Conformational selection is present in ligand binding to cytochrome P450 19A1 lipoprotein nanodiscs

Abstract

Cytochromes P450 (CYPs) display remarkable plasticity in their ability to bind substrates and catalyze a broad array of chemical reactions. Herein we evaluate binding of androstenedione, testosterone, and 7-hydroxyflavone to CYP19A1, also known as aromatase, in phospholipid nanodiscs by stopped-flow UV-vis spectroscopy. Exponential fitting of the kinetic traces supports the possibility of a multi-step binding mechanism. Subsequent global fitting of the data to the solutions of the coupled differential equations describing the fundamental mechanisms of induced fit and conformational selection, consistently support presence of the latter. To our knowledge, this is the first discrimination of conformational selection from induced fit for a monodisperse CYP in a native-like membrane environment. In addition, 7-hydroxyflavone binds to CYP19A1 nanodiscs with comparable affinity to the substrates and induces an unusual spectral response likely attributable to hydrogen bonding to, rather than displacement of the heme-coordinated water molecule.

1. Introduction

Site-specific ligand binding to biological macromolecules are the fundamental elements of a myriad of processes required for the maintenance of life. These events constitute the basic units of metabolic networks and mediate distant communications between cells and tissues. Understanding ligand binding mechanisms and identification of the critical interactions are the bases for the design of chemical probes and discovery of novel therapeutics. The original model for specific ligand binding is the so-called „lock and key‟ concept described by Emil Fisher (1) where the enzyme (E) binds to the ligand (L) in one reversible step (OS). (Equation 1)

$$E\underset{k_{off}}{\stackrel{k_{on}[L]}{\rightleftharpoons }}EL$$ (Eq.1)

This model assumes that the enzyme and the ligand complex, EL, are essentially static entities. Though it is widely appreciated that proteins undergo conformational transitions over timescales spanning several orders of magnitude. The OS model is adequate if the transitions between ensembles occur on timescales that are much shorter than those of binding, as the ligand perceives that it is interacting with a single ensemble. However, conformational transitions in proteins typically occur on similar or longer timescales than those required for ligand binding, necessitating the incorporation of at least a second step in the kinetic scheme. If this additional step is assumed to be a conformational transition, it could occur before or after the ligand binding step. Eigen originally envisioned the scenario where the conformational transition precedes ligand binding as pre-equilibrium (2), known better known today as conformational selection (CS). In its most simple form, the ligand preferentially binds to, or singles-out, one of at least two ensembles (E₁ or E₂). (Eq. 2)

$$E_1\underset{k_{-1}}{\stackrel{k_1}{\rightleftharpoons }}{E}_2\underset{k_{off}}{\stackrel{k_{on}\left[L\right]}{\rightleftharpoons }}EL$$ (Eq.2)

Alternatively, it can be assumed that there is only a single relevant ensemble present, and the conformational transition occurs following, or is induced by, ligand binding. This mechanism, originally envisioned by Koshland (3,4), is known as induced fit (IF). Because the conformational transition follows ligand binding, introduction of a transient encounter complex (EL*) that subsequently relaxes to the terminal bound state (EL) is required. (Equation 3)

$$E\underset{k_{off}}{\stackrel{k_{on}\left[L\right]}{\rightleftharpoons }}E{L}^{*}\underset{k_{-1}}{\stackrel{k_1}{\rightleftharpoons }}EL$$ (Eq.3)

The prevalence of CS or IF in biological macromolecules has been extensively debated and several experimental approaches have been developed to resolve the conundrum. Vogt and Di Cera (5–7) have proposed a remarkably simple, yet rigorous approach to extract such mechanistic information from stopped-flow kinetics measurements. They showed that a straightforward diagonalization of the rate constant matrix for each mechanism yielded both „fast‟ and „slow‟ relaxations, the latter corresponding to the k_obs values from the multi-exponential fits of the data. A decreasing or flat trend in k_obs with increasing [L] was shown to be diagnostic of CS, however neither can be ruled out if the trend is increasing. Prior to these studies, increasing k_obs with [L] was accepted as a hallmark of induced fit; however, in view of the mechanistic ambiguity of this trend, the corollary is that CS is likely more common in site-specific ligand binding to proteins than previously appreciated. (5)

