Mechanistic Studies of Bioorthogonal ATP Analogues for Assessment of Histidine Kinase Autophosphorylation

Abstract

Phosphorylation is an essential protein modification and is most commonly associated with hydroxyl-containing amino acids via an adenosine triphosphate (ATP) substrate. The last decades have brought greater appreciation to the roles that phosphorylation of myriad amino acids plays in biological signaling, metabolism, and gene transcription. Histidine phosphorylation occurs in both eukaryotes and prokaryotes but has been shown to dominate signaling networks in the latter due to its role in microbial two-component systems. Methods to investigate histidine phosphorylation have lagged behind those to study serine, threonine, and tyrosine modifications due to its inherent instability and the historical view that this protein modification was rare. An important strategy to overcome the reactivity of phosphohistidine is the development of substrate-based probes with altered chemical properties that improve modification longevity but that do not suffer from poor recognition or transfer by the protein. Here, we present combined experimental and computational studies to better understand the molecular requirements for efficient histidine phosphorylation by comparison of the native kinase substrate, ATP, and alkylated ATP derivatives. While recognition of the substrates by the histidine kinases is an important parameter for the formation of phosphohistidine derivatives, reaction sterics also affect the outcome. In addition, we found that stability of the resulting phosphohistidine moieties correlates with the stability of their hydrolysis products, specifically with their free energy in solution. Interestingly, alkylation dramatically affects the stability of the phosphohistidine derivatives at very acidic pH values. These results provide critical mechanistic insights into histidine phosphorylation and will facilitate the design of future probes to study enzymatic histidine phosphorylation.

Graphical Abstract

INTRODUCTION

Phosphorylation of proteins is a ubiquitous post-translational modification (PTM) in eukaryotes and prokaryotes that leads to the control of a vast range of events from cell signaling to transcription.^1–3 This process is achieved by protein kinases and results from the transfer of a phosphate group in adenosine triphosphate (ATP) to a protein substrate on an amino acid side chain. While the phosphorylation of tyrosine, threonine, and serine has been extensively studied, much less attention has been paid to histidine phosphorylation. Aside from historically being considered a minor phosphorylation event, although we now know that phosphohistidines are found in 6% of phosphorylated proteins,⁴ the instability of the phosphohistidine (pHis) species has rendered its study difficult.¹ The P–N bond in phosphoramidates, such as pHis, is acid labile but alkali stable. Under acidic conditions, the pHis is protonated and, thus, more prone to hydrolysis; P–N bond hydrolysis is highly favorable with a ΔG° of −12 to −13 kcal·mol⁻¹ for pHis. This is in contrast to the ΔG° of −6.5 to −9.5 kcal·mol⁻¹ for loss of the phosphate from phosphohydroxyamino acids such as serine, threonine, and tyrosine, which are commonly found in mammalian kinase biology.⁵ Consequently, methods applied to the study of eukaryotic kinases, which often involve acid treatment, are not suitable.⁵

Histidine phosphorylation plays crucial roles in prokaryotes in two-component signal transduction systems (TCSs), phosphorelays, and in metabolism by helping the organisms to respond to their continuously changing environment and in eukaryotes in various cellular mechanisms such as phagocytosis and cell cycle events.^1,6–10 The TCSs are particularly of interest, as they are involved in numerous bacterial processes, including growth and cell maintenance, but also bacterial virulence (biofilm formation, toxin production, etc.), antibiotic resistance, and communication, making them an attractive target for antibiotic development.^11,12 There are up to 150 different TCSs per organism, responding to various stimuli.⁷ Therefore, methods to globally study TCSs are critical to understanding their activity, function, and the stimuli which trigger them, as most remain unknown to date.⁷

A typical TCS consists of a histidine kinase (HK) and a cognate response regulator (RR). The HK is the first responder to a stimulus through recognition by its periplasmic sensor, which leads to a signaling cascade. The cytoplasmic domain becomes poised to initiate an autophosphorylation event and binds to ATP in the catalytic domain (CA) of the HK. Subsequently, the γ-phosphate of ATP is transferred to the conserved histidine residue on the dimerization histidine phosphotransfer (DHp) domain and then, to a conserved aspartate on the cognate RR. The phosphorylated RR can initiate a downstream cellular response (Figure 1A).^13,14 Histidine phosphorylation has been observed on both the τ-(tele) and the π- (pros) positions of this heterocycle in nature (Figure 1B).² Phosphorylation on the τ-position is thermodynamically favored, yielding the most stable isomer.^5,15

Figure 1.

