Temporal Resolution of Activity-Related Solvation Dynamics in the TIM Barrel Enzyme Murine Adenosine Deaminase

Abstract

Murine adenosine deaminase (mADA) is a prototypic system for studying the thermal activation of active site chemistry within the TIM barrel family of enzyme reactions. Previous temperature-dependent hydrogen deuterium exchange studies under various conditions have identified interconnected thermal networks for heat transfer from opposing protein-solvent interfaces to active site residues in mADA. One of these interfaces contains a solvent exposed helix-loop-helix moiety that presents the hydrophobic face of its long α-helix to the backside of bound substrate. Herein we pursue the time and temperature dependence of solvation dynamics at the surface of mADA, for comparison to established kinetic parameters that represent active site chemistry. We first created a modified protein devoid of native tryptophans with close to native kinetic behavior. Single site-specific tryptophan mutants were back inserted into each of the four positions where native tryptophans reside. Measurements of nanosecond fluorescence relaxation lifetimes and Stokes shift decays, that reflect time dependent environmental reoroganization around the photo-excited state of Trp*, display minimal temperature dependences. These regions serve as controls for the behavior of a new single tryptophan inserted into a solvent exposed region near the helix-loop-helix moiety located behind the bound substrate, Lys54Trp. This installed Trp displays a significantly elevated value for $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$; further, when Phe61 within the long helix positioned behind bound substrate is replaced by a series of aliphatic hydrophobic side chains, the trends in $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ mirror the earlier reported impact of the same series of function-altering hydrophobic side chains on the activation energy of catalysis, $\mathit{Ea}(k_{\mathrm{cat}})$.The reported experimental findings implicate a solvent initiated and rapid (>ns) protein restructuring that contributes to the enthalpic activation barrier to catalysis in mADA.

Graphical Abstract

Introduction

The classical, textbook description of the origins of enzyme catalysis has been largely derived from static models of protein structures. However, proteins in solution are in constant motion, a result of thermally activated fluctuations that span a wide range of distances and time scales^1–8. The different classes of protein dynamics include large scale conformational changes, domain shifts and local atomic vibrations that are associated with time scales ranging from femtoseconds to seconds^{3, 9, 10}. A major, ongoing challenge is the design of experimental approaches that are capable of linking rapid and transient protein motions to enzymatic activity. The success of such an endeavor depends on a variety of biophysical techniques that begin with high resolution static protein structures^11–13 leading to time-resolved methods such as NMR spectroscopy^14–17, optical spectroscopies^18–23, and molecular dynamics simulations^24–27. Recently, temperature-dependent hydrogen-deuterium exchange mass spectrometry (TDHDX-MS) emerged as a broadly applicable experimental probe that provides a three-dimensional map of protein thermal flexibility that can be correlated to enzyme function^28–32. The application of TDHDX to different enzyme classes has shown the importance of reaction and site-specific thermal networks that connect a respective enzyme active site to a discrete protein-water interface^{28, 30–32}. In this context, it has become pivotal to probe the time dependence of motions at solvent exposed side chains that have been implicated in functionally relevant structural dynamics.

The TIM barrel enzyme murine adenosine deaminase (mADA), catalyzing the reaction shown, Figure 1A, has become an important paradigmatic system for understanding the role of protein scaffold motions in metal-based nucleophilic catalysis ^33–35. Phe61 was originally targeted as one of four hydrophobic side chains that line a shared face of a long α−helix to generate a hydrophobic wall positioned directly behind the purine ring of substrate (Figure 1B)³⁰. Mutating the bulky phenylalanine to smaller hydrophobic side chains (F61X) within apo-protein was found to perturb the native protein equilibrium state, leading to relatively small impacts on $k_{\mathit{cat}}$ and correlated alterations in the Ea (activation energy) for $k_{\text{cat}}$ and the rate constant for HDX-MS, $k_{\mathit{HDX}}$³⁰. A similar spatial resolution of protein flexibility has seen for mADA in the presence of a tightly bound transition state analog^{30, 31}. As summarized in Figure 1C, two opposing and embedded thermal networks are proposed as networks for solvent activation/slaving³⁶ in the thermal initiation of mADA activity.

Figure 1.

Catalytic mechanism, structure, and thermal networks of mADA. A. Proposed catalytic mechanism of mADA in which the active site base His238 converts zinc-bound water to hydroxide; rate limiting attack of hydroxide ion at C-6 of substrate, concomitant with proton transfer from Glu217 to N-1 of the purine ring, leads to the tetrahedral adduct (see details in references^{33, 34, 37}). B. The overall structure of mADA in complex with 1-deaza-adenosine (PDB:1ADD) is shown where helices are colored red, sheets are colored yellow, and turns are green. To the right, key features are highlighted. These include the hydrophobic wall behind bound substrate comprised of Leu58, Phe61, Leu62 and Phe65 (red) and active site residues. The latter include ligands (His15, His17, His214) to the active site zinc ion (magenta sphere), Glu217, and His238 (dark grey sticks) that are assigned as active site acid and base, respectively. C. The thermal network comprised of regions from peptides 46–62, 86–97, 260–267, 230–248, and 268–290 detected for mADA by TDHDX in the presence of a transition state analog pentostatin (2′-deoxycoformycin), is shown in red³¹. The heat map color index to the right indicates the magnitude and direction of changes in ΔEa_HDX ³¹. Note that this defined thermal network for the complex of enzyme with a tight binding inhibitor is highly similar to the network derived from a comparison of apo-WT to apo-Phe61A (using the control F61I to correct for mutation that changes $k_{\text{cat}}$ but leaves Ea unperturbed, Table S1 Figure S1)³⁰.

