Enzyme I initiates a series of phosphotransfer reactions during sugar uptake in the bacterial phosphotransferase system. Here, we have isolated a stable recombinant C-terminal domain of Enzyme I (EIC) of Escherichia coli and characterized its interaction with the N-terminal domain of Enzyme I (EIN) and also with various ligands. EIC can phosphorylate EIN, but their binding is transient regardless of the presence of phosphoenolpyruvate (PEP). Circular dichroism and NMR indicate that ligand binding to EIC induces changes near aromatic groups but not in the secondary structure of EIC. Binding of PEP to EIC is an endothermic reaction with the equilibrium dissociation constant (KD) of 0.28 mM, whereas binding of the inhibitor oxalate is an exothermic reaction with KD of 0.66 mM from calorimetry. The binding thermodynamics of EIC and PEP compared to that of Enzyme I (EI) and PEP reveals that domain–domain motion in EI can contribute as large as ∼−3.2 kcal/mol toward PEP binding.
Enzyme I (EI) is a ∼130-kDa dimeric protein (575 aa) that initiates a cascade of phosphotransfer reactions in the bacterial phosphotransferase system.1−3 EI catalyzes two reversible reactions: an Mg2+-dependent autophosphorylation reaction using phosphoenolpyruvate (PEP) as a substrate, and a phosphotransfer reaction to histidine-containing phosphocarrier protein, HPr. EI comprises a stable N-terminal domain (EIN) which has a catalytic αβ domain and an HPr-binding α domain, and a PEP-binding C-terminal domain (EIC).4 First, EIC binds to PEP and phosphorylates the active site His189 in the αβ domain of EIN, and then the phosphoryl group is transferred to HPr bound on the α domain of EIN. EIN is a highly stable protein and the structure has early been determined by X-ray crystallography and NMR spectroscopy.5, 6 On the other hand, EIC is prone to proteolysis, which made it difficult to isolate recombinant EIC protein for biochemical and biophysical characterization.7 EIC is responsible for the dimerization of EI, whereas EIN exists exclusively as a monomer. Crystal structures of EI have suggested the presence of multiple conformational states during the enzyme reactions, but it is not clear how the domain motions evolve with the substrate binding and phosphotransfer reactions.8–10 Characterizing the interaction of isolated EIC with its ligands and also with EIN would provide useful information to understand the mechanism underlying the domain motions.
There have been efforts for isolation and biophysical characterization of recombinant EIC. Erni group determined the first crystal structure of EIC of a thermophilic bacterium, Thermo tengcongensis,11 and then Roseman group successfully isolated EIC of Escherichia coli using an intein system and characterized its interaction with the ligands by spectroscopic and kinetic analysis.12 Isolation of EIC of E. coli, however, still suffered from proteolysis during expression, and protein ligation was the only method of choice.12 Here we have expressed multiple constructs of EIC of E. coli varying the linker region between EIN and EIC, and obtained a construct that expresses stable EIC in an intact form. We have characterized the interaction of EIC with its substrates and inhibitor, and also with EIN using calorimetry, circular dichroism (CD), and NMR spectroscopy.
Results and Discussion
We designed three EIC constructs varying the linker region between EIN and EIC into a pET11a vector and examined the expression of the recombinant proteins. The linker region forms a long helix from residue Thr232 to residue Lys257, followed by a short β bridge (Ala261-Val269) that connects the linker helix to the (αβ)8 barrel fold of EIC. The EIC231–575 construct contains the linker helix followed by EIC, whereas the EIC260–575 construct lacks the whole linker helix. Finally, the EIC270–575 construct contains only the (αβ)8 barrel fold of EIC with both the linker helix and the β bridge removed. Figure 1(A) shows the three-dimensional structure of EI in a ribbon diagram, where EIN is colored in gray, EIC in blue, and the linker helix and the β bridge in red. Only one subunit of the EI dimer is shown for visual clarity, and the N-terminal residues of individual EIC constructs in this study are shown as a space-filling model in orange. When the constructs were transformed into an E. coli strain BL21star(DE3) and overexpressed in Luria Broth media, EIC231–575 and EIC270–575 appeared as single bands in the sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) [Fig. 1(B)]. EIC260–575 exhibited a lower expression level and an additional band with a slightly lower molecular weight could be observed implying a proteolytic cleavage. We then checked whether the expressed proteins formed a soluble fold or an inclusion body. EIC270–575 mostly formed an inclusion body likely due to misfolding, whereas EIC231–575 largely remained as a soluble form [Fig. 1(B)]. The absence of proteolysis of EIC270–575 is possibly due to its formation of an inclusion body before proteolysis took place during the overexpression. Our result is consistent with previous studies,12, 13 and indicates that keeping the linker helix is critical to confer proteolytic stability on EIC, while the short β bridge is important for proper folding of EIC. EIC231–575 was stable during the purification using chromatography at a room temperature, and used for all experimental measurements hereafter.