Autophosphorylation cascade. (A) A stimulus is received by the extracellular domain causing a conformational change in the HK through the transmembrane domain (TM) and, consequently, the autophosphorylation event. First, ATP binds to the catalytic domain (CA) in the ATP-binding pocket. Next, the γ-phosphate is transferred to the conserved histidine on the dimerization and histidine phosphotransfer (DHp) domain. The phosphoryl group is transferred to an aspartic acid on the response regulator (RR), which consequently binds to DNA initiating a cellular response. (B) The structures of ATP and the resulting phosphorylated histidine with the transferred phosphate boxed in pink. Phosphorylation at the τ position is thermodynamically favored.

A wide array of methods has been developed to study HK activation. Genetic methods have been employed to assess the function of HKs either by knockout of individual proteins, mutations in specific HKs, or by assessing HK expression via quantitative determination of mRNA levels.^16–20 However, these processes are lengthy, tedious, and cannot easily pinpoint the specific HK(s) being activated. Recently, Lemeer and coworkers investigated the levels of histidine phosphorylation in Escherichia coli by mass spectrometry after phosphate enrichment on an Fe³⁺-IMAC column. They found that at acidic pH (2.3), hydrolysis of most pHis species could be prevented.²¹ Nevertheless, these conditions lead to an incomplete profile of activated HKs, perhaps due to the low abundance of these species and/or hydrolysis of some pHis species in the acidic buffer solution. They detected two HKs out of a possible 29 from E. coli, ArcB and DcuS (7% coverage). In addition, Muir and coworkers have reported a unique mass spectrometry fragmentation mechanism for pHis identification, although no HKs were seen from E. coli.²² Enrichment techniques with antibodies have been implemented to globally study kinomes. Marletta and coworkers designed semisynthetic epitopes of pHis species for further detection by alkylated thiophosphate-specific antibodies.^23,24 Nonhydrolyzable analogues of pHis were successfully used to raise pHis antibodies, enabling the visualization of KinB HK phosphorylation in vitro and immunoenrichment of pHis peptides from E. coli lysates.^15,25 Hunter and coworkers produced antibodies specific to τ-pHis and π-pHis.^{10,15,23–25} So far, no antibody has been able to isolate a pHis in a HK from a complex mixture. Moreover, the low abundance of phosphorylated HKs in the bacterial proteome (tens to hundreds of copies per cell of each HK) is a hindrance to their detection.²⁶ Consequently, probes that stabilize the phosphoramidate bond and enable the enrichment of the pHis species are needed.

We sought to design and synthesize activity-based probes inspired by ATP wherein we could modify the γ-phosphate group to contain functionalities with disparate chemical properties but that we still anticipated would mimic the native ATP substrate of the HKs. It has previously been shown that ATPγS is accepted by HKs and that the P–N bond is stabilized due to the lower electronegativity of the sulfur atom and the resulting decrease in electrophilicity of the thiophosphate.⁸ We developed the first nonradioactive ATP-based probe, BODIPY-FL-ATPγS (B-FL-ATPγS), which enables direct visualization of the HKs.²⁷ We have recently shown that B-FL-ATPγS is turned over with lower catalytic efficiency compared to ATP by nearly 2 orders of magnitude, not due to the presence of thiophosphate, but likely because of the bulky BODIPY group that may disrupt substrate recognition.²⁸ Therefore, we sought to decrease the bulk of the transferred group. We postulated that ATP-based probes with a terminal γ-amino-, γ-thio-, and γ-hydroxyphosphate would all be accepted as pseudosubstrates of the HKs. We anticipated that the decreased electronegativity of N in comparison to O could provide similar advantages as the S atom used in our previous probe design. Indeed, Pflum and coworkers have described the use of ATP-based probes containing a terminal aminophosphate derivative for the study of mammalian kinases.^29,30 Kraatz and coworkers have utilized a 5′-γ-ferrocenyl-adenosine triphosphate, another aminophosphate analogue, for monitoring histidine kinase activity electrochemically.³¹