The present study focuses on the temporal and temperature analysis of the environmental reorganization energy of Stokes shifts emanating from attached fluorescence probes in mADA, for direct comparison to the time and temperature properties of active site chemistry. Given that the energy required for the transformation of reactant to product must reside in the solvent bath, we sought to position fluorophores at or near the protein-water interface. In an earlier study of a different enzyme system, this was achieved through the covalent attachment of an extrinsic fluorophore that was non-perturbing with regard to protein structure^{38, 39}. However, similar attempts with mADA were unsuccessful, and we turned to the site-specific insertion of tryptophan as an intrinsic optical probe. After screening multiple positions for the best placement of a single Trp, we pursued the comparative behavior for four control Trp positions to Lys54Trp (K54W); the latter is located at a solvent interface close to the hydrophobic wall that resides behind bound substrate. Detailed measurements of temperature-dependent Stokes shifts for K54W as a function of the previously characterized F61X series indicate a strong correlation between activation energies for millisecond enzyme activity and the observed nanosecond, light-induced environmental reorganization at Trp54. Analogous to recently completed studies of the role of a thermal network in the H-tunneling enzyme soybean lipoxygenase³⁹, a rapid (> nanosecond) protein restructuring in mADA is ascribed a role in the distal activation of zinc bound water/hydroxide addition to a purine base, Figure 1A.

Results

Steady-state kinetic data serve as a screen for suitable single tryptophan mADA variants as optical probes of protein dynamics.

Dynamic fluorescence Stokes shifts (DFSS) provide a time-resolved spectroscopic method to monitor rapid changes in environmental solvation following the photoexcitation of a local probe^40–42. DFSS measurement involves exciting the target molecule with a short laser pulse and then monitoring the time-dependent changes in the fluorescence emission wavelength, λ_max. The shift in emission λ_max results from a change in the local environment of the molecule, which may arise from induced changes in solvent polarity or refractive index or an overall conformational change of the molecule. Combined with fluorescence lifetime measurements, time, temperature, and mutation-dependent Stokes shifts can offer significant insight into the relationship of remote environmental changes to chemical transformations occurring within enzyme active sites.

An original design of this project was to utilize a well-established fluorophore probe Badan (6-bromoacetyl-2-dimethylaminonaphthalene) that proved extremely effective in studies of DFFS in the “extreme tunneling” enzyme soybean lipoxygenase^{38, 39}. However, preliminary results with mADA indicated significant quenching of fluorescence for covalently attached Badan by nearby tryptophan and tyrosine residues, hindering its further use. We thus refocused studies on the intrinsic optical fluorophore tryptophan. Native mADA contains four Trp residues (W117, W161, W264, and W272) (Figure 2, left), which are either solvent exposed or in regions that are separate from the identified thermal networks. When all four tryptophan residues were replaced with phenylalanine to generate an all-tryptophan knockout mutant (WT’), neither catalytic activity nor activation energies for catalysis were significantly altered (Table 1). This renders the tryptophan knockout mADA a perfect parent variant for the installation of single tryptophan probes at targeted positions.

Figure 2.

Four native tryptophan residues in the structure of WT mADA (PDB: 1ADD) (left) and the positions of insertion of single tryptophans into the thermal network regions of protein (right). The four native tryptophan residues (shown in cyan stick) are Trp117, Trp161, Trp264, and Trp272. Positions screened for insertion of suitable single tryptophan variants (shown in blue stick) included Lys54, Tyr67, Tyr68, Ile243, Tyr249, and Phe283. Structures in this figure share the same orientation as in Figure 1.

Table 1.

Kinetic data for WT and F61A within a series of single tryptophan mutants.

WT			F61X
Protein variant	$k_{\text{cat}}$ at 30 °C, s−1	Ea, kcal/mol	Protein variant	$k_{\text{cat}}$ at 30 °C, s−1	Ea, kcal/mol
WT	199(8)a	11.0 (1.6)a	F61A	12(1)a	16.3(1.0)a
WT’	178(13)	10.4 (2.6)	F61A’	10(2)	16.8(2.4)
WT’-W117in	190(20)	10.7(1.7)	F61A’-W117in	9(2)	16.0(1.7)
WT’-W161in	188(33)	10.1 (1.6)	F61A’-W264in	9(1)	15.8(2.1)
WT’-W264in	179(28)	10.9 (1.2)	F61A’-W272in	9(1)	17.0(1.2)
WT’-W272in	192(36)	9.9 (2.1)	F61A’-K54Win	1.5(0.2)	15.6(3.2)
WT’-K54Win	25(5)	10.0 (1.5)	F61V’-K54Win	1.7(0.2)	13.0(1.4)
WT’-Y67Win	9(1)	N.D.	F61L’-K54Win	1.4(0.1)	12.5(1.3)
WT’-Y68Win	0.2(0.1)	N.D.	F61I’-K54Win	1.2(0.1)	11.2(1.6)
WT’-Y249Win	1.8(0.2)	N.D.	F61G’-K54Win	0.11(0.01)	18.9(3.1)
WT’-F283Win	0.01(0.01)	N.D.
WT’-I243Win	0.05(0.01)	N.D.

The designation WT’ indicates protein devoid of native tryptophans that has been subjected to either restoration with a single native Trp or a single site Trp insertion elsewhere. The designation F61X’ refers to a single site mutant-bearing parent protein that is derived from WT’ and has Trp insertion either at one of the four native tryptophan sites or as K54W. Note that in every case, the $\mathit{Ea}(k_{\mathit{cat}})$ values for F61X’-K54Trpin are within experimental error of the $\mathit{Ea}(k_{\mathit{cat}})$ values for the corresponding WT series (Table S1). Activation energies are calculated based on kinetic data collected between 10 and 40 °C. N.D. indicates not determined. Values in brackets refer to errors. Mutants underlined were subjected to Stokes shift studies.

^aFrom data previously reported in reference³⁰.