EIN can switch between two conformational states which differ in their domain–domain orientation of the α and αβ domains. The first conformational state is optimal for the phosphotransfer reaction between EIN and HPr, and the second conformational state is relevant for the autophosphorylation reaction between EIN and EIC.8–10 Isolated EIN adopts predominantly the first conformational state in solution, but the interaction with EIC requires a switch to the second conformational state.5 Exchange between the two conformational states entails a large domain–domain motion between α and αβ domains within EIN, which would cause chemical shift perturbation at the interface between the domains. In light of this, we monitored spectral changes in the 1H−15N correlation spectra of EIN titrating with EIC to investigate their interactions.
We titrated 0.2 mM of 2H,15N-EIN with up to 0.8 mM of EIC231–575, but could not observe significant chemical shift changes or line broadening. The lack of chemical shift perturbation and line-broadening implies that the intrinsic binding between separate EIN and EIC would be extremely weak [black and green cross-peaks in Fig. 2(A)]. Weak affinity between covalently-linked domains has been observed in other phosphotransferase system (PTS) proteins when the connecting linker is removed.14 Domains of weak affinity easily achieve fast dissociation kinetics, but require high local concentrations for fast association. A covalent linker connecting the domains makes the association a unimolecular reaction which assures fast association kinetics defined by the size of the linker.15 This can be a useful strategy for PTS in which rapid association and dissociation coupled with phosphotransfer reactions are important for prompt sugar uptake and phosphorylation in bacteria.
We examined the functional complementarity between EIC and EIN for the phosphotransfer reaction. It has been reported that EIC can phosphorylate EIN from a biochemical assay using 32P-PEP as a substrate.12, 16 When we added PEP to the mixture of EIN and EIC231–575 after titration, we could immediately observe the chemical shift changes due to the phosphorylation reaction of EIN [red cross-peaks in Fig. 2(A)]. Characteristic chemical shift changes could be observed for Gly10, Gly178, Gly185, and Gly204 among others, and these cross-peaks returned to their original position when EIN was dephosphorylated. The residues with chemical shift perturbation are drawn as space-filling models on the three-dimensional structure of EIN to visualize their spatial distribution [Fig. 2(B)]. The chemical shift changes were consistent with the phosphorylated EIN [blue cross-peaks in Fig. 2(A)], which was obtained using EIN mixed with HPr, EI, and PEP, where the phosphoryl group was transferred from EI to HPr and finally to EIN.17, 18 Our result shows that separate EIN and EIC formed a functional complex for the phosphotransfer reaction in vitro and the two domains would readily dissociate after the phosphotransfer reaction. There was no line-broadening in the heteronuclear single quantum coherence (HSQC) spectra of 15N-EIN with EIC, regardless of the presence of PEP, implying that the dynamics of EIN little changed by EIC and ligands in the experimental condition. We exclude the possibility that contaminating EI and HPr in the preparation of EIC or EIN have caused the phosphorylation reaction in our study. First, EIN was not phosphorylated when we added PEP to EIN without EIC, so the preparation of EIN was free of contaminating EI and HPr. Second, we monitored the phosphorylation of the A domain of the mannitol transporter, Enzyme IIMtl (IIAMtl), which could be phosphorylated by EI and HPr. If contaminating EI and HPr in the preparation of EIC caused the phosphorylation of EIN, they would also phosphorylate IIAMtl in the same manner. Phosphorylation of 15N-IIAMtl can be immediately identified by the chemical shift changes in the 1H−15N correlation spectrum. When we incubated 15N-IIAMtl with EIC using the same buffer composition as was used to phosphorylate EIN, the spectra clearly indicated that IIAMtl remained in the unphosphorylated state (Supporting Information Fig. 1S). Taken together, the phosphorylation of EIN resulted from the direct interaction between EIN and EIC, rather than from contaminating EI and HPr. We note that EIN1–249 and EIC231–575 share part of the linker helix between 231 and 249. As the two domains could form a functional complex for the phosphotransfer reaction, precise orientation of the linker helix may not be critical in the reaction complex between EIN and EIC.