Here, we sought to establish an understanding of the molecular requirements for the design of efficient HK probes based on the ATP scaffold. We investigated three ATP derivatives, both experimentally and computationally, that possess γ-phosphates conjugated to a terminal propargyl group, γ-propargyl-ATP (Probe O), γ-[(propargyl)-thio]-ATP (Probe S) and γ-[(propargyl)-imido]-ATP (Probe N; Figure 2). The terminal alkyne abrogates the need for a bulky tag, such as BODIPY, and has versatile uses. We hypothesized that the small size of the alkyne would minimize steric clash with the HK during the initial binding to the CA domain and improve reaction rates with the nucleophilic histidine, leading to overall increased protein labeling. The alkyne moiety also enables the installation of a fluorophore for gel-based visualization or a biotin affinity tag for pulldown assays through a Cu(I)-catalyzed alkyne–azide cycloaddition (CuAAC) reaction.^32–34

Figure 2.

Structures of ATP-based probes containing a terminal propargyl phosphate; γ-propargyl-ATP (Probe O), γ-[(propargyl)-thio]-ATP (Probe S), and γ-[(propargyl)-imido]-ATP (Probe N) and their corresponding pHis after the autophosphorylation event. The ATP-based probes bear an alkyne-modified phosphate, thiophosphate, or aminophosphate.

RESULTS AND DISCUSSION

To provide clues useful for the development of efficient HK probes, we first determined the kinetic parameters for the γ-modified ATP-based molecules (Figure 2). Probe S was synthesized from ATPγS and propargyl bromide via nucleophilic substitution (Supplementary Scheme 1). Probes N and O are commercially available. The three probes were tested for recognition with a model HK, HK853 from Thermotoga maritima, which is constitutively active²⁷ and demonstrated to autophosphorylate on the τ-position.^35,36 This construct does not need to be activated by its stimulus, which is unknown, to bind to ATP and perform histidine phosphorylation. Throughout the experiments, we evaluated and quantified pHis formation using a fluorescence gel-based assay. In short, HK853 was incubated with Probe O, S, or N.^28,37 After a denaturing quench, the alkyne handle was reacted with TAMRA-PEG3-azide (TAMRA-N₃) through a CuAAC for 30 min (Supplementary Figure 1), and the sample was separated on an SDS-PAGE gel, followed by in-gel fluorescence quantification (Figure 3, Supplementary Figure 2 and Table 1).³⁸

Figure 3.

Comparison of Michaelis–Menten plots of HK853 autophosphorylation with Probes O, S, and N. HK853 (2 μM) was incubated with various concentrations of the probes (1 to 60 μM) at 25 °C. The reactions were quenched at 15 min, and TAMRA-N₃ was conjugated by CuAAC to the propargyl handle (30 min). Formation of the pHis species was quantified by gel-based analysis (n = 2, error bars are standard error of the mean).

Table 1.

HK853 Autophosphorylation Relative Kinetic Parameters for Probes O, S, and N^a

Probe	Binding Affinity Km (μM)	Turnover Apparent kcat (s−1)	Catalytic Efficiency Apparent kcat/Km (M−1·s−1)b
Probe O	4.8 ± 1.2	1.5 ± 0.1	0.31 (100%)
Probe S	29 ± 12	2.1 ± 0.6	0.070 (22%)
Probe N	6.9 ± 3.3	0.15 ± 0.03	0.022 (7%)

^aK_m values are absolute values and obtained using Michaelis–Menten plots. Apparent turnover (k_cat,app) values were obtained from Figure 3. Apparent catalytic efficiency (k_cat/K_m) values were calculated based on K_m and k_cat,app values. Percentages were calculated against the catalytic efficiency value for Probe O.

^bPercentage relative to Probe O catalytic efficiency.

