻¹mM⁻¹ with EGFR). With only modest reduction in k_cat (Table 1, <2-fold for PKA and <3-fold for EGFR), differences in catalytic efficiency with ATP-HexBr and ATP-AFS reflected a substantial increase in K_M compared to ATP (4−16-fold for PKA and 1.2–9-fold for EGFR). Similarly, while the k_cat values were only modestly reduced for ATP-ArN₃, ATP-BP, and ATP-HexBr (Table 1, <2-fold for PKA and <3-fold for EGFR), K_M values were reduced to an even greater extent than ATP-HexBr and ATP-AFS (~15-fold for PKA and 5 to 10-fold for EGFR), resulting in the poorest efficiencies among the analogs. Nonetheless, all analogs were catalytically competent and viable for kinase-catalyzed crosslinking.

Table 1:

NADH-linked kinase kinetics assay.^*

Cosubstrate	PKA				cEGFR
	KM (μM)	kcat (s−1)	kcat/KM (s−1 mM−1)	ratio	KM (μM)	kcat (s−1)	kcat/KM (s−1 mM−1)	ratio
ATP	20 ± 3	8.5 ± 0.4	430 ± 80	–	220 ± 20	2.9 ± 0.1	13 ± 2	–
ATP-HexBr	78 ± 2	6.6 ± 0.1	85 ± 3	5	260 ± 20	1.8 ± 0.1	6.9 ± 1.0	2
ATP-AFS	140 ± 10	7.5 ± 0.2	52 ± 5	8	600 ± 90	0.9 ± 0.1	1.5 ± 0.4	9
ATP-ArN3	300 ± 20	6.9 ± 0.2	23 ± 3	19	1400 ± 400	1.3 ± 0.2	0.9 ± 0.4	14
ATP-BP	300 ± 20	7.2 ± 0.3	23 ± 4	19	1200 ± 300	0.9 ± 0.1	0.8 ± 0.3	16
ATP-MAc	290 ± 20	4.7 ± 0.2	16 ± 2	27	1900 ± 400	2.0 ± 0.2	1.1 ± 0.3	12

^*Michaelis-Menten constant (K_M), turnover number (k_cat), and catalytic efficiency (k_cat/K_M) are shown for PKA and cEGFR. The ratio was calculated by dividing the k_cat/K_M of ATP by the k_cat/K_M of each ATP analog. Kinetic values were calculated from at least 3 independent trials for each ATP analog/kinase pair, with plots shown in Fig. S1.

Given that the kinetics assay measured consumption of ATP, and the ATP analogs can hydrolyze to ATP in solution,²¹ a concern was that degradation of the analogs was complicating the kinetics assay. To address this concern, analog purity was evaluated by this layer chromatography (TLC) analysis and then compared to the kinetics measurements. No correlation was observed between analog purity (70–95%) and V_max or K_M values (Figure S2), which indicated that the kinetics measurements were independent of analog purity. A second concern was that covalent reaction of the electrophilic ATP analogs with the kinase could complicate kinetics measurements. To assess if kinetics were impacted by electrophilic crosslinking, ATP-AFS was preincubated with beta-mercaptoethanol (BME) or dithiothreitol (DTT) reducing agents to quench the aryl fluorosulfate electrophile prior to addition of PKA and kinetics analysis. Kinetics measurements were similar regardless of quenching, with only 1.2-fold difference in catalytic efficiencies (Figure S3, k_cat/K_M values of 52, 42, and 42 s⁻¹mM⁻¹ with untreated, BME-quenched, and DTT-quenched ATP-AFS, respectively), which indicates that catalysis was independent of the nature of the crosslinker.

The kinetics data with the five ATP-crosslinkers is consistent with prior studies with other γ-phosphoryl modified ATP analogs. Kinetics analysis using ATP-dansyl and PKA demonstrated 18-fold reduction in catalytic efficiency compared with ATP.²⁰ In addition, ATP-biotin and PKA displayed a 7-fold reduction in catalytic efficiency (k_cat/K_M) compared to ATP.⁷ In fact, a more comprehensive kinetics analysis with ATP-biotin and 24 human kinases documented 1.1 to 16-fold reduction in efficiencies compared to ATP.⁷ Combined with the 5 to 27-fold reduced catalytic efficiencies observed here with the five ATP-crosslinkers, the data indicate that all kinases and γ-phosphoryl modified ATP analogs, regardless of the attached modification, demonstrate similar efficiencies.