Our initial investigation was focused on the solvation dynamics of the native individual tryptophan residues. To proceed, we back-inserted single tryptophans into their original positions, generating four single tryptophan mADA constructs (WT’-W117in, WT’-W161in, WT’- W264in, and WT’-W272in. Kinetics experiments were conducted to compare the $k_{\text{cat}}$ and Ea of these single Trp variants with parameters for the WT mADA. As shown in Table 1, in all cases both $k_{\text{cat}}$ and Ea are essentially unchanged. Thus, these Trp residues that are located largely outside of the intrinsic thermal networks were reserved as controls for successive studies. To search for the ideal position of a single Trp probe within the thermal network regions, we screened a variety of possible mutation sites. These included regions represented by peptides 46–62 and 63–74 located within the left-hand side of the inferred thermal network and regions represented by peptides 230–248 and 268–290 in the opposing right-hand side of the thermal network (Figure 2, right)³⁰. In total, six new single tryptophan mADA variants were generated (WT’-K54Win, WT’-Y67Win, WT’-Y68Win, WT’-I243Win, WT’-Y249Win, and WT’-F283Win). Kinetic data showed that five of these six mutants exhibit significant catalytic activity loss (from a 20 to 20,000-fold reduction in $k_{\text{cat}}$). The remaining variant, WT’- K54Win, was chosen for further study as assessed by an acceptable level of catalytic impairment (8-fold reduction in $k_{\text{cat}}$) and, importantly, an activation energy for $k_{\text{cat}}$ that is within experimental error of WT-mADA (Table 1).

As described in earlier studies, the identification of embedded thermal networks follows from analyses of site-specific mutants that contribute a relatively modest impact on $k_{\text{cat}}$ while introducing significant changes/trends to the enthalpy of activation for $k_{\text{cat}}$³⁰. This is generally achieved through the conversion of a targeted hydrophobic side chain to a series of hydrophobic side chains of reduced size and altered side chain structure. As introduced above, Phe61 had been targeted in mADA based on its positioning in close proximity to C-6 of a bound ground state analog 1-deaza-adenosine (van der Waals distance of 3.7 Å, Figure 1B). In this study, mutagenesis at Phe61 was pursued within the Trp-free enzyme, WT’, designated Phe61A’, as well as within the single site, native Trp variants of mADA (Table 1). Each F61A’-single Trp variant analyzed (F61A’-W117in, F61A’-W264in, and F61A’-W272in) displays an Ea that is within error of native (four Trp) F61A enzyme behavior. With our primary interest in K54Win as a probe of solvent induced dynamics, the second site mutation at Phe61 was extended from F61A’-K54Win to the full series of hydrophobic side chains (F61I’-K54Win, F61L’-K54Win, F61V’-K54Win, and F61G’-K54Win). Analogous to mutational effects on the catalytic parameters of WT mADA, these mutations at position 61 of K54Win produce a regular increase in $\mathit{Ea}(k_{\mathrm{cat}})$ that correlates with the Ea increase caused by reduction in the volume of inserted side chain³⁰. There appear to be no significant synergistic effects on rate at 30 °C between the two sites of mutation, with an approximately 17 to 20-fold decrease in $k_{\text{cat}}$ relative to WT’-K54Win arising from F61X mutation, analogous to the impact of the series of F61X on $k_{\text{cat}}$ for WT. The sole exception is F61G’-K54Win, which causes another order of magnitude reduction in the magnitude of $k_{\text{cat}}$.

Steady-state fluorescence.

As an intrinsic fluorophore, tryptophan can be an ideal local optical probe with a fluorescence emission λ_max that is sensitive to the local microenvironment. For example, the Trp λ_max in proteins ranges from ∼308 nm (azurin) to ∼355 nm (e.g., glucagon) and roughly correlates with the degree of solvent exposure of the chromophore⁴³. From the kinetic characteristics of single Trp variants in Table 1, five were chosen for detailed analyses of fluorescence lifetimes and time and temperature dependences of Stokes shifts. These include each of the control mutants (WT’-W117in, WT’-W161in, WT’-W264in, and WT’-W272in) and the newly created single Trp mutant WT’-K54Win. Steady-state fluorescence data were also collected for double mutants in which the native single tryptophan variants contain a second site of mutation F61A (F61A’-W117in, F61A’-W264in, and F61A’-W272in) and K54Win that contains a suite of F61X (F61I’-K54Win, F61L’-K54Win, F61V’-K54Win, F61G’-K54Win, as well as F61A’-K54Win). The excitation wavelength was set at 291 nm, and emission spectra were collected between 300 and 400 nm. As shown in Figure 3 for WT’-W117in and WT’-K54Win (see the full set of analyzed emission λ_max data for variants in Supporting Information Table S2), the peak emission varies from 335 to 350 nm. The emission λ_max is found to be independent of temperature and similar with or without the additional mutation at Phe61.

Figure 3.

Temperature dependence of steady-state fluorescence emission spectra (λ_exc = 291 nm) of (A) WT’-W117in and (B) WT’-K54Win in 50 mM potassium phosphate buffer (pH 7.3).

Time-resolved fluorescence lifetime decays for mADA single tryptophan mutants.

Time dependent lifetime decay traces were subsequently determined for each of the targeted single Trp variants before and after insertion of a second mutation at Phe61. Data were collected between 320 and 380 nm with an interval of 10 nm and at seven temperatures between 10 and 40 °C. The decay traces were fitted to multiexponential equations to obtain the corresponding lifetimes (τ) and amplitudes (α). The fluorescence decay data fitting can be grouped into two categories. The first group includes variants where the tryptophan probe is inserted into one of the original positions in the mADA amino acid sequence (W117, W161, W264, and W272). Fluorescence decay data for these single tryptophan variants are best fit to a biexponential equation (lifetimes of τ₁ and τ₂, and amplitudes of α₁ and α₂.). The second group contained the tryptophan probe substituted for Lys54 within a loop defined as a thermally activated region. In this case, decay data are better fit to a three-exponential function (lifetimes of τ₁, τ₂, and τ₃ and amplitudes of α₁, α_2, and α₃). The average relaxation lifetimes (τ_ave) at each temperature for all the protein samples are almost constant at ca. 3 ns. Fluorescence decay lifetime data collected as a function of temperature are compiled in the Supporting Information (Dataset). Representative Arrhenius plots of the relaxation lifetimes are shown below (Figure 4) for WT’-W117in and WT’-K54Win at an emission λ_max of 350 nm (full set of Arrhenius plots for all single Trp variants are compiled in Figure S2). A good linear relationship is observed between the fluorescence decay rates ln(1/τ) and 1/T. The weighted average fluorescence lifetimes (1/τ_ave) also follow the Arrhenius equation. As summarized in Table S3, the magnitudes of the activation energies (Ea) are centered at 1 kcal/mol for all of the native Trp insertions and at 2 kcal/mol for K54Win and its variants.

