Multiple Drug Binding Modes in Mycobacterium tuberculosis CYP51B1

Abstract

The Mycobacterium tuberculosis (Mtb) genome encodes 20 different cytochrome P450 enzymes (CYPs), many of which serve essential biosynthetic roles. CYP51B1, the Mtb version of eukaryotic sterol demethylase, remains a potential therapeutic target. The binding of three drug fragments containing nitrogen heterocycles to CYP51B1 is studied here by continuous wave (CW) and pulsed electron paramagnetic resonance (EPR) techniques to determine how each drug fragment binds to the heme active-site. All three drug fragments form a mixture of complexes, some of which retain the axial water ligand from the resting state. Hyperfine sublevel correlation spectroscopy (HYSCORE) and electron-nuclear double resonance spectroscopy (ENDOR) observe protons of the axial water and on the drug fragments that reveal drug binding modes. Binding in CYP51B1 is complicated by the presence of multiple binding modes that coexist in the same solution. These results aid our understanding of CYP-inhibitor interactions and will help guide future inhibitor design.

Graphical Abstract

1. Introduction

Mycobacterium tuberculosis (Mtb) is the causative agent of tuberculosis (TB), an infectious disease that has afflicted humans for centuries. TB remains a threat and a major cause of death today. Current treatments are complicated by the rise of resistant strains of Mtb, which has caused renewed urgency in TB research and drug development. A major recent advance in TB research was the sequencing of the Mtb genome, which revealed new potential drug targets. One of the most noteworthy aspects of the genome is that Mtb encodes 20 different cytochrome P450 enzymes (CYPs) [1]. CYPs are heme-containing monooxygenase enzymes responsible for essential biosynthetic and detoxification tasks in nearly all living organisms. CYP51B1 in Mtb was the first member of the CYP51 sterol demethylase family to be found in prokaryotes [1]. CYP51B1 catalyzes the oxidative demethylation of dihydrolanosterol and obtusifoliol, which is similar to the role and function of eukaryotic CYP51 [2, 3]. While a complete sterol biosynthesis pathway is not known in Mtb, CYP51B1 remains a potential drug target. The first effective drugs against Mtb, isoniazid, pyrazinamide, and rifampicin, came out in the 1950s [4]. These remained the most effective drugs for decades and are still among the few treatment options available. A robust understanding of how drugs bind to and inhibit CYPs will help to target Mtb CYPs in the development of new and effective treatments.

CYPs contain an active site comprised of a ferric, low-spin heme coordinated to a conserved cysteine residue (Figure 1, left). The sixth axial ligand present in the resting state is a water molecule, but it is assumed to be either displaced or replaced upon drug binding. In the traditional drug-binding paradigm, nitrogen-containing heterocycles are considered to act as CYP inhibitors because they replace the axial water by direct coordination to the heme (Figure 1, middle). These complexes are typically thought to “trap” the heme in the low-spin state, thus preventing the one electron reduction that begins the catalytic cycle. Substrates, on the other hand, displace the axial water without replacing it, which leaves the heme in a high-spin state that is more easily reduced and serves as the entry point to the catalytic cycle [5]. In some cases, the drug does not fit into the traditional substrate/inhibitor paradigm, and instead of displacing or replacing the axial water, it interacts with the heme via a hydrogen bonding network (Figure 1, right). This complex, which we refer to as water-bridged, is only observed in a few crystal structures and is considered rare [6–10]. The water-bridged binding mode is not necessarily associated with substrates or inhibitors; in fact, it has been observed with both in a variety of CYP isoforms [11]. While the water-bridged complex is still low-spin, it is not necessarily catalytically inactive. For example, in CYP3A4, the major human isoform responsible for drug metabolism, we found that a potential estradiol-based drug, 17α-(2H-2,3,4-triazolyl)-estradiol (17-click), formed a water-bridged complex that was still catalytically active, producing a library of altered drug products [12].

Figure 1:

Comparison of low-spin CYP binding modes. Left: resting state enzyme; middle: PPT has replaced the water as the axial ligand, resulting in a directly-coordinated complex; right: PPT forms a hydrogen bond to the axial water, resulting in a water-bridged complex.