We also examined whether the transient complex between EIN and EIC could be captured when PEP was bound to EIC. As EIN released from EIC after the phosphorylation reaction, we used EIN(H189Q) mutant where the active site His189 was replaced by a isosteric glutamine residue that could not be phosphorylated. The NMR spectra of 0.2 mM15N-EIN(H189Q) mixed with 0.8 mM EIC231–575, 4 mM MgCl2, and 10 mM PEP did not show noticeable chemical shift changes (data not shown). Using the equilibrium dissociation constant (KD ∼0.28 mM) of EIC and PEP from isothermal titration calorimetry (ITC) study in Table I (see below), more than 95% of EIC is bound to PEP in this concentration. We infer that the conformational state of EIN that binds to EIC is thermodynamically unstable, so that the phosphotransfer reaction complex between EIN and EIC represents a metastable transient state.
Table I. Equilibrium Dissociation Constants (KD), Binding Enthalpies (ΔH), and Entropies (ΔS) Obtained from Least-Squares Fitting of One Set of Sites Binding Model to the Experimental ITC Data for the Interaction Between EIC and Its Ligands at pH 7.4 Under Different Buffer Conditions
Buffer (pH 7.4)
0.28 ± 0.14
1.76 ± 0.98
0.24 ± 0.08
0.99 ± 0.15
0.17 ± 0.05
0.54 ± 0.08
0.66 ± 0.24
-0.63 ± 0.26
0.42 ± 0.05
-0.94 ± 0.12
0.44 ± 0.14
-0.62 ± 0.20
Far UV CD spectroscopy (200–250 nm) provides information on the content of secondary structures, and near UV CD spectroscopy (250–350 nm) provides information on the tertiary structures from asymmetric environments of aromatic chromophores. The far UV CD spectrum of EIC231–575 showed the presence of significant secondary structures, where two minima at 208 and 222 nm were characteristic signatures of an α-helix [Fig. 3(A)]. When we added the substrates, PEP and pyruvate, or the inhibitor oxalate to EIC231–575, we could not observe significant changes in the far UV CD spectra of EIC231–575, indicating that ligand binding did not largely affect the secondary structures of EIC [Fig. 3(A)]. On the other hand, the near UV CD spectra of EIC231–575 exhibited noticeable differences between the free and the ligand-bound states between 260 and 290 nm [Fig. 3(B)]. As the near UV CD spectra imply that ligand binding affected aromatic residues, we examined the 1D 1H-NMR spectra of EIC231–575 in the aromatic region upon ligand binding. To remove amide proton signals EIC231–575 and its complex with ligands were prepared in the same buffer using D2O. The 1H-NMR spectra of EIC231–575 revealed manifest chemical shift changes in the aromatic region, where the changes were more pronounced in the binding to PEP or oxalate than the binding to pyruvate [Fig. 3(C)]. 1H-NMR spectra of methyl region of EIC231–575 showed modest chemical shift changes upon ligand binding [Fig. 3(D)]. EIC231–575 has 2 tryptophan, 6 tyrosine, 13 thirteen phenylalanine residues, and none of the aromatic residues are within 6 Å from the bound oxalate in the crystal structure (PDB code: 2HWG). However, the crystal structure of EIC from Thermoanaerobacter tengcongensis recently showed a conformational change of Phe354 upon binding to PEP.19 As Phe354 is conserved in EIC of E. coli, it likely contributed to the change in the CD and NMR spectra in this study. In addition, Tyr459 and Tyr474 are the nearest neighboring aromatic residues to the PEP binding site, which may have contributed to the spectroscopic changes observed in this study.