We assessed the following kinetics parameters for all three probes: binding affinity (K_m), turnover (k_cat), and catalytic efficiency (k_cat/K_m). Here, we present apparent k_cat and V_max values as the alkyne probes do not provide a direct read-out like radioactive (i.e., [γ−33P]-ATP) or fluorescent probes (i.e., B-FL-ATPγS) due to the CuAAC step. Indeed, we found that in addition to the CuAAC reaction not providing a quantitative yield of the triazole product (Supplementary Figure 3),³⁹ the resulting pHis species are somewhat unstable under the CuAAC conditions when we compared the signal intensity resulting from the labeling of HK853 by B-FL-ATPγS to that of the labeled enzyme subsequently exposed to CuAAC conditions (signal decreased ~50%; Supplementary Figure 4).

To obtain K_m values, we ran individual assays with increasing concentrations of Probes O, S, or N (Table 1, Supplementary Figures 5–7). Probe N gave a weak fluorescent signal; thus, we increased the protein concentration to 3 μM of HK853 for Probe N instead of 1 μM, which was used for Probes O and S, to improve the signal-to-noise ratio and enable us to quantify the fluorescent bands. Nonetheless, quantification of the bands generated by the formation of the aminophosphohistidine was difficult due to their extremely low intensity. We determined the linear range of the experiments through time-dependent assays for each probe (Supplementary Figures 8–10). A reaction time of 15 min enabled us to observe quantifiable labeling while staying in the linear range for a protein concentration of 1–3 μM. The HK853 pHis product formed with Probe N was approximately one order of magnitude lower than with Probes O and S (Figure 3), which is consistent with the fact that the nitrogen–phosphorus–nitrogen linkage resulting from the formation of the aminophosphohistidine is less stable than a nitrogen–phosphorus–oxygen linkage.⁴⁰ Therefore, the aminophosphohistidine product is likely more susceptible to hydrolysis, limiting the observed labeling.

We found that Probe N has a K_m of 6.9 μM, which is comparable to Probe O (K_m = 4.8 μM), ATP (K_m = 5.2 μM), and ATPγS (K_m = 6.3 μM). We attribute this to the similar atom sizes of O (r = 0.066 nm) and N (r = 0.070 nm), which enables good recognition of these probes by the CA domain of HK853. In contrast, Probe S has a lower binding affinity for HK853 (K_m = 29 μM) than Probe O or Probe N. This result was surprising as we previously determined that B-FL-ATPγS, with a large BODIPY group on the terminal phosphate, has a lower binding affinity (K_m = 6.3 μM),²⁸ perhaps implying productive interactions between the protein and BODIPY. The apparent turnover of Probe S (k_cat,app = 2.1 s⁻¹) is similar tothat of Probe O (k_cat,app = 1.5 s⁻¹) and significantly higher than that of Probe N (k_cat,app = 0.15 s⁻¹; Table 1). Given that ourprevious work demonstrated that ATP is turned over ~6× faster than ATPγS,²⁸ this result may indicate that the Probe S-derived pHis species is stabilized in comparison to the product formed by autophosphorylation with Probe O (i.e., more product is detected because the species is more stable).

Although Probes O and S have similar apparent turnover rates (1.5 and 2.1 s⁻¹, respectively), the binding of Probe O is superior, and thus, this molecule has the highest apparent catalytic efficiency of all three probes (Table 1; Probe S is 22% relative to Probe O). Probe N has the lowest catalytic efficiency (7% relative to Probe O). Next, we compared the labeling of HK853 obtained with Probes O, S, and N and BFL-ATPγS (Supplementary Figure 11).²⁷ We observed that BFL-ATPγS yields a signal about 65% lower than those of Probes O and S. In total, these results indicate that molecules containing the native γ-O-phosphate, but modified with an alkyl group, are likely to provide the best compromise between substrate turnover and product stability, and thus, the most efficient means to study the HKs.