The cytochromes P450 (CYPs) are heme-containing monooxygenases that catalyze the oxidations of nearly all drugs and environmental xenobiotics to which humans are exposed.(8–10) CYPs achieve substrate oxidation through a complex catalytic cycle initiated by substrate binding that induces a spin and redox potential shift of the heme iron, followed by reduction and O₂ binding and activation.(11) The Soret band of the heme chromophore is very sensitive to changes in the Fe coordination environment and spin state induced by ligand binding. Accordingly, both the extent of active site occupation and the nature of heme-ligand interactions can be conveniently obtained through UV-vis spectroscopy. The human genome codes for 57 CYPs that can be partitioned into two groups based on their substrate specificity. Those involved in endobiotic metabolism-i.e., steroid hormone, bile acid, and eicosanoid biosynthesis-display high substrate specificity. Conversely, those CYPs involved in the metabolism of xenobiotics are catalytically promiscuous, accounting for ~75% of drug oxidations.(8) Hence, knowledge about the propensity for a drug to cleared by a CYP-dependent pathway is critical to optimizing pharmacokinetics, patient dosing, and avoiding adverse effects in the clinic.

Many CYPs involved in xenobiotic metabolism bind multiple substrates and display homotropic and heterotropic cooperativity.(12–19) The remarkable plasticity displayed by these enzymes to catalyze such a broad array of challenging reactions has led to significant interest in the mechanisms used to bind substrates. Nevertheless, there is still little known about how these enzymes discriminate and bind ligands. Crystal structures with different ligands display multiple protein conformations, and NMR has revealed the presence of conformational heterogeneity in several CYPs.(20–27) Two comprehensive kinetics studies that paired an evaluation in k_obs trends with global fitting concluded that CS was present in several CYPs.(28,29) As in many studies of the membrane associated forms, the enzymes were engineered for solubility and the membrane ignored. Due to the hydrophobicity of domains intended to interact with membranes in their native environments, these enzymes are rarely monodispersed in solution and the underlying equilibria have the potential to confound the interpretation of kinetics results.

Herein we describe a kinetic study of ligand binding to a monodisperse CYP in a native-like membrane environment. To this end, we have adopted CYP19A1 (19A1 hereafter) embedded in phospholipid nanodiscs (NDs) as a model. 19A1 catalyzes the conversion of androgens to estrogens in three sequential oxidation reactions with a requirement of three equivalents each of O₂ and NADPH.(30–32) The primary substrates for 19A1 are androstenedione (ASD) and testosterone (TST), that are converted to estrone and 17β-estradiol, respectively. (Scheme 1) 19A1 is among the most primordial of the human CYPs(20) and to the extent that the fundamental dynamics are conserved, those coupled to ligand binding are broadly relevant, especially to the understanding of the more recently evolved and promiscuous enzymes.(33,34) Finally, inhibitors of 19A1 have proven extremely valuable as pharmacotherapy for estrogen-dependent cancers and gynecological disorders.(35–39) The ND platform provides the native-like membrane environment and because they harbor a single enzyme per disc, are intrinsically monodisperse.(40–43) These circumstances increase confidence that the rates are attributable to the ligand binding properties and conformational transitions of the monomeric enzyme, rather than mixtures of oligomers. Our studies support that CS is the dominant mechanism used by isolated, membrane-associated 19A1 to bind ligands.

Scheme 1.

Structures of androstenedione (ASD), testosterone (TST), and 7-hydroxyflavone (7HF).

2. Materials and Methods

2.1. Protein expression and purification.

19A1 used in these studies was identical to the form originally described by Kagawa and coworkers with arginine at position 264.(44) A codon-optimized gene for 19A1 was inserted between the NdeI and HindIII sites of pCWOri (pCW). DH5α cells were transformed with the resulting plasmid, pCW-19A1, as well as with pBB542 encoding the groES/L, dnaK, and dnaJ chaperones. (45) Transformed colonies were used to inoculate 100 ml of Luria-Bertani media containing 100 μg/ml ampicillin and 50 μg/ml spectinomycin. Cultures were grown overnight at 37 °C and shaking at 220 rpm. 15 ml of the overnight cultures were used to inoculate 1 L of Terrific Broth containing 100 μg/ml ampicillin, 50 μg/ml spectinomycin, and 10 μM of ASD. These were grown at 37 °C and shaking at 220 rpm until the O.D. at 600 nm reached 0.5–0.7. Following the addition of 1 mM δ-aminolevulinic acid, reduction of the temperature to 28 °C, and reduction of shaking to 180 rpm, cultures were grown for an additional hour. Cultures were then induced with 1 mM IPTG and allowed to grow for 48 h at 28 °C and shaking at 180 rpm. Cells were harvested by centrifugation at 4000×g for 20 min at 4 °C. Cell pellets were used immediately or stored at −80 °C until required.