We measured the equilibrium binding between EIC231–575 and its ligands at 25°C using ITC. It is known that the dimeric form of EI is the active form and the monomer–dimer equilibrium can vary with experimental conditions.12, 20, 21 Previously, the KD for the dimerization of EIC256–575 in the presence of 5 mM MgCl2 and 100 mM NaCl at pH 7.5 and 25°C was reported as ∼6 nM from analytical ultracentrifugation.12 When we loaded a serial dilution of EIC231–575 from 200 μM to 2 μM onto the Superdex 200 size exclusion column, the elution volume was constant, indicating that EIC231–575 was a dimer in the given concentration range. If we assume that >90% of 2 μM EIC231–575 is dimeric based on the size exclusion chromatography, we can calculate the upper limit of KD as ∼10 nM. We also carried out the size exclusion chromatography of EIC with ligands, and the elution volume did not change, hence the ligand-bound EIC was also a dimer. In the current ITC measurement, we used 0.25–0.5 mM of EIC in the cell and titrated with 3–10 mM of ligands. At this concentration range, more than 80% of EIC is predicted to be dimeric during the titration. The measured KD between EIC231–575 and PEP was 0.28 mM, which was close to the KD values (0.22–0.28 mM) obtained from previous spectroscopic studies [Fig. 4(A)].12 The KD between EIC231–575 and oxalate was 0.66 mM, so the binding was 2.4 times weaker than the binding to PEP [Fig. 4(B)]. The stoichiometry of the binding of EIC231–575 to the ligands indicated that each monomeric subunit of the EIC dimer bound to one molecule of the ligand with a comparable affinity.
We note that the apparent binding enthalpy (ΔHbinding) measured by ITC is the sum of the intrinsic reaction enthalpy (ΔHreact; independent of the buffer) and the ionization enthalpy (ΔHionization) from the buffer, as was described by Freire and coworker.22 ΔHreact can be obtained by performing ITC in buffers with different ionization enthalpies. We carried out titration experiments at pH 7.4 using sodium phosphate and 4-(2-hydroxyethyl(-1-piperazineethanesulfonic acid (HEPES) buffers as well as Tris-HCl buffer, which exhibited a wide range of the ionization enthalpies at 25°C (0.98 kcal/mol, 5.03 kcal/mol, and 11.16 kcal/mol for phosphate, HEPES, and Tris-HCl buffer, respectively).23 When ΔHbinding is plotted as a function of ΔHionization, the slope of the linear regression gives the number of protons released by the buffer (taken by the protein) upon binding and the y-intercept is ΔHreact.22 From the linear regression, the ΔHreact of EIC231–575 for binding to PEP was 0.41 kcal/mol, and the ΔHreact for binding to oxalate was −0.75 kcal/mol [Fig. 4(C,D)]. Also, the slope of the linear regression indicated that ligand binding of EIC231–575 involved little changes in the ionization states at the binding site. It is notable that the interaction between EIC231–575 and PEP was an endothermic reaction, while the interaction between EIC231–575 and the inhibitor oxalate was an exothermic reaction (Table I). Binding to PEP was driven by a favorable entropic contribution, so that hydrophobic interaction was important in the binding between EIC and PEP. On the other hand, binding to oxalate was driven by both enthalpic and entropic contributions, so that electrostatic interaction was important in the binding between EIC and oxalate. We also carried out the titration between EIC231–575 and pyruvate, but could not observe a reliable binding isotherm due to small measured heat and weak binding. The weak binding between EIC and pyruvate was not unexpected as the Michaelis constant (Km) for EI and pyruvate was reported to be ∼2 mM, which was 10 times larger than the Km for EI and PEP.24