To gain insight into the observed kinetic parameters, we developed a computational model for the active site of HK853 shown in Supplementary Figure 15, which includes ATP, Probe S, Probe O, or Probe N and five catalytic amino acids (His243, Asp244, Asn343, Arg392, and Asn347). These five amino acids are required for the autophosphorylation to occur and maintain ATP in the correct position within the protein.³⁵ We used DFT to compute the autophosphorylation reaction barriers and reaction thermodynamics (the free energy difference between product and reactant) using the B3LYP⁴¹ exchange correlation functional with D3 dispersion correction,⁴² 6–311+G(d,p) basis set^43–47 and SMD solvation model⁴⁸ as shown in Table 2. To test the sensitivity of our calculations to basis set and exchange correlation functional we tested other levels of theory, which are provided in Supplementary Tables 1 and 2. While the absolute values of these reaction barriers are sensitive to the level of theory, we see a steady trend at all levels of theory. These calculations consistently show that HK853 autophosphorylation has the lowest reaction barrier with ATP, followed by Probe S, Probe O, and then Probe N. The reaction thermodynamics for all the substrates favor their corresponding products (negative value corresponds to thermodynamically favorable). The computed reaction barriers correspond to the turnover rates given in Table 1 that were measured at V_max, where the enzyme is saturated with the substrate and all the catalytic sites are occupied. The computational and experimental results are in good agreement, with both suggesting that autophosphorylation is fastest with Probe S, followed by Probe O, and Probe N (experimental values of 2.1, 1.5, and 0.15 s⁻¹, respectively).

Table 2.

Reaction Barriers of HK853 Autophosphorylation from DFT Study for Probes O, S, and N^a

probe	ΔG‡(TS1)	ΔG(IM2)
ATP	11.6	−9.4
Probe O	16.9	−6.4
Probe S	13.5	−7.5
Probe N	19.3	−9.3

^aReaction barriers ΔG^‡(TS1) and thermodynamics ΔG(IM2) are given in kcal·mol⁻¹. Computational details and optimized geometries are given in the Supporting Information.

During the phosphotransfer, the distance between the terminal γ-phosphorus of ATP and the τ-nitrogen of the histidine is 4.39 Å, as determined by our theoretical model. This value is smaller than the 4.9 Å threshold set by Mildvan to indicate when a reaction has associative character.⁴⁹ Therefore, the autophosphorylation is partially associative and within the range of a S_N2 reaction. Consequently, the reaction barriers are primarily determined by steric effects.⁵⁰ By comparing the structure of TS1 given in Supplementary Figure 16, we found that Probe S takes on an orientation that is different from that of Probe O or Probe N. Because the P–S (2.15 Å) bond is longer than the P–O (1.64 Å) or P–N (1.70 Å) bonds, Probe S is able to reorient itself toward the negatively charged Asp (Asp244), resulting in a lower transition state barrier and faster turnover rate. The major difference between Probe O and Probe N is the hydrogen on the nitrogen atom, which increases the hindrance between Probe N and the attacking nitrogen from the histidine residue, impeding the reaction and yielding the slowest turnover. Thus, the autophosphorylation rate correlates with the steric hindrance of the probe.

Lasker et al. demonstrated that thiophosphohistidine derivatives are less susceptible to hydrolysis than their pHis counterparts due to the lower electronegativity of sulfur compared to oxygen.⁸ To corroborate this result with the propargyl probes, we conducted hydrolysis studies on the phosphorylated HKs. Due to the poor autophosphorylation efficiency of HK853 with Probe N and low resulting signal, we focused on Probes O and S for the degradation studies. After an HK autophosphorylation period of 15 min with Probes O or S, ADP (1.1 mM) was added to quench the reaction. The acidity of the medium was adjusted to the desired pH value, and the reactions were incubated for 3 h. The samples were neutralized before the CuAAC reaction was performed. The remaining amounts of phosphorylated HKs were quantified by in-gel fluorescence detection and compared to a control sample (represent 100% phosphorylated HK; reaction medium pH 7 and no 3 h incubation period; Supplementary Figure 12). As expected, the Probe S-derived pHis species (orange) is more stable than the Probe O-derived pHis species (green) over the studied pH range, and both become less stable as the pH decreases (3–7 range; Figure 4 and Supplementary Figure 14). Interestingly, product stability improved at pH 1 for both Probes O and S. This nonlinear relationship between degradation and pH has been previously reported for other phosphorylated histidine-containing proteins when native ATP was used.^5,51 The amount of degradation observed with both Probe O and S at pH 1 was less (~20–30% unphosphorylated HK853) than at pH 7 (~40% unphosphorylated HK853; Figure 4, Supplementary Figure 12). This result indicates that strongly acidic conditions may enable the study of HKs with alkylated ATP-based probes.