Cell pellets were resuspended in lysis buffer (0.1 M potassium phosphate, 10 % glycerol, 0.5 mM EDTA, 10 μM ASD; 5 ml/g cells) and stirred on ice for 30 min with 1 mg/ml lysozyme and complete protease inhibitor cocktail (Roche). 1% Tergitol NP-10, 2 mM β-mercaptoethanol (ME), and 1 mM phenylmethylsulfonyl fluoride were then added and the suspension was allowed to stir for an additional 30 min. Cells were disrupted using a Misonix Sonicator 3000 and the resulting cell debris was pelleted by centrifugation at 45,000×g for 1 h at 4 °C. The supernatant was loaded onto a HisPrep 16/10 Ni²⁺-NTA column (GE Healthcare Life Sciences) previously equilibrated with Buffer A (0.1 M potassium phosphate [pH 7.4], 0.1 M NaCl, 20% glycerol, 10 μM ASD, and 2 mM ME). The column was sequentially washed with five bed volumes each of buffer A containing 0.2 % Tergitol NP-10, buffer A containing 0.2 % sodium cholate, and buffer A containing 10 mM L-histidine. 19A1 was eluted with buffer A containing 75 mM L-histidine. Red-colored fractions were pooled and dialyzed overnight against 19A1 storage buffer (0.05 M potassium phosphate, 0.1 M NaCl, 10% glycerol, 2 mM ME, 0.5 mM EDTA, and 10 μM ASD). SDS-PAGE analyses of purified 19A1 revealed dominant bands for 19A1 as well as for GroEL; however, the latter contaminant did not interfere with, and is removed during the preparation of the NDs. (Supporting Figure S1)

Membrane scaffold protein 1D1 (MSP1D1) was expressed in BL21 Gold (DE3) E. coli and purified as previously described. (46) TEV protease was used to cleave the N-terminal 7-His tag from MSP1D1 to yield MSP1D1(−). (47) Residual un-cleaved MSP1D1 was removed from the cleaved protein, MSP1D1(−), by capture on a HisPrep 16/10 Ni²⁺-NTA column.

2.2. Assembly of 19A1 NDs.

Assembly of 19A1-NDs utilized a 19A1:MSP1D1(−):POPC ratio of 0.1:1:65. A dried POPC film was solubilized with a 2:1 ratio of cholate:POPC with the addition of a 0.1 M sodium cholate/0.1 M NaCl followed by the addition of 19A1 (~40–50 μM) and MSP1D1(−) (~150 μM) in their respective storage buffers. As necessary, additional sodium cholate/NaCl solution was added to ensure a final cholate concentration of at least 20 mM in the final assembly mixture. Finally, 5 mM ME and 1 μM ASD were added. The assembly mixture was rocked for 3 h at 4 °C. ND assembly was initiated by the addition of 1 g/ml of washed Amberlite XAD-2 beads and rocked for an additional 5 h at 4 °C. The Amberlite was filtered and loaded onto a 5 ml Bio-Rad Profinity Ni²⁺ column previously equilibrated with buffer B (0.05 M Tris-HCl [pH 7.4)] 0.1 M NaCl, 2 mM ME). The column was subsequently washed with five bed volumes of buffer B containing 15 mM imidazole. 19A1-NDs were eluted with buffer B containing 200 mM imidazole. If not immediately subjected to size-exclusion chromatography, 19A1-NDs were spiked with 10% glycerol and stored at −80 °C until required. High-MW aggregates were removed by analytical size-exclusion chromatography using a Superdex 200 Increase 10/300 GL column (GE Healthcare Life Sciences) previously equilibrated with ND buffer (0.04 M Tris-HCl [pH 7.4], 0.1 M NaCl, 1.5 mM TCEP). Representative size-exclusion chromatography and the results of SDS-PAG analysis of 19A1-NDs are provided in the Supporting Information, Figure S1.