Figure 4.

Arrhenius plots of fluorescence decay rates for (A) WT’-W117in (1/τ₁, 1/τ₂, and 1/τ_ave) and (B) WT’-K54Win (1/τ₁, 1/τ₂, 1/τ₃ and 1/τ_ave).

Time-dependent Stokes shifts analysis.

Stokes shifts probe the ability of the local environment surrounding an inserted probe to rearrange during the lifetime of its excited state. Electronic excitation of tryptophan induces a distorted electric field and protein motions such as side chain rotations and translations and backbone fluctuations are expected to arise over multiple time scales in accommodation of the light-induced change in local electrostatic environment. Further, for fluorophores placed near the protein/solvent interface, nanosecond solvent dynamics are expected to directly influence the magnitude and time dependence of the induced red shifts in emission λ_max ^{36, 44, 45}. This phenomenon is illustrated in Figure 5A where PRO represents mADA. Following photoexcitation, the fluorescence lifetime of the inserted Trp is measured at different emission wavelengths (320 to 380 nm in intervals of 10 nm), leading to a predicted shift in the peak emission wavelength to the red as the excited state Trp* undergoes time-dependent stabilization (Figure 5B). These data, together with normalized fluorescence intensities, are the basis for the construction of time-resolved and temperature-dependent emission spectra (TRES) (Figure 5C). The latter are fitted to exponential decay curves to yield excited state Stokes shift lifetimes, amplitudes and total Stokes shifts. Extension of measurements to multiple temperatures provides activation energies (Ea) for Stokes shift relaxations (Figure 5D), for direct comparison to the temperature dependences of catalytic rate constants (Table 1).

Figure 5.

Stokes shift measurements: fundamental basis and data processing procedures. (A) Illustrated for mADA (represented as PRO), where a single tryptophan has been inserted at the protein surface and undergoes restructuring of both protein and associated water molecules subsequent to photoexcitation. Free waters and protein bound waters are shown in red and pink, respectively. Configuration for free water and protein/surface bound water are similar at ground state (t~0 ns). For the excited state configuration (t~10ns), both free water and protein/surface bound waters have moved into an altered configuration (B) Using fluorescence lifetime data collected at multiple emission wavelengths, spectra are seen to shift to longer wavelengths (smaller wave numbers) with some decrease in the fluorescence intensity. (C) These data allow the construction of time-resolved emission spectra (TRES) as a function of temperature. (D) Fitting of data in (C) according to the Arrhenius equation leads to the activation energy, $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$, for individual and/or average lifetimes.

The time-resolved emission spectra (TRES) of the single tryptophan variants in mADA were first analyzed at 30 °C. Spectral and Stokes shift decay curves for WT’-W117in, WT’-K54Win, and F61G’-K54Win are shown in Figure 6 (see Figure S3 for the full set of TRES data for all variants). In the case of single tryptophan proteins in the control group, Stokes shift data were best fit as single exponential decay functions (with or without insertion of F61A), whereas best fits for WT’-K54Win and its inserted series of F61X are bi-exponential. The total Stokes shifts are summarized in Table S4. For the four WT single tryptophan proteins in the control positions (WT’-W117in, WT’-W161in, WT’-W264in, WT’-W272in, the total Stokes shift is around 600–650 cm⁻¹. Insertion of F61A into the single tryptophan control proteins produces variable changes, with F61A’-W117in similar to WT’-W117in, and F61A’-W264in and F61A’-W272in producing larger total shifts of around 750 cm⁻¹. The latter are similar to the solvent-exposed thermal network position mutant WT’-K54Win and F61A’-K54Win (800 cm⁻¹) with shifts between 800 to 1000 cm⁻¹ in the remainder of the F61X’-K54Win series. This range of observable Stokes shifts reflects the microenvironmental differences among the variants.

Figure 6.

Representative time-resolved emission spectra (left) and Stokes shift decay curves (right) for WT’-W117in (top), WT’-K54Win (middle), and F61G’-K54Win (bottom) in 50 mM potassium phosphate buffer pH 7.30 at 30 °C. The time scale for each reconstructed spectrum varies from 0 (black trace) to 10 ns (green trace).

The Stokes shift decay lifetimes (τ, reciprocal of Stokes shift rate constants) at 30 °C for all single site tryptophan variants are reported in Table 2. The single exponential decays of the native variants WT’-W117in, WT’-W161in, WT’-W264in, and WT’-W272in yield average lifetimes of 2–4 ns. The biexponential decays for WT’-K54Win are approximately equal in amplitude, with decay times of 2 ns (τ₁) and ∼9 ns (τ₂) (Table 2). Following insertion of the F61A mutation, the single site native tryptophan variants (F61A’-W117in, F61A’-W264in, and F61A’-W272in) show a single decay pattern with lifetimes of 3–4 ns, which is slightly longer than the parent single tryptophan variants. Turning to the F61X’-K54Win series of mutants, biexponential decays range from 0.6–2 ns for τ₁ and 4–7 ns for τ₂, where the latter shows a trend with the size/shape of the inserted side chain. In the case of τ_1, where it is possible to compare native Trp insertion to WT’-K54Win, with and without mutation at Phe61, the magnitude of the short lifetime is seen to be similar for WT’-K54Win and F61A’-K54Win and to become shorter for the remainder of the F61X’-K54Win series; however, in this case there is no regular trend between τ₁ and the size/shape of the inserted aliphatic side chain (Table 2).

Table 2.