Figure 4.

Hydrolysis studies of phosphorylated HK853 species at various pH values. HK853 (2 μM) was incubated with Probe O or Probe S for 15 min. After 3 h of degradation, the media were neutralized, and TAMRA-N₃ was conjugated by CuAAC to the propargyl handle (30 min). Formation of pHis was quantified by fluorescence gel-based analysis. HK853 (5 μM) was incubated for 30 s with [γ−33P]-ATP (60 μM). After 3 h of degradation, the media was neutralized. Formation of pHis was quantified by gel-based analysis and phosphorimaging. Samples that did not undergo acidification or the 3 h incubation period were “100% phosphorylated” controls, to which all samples were compared. Plot indicates the percent of phosphorylated HK remaining in comparison to the 100% control for each probe or ATP as appropriate. n ≥ 3, error bars show standard deviation. Statistical analysis to compare two substrates at a specific pH performed with an unpaired t test (**p ≤ 0.01, *p ≤ 0.05).

To investigate the effect of γ-alkylation on the stability of the pHis derivatives, we conducted similar studies using the radiolabeled version of ATP, [γ−33P]-ATP (radiolabeled ATPγS is no longer commercially available). As the pH became more acidic, the native pHis species was less stable (blue, Figure 4 and Supplementary Figure 13), consistent with previous reports.⁸ Comparison of the amount of dephosphorylation observed with the Probe O-derived pHis (green) to that of ATP-derived pHis (blue) indicates that the levels of phosphorylated HK853 are slightly higher with ATP in the pH range of 3–7 (by ~10–15%). Thus, the alkynyl substituent does not improve the stability of the probe-derived pHis, except at low pH (pH 1), where the alkylated pHis is dramatically more resistant to hydrolysis.

To understand the origin of the increased stability of alkylated pHis species at low pH values, we calculated the reaction thermodynamics of pHis and pHis derivative hydrolysis. At RT, a change of reaction thermodynamics by 1 kcal·mol⁻¹ leads to a factor of ~5 change in equilibrium constant via ΔG = −RTln K_eq, where ΔG is the free energy difference between reactants and products (reaction thermodynamics), R is the gas constant, T is the temperature, and K_eq is the equilibrium constant. As the amount of unphosphorylated HK853 goes from ~20% at pH 7 to ~70% at pH 1 for ATP, this suggests that differences in the reaction energies are on the order of ~1 kcal·mol⁻¹. Therefore, we performed both double-hydride DFT⁵² and DLPNO–CCSD(T)⁵³ calculations, which should provide high enough accuracy to resolve such small energy differences. Due to the computational cost of these methods, we built a simple model in which the histidine is replaced with an imidazole. To determine why the equilibrium is shifting as a function of pH, we performed calculations to (1) establish the protonation state of pHis and derivatives (hydrolysis reactants) as a function of pH, (2) establish the protonation states of inorganic phosphate, propargyl phosphate, and propargyl thiophosphate as a function of pH (hydrolysis products), (3) establish the reaction barriers as a function of pH, and (4) establish the reaction thermodynamics as a function of pH.

First, we calculated the protonation states for pHis and derivatives at various pH values. Supplementary Tables 3 and 4 show the calculated absolute pK_a values using the thermodynamic cycle and direct approaches, respectively. To obtain more accurate pK_a values, we also calculated the relative and scaled pK_a values. Our results, summarized in Supplementary Figure 18, suggest that the pHis species resulting from autophosphorylation with Probe O or Probe S have no protons bound to the oxygens of phosphorus in the range of pH 1 to 7, whereas with ATP, the pHis species is protonated at low pH, below pH 4.0–6.3, depending on the level of theory. We next calculated the protonation states for the hydrolysis products (Supplementary Figure 19). At pH 3 to 7, inorganic phosphate has two protons, while at pH 1, it has three protons. Hydrolysis of the pHis species derived from Probe O or S yields propargyl phosphate or propargyl thiophosphate, respectively. These species possess no protons on the phosphate oxygen atoms at pH 7, one proton in the pH range of 2–6, and two protons at pH 1.