2.3. UV-visible Absorbance Analysis of Ligand Binding.

UV-vis absorbance spectra were measured using an OLIS (Bogart, GA) Cary-14 Conversion spectrophotometer. Spectra of 19A1-NDs were measured with ND buffer as the reference. When titrating increasing concentrations of ligands, an equivalent volume of ethanol was added to the reference cuvette. Absorbance spectra for n ligand concentrations were organized into the matrix A(λ,n) and subject to singular value decomposition (SVD). (48) SVD performs the decomposition A = USV^T where U are the basis spectra, is a diagonal matrix of singular values, and are the spectral amplitudes. Difference spectra (spectra of protein and ligand less that of the ligand-free protein) could be reconstructed from the second singular value, basis spectrum and amplitude vectors. Assigning ΔA = SV^T, the K_d was determined by fitting these values to the equation for one-site binding. SVD and fitting of the spectral amplitudes to a single-site binding model were performed in MATLAB.

2.4. Stopped-flow UV-vis Spectroscopy.

Stopped-flow measurements were made with an OLIS RSM1000 spectrophotometer equipped with a dual-syringe stopped-flow system maintained at 25 °C. One syringe contained 2 μM or 4 μM 19A1-NDs while the other contained ASD (0.5–16 μM), TST (1–20 μM), or 7HF (10–60 μM; Scheme 1) in ND buffer. Spectra were recorded from 330 to 580 nm at a scan rate of 1 ms⁻¹ for 5.9 seconds. Representative stopped-flow UV-vis spectra obtained with the ligand concentrations used for global fitting are provided in the Supporting Information, Figures S2–S4.

2.5. Global fitting of stopped-flow data.

For each ligand concentration, data from at least three shots were averaged and fit to one or the sum of two or three of exponentials,

$$y(t)=A_0+\sum _{i=1}^n{A}_i{e}^{-k_{i}t}$$ (Eq.4)

with n=1–3, A_i are the amplitudes, and k_i are the observed rate constants (k_obs). Data sets were also globally fit to solutions of the ordinary differential equations for OS (Equation 1), IF (Equation 2), CS from two states (Equation 2), and CS followed by IF (Equation 5)

$$E_1\underset{k_{-1}}{\stackrel{k_1}{\rightleftharpoons }}{E}_2\underset{k_{off}}{\stackrel{k_{on}\left[L\right]}{\rightleftharpoons }}E{L}^{*}\underset{k_{-2}}{\stackrel{k_2}{\rightleftharpoons }}EL$$ (Eq.5)

All global fitting was performed in MATLAB. To this end, multiple initial rate constants were defined using the multi-start algorithm included in the Global Optimization Toolbox. For mechanisms involving CS, initial concentrations of each ensemble were calculated using k₁, k₋₁, and the total protein concentration. Otherwise, the initial concentration was equal to the total protein concentration. In turn, the coupled differential equations were solved using the ode15s solver. The cycle was iteratively repeated to obtain a least-squared fit using the constrained nonlinear optimizer fmincon. Upper bounds of 100 s⁻¹ and 10⁸ M⁻¹·s⁻¹ were placed on first- and second-order rate constants, respectively. Goodness of fit is reported as ‘global’ χ² normalized to the number of ligand concentrations evaluated in each fit. For comparisons with values obtained from equilibrium titration experiments, K_d values were calculated using rate constants derived from the global fits using equations 6–8.

$$K_d^{IF}=\frac{[L][E]}{[EL]}=\frac{k_{off}{k}_{-1}}{k_{on}\left(k_1+k_{-1}\right)}$$ (Eq. 6)

$$K_d^{CS}=\frac{[L]\left(\left[E_1\right]+\left[E_2\right]\right)}{[EL]}=\frac{k_{off}\left(k_{-1}+k_1\right)}{k_{on}{k}_1}$$ (Eq. 7)

$$K_d^{CS+IF}=\frac{[L]\left(\left[E_1\right]+\left[E_2\right]\right)}{[EL]+\left[E{L}^{*}\right]}=\frac{k_{off}{k}_{-2}\left(k_1+k_{-1}\right)}{k_1{k}_{on}\left(k_2+k_{-2}\right)}$$ (Eq. 8)

3. Results

3.1. Equilibrium binding to 19A1 NDs.

The equilibrium binding affinities of ASD, TST, and 7-hydroxyflavone (7HF) were measured by monitoring changes in the Soret bands induced by distinct environmental changes of the heme characteristic of the incoming ligand. Structures of these ligands are illustrated in Scheme 1. All ligands bound with comparably high affinity and the substrates (ASD and TST) produced difference spectra and K_d values consistent with those previously reported data obtained with soluble enzyme.(49) Increasing concentrations of ASD and TST shifted the Soret from 419 nm to 388 nm, producing what are known as ‘type I’ difference spectra. (Figure 1A, B) Such spectra are attributable to conversion of the heme from a hexacoordinate, low-spin state to a pentacoordinate, high-spin state coupled to the displacement of the Fe³⁺-coordinated water. Binding isotherms for ASD and TST were best fit to a single-site binding model with K_d values that were identical within standard deviation.