Stokes shift decay lifetimes (τ₁, τ₂) and amplitudes (α₁ and α₂) for all mADA single tryptophan variants at 30 °C.

Protein	τ1(ns)	α1	τ2 (ns)	α2
WT’-W117in	1.8(0.4)	100	-	-
WT’-W161in	1.6(0.3)	100	-	-
WT’-W264in	3.1(1.8)	100	-	-
WT’-W272in	3.7(0.3)	100	-	-
WT’-K54Win	1.9(0.5)	0.44	9.1(1.0)	0.56
F61A’-W117in	2.6(1.2)	100	-	-
F61A’-W264in	3.9(.02)	100	-	-
F61A’-W272in	3.6(0.1)	100	-	-
F61A’-K54Win	1.9(0.1)	0.51	5.9(2.8)	0.49
F61I’-K54Win	0.8(0.1)	0.48	5.6(1.6)	0.52
F61L’-K54Win	1.4(0.8)	0.49	6.1(0.2)	0.51
F61V’-K54Win	0.9(0.4)	0.53	7.4(2.2)	0.47
F61G’-K54Win	0.6(0.1)	0.55	4.4(0.2)	0.45

Temperature dependence of Stokes shifts.

The time-dependent Stokes shifts track how the environment surrounding the photo-excited probe relaxes upon the induction of a distorted electric field⁴⁶. This relaxation process occurs in a thermally controlled manner that can be expected to depend on the position of the probe within the protein of interest^{21, 22}. To reveal the temperature-dependent behavior of the Stokes shift for each of the mADA constructs, we extended the Stokes shift lifetime measurements to seven temperatures from 10 to 40 °C with 5 °C increments. The resulting temperature-dependent Stokes shift rate constants (1/τ) were fitted to the Arrhenius equation to obtain the activation energy for Stokes shifts, $\mathit{Ea}(k_{\mathrm{Stokes}\mathrm{shift}})$. For the proteins that are fit to two exponentials, activation energies for both the fast rate constant (1/τ₁) and the slow rate constant (1/τ₂) are provided.

We first inspected Ea values for Stokes shift for each of the single site control tryptophan variants that are characterized by τ values of ca. 1.4–4.0 ns between 10–40 °C (see Table S5 and Figure S4 for full data sets). As illustrated in Figure 7 for WT’-W117in and tabulated in Table 3, the experimental $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ values are all small, within experimental error of zero in most instances and indicating little or no relationship to the activation energy of 10 kcal/mol for the chemical reaction catalyzed by mADA (Table 1). This conclusion is supported by the failure to observe altered values for $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ in the presence of F61A (Table 3), a mutation that elevates the $\mathit{Ea}(k_{\mathit{cat}})$ by ca. 5 kcal/mol relative to the WT’ constructs. We note that WT’-Trp272in, a residue in the connector regions between the two major thermal networks of mADA (Figure 1C and Figure S5) may indicate an elevation in $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ in the presence of F61A (Table 3), however, the reported values have large errors and are within experimental error of each other. As a frame of reference, the distance from each of the four native Trp residues to Phe61 is 11 Å, 15 Å, 10 Å, and 16 Å for W117, W161, W264, and W272, respectively

Figure 7.

Comparative Arrhenius plots of temperature dependent Stokes shift studies for WT’-W117in, WT’-K54Win, F61A’-K54Win, F61L’-K54Win, F61V’-K54Win, F61I’-K54Win, and F61G’-K54Win in 50 mM potassium phosphate buffer pH 7.30. Stokes shift decays rates from 10−40 °C were fit to biexponential functions except for WT’-W117in which is fitted to a single exponential decay function. See Figure S4 for Arrhenius plots for all single tryptophan mutants.

Table 3.

Activation energies of Stokes shift decays, $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$, for single control Trp variants of WT’ as well as their F61A’ mutations.

Single tryptophan mutant	Ea: 1/τ, kcal/mol
WT’-W117in	0.8(1.8)
WT’-W161in	0.2(0.5)
WT’-W264in	0.6(1.2)
WT’-W272in	0.5(1.2)
F61A’- W117in	0.1(0.8)
F61A’-W264in	1.0(0.6)
F61A’-W272in	1.6(0.5)

We next turned to a detailed investigation of the properties of K54W inserted onto the loop directly adjacent to the left-hand side of the identified thermal network (cf. Figure 2). As shown in Figure 7 and summarized in Table 4, the biexponential fits to TRES curves for WT’-K54Win and the F61X’-K54Win mutants exhibit strikingly different behavior. First, the magnitude and trends of the temperature dependences of 1/τ₁ for all six K54Win constructs are similar to the data obtained with the single tryptophan controls (Table 3): these are small values, within experimental error of zero, and independent of the activity impactful F61X mutations. By contrast, the slower transients, represented by 1/τ₂ values of 0.1 to 0.25 ns⁻¹ (See Table S5) yield elevated values for $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ that further increase upon insertion of F61X in a fashion that parallels the trends in activation energies formerly measured³⁰ for $k_{\text{cat}}$,$\mathit{Ea}(k_{\mathit{cat}})$.

Table 4.

Comparison of activation energies of Stokes shift decays, $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$, for WT’-K54Win to the F61X’ series within this single tryptophan variant.

Single tryptophan mutant	Ea: 1/τ1, kcal/mol	Ea: 1/τ2, kcal/mol
WT’-K54Win	0.2(1.0)	2.4(0.6)
F61I’-K54Win	0.2(0.6)	2.8(1.2)
F61L’-K54Win	0.6 (1.0)	3.6(0.4)
F61V’-K54Win	0.4 (0.7)	4.4(1.2)
F61A’-K54Win	0.01(1.50)	5.1(1.0)
F61G’-K54Win	0.07 (0.70)	8.3(1.1)

Discussion

The relationship between the time scales of solvent exposed Stokes shifts for probes covalently attached to the surface of enzymes is only beginning to be understood, aided by available models from the literature^47–54. One such model, proposed by Nandi and Bagchi, defines a mixture of rapidly exchanging bound and free waters in the vicinity of the protein as “biological waters” and these are differentiated from bulk water based on their slower relaxation time scales⁴⁶. From the present study, the nanosecond biexponential Stokes shift decays for F61X’-K54Win offer a clear distinction between a fast (τ₁~1 ns) component that appears independent of the mutation at Phe61 and a slower (τ₂~3–10 ns) component that is highly sensitive to the size and shape of the side chain at position 61. In this context, we propose that the faster relaxation rate most likely derives from free water and that the slower relaxation rate comes from environmental reorganization involving a combination of the protein surface and its bound waters. Similar nanosecond Stokes shift decays have been reported for probes attached to the active site of glutaminyl-tRNA synthetase⁴⁴ and to a substrate binding tunnel at the protein surface of haloalkane dehalogenase⁴⁵.