To calculate the hydrolysis reaction barrier as a function of pH, we analyzed this reaction considering either one or two water molecules (Supplementary Figures 20 and 21, respectively). These data indicate that as the pH is lowered from 7 to 3–5, the reaction barrier for degradation of the native pHis species decreases, correlating to a higher hydrolysis rate. From pH 3–5 to 1, the free energy of pHis hydrolysis again decreases, agreeing with the experimental trend of diminished pHis stability as the pH decreases. However, for the Probe O- and Probe S-derived pHis species, the reaction barriers are identical over the investigated pH range, as the protonation states for the phosphorylated species do not change. Because the reaction barrier is not changing with pH, differences in the reaction thermodynamics must be governing pHis stability. Therefore, pHis stability must be a function of the reaction thermodynamics of hydrolysis, which varies depending on the pH, as the products of hydrolysis have different protonation states.

Finally, we calculated the reaction thermodynamics as a function of pH. Data for the Probe O-derived pHis species are presented in Figure 5. As stated above, the hydrolysis reactant is identical in the range of pH 1–7. However, the product (propargyl phosphate) goes from being fully deprotonated at pH 7 to monoprotonated at pH 3–5 to doubly protonated at pH 1. This protonation state change leads to differences in the reaction thermodynamics, where an initial lowering of the pH from 7 to 3–5 decreases the reaction thermodynamics; however, a further decrease to pH 1 leads to an increase in the reaction thermodynamics.

Figure 5.

Calculated reaction thermodynamics in kcal·mol⁻¹ (on right) for the hydrolysis of propargyl phosphoimidazole (model of phosphorylated HK853) at pH 1 to 7. Reaction thermodynamics were calculated by taking the average and standard error of double-hydride DFT calculations, and DLPNO–CCSD(T) corresponds to data sets TC1 and TC3 in Supplementary Figure 23.

Hydrolysis reaction thermodynamics as a function of pH for ATP, Probe O, and Probe S at different levels of theory are provided in Supplementary Figures 22–24, and we summarize these findings in Table 3. These data are the average and standard error of the double-hydride DFT calculations and high-level DLPNO–CCSD(T), where the geometries were optimized using three different DFT functionals (correspond to data set TC1 to TC3 in Supplementary Figures 22–24, six total calculations for each pH). At pH 7, the propargyl pHis species is substantially more stable, in contrast to the experimental results. This is due to the fact that the reaction produces two solvated protons (Supplementary Figures 23 and 24); however, because the calculation of the solvation free energy of a proton is extremely difficult, we have used the experimental solvation free energy of a proton, which could lead to an overestimation of the hydrolysis free energy. In contrast, hydrolysis of the native pHis does not involve a solvated proton. Therefore, we calculated a significantly lower hydrolysis free energy for this pHis species. Despite this, the trends in the data are meaningful, particularly for the lower pH steps, where the number of solvated protons is balanced between the phosphorylated species resulting from autophosphorylation with ATP, Probe O, and Probe S.

Table 3.

Average and Standard Error of Calculated Reaction Thermodynamics for the Hydrolysis of Phosphoimidazole, Propargyl Phosphoimidazole and Propargyl Thiophosphoimidazole^a

probe	ΔG (pH ~ 1)	ΔG (pH ~ 3–5)	ΔG (pH ~ 7)
ATP	−5.8 ± 0.3	−5.3 ± 0.1	−4.5 ± 0.4
Probe O	−4.8 ± 0.2	−6.5 ± 0.1	14.9 ± 0.6
Probe S	−2.5 ± 0.6	−6.2 ± 0.3	13.8 ± 0.5

^aResults given in kcal·mol⁻¹. Detailed computational results are provided in Supplementary Figures 22–24.

For the pHis species formed with ATP, the reaction thermodynamics only decrease with increasing pH. However, for the Probe S-derived pHis species, we see the same trend that we saw with Probe O, which is an initial reduction in free energy of hydrolysis when the pH decreases from 7 to 3–5, and then an increase in the free energy when the pH goes from 3–5 to 1. These results match the experimental trend shown in Figure 4 in that the highest stability is observed at pH 1 and pH 7 for the Probe O- and Probe S- derived pHis species. The most straightforward interpretation of these results is that the pH is modulating the protonation state of the hydrolysis product, which in turn modulates the equilibrium of the hydrolysis reaction.