Prior work documented the critical role of the polyethylene glycol linker in each ATP analog to position the bulky modification (crosslinker, biotin, dansyl) outside the active site of the kinase.²² Because the ATP-crosslinker analogs contain similar polyethylene glycol linkers, only modest differences in catalytic efficiency compared to ATP were observed, which implies that the attached group has only minimal influence on kinetics. For example, ATP-ArN₃ and ATP-AFS are structurally similar with identical polyethylene glycol linkers and para-substituted aryl groups (Figure 1B) and showed only 2-fold difference in catalytic efficiencies (Table 1, 52 s⁻¹mM⁻¹ and 23 s⁻¹mM⁻¹ with PKA). As another example, ATP-MAc contains the longest polyethylene glycol linker of the five ATP-crosslinker analogs (Figure 1B), yet demonstrated the poorest catalytic efficiency (Table 1, 16 s⁻¹mM⁻¹ with PKA). Given that enzymes generally accelerate biological reactions by several orders of magnitude,²³ the <5-fold difference seen among the ATP-crosslinker analogs is modest and acceptable.

With the compatibility of the five analogs established using kinetics assays, the crosslinking capabilities of the analogs were surveyed with both PKA and EGFR. We note here that PKA lacks an active site Cys and should be unable to crosslink with ATP-MAc, whereas EGFR contains an active site Cys compatible with ATP-MAc crosslinking.¹⁵ Recombinant PKA or EGFR was incubated with each analog, with the expectation that high molecular weight crosslinked complexes due to autophosphorylation would be observed after SDS-PAGE separation and western blotting. Considering PKA, higher molecular weight crosslinked complexes consistent with crosslinking of several PKA proteins due to autophosphorylation were robustly observed with UV-dependent ATP-BP and ATP-ArN₃ (Figure 2A, lanes 1 and 2) as compared to the uncrosslinked PKA (43 kDa) control (Figure 2A, lane 6). The electrophilic analogs showed less efficient crosslinking (Figure 2A, lanes 3–5) compared to UV-induced analogs (Figure 2A, lanes 1 and 2). Among the electrophilic analogs, ATP-AFS showed the most crosslinked complexes compared to the uncrosslinked control (Figure 2A, compare lanes 5 and 6). ATP-HexBr did not generate crosslinked complexes above background (Figure 2A, compare lanes 4 and 6), which suggests poor crosslinking efficiency. Higher molecular weight complexes were also absent with ATP-MAc compared to the uncrosslinked PKA (Figure 2A, compare lanes 3 and 6), which was expected because PKA does not have a Cys near the active site.

Figure 2.

Kinase-catalyzed crosslinking with the five ATP-crosslinkers. PKA (A) or EGFR (B) were incubated without (lane 6) or with each ATP analog (10 mM; ATP-BP, lane 1; ATP-ArN₃, lane 2; ATP-Mac, lane 3; ATP-HexBr, lane 4; and ATP-AFS, lane 5) for 2 hours at 31°C. After crosslinking reactions, proteins were separated via SDS-PAGE and visualized by western blot analysis with PKA (α-PKA, A) or EGFR (α-EGFR, B) antibodies. Arrows indicate the PKA (43 kDa) or EGFR (90 kDa) monomers, whereas brackets indicate crosslinked complexes. Molecular markers are indicated to the left of the gel images. Additional independent trials are shown in Figure S4 and S5.

With EGFR, higher molecular weight complexes consistent with crosslinking of several EGFR proteins due to autophosphorylation were robustly seen with UV-dependent ATP-BP and ATP-ArN₃ (Figure 2B, lanes 1 and 2) when compared to the uncrosslinked EGFR (90 kDa) control (Figure 2B, lane 6), similar to PKA. A reactive cysteine is present near the active site of EGFR, allowing for ATP-MAc crosslinking. As expected, modestly elevated crosslinking was observed with ATP-MAc compared to the uncrosslinked control (Figure 2B, compare lane 3 and 6). The reduced crosslinking with ATP-MAc compared to UV-inducible analogs (Figure 2B, compare lanes 1 and 2 to lane 3) is likely due to reliance on only a single low abundance Cys amino acid for crosslinking. Higher molecular weight crosslinked complexes were seen with ATP-AFS and ATP-HexBr (Figure 2B, 4 and 5) when compared to uncrosslinked control (Figure 2B, lane 6). Among the electrophilic analogs, ATP-HexBr generated the most complexes with EGFR (Figure 2B, lane 4 compared to lane 6). Considering both PKA and EGFR, the most robust crosslinking was generated with the UV-activated analogs ATP-BP and ATP-ArN₃.

Comparing the crosslinking and kinetics data of each ATP-crosslinker, no correlation is observed between catalytic efficiency and crosslinking ability. For example, among the five ATP-crosslinkers, the most robust crosslinking was observed with ATP-ArN₃ and ATP-BP (Figure 2, lanes 1 and 2), yet the best kinetic efficiencies were observed with ATP-HexBr (Table 1). Similarly, despite the high catalytic efficiency of ATP-HexBr (Table 1), relatively low levels of crosslinking were observed with ATP-HexBr. The combined data suggest that crosslinking can occur with any kinetically viable ATP-crosslinker, with crosslinking efficiency dictated by other factors.