Figure 1.

Binding isotherms for 19A1 binding to ASD (A), TST (B), and 7HF (C) with the corresponding difference spectra inset. 19A1 concentrations were 1 μM in the experiments with ASD and TST and 2 μM in the experiments with 7HF. Binding data obtained with ASD, TST, and 7HF were best fit to a single-site binding model. Fits are depicted with solid black lines, while the upper and lower bounds of 95%confidence interval are depicted with hashed lines. Reported K_d values are mean of three each titration experiments ± standard deviation.

Several flavone natural products have been identified as inhibitors of 19A1, with one of the most potent in catalytic turnover experiments being 7HF.(50,51) A type I spectrum was expected akin to the substrates; however, an unusual spectral response was observed. (Figure 1C) Instead of the expected type I difference spectrum, titration of 7HF into 19A1 NDs resulted in difference spectra with maxima and minima at 413 nm and 432 nm respectively. The resulting binding isotherm was likewise best fit a one-site binding model with a K_d of 0.22 μM.

3.2. Stopped-flow UV-vis analyses of binding.

Multiple concentrations each of ASD, TST, and 7HF were rapidly mixed using a stopped-flow apparatus and the magnitude of the shifts in the Soret bands monitored every ms. Averaged datasets were fit to one or sums two or three exponential functions (Eq. 4); however, two exponentials were minimally sufficient to fit the data with random, normally distributed residuals. Such behavior is consistent with a multi-step binding mechanism. Single-exponential fits yielded poor fits with clear patterns in the residuals, especially at early time points. The ‘fast’ and ‘slow’ k_obs values extracted from the double exponential fits for each ligand and concentration are plotted in panels A of Figures 3–5. The trends for k_obs are consistently increasing with [L]; hence nor CS or IF can be unambiguously ruled out for any of the ligands. To delineate the dominant mechanisms involved in binding these ligands, we adopted a sophisticated global fitting approach.

Figure 3.

Parameters from exponential fits in Figure 2A and global fitting of stopped flow data for ASD binding to 19A1-NDs. k_obs (s⁻¹) (A) for the fast (blue) and slow (orange) phases of the exponential fits. Reported values are the mean ± standard deviation of at least three stopped-flow experiments. Inset in panel (A) are raw spectra collected every ms for 3 s with 2 μM ASD and 1 μM 19A1-NDs. (B) Global fitting to an IF model (Eq. 3). (C) Global fitting to a CS model (Eq. 2). (D) Global fitting to a CS-IF model (Eq. 5) In B-C, concentrations of ASD after mixing are 0.25 μM (blue), 0.5 μM (orange), 1 μM (yellow), 2 μM (purple). Global fitting results are depicted with solid black lines. An extinction coefficient ε^{388nm−420nm} = 0.053 μM⁻¹ cm⁻¹ was used to calculate the concentration change from difference spectra.

Figure 5.

Parameters from exponential fits in Figure 2C and global fitting of stopped flow data for 7HF binding to 19A1-NDs. k_obs (s⁻¹) (A) for the fast (blue) and slow(orange) phases of the exponential fits. Reported values are the mean ± standard deviation of at least three stopped-flow experiments. Inset in panel (A) are raw spectra of collected every ms for 1.5 s with 30 μM 7HF and 2 μM 19A1-NDs. (B) Global fitting to an IF model (Eq. 3). (C) Global fitting to a CS model (Eq. 2). (D) Global fitting to a CS-IF model (Eq. 5). In B-C, concentrations of 7HF after mixing are 5 μM (blue), 10 μM (orange), 15 μM (yellow), 20 μM (purple), and 30 μM (green). Global fitting results are depicted with solid black lines. An extinction coefficient ε^{432nm−413nm} = 0.047 μM⁻¹ cm⁻¹ was used to calculate the concentration change from difference spectra.

To this end, data were globally fit to the systems of coupled differential equations describing OS, CS, IF, and CS followed by IF to distinguish the preferred mechanism. In view of the double exponential fits supporting a multi-step mechanism, global fits to the OS model are reserved for the Supporting Information, Figure S5. Each fit to the OS model yields consistently larger global χ² than the multiple step mechanisms. As shown in Figure 2, panels A and B, the steady state concentrations of [EL] initially increase, then decrease for the two highest concentrations of ASD and TST. This is possibly due to aggregation at the higher concentrations of these ligands, reducing their effective concentration for binding to the enzyme. Another possibility is that they accumulate within and disrupt the integrity of the ND bilayer, somehow attenuating access to the 19A1 active site. Regardless of the physical basis of these observations, such scenarios are not accounted for in minimal kinetic models; hence we have excluded data obtained with highest concentrations of ASD and TST from the global fits.