In deciphering the implication of the comparative $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ to $\mathit{Ea}(k_{\mathit{cat}})$, it’s essential to understand the scope and limitation of standard models of enzyme catalysis. Studies of the origin of enzyme catalysis have been largely focused on the free energy of reaction, represented by ΔG^‡, in particular the reduction in ΔG^‡ that is accompanied by an increase in reaction rate. In the work herein and related studies, an effort is underway to understand how the enzyme proceeds from its ground state to excited state structures. This is being done in the context of the recognition that ca. 98–99% of all known enzyme catalyzed reactions undergo thermal (as opposed to photochemical) initiation of their reactions. Many treatises stress the importance of protein sampling among large conformational ensembles that are in rapid equilibrium with each other ^55–58. This model provides an overall interpretation for altered enzyme reactivity with shifted active conformation percentage but fails to stress the direct connection of site-specific residue perturbation with active site chemistry. As pursued herein and elsewhere^28–32, a decomposition of ΔG^‡ into its component ΔH^‡ (and consequently TΔS^‡) offers a way of correlating thermally induced dynamics within a protein scaffold (as determined using an inserted spectroscopic probe) with the required thermal activation of active site chemistry. This serves as the evidence for thermal activation of enzymes through regionally restricted protein motifs (see below). The model posited is within the context of established conformational ensemble models, with the important additional component of a role for site specific thermally activated protein restructuring that takes place on rapid time scales and over significant distances.

Identifying and characterizing the properties of dynamical activation of enzyme activity.

The ease of definition of the contribution of protein dynamics to enzyme function is dependent on both the time scale and spatial resolution of the targeted motions. For movements that occur at longer times and involve large segments of a protein, as occurs during protein conformational changes and/or protein domain shifts, direct detection is often possible through X-ray crystallography or cryogenic electron microscopy^59–61. Motions that occur on very fast time scales, ranging from nanoseconds to femtoseconds, can be shown to be correlated with function through computation, but these are much more challenging to detect experimentally^{62, 63}. In the last several years, this laboratory has chosen to focus on the temperature dependence of function-linked enzyme motions^{28–32, 38}.

The introduced methodology is multi-tiered, with temperature-dependent hydrogen deuterium exchange mass spectrometry (TDHDX-MS) offering a method for the spatial resolution of activity related protein dynamics^64–69. As has been described in several original research papers and reviews^{30, 32, 70}, the combination of TDHDX-MS with site-specific mutagenesis, leads to a simplification in the interpretation of the kinetic parameter k_(HDX), enabling the construction of heat sensitive maps for protein flexibility that is central to the catalytic behavior of native enzymes. At this juncture a growing list of different enzyme classes have been categorized with regard to the presence of spatially resolved thermal networks that connect site-specific protein surfaces to active site residues over long distances that are in the range of 10–20 Å^{28–30, 32}. The enzyme mADA was the first member of the TIM barrel family of enzymes to be characterized in this manner.

Once a thermal network has been identified, the next challenge is to measure the time dependence of functionally relevant thermally activated motions. Fluorescence spectoscopy is ideally suited to the detection of rapid and cooperative motions, utilizing either intrinsic or extrinsic probes^{21, 22, 42, 71}, with different information being detected depending on the position of the fluorescent probe. The insertion of chromophores at the protein surface provides a particularly useful probe for understanding the role of solvent in the initiation of active site chemistry³⁶. Once again, the availabity of protein mutants that have been shown to differ with regard to their enthalpies of activation for catalysis is essential, providing a basis for determining whether the behavior of an inserted probe is relevant to an active site transformation. Appropriate controls include the insertion of the spectrocopic probe at a variety of diferent protein surface positions (as reported herein and for soybean lipixygenase, SLO³⁷) or the interrogation of variants containing mutations outside of the inferred thermal network (in the case of SLO³⁸).

In a study of SLO, Stokes shifts analyses of an extrinsic probe appended to a surface introduced cysteine residue were shown to produce an activation energy for the environmental rearrangement around the surface fluorophore, $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ that is identical to the $\mathit{Ea}(k_{\mathit{cat}})$ for C-H activation of substrate via quantum mechanical tunneling³⁷. This identity of Ea values was subsequently shown to be maintained when WT SLO was converted to a series of single site mutants at hydrohobic side chains identified as residing within the thermal network³⁸. From such data, a shared, temperature dependent protein conformational change was concluded to connect the time dependent environmental reorganization around the appended fluorescent probe and the active site bound substrate over a distance of ca. 20 Å; further, this reorganization was required to occur on a time scale that exceeded the fastest measured rate constant (i.e., >nanosecond, the time of experimental Stokes shift measurments).

In studies of SLO, it was further noted that the magnitude of the individual rate constants $k_{(\text{Stokes}\text{shifts})}$ and $k_{(\text{cat})}$ differ by 10⁶-fold, in contrast to the 1:1 identity of their respective Ea values ^37,38. A formalism for such disparate experimental rate constants was proposed, that combines a common, rapid and temperature dependent protein restructuring (represented by $k_{\mathit{int}}$,) with the probability that the activated region of protein will be able to support productive barrier crossings (${\text{P}}_{\text{conf}}$). The final rate constant is expressed as below:

$$k_{obs}={\text{P}}_{\text{conf}}\times k_{\mathit{int}}$$ (1)

where ${\text{P}}_{\text{conf}}$ is much larger in the case of surface Stokes shifts than active site bond cleavage, reflective of a smaller entropic barrier for the environmental reorganization around a solvent exposed fluorescent probe.