The qualitative explanation for the surprising increase in stability at pH 1 is that a change in reaction thermodynamics for hydrolysis is due to differences in acidity. Supplementary Figure 25 shows the calculated reaction thermodynamics for protonation of the hydrolysis products. The reaction thermodynamics for protonation of phosphate decreases by −0.5 ± 0.3 kcal·mol⁻¹, while the reaction thermodynamics for protonation increases by 1.8 ± 0.3 for the propargyl pHis and 3.6 ± 0.3 kcal·mol⁻¹ for the propargyl thio-pHis. Therefore, while a decreased pH is forcing the products into higher protonation states (forcing the equilibrium to the right in Supplementary Figure 25), this leads to a lowering of the reaction thermodynamics in hydrolysis (decrease in stability) for the pHis species formed with ATP and an increase in the reaction thermodynamics (increase in stability) for pHis species formed from Probe S and Probe O.

Differences in O–H bond formation energies drive the hydrolysis reaction. These are the greatest for phosphate, followed by propargyl phosphate, and propargyl thiophosphate. This bond energy is directly related to the pK_a values as in Supplementary Figure 25. Propargyl thiophosphate is more acidic than propargyl phosphate, and phosphate is the least acidic species, which is due to the presence of substituents with differing acidities on the phosphorus atom (hydroxyl for phosphate, propargyl for propargyl phosphate and both a sulfur atom and a propargyl group for propargyl thiophosphate). To evaluate the influence of the propargyl and hydroxyl moieties on the acidity of their corresponding pHis species, we utilized the experimental pK_a values of their respective conjugate acid, propargyl alcohol (13.55)⁵⁴ and water (14.0). Propargyl alcohol is more acidic than water, making it unsurprising that propargyl phosphate is also more acidic than the phosphate counterpart. The thiopropargyl group stabilizes the negative charge on propargyl thiophosphate more efficiently than the propargyl group due to the larger size of the sulfur atom compared to the oxygen atom. Therefore, our results suggest that the stability of the phosphorylated HK species is related to the stability of the hydrolysis products, particularly the reaction thermodynamics of the product in solution, which can be controlled through pH.

CONCLUSION

Toward our goals of investigating the bacterial kinome, understanding the individual roles of the bacterial HKs, and studying HK phosphorylation, we experimentally and computationally explored the molecular characteristics required to design efficient ATP-based probes for these proteins. Overall, γO-phosphate-containing molecules yield the most phosphorylated protein; even so, thiophosphate-based probes are turned over more efficiently, likely due to improved reaction sterics. Amino-based probes label the HKs only poorly. The stability of the phosphorylated species to hydrolysis is also an important parameter to consider when designing HK probes and ultimately depends upon the stability of the hydrolysis products at certain pH values due to the solvation free energy of the phospho-derivative products. Our studies demonstrate that thiophosphorylated histidine species are the least susceptible to hydrolysis and that alkylation confers added stability to the phosphorylated histidine at very acidic pH values. Together, these mechanistic studies provide the molecular foundations for the design of future HK probes and will assist in the tuning of ATP-based molecules to improve enzymatic histidine phosphorylation for downstream applications such as phosphoproteomic analysis.

MATERIALS AND METHODS

Synthesis and characterization of Probe S and experimental and theoretical protocols are discussed in detail in the Supporting Information.

General Procedure for Labeling of HK853 with Probe O, S, or N.

Pure protein HK853 was incubated with Probe O, S, or N in reaction buffer (25 μL). Samples were quenched with PBS containing 10% SDS and 10% triton X-100 (2.5 μL) at 15 min. The solutions were “clicked” with TAMRA-N₃ (1 μL of 5 mM stock solution), TCEP (0.8 μL of 50 mM stock solution), TBTA (0.4 μL of 10 mM stock solution), and CuSO₄ (0.8 μL of 50 mM stock solution) for 30 min. Samples were mixed with 4 × SDS-PAGE loading buffer and then, run on a 12% SDS-PAGE gel. The gels were scanned for gel-based analysis using ImageJ software (NIH).