Figure 2.

Double-exponential fitting of stopped-flow UV-vis data obtained with ASD (A), TST (B), and 7HF (C) and 19A1-NDs. ASD concentrations in (A) were 0.25 μM (dark blue), 0.5 μM (orange), 1 μM (yellow), 2 μM (purple), 4 μM (green), and 8 μM (light blue). TST concentrations in (B) were 0.5 μM (dark blue), 1 μM (orange), 2 μM (yellow), 4 μM (purple), 6 μM (green), and 10 μM (light blue). 7HF concentrations in (C) were 5 μM (dark blue), 10 μM (orange), 15 μM (yellow), 20 μM (purple), and 30 μM (green).

Global fits of the stopped-flow data are illustrated in Figures 3–5 and the rate constants summarized in Table 1. In Figure 3B–D, fits to the ASD data are not superb, possibly due to the onset of aggregation even at the low concentrations used. Nevertheless, the fit to the CS model (χ²= 15.9) is superior to that for IF model (χ²= 18.6). Notably, despite the greater flexibility of the combined CS+IF model (six vs. four rate constants), the addition of the IF step did not improve the quality of the fit (χ²= 18.6). Global fitting of the TST data are substantially improved (Figure 4B–D). The fit to the CS model (χ²= 4.7) is slightly better than that for IF (χ²= 5.0). While the quality of the CS and IF fits are similar, it should be noted that the values of k_off and k₁ in the IF fit were at the upper bound of what was considered physically realistic (100 s⁻¹). Combining CS with IF resulted in a fit that was significantly improved (χ²= 2.7) compared to either of the pure mechanisms with reasonable rate constants. Hence it is impossible to exclude the possibility of a combined mechanism in the binding of TST by 19A1. The global fit of the data obtained with 7HF to the CS model (χ²= 2.7) is visibly superior to IF (χ²= 11.3). (Figure 4B, C). The additional flexibility offered by the combined model does not improve the quality of the fit. (Figure 4D). In summary, global fitting to supports the possibility that CS is present in ligand binding to 19A1; however, the possibility that a combined mechanism is operative for TST cannot be excluded.

Table 1.

Global-fitted rate constants to the IF, CS, and CS+IF models.

Model	kon (M−1 s−1)	koff (s−1)	k1 (s−1)	k−1 (s−1)	k2 (s−1)	k−2 (s−1)
ASD
IF	1.1×107	100	100	2.9
CS	7.8×106	1.9	3.7	0.9
CS+IF	1.1×107	100	67	0	100	2.9
TST
IF	3.2×107	100	7.0	0.5
CS	2.0×106	0.3	0.3	0.05
CS+IF	2.2×107	100	0.75	0.13	12.9	0.5
7HF
IF	2.3×106	100	89	6.4
CS	2.2×106	5.4	4.8	2.4
CS+IF	3.6×106	100	5.7	1.7	100	8.0

Figure 4.

Parameters from exponential fits in Figure 2B and global fitting of stopped flow data for TST binding to 19A1-NDs. k_obs (s⁻¹) (A) for the fast (blue) and slow (orange) phases of the exponential fits. Reported values are the mean ± standard deviation of at least three stopped-flow experiments. Inset in panel (A) are raw spectra of collected every 2 ms for 5.9 s with 4 μM TST and 1 μM 19A1-NDs. (B) Global fitting to an IF model (Eq. 3). (C) Global fitting to a CS model. (Eq. 2); (D) Global fitting to a CS-IF model (Eq. 5). In B-C, concentrations of TST after mixing are 0.5 μM (blue), 1 μM (orange), 2 μM (yellow), 4 μM (purple). Global fitting results are depicted with solid black lines. An extinction coefficient ε^{388nm−420nm} = 0.057 μM⁻¹ cm⁻¹ was used to calculate the concentration change from difference spectra.