One factor likely dictating robustness of the crosslinking reaction is the reactivity of the crosslinking group. For example, ATP-ArN₃ and ATP-BP contain aryl azide and benzophenone groups, respectively, that become highly reactive with UV light and could effectively generate crosslinked complexes (Figure 2A and 2B, lanes 1 and 2). In the case of ATP-ArN₃, the aryl azide crosslinker reacts with a wide variety of functional groups due to UV-mediated generation of a highly reactive nitrene species that can insert into even C-H bonds.²⁴ Given their wide reactivity, the use of aryl azides for protein crosslinking has been well-established.²⁵ Similarly, the benzophenone in ATP-BP generates with UV irradiation a highly reactive carbene, which also inserts across C-H bonds.²⁴ Comparing the two UV-activated crosslinkers, nitrene formation by aryl azide is a terminal reaction in which the nitrene must react with another compound to create a stable adduct. In contrast, the UV-induced excited state of benzophenone can either return to the ground state or access many resonance structures, which might decrease crosslinking.¹⁶ Longer irradiation times are often required to produce crosslinked complexes with benzophenone.²⁶ However, advantages of benzophenone include reversibility that avoids reaction with water and insensitivity to ambient light, unlike aryl azides.^{26, 27} While previous findings demonstrated that ATP-ArN₃ is a superior photocrosslinker to ATP-BP,¹² both ATP-ArN₃ and ATP-BP were similarly effective in crosslinking here, which expands the toolbox of compounds available for crosslinking.

In contrast to the UV-inducible ATP-ArN₃ and ATP-BP analogs that can react with any amino acid side chain or backbone,^{24, 26} the electrophilic analogs ATP-MAc, ATP-AFS, and ATP-HexBr react only with nucleophilic functional groups on amino acid side chains, including Cys, Lys, His, and Tyr.¹⁶ In fact, electrophilic crosslinking is often described as “proximity-enabled” due to the need for close proximity to the nucleophilic amino acid side chains. Electrophiles also react on the minute to hour time scale, unlike UV-activated crosslinkers that have ns to μs half-lives.¹⁶ Given the amino acid and spatial requirements of electrophilic crosslinkers, the electrophilic ATP-crosslinker analogs generated fewer covalently crosslinked complexes compared to the UV-inducible analogs (Figure 2). As a future direction, additional studies are needed to explore the linker at the γ-phosphoryl of the electrophilic ATP-crosslinker analogs. The length and composition of the linker likely influence crosslinking by dictating proximity of the electrophilic crosslinker to nucleophilic amino acid side chains. A toolbox of electrophilic ATP-crosslinking analogs with different linker lengths would be helpful for optimal crosslinking with any kinase.

Although the UV-inducible ATP analogs, ATP-ArN₃ and ATP-BP, generated the most robust crosslinking, some proteins are influenced by UV light. As mentioned previously, DNA-PK is a protein kinase that is stimulated by UV light. Similarly, c-Jun N-terminal kinase (JNK1) is a UV-inducible kinase.²⁸ Given their sensitivity, UV irradiation during crosslinking might hyperactivate DNA-PK and JNK1 to generate super-physiological crosslinking, which could overemphasize substrates of these kinases. Beyond activation of select kinases, UV light can induce stress responses in cells that can trigger kinase-mediated signaling, including p38 or ERK MAP pathways, which could influence phosphorylation-mediated events.²⁹ When UV light is a confounding factor, the electrophilic ATP-crosslinkers become useful options. If the study requires the inclusion of all cellular kinases, then ATP-AFS is an ideal due to the broad reactivity with different amino acid side chains, which likely enhanced crosslinking (Figure 2, lane 4). As an alternative, ATP-MAc is a practical option when focusing on kinases of interest containing a reactive Cys residue near the active site.¹⁵

In summary, five ATP-crosslinker analogs were experimentally compared. Kinetics demonstrated all ATP analogs acted as kinase cosubstrates, which is a prerequisite of crosslinking. In kinase-catalyzed crosslinking reactions, the UV-activated analogs, ATP-BP and ATP-ArN₃, generated the most robust crosslinked complexes. However, where UV light would complicate the cellular system, ATP-AFS utilizes proximity-enabled crosslinking with a variety of nucleophilic amino acid side chains. By defining the optimal ATP-crosslinker for various scenarios, the K-CLIP or K-CLASP methods can be effectively applied to study kinase biology, discover kinase-substrate pairs, or probe phosphorylation-mediated protein interactions.

Highlights.

• ATP-crosslinkers promote monitoring of kinase-substrate relationships

• Five ATP-crosslinker analogs were compared in kinase-catalyzed crosslinking

• UV-inducible ATP-arylazide and ATP-benzophenone generated the most crosslinking

• Where UV must be avoided, ATP-arylfluorosulfate gave the most crosslinking