Previous ITC study reported that the KD between the active-site mutant EI(H189A) and PEP was ∼1 μM, so that the binding between EI and PEP was about two orders of magnitude stronger than the binding between EIC231–575 and PEP.25 The binding between EI(H189A) and PEP was also an endothermic reaction, where the free energy of the binding (ΔG) was −8.1 ± 0.1 kcal/mol, the binding enthalpy (ΔH) was 3.9 kcal/mol, and the binding entropy (ΔS)was 40.8 cal/mol/K in the phosphate buffer.25 These thermodynamic parameters compare to ΔG of −4.9 ± 1.4 kcal/mol, ΔH of 0.54 kcal/mol, and ΔS of 19.1 cal/mol/K for EIC231–575 and PEP in this study (Table I). The larger binding affinity between EI and PEP compared to that of EIC and PEP suggests the contribution of domain–domain motions in EI toward PEP binding, which may be as large as −3.2 ± 1.4 kcal/mol of ΔΔG (ΔΔG = ΔGEI:PEP − ΔGEIC:PEP). The large negative ΔΔG mainly results from favorable contribution of binding entropy, suggesting that large hydrophobic interactions are involved in the domain–domain motions in EI upon binding to PEP. The domain motions that potentially contribute to the observed ΔΔG include the conformational switching within EIN that could expose buried interfaces between α and αβ domains, and also the interaction between the αβ domain of EIN and EIC.
It has been reported that oxalate can inhibit the autophosphorylation reaction of EI using PEP with the inhibition constant (KI) of 0.02 mM.26 The large difference between KI and KD of EIC and oxalate can be explained by the mechanism of inhibition. Oxalate inhibits the autophosphorylation reaction by forming a tight complex with phosphorylated EI, where the binding would be stabilized significantly by the association between EIN and EIC sandwiching the inhibitor oxalate.9 Hence, the KD of EIC and oxalate in this study accounts for the intrinsic binding of the inhibitor when the domain–domain association does not take place.
In conclusion, we show that separate EIN and EIC domains form a functional reaction complex which is very weak regardless of the presence of ligands. The substrate PEP binds to EIC stronger than the inhibitor oxalate, and the two ligands exhibit an opposite sign of binding enthalpy. Finally, comparison of PEP binding to EI and EIC suggests that apparent equilibrium binding between multi-domain proteins and ligands can be significantly affected by domain–domain motions.
EI, Enzyme I; EIC, C-terminal domain of Enzyme I; EIN, N-terminal domain of Enzyme I; HPr, histidine-containing phosphocarrier protein; HSQC, heteronuclear single quantum coherence; PEP, phosphoenolpyruvate; PTS, phosphotransferase system.
Materials and Methods
Cloning, expression, and purification of EIC and EIN
EIC231–575, EIC260–575, and EIC270–575 were cloned into a pET11a vector (Novagen) without tags, and the new constructs were verified by DNA sequencing. The plasmids were transformed into an E. coli strain BL21star(DE3), and then grown in Luria Bertini or minimal media with 13C6-glucose and 15NH4Cl as the sole carbon and nitrogen sources in D2O. The protein expression was induced with 1 mM isopropyl-β-D-thiogalactopyranoside at an A600 of ∼0.8, and harvested by centrifugation after 4 h of induction. The cell pellet was resuspended with 50 mL (per liter of culture) of 50 mM Tris, pH 8.0, 2 mM β-mercaptoethanol, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and a protease inhibitor cocktail tablet (Sigma-Aldrich, S8820 SIGMAFAST). The suspension was lysed by three passages through Emulsiflex (Avestin, Canada) after homogenizing and centrifuged at 24,000g for 20 min at 4°C. The supernatant fraction was filtered and loaded onto a DEAE anion exchange column (20 mL; GE Healthcare) with a 400-mL gradient of 1M NaCl. The fractions containing EIC were purified by size exclusion on a Superdex 200 column (GE Healthcare) equilibrated with 20 mM Tris-HCl, pH 7.4, 200 mM NaCl, and 2 mM β-mercaptoethanol, and then on a monoQ anion exchange column (8 mL; GE Healthcare) with a 160-mL gradient of 1M NaCl.