As described previously, in CS the rate of interconversion between ensembles is invariant to the ligand. However, global fits to the CS models indeed resulted in differing values for k₁ and k₋₁. This is attributable to the covariance of the parameters as well as the simplicity of the models that, at best, capture only the most fundamental elements of the binding mechanism. To illustrate the resilience of our mechanistic assignment to variations in the rate constants for conformational interconversion, we interchanged those from the best fits obtained for TST and 7HF. Holding k₁=4.8 s⁻¹ and k₋₁= 2.4 s⁻¹ constant while refitting the TST data to the combined CS and IF model resulted in the rate constants k_on = 3.0 × 10⁶ M⁻¹ s⁻¹, k_off = 5.2s⁻¹, k₂ = 100 s⁻¹, and k₋₂ =7.4 s⁻¹. Conversely, holding k₁ = 0.75 s⁻¹ and k₋₁ = 0.13 s⁻¹ constant and refitting the 7HF data to the CS model resulted in rate constants k_on = 1.2 × 10⁶ M⁻¹ s⁻¹, k_off = 3.0 s⁻¹. In the case of TST, this restrained fitting resulted in a χ² = 4.5, albeit narrowly, is still superior to the results obtained by free fitting both the pure CS and IF models. For 7HF, restraining the CS fit resulted in χ² = 9.0, that while higher than that of the free fit to CS, is superior to that obtained for pure IF. In summary, the conclusion that CS plays a major role in ligand binding to 19A1 is resilient to the extent of the variation observed for k₁ and k₋₁.

Finally, expressions relating K_d to the set of rate constants for each model were derived (Equations 6–8) and used to calculate these values for comparison to the equilibrium binding experiments. These values are inset into panels B-D Figures 3–5. For ASD (0.27–0.30 μM) and TST (0.19–0.20 μM), the values are similar to those obtained from the equilibrium binding experiments (0.14 and 0.17 μM, respectively). Despite the excellent quality of the fit obtained with the 7HF data to the CS model, the K_d values calculated using the rate constants from this and other models were greater by an order than those measured in equilibrium titrations. A likely source of the discrepancy is the huge gap in the data collection time scales in the two experiments. In the stopped-flow experiment, the data is collected over a few seconds, while the equilibrium binding study could extend to tens of minutes or hours. While the simplified models applied herein may be sufficient for the events occurring on the short timescales, they do not capture slow, ligand induced conformational changes that result in the higher affinity measured in the equilibrium binding experiments.

4. Discussion

Distinguishing between CS and IF requires a model system free of confounding features. Membrane proteins present particular challenges because they are rarely monodisperse, with the resulting oligomers possibly having multiple available binding sites. Delineating the dominant mechanism in CYPs presents both challenges, as many are peripheral membrane proteins and several, especially those involved in xenobiotic metabolism, cooperatively bind multiple substrates. 19A1-NDs proved to be convenient model to circumvent these challenges. Binding isotherms for ASD, TST, and 7HF show that these ligands bind to a single specific site on 19A1 and incorporating the enzyme into NDs ensures that it remains monodisperse. Unfortunately, exponential fits of the stopped-flow data did not yield a flat or decreasing trends in k_obs with increasing [L] that would have been diagnostic for the presence of CS. (5–7) Our observations contrast those of two previous studies that surveyed several human CYPs that, while lacking any consideration of the membrane, demonstrated decreasing trends in k_obs vs. [L]. (28,29) Since we did not observe such a diagnostic trend, so we resorted to a more computationally intensive global fitting of the data to the coupled differential equations of the models. These fits favored the presence of CS in the binding of ASD, TST, and 7HF to membrane-associated 19A1.

It is generally accepted that ASD is the preferred substrate by 19A1 to TST. Enzyme kinetics studies of 19A1 purified from the human placenta have measured K_m values of 0.06 μM and 0.21 μM for ASD and TST respectively.(52) ASD’s K_d value for binding to 19A1-NDs aligns with this K_m and is very similar to the K_d measured for a soluble form of the human enzyme (0.13 μM).(49) However, the value we measure for TST is, within error, the same as that measured ASD. While equilibrium binding affinities do not offer any insight into the substrate preference, there is a notable difference in the kinetics of ASD and TST binding. Because of the poor global fit obtained for ASD, it is unreliable to compare on and off rates for the two ligands. However, based on differences in the k_obs from the fast phase of the exponential fit, ASD binding is ~10-fold faster than TST. Estrogens produced by 19A1 are critical for the differentiation, regulation, and normal physiology in a variety of tissues such as ovaries, bone, brain, testes, ovaries, and prostate, breast, and adipose.(53) The biosynthetic precursor, ASD, is converted to TST by 17β-hydroxysteroid dehydrogenase. Hence, both substrates are expected to be present in the aforementioned tissues to compete for aromatization by 19A1. However, it is unknown whether the apparent order of magnitude difference in the binding rate between ASD and TST is physiologically relevant.