Linking the activation energy for Stokes shifts to the activation for catalysis in mADA.

For the present work, we extended studies of a tunneling-based enzyme such as SLO to a process where thermally activated motions within a protein scaffold will control a very different and potentially more complex active site chemical transformation. Initial studies of protein dynamics within the TIM barrel family of enzymes found the attachment of an extrinsic fluorophore Badan to the protein surface to be unsuited to detailed studies of protein dynamics. For example, covalent attachment of the fluorophore Badan to engineered surface cysteine residues in the enzyme enolase led to disruption of the protein structure⁷² while application of a similar strategy in the mADA system was compromised by both protein instability and tryptophan quenching (data not shown). That said, the fact that there are only four tryptophan side chains in mADA proved fortuitous and greatly simplified subsequent studies. As shown, it has been possible to create a Trp-free variant of mADA (Trp to Phe conversions in each case) that preserves enzyme activity (Table 1).

Using this Trp-free variant as a template, five single Trp variants have been studied in detail, either by reinserting Trp at each of the native Trp positions or by creating a new Trp position by replacement of a protein surface lysine (K54Win). These two strategies produce very different behaviors, with the reinsertion of each native Trp leading to almost unchanged properties of $k_{\text{cat}}$ and $\mathit{Ea}(k_{\mathit{cat}})$ (Table 1), together with single exponential decays accompanied by very small and essentially invariant activation energies for the experimental Stokes shifts, Ea (Stokes shifts), either in the absence or presence of the $\mathit{Ea}(k_{\mathit{cat}})$ altering mutation F61A (Table 3). These constructs provide robust controls for results from the engineered single Trp variant WT’-K54Win.

As seen (Table 4), the choice of Lys54 as a site for installation of a surface Trp has produced a different behavior, with the onset of a biexponential decay pattern in the Stokes shift relaxation rates. The faster of these two processes, 1/τ₁, is highly similar to the behavior of the single Trp controls, while the slower rate of the two processes, 1/τ₂, is observed to be quite sensitive to the insertion of the F61X series. Examination of a plot of $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ for 1/τ₂ vs $\mathit{Ea}(k_{\mathit{cat}})$ (Figure 8A) illustrates the linear relationship among the data points and implicates a significant correlation between distal protein motions surrounding the exposed fluorescent probe and the active site motions that contribute to catalysis. One notable feature of Figure 8A is an intercept of 6.6±(1.0) kcal/mol on the x-axis where $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ is zero, representing the fraction (0.66) of the total activation barrier for WT’ (10.0 kcal/mol) that cannot be attributed to activation of the helix-loop-helix motif. The structure in Fig.1C provides a possible explanation for this parsing of $\mathit{Ea}(k_{\mathit{cat}})$ into different origins. As illustrated, the right-hand side of the mADA structure contains a separate network that originates at an opposing protein/solvent interface, to reach past the inferred catalytic acid and base side chains and the active site Zn²⁺ toward bound substrate. In this regard, we have been screening for a suitable dynamical probe within this alternate region of mADA that both retains enzyme stability and maintains $k_{\mathit{cat}}$ values reasonably close to WT. The success of such experiments will allow us to test the degree to which thermal activation of the active site of mADA via the opposing regions of the network can be detected and, further, whether the thermal activation of two opposing networks in a protein will behave in an additive manner.

Figure 8.

Comparative activation energies for the fluorescence Stokes shift decay rates 1/τ₂ ($\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$)at K54Win and for catalysis ($\mathit{Ea}(k_{\mathrm{cat}})$). The plot in (A) emphasizes the x-intercept of 0.66 (±1.0) kcal/mol, whereas the plot in (B) emphasizes the slope of 0.65(±0.06), R squared = 0.96. The values shown are the average of three biological replicates with the indicated error bars. The labels WT, F61A, F61V, F61G, F61L, F61I represent WT’-K54Win, F61A’-K54Win, F61V’-K54Win, F61G’-K54Win, F61L’-K54Win, and F61I’-K54Win, respectively.

In an effort to simplify the presentation of the contribution of thermal activation within the K54Win-containing helix-loop-helix to the temperature dependence of either Stokes shifts or enzyme turnover we replotted the impact of mutation-induced changes to $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ in relation to the portion of the total $\mathit{Ea}(k_{\mathit{cat}})$ that can be assigned to the left hand segment of the thermal network, [$\mathit{Ea}(k_{\mathit{cat}})$ – Intercept] (Figure 8B). The slope of 0.65(±0.06) means that the environmental reorganization around the excited state dipole of the photoactivated K54W is somewhat better accommodated than the reorganization required for chemical reactivity within the buried active site. A specific role for free water (Figure 5A) that selectively affects the stabilization of Trp* with a significantly smaller $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ offers a possible explanation for the net reduced $\mathit{Ea}(k_{\mathit{Stokes}\mathit{shift}})$ relative to the value representing $k_{\text{cat}}$.

We conclude that while the broadly based dynamical properties of mADA are similar to the earlier studied SLO³⁹, the details of mADA are distinct from the deep tunneling reaction catalyzed by SLO. In the latter case, the catalyzed C-H activation process has been shown in all constructs to share an identical activation energy with the distal protein reorganization that governs the measured Stokes shift relaxation process ³⁹. Clearly, the chemical reaction of mADA is quite different from that of SLO, where the primary chemical coordinate involves wave function overlap between an aligned substrate hydrogenic donor and active site iron hydroxide. The reaction mechanism of mADA requires the coordination of a nucleophilic addition reaction that is dependent on an active site zinc bound water and its other metal ligands, the participation of both acidic and basic side chains, as well as a productive positioning of the reactive carbon of substrate. Further, TDHDX studies of SLO implicated only a single solvent interface in the thermal initiation of active site chemistry²⁸, As discussed above and in the previous study³⁰, the mADA is proposed to be dependent on a minimum of two opposing protein solvent interfaces.