EIN (residue 1–249) was cloned into a pET11a vector (Novagen) without tags, and the active site His189 of EIN was mutated to Gln using the QuickChange kit (Stratagene). The plasmids of wild-type EIN and EIN(H189Q) were individually transformed into an E. coli strain BL21star(DE3), and then grown in either Luria Bertini or minimal media with 15NH4Cl as the sole nitrogen source in H2O or D2O. EIN proteins were expressed and purified as previously described.17
NMR spectroscopy of EIN with EIC
NMR spectra were recorded at 30°C on a Bruker Avance 600 MHz or 900 MHz spectrometer equipped with an x,y,z-shielded gradient triple resonance probe. To examine EIN upon binding to EIC, 0.8 mM unlabeled EIC231–575 was titrated into 0.2 mM2H,15N-labeled EIN with or without 10 mM PEP in 20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA, 2 mM β-mercaptoethanol, and 4 mM MgCl2. In addition, 0.2 mM13C,15N-labeled EIN(H189Q) and 0.8 mM unlabeled EIC231–575 were used with or without 10 mM PEP in 20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA, 2 mM β-mercaptoethanol, and 4 mM MgCl2. Spectra were processed using the NMRPipe program and analyzed using the NMRView program.27, 28 For 1D 1H-NMR, 0.3 mM unlabeled EIC231–575 were used with 5 mM of PEP, oxalate, and pyruvate in 20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA, 2 mM β-mercaptoethanol, and 4 mM MgCl2. The samples for 1D 1H-NMR were first prepared in H2O-based buffer, lyophilized, and then dissolved in the same volume of D2O.
Circular dichroism of EIC with PEP, oxalate, and pyruvate
CD spectroscopy was conducted at 25°C using a with Chirascan™-plus CD spectrometer. Wave scans were acquired by sampling data at 1 nm intervals between 200 and 250 nm for far UV CD measurement and between 250 and 350 nm for near UV CD measurement. Far UV CD spectroscopy was carried out using 10 μM of EIC231–575 and 4 mM of substrates in 50 mM sodium phosphate, pH 7.5, 2 mM β-mercaptoethanol, and 2 mM MgCl2 using a 1-mm quartz cuvette. Each far UV CD spectrum was obtained from an average of three scans and the results were presented as mean residue ellipticity (deg cm2/dmol) at each wavelength. Near UV CD spectroscopy was carried out using 40 μM of EIC231–575 and 10 mM of substrates in 20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 2 mM β-mercaptoethanol, and 4 mM MgCl2 using a 10-mm quartz cuvette. Each near UV CD spectrum was obtained from an average of three scans and the results were presented as molar ellipticity (deg cm2/dmol) at each wavelength.
Isothermal titration calorimetry of EIC with PEP and oxalate
ITC experiments were performed at 25°C using an ITC200 calorimeter (GE Healthcare). Twenty consecutive 2 μL aliquots of 3 mM PEP or 10 mM sodium oxalate were injected into the cell containing 0.25 or 0.5 mM EIC231–575 in 20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 2 mM β-mercaptoethanol, and 4 mM MgCl2. The ITC experiments were also carried out in 20 mM sodium phosphate, pH 7.4, and 20 mM HEPES, pH 7.4, replacing the 20 mM Tris-HCl, pH 7.4 with the same buffer composition otherwise, to correct the binding heat by buffer contribution. The duration of each injection was 4 sec, and injections were made at intervals of 180 sec. Dilution heats of the substrates into the protein solution were subtracted from measured heats of the binding. ITC titration data were analyzed using the Origin software provided with the instrument.
Authors thank the high field NMR facility at Korea Basic Science Institute for NMR experiments.