The difference spectrum observed for the 7HF complex is unusual and indicates that it exploits unique interactions within the 19A1 active site. To our knowledge, such a difference spectrum has only been observed before in CYP46A1 when binding 24S-hydroxycholesterol and bicalutamide.(54) In the crystal structure it was clear that bicalutamide did not displace the Fe³⁺-coordinated water. Instead, the bicalutamide nitrile accepted an apparent hydrogen bond from the heme-coordinated water. This interaction likely tunes the strength of the Fe³⁺-OH₂ bond relative to that of the ligand free enzyme, giving rise to the distinctly blue-shifted Soret band. The difference spectrum observed for the 7HF-19A1 complex has identical features to those observed in CYP46A1. We posit that 7HF likewise does not displace the Fe³⁺-coordinated water from 19A1, and instead stabilizes Fe³⁺-OH₂ coordination through hydrogen-bonding to one of the three 7HF oxygen atoms.

Evidence for water-bridged heme-ligand complexes in CYPs is not limited to CYP46A1, rather there is precedence for this binding mode in other human drug-metabolizing and bacterial forms. Hyperfine sublevel correlation spectroscopy (HYSCORE) and electron-nuclear double resonance spectroscopy reveal that azole complexes of Mycobacterium tuberculosis CYP51B1 form a mixture of direct N-coordination as well as hydrogen-bonded and water-bridged complexes.(55) In the human drug-metabolizing enzymes CYP2C9 and CYP3A4, 1,2,3-triazole containing ligands were likewise shown to form water-bridged complexes using the same approach. In CYP2C9, while 4-(3-phenylpropyl)-1H-1,2,3-triazole induces a “type II” UV-vis spectral signature consistent with direct coordination of a heterocyclic N atom to the heme iron; HYSCORE reveals that the axial water molecule is maintained and bridges the interaction with the 1,2,3-triazole functionality.(56) A similar observation was made with a 1,2,3-triazole analog of ethinyl estradiol (EE) and CYP3A4.(57) In this case, the 1,2,3-triazole-EE analog was turned over at a similar rate to EE, albeit with different regioselectivity. This observation supports that such water-bridged complexes can be catalytically competent. Whether a water-bridged complex of 7HF in the active site of 19A1 could be subject to catalytic turnover is intriguing and warrants further study.

5. Conclusions

In summary, we describe both equilibrium and stopped-flow UV-vis studies of ASD, TST, and 7HF to 19A1-NDs. The two substrates bind to 19A1 with comparable affinity; however, ASD was observed to bind faster to 19A1 by an order of magnitude. Kinetic parameters derived from exponential fitting of the stopped flow data were ambiguous to whether CS or IF better described the binding mechanism. However, global fitting of the stopped-flow data supported the presence of CS in the binding mechanisms of these ligands. Specifically, binding of both ASD and 7HF to 19A1 was better described by CS alone, while a mechanism combining both CS and IF was superior for TST. In evaluating the binding mechanism for 7HF, this ligand was found to induce an unusual blue-shifted difference spectrum akin to that observed for the CYP46A1-ligand complexes, suggesting that the unusual spectrum arises from a water-bridged 7HF-heme complex.

Synopsis.

Cytochrome P450 19A1, also known as aromatase, catalyzes the rate determining step in the biosynthesis of estrogens from androgens. Using aromatase in nanodiscs, stopped-flow UV-vis spectroscopy and global fitting supports the presence of conformational selection in the mechanism of ligand binding.

Highlights (Limited to 85 characters with spaces).

• CYP19A1-ligand interactions were evaluated by stopped-flow UV-vis spectroscopy.

• Global fits support that conformational selection is present in the binding mechanisms.

• 7-hydroxyflavone induces an unusual spectral shift when bound to aromatase.

Abbreviations

ASD

Androstenedione

CS

Conformational Selection

EE

Ethinyl estradiol

7HF

7-Hydroxyflavone

HYSCORE

Hyperfine sublevel correlation spectroscopy

IF

Induced Fit3

ME

β-Mercaptoethanol

ND

Nanodisc

OS

One reversible step

SVD

Singular-Value Decomposition

TCEP

Tris(carboxyethyl) phosphine

TST

Testosterone