The detailed structural relationship between the surface exposed tryptophan in K54Win and the active site Phe61 is provided in Figure 9. As shown, the Trp fluorophore at position 54 is situated in the middle of a loop that belongs to the solvent-exposed helix-loop-helix motif. As noted above, the thermally activated environmental reorganization near Trp54 that accompanies electronic excitation during Stokes shifts measurements is likely dependent on both movements of free water molecules close to the exposed Trp and local intrinsic motions within the exposed loop entity at the protein/solvent interface. The latter is expected to be dynamically coupled to the adjacent internal helix structure^{26, 73}.

Figure 9.

Representation of a solvent-exposed helix-loop-helix motif (labelled red), derived from residues 46–74 in mADA (PDB: 1ADD). The interior helix provides a hydrophobic wall comprised of Leu58, Phe61, Leu62 and Phe65 (red stick). This is located behind the substrate and is directly connected to the inserted tryptophan at position 54 (shown in yellow stick) via the adjacent loop. The ground state analog 1-deaza-adenosine is shown in grey stick, zinc ion is magenta and the three histidine ligands to zinc ion are in dark grey stick representation.

While the observed correlation between Ea for $k_{(\mathit{cat})}$ and k_{(Stokes Shifts)} could have been fortuitous, this seems highly unlikely given the large number of accompanying controls involving insertion of a Trp probe at four other positions. Notably, the long internal amphiphilic helix displays outward pointing residues that are hydrophilic and inward pointing residues that are hydrophobic (Leu58, Phe61, Leu62 and Phe65). Directional heat flow from solvent to the active site of mADA can be envisaged as resulting from propagated changes in side chain and backbone vibrational modes that alter the relationship of the Phe61-containing helix to the back face of bound substrate, inducing alterations in distance and charge distributions between the reactive C-6 of the purine ring of substrate and the attacking nucleophilic zinc-bound water/hydroxide ion. The entire long range (~20 Å) thermal activation process is expected to occur on a time scale faster than nanoseconds, based on the observed ns correlated Stokes shift measurements (Figure 8). The regional nature of the productive thermal activation of enzymes is intimately linked to the highly anisotropic nature of protein/solvent interfaces for proteins in general^{74, 75}, and for m-ADA in the present case.

Finally, a full atomistic description of protein thermal network function (and design) will depend on better understanding of the mechanisms of long-range heat transfer in proteins. Rapid and long-range heat flow has been characterized in model peptides and a few proteins and is generally understood in the context of coupled anharmonic vibrational modes that are of low frequency^76–85. The speed of heat transmission via such vibrational excitation has been estimated to be as high as ca.15Ų/ps^{86, 87}, compatible with the length of interrogated thermal networks. The molecular mechanism for the initiation of such heat flow for a protein that is at thermal equilibrium with its solvent bath is less obvious. The proximity of a remote mobile loop that can access multiple conformations is seen for both the SLO and mADA thermal networks, suggesting one working hypothesis. This involves solvent collisions with an extended loop that induces its movement toward a proximal, structured region of protein that is capable of effective vibrational excitation that extends from the protein surface to a more buried active site. For each rapid protein restructuring (>nanosecond in the present case) there will be ongoing competition between productive vibrational energy transfer that enables femtosecond barrier crossings and dissipation. The generation of enzymatic rate constants in the range of 10³ s⁻¹, can be viewed as arising from such very rapid protein restructuring events (e.g., 10¹² s⁻¹) that are extremely rare events (ca. 10⁻⁹). According to such a model, an increase in temperature leads to a higher heat content within the solvent bath and, hence, a greater likelihood of generating the activated protein conformation that initiates reaction.

Looking to the future.

In recent years, with the advances in instrumentation, researchers are increasingly able to understand the fundamental basis of how enzymes work. The growing evidence that remote thermal activation of enzyme active site chemistry is site-specifically encoded within a protein scaffold has ramifications for modern applications of enzyme inhibitor design^84–86, protein engineering^{88, 89} and de novo design^{90, 91}. In this work, we present a detailed analysis of the thermal activation of the prototypical TIM barrel enzyme mADA. Through two complementary methodologies, we were first able to locate regions attributable to thermal activation via TDHDX ^27,28, and then use this information to guide studies of the time-dependent behavior of heat transfer via detailed Stokes shift analysis (the present work).

Time-resolved optical spectroscopy is a powerful tool for investigating chemical dynamics in condensed phases. The fluorescence Stokes shift reports on the response of the solvent environment to light-induced changes in electrostatic charge within a suitable probe. For a protein surface-bound chromophore, the source of the reorganizational process is expected to arise from nearby water molecules along with bound water and charged/polar side chains at the protein/water interface. This study provides the first temporal resolution of nanosecond protein surface motions that can be correlated with millisecond enzyme activity within a canonical TIM barrel enzyme. The observed behavior for mADA is complementary to the well described SLO system, with differences that appear due to the nature and placement of the interrogated chromophore and the role of a more extended thermal network in mADA. Overall, the emerging data and model extend the default role of broadly distributed and rapidly equilibrating protein conformational landscapes as the dominant dynamical behavior impacting enzyme function, providing support for a model that incorporates site specific, rapid (>nanosecond) and long range protein dynamics to initiate enzyme catalysis³⁹. These properties have analogy to ongoing discussions in the literature regarding the role of discrete protein networks in the origins of protein allostery^92–94. The spatial and temporal resolution of such functionally impactful protein motions in mADA, together with other proteins (e.g., enolase³², SLO^{28, 39}, and catechol-O-methyltransferase (COMT)²⁹), has the potential to produce an in-depth understanding of the atomistic mechanisms of the thermal activation of enzymes. The long-range goal of building a library of identified thermal networks within protein scaffolds could be transformative for both rational design and redesign of protein catalysts.