In the case of Mtb CYPs, a few crystal structures show drugs, substrates, and substrate analogs interacting with the heme through active site waters [7–10, 13–15]. Recently, a water-bridged complex was identified in the Mtb CYP121 with the substrate cyclodityrosine (cYY) [9]. The H-bond network involves two ordered water molecules which are present in other CYP121 ligand bound crystal structures, and the authors suggest that such H-bond networks could position substrates or drugs in the active site for metabolism [9]. The drug fluconazole does not always displace the axial water in Mtb CYP121; sometimes, it forms a hydrogen-bonding network with the heme [8]. A water-bridged complex of Mtb CYP125A1 with the pyridine-based inhibitor α-[(4-methylcyclohexyl)carbonyl amino]-N-4-pyridinyl-1H-indole-3-propanamide (LP10; structure in section S1) has also been observed and studied using electron paramagnetic resonance (EPR) and visible and near-infrared magnetic circular dichroism (MCD) spectroscopy [5].

A few water-bridged complexes have been characterized in Mtb CYPs, but the extent to which these complexes exist in other isoforms and with other drugs is not known. We have found that water-bridged complexes oftentimes coexist in frozen solution with directly-coordinated complexes for several CYPs, producing a mixture of binding modes for the same drug that are not captured in the crystal structure databases [11]. This mixture of binding modes makes the question of their catalytic competency important because water-bridged complexes have been observed with both inhibitors and substrates. Mixtures of binding modes are common in the human CYPs we have studied, and this study suggests that a similar mixture of binding modes in solution may exist with CYP51B1 with nitrogen heterocycles that would traditionally be expected to replace the axial water and directly coordinate to the heme.

Here, we use EPR to characterize the binding of three nitrogen-containing heterocycles that traditionally would be considered inhibitor-like ligands to CYP51B1. These three compounds will be referred to as “drugs” henceforth for convenience. Continuous wave (CW) EPR and hyperfine sublevel correlation spectroscopy (HYSCORE) are used to directly probe the protons on the axial water and distinguish drugs that replace the axial water and are directly-coordinated from those that form a water-bridged complex [11]. In addition, we use electron-nuclear double resonance (ENDOR) spectroscopy to show that the drugs occupy the active site and are in close proximity to the heme. All three drugs bind as a mixture of water-bridged and directly-coordinated complexes. The EPR and HYSCORE spectra both show that the axial water is perturbed and reoriented with respect to the heme in the water-bridged complexes. This study suggests that water-bridged complexes may not be rare in Mtb CYPs, and supports the suggestion from CYP121 that the axial water can play a structural role in drug binding [9]. These multiple complexes and their possible individual activities have implications for drug design and need to be considered in the design of effective inhibitors for CYP51B1 and the other Mtb CYP isoforms.

2. Materials and Methods

2.1. Protein expression and purification

CYP51B1 was expressed and purified as previously described [2]. Purified protein was dialyzed against 10 mM Tris-HCl (pH 7.5) buffer and concentrated to 324 μM. EPR samples consisted of 50 μL protein in buffer with 20 % glycerol added (v/v) as a cryoprotectant. The compounds in this study, 1,2,3-triazole (1,2,3-TRZ), 17α-(2H-2,3,4-triazolyl)-estradiol (17-click), and 4-(3-phenylpropyl)-1H-1,2,3-triazole (PPT), were chosen because they were found to produce a type II optical difference spectrum typical of many existing azole drugs [5, 12, 16]. The structure of each of these drugs is shown in section S1. Drug concentrations for the EPR samples were nominally saturating at room temperature (10 × K_D). Samples were made, transferred to 3 mm quartz EPR tubes, and immediately frozen and stored at 77 K in liquid nitrogen.

2.2. Continuous Wave (CW) EPR measurements and simulations

CW EPR measurements were made on a Bruker ELEXSYS E540 X-band spectrometer with an ER 4102 ST resonator and a quartz liquid nitrogen insertion dewar. Spectra were recorded at 77 K with a nominal microwave frequency of 9.45 GHz, a modulation amplitude and frequency of 5.0 G or 10.0 G and 100 kHz, respectively, and a microwave power of 3.34 mW or 6.64 mW. CW EPR simulations were made using the EasySpin toolbox in MATLAB (Mathworks, R2018a) [17]. Simulations included g-values, g-strains, and weights for each component in the EPR spectrum. The g-value in EPR describes the peak position with respect to the microwave frequency and magnetic field, the g-strains are related to line widths and account for Gaussian distributions of the g-values, and the weights describe each spectral component’s relative contribution to the overall spectrum [17]. With the exception of the drug-free spectrum, each spectrum is the sum of several overlapping spectra. In these cases, the overall spectrum was simulated with one component using the drug-free parameters first, and additional components were added as necessary to match the experimental spectrum. In some cases, it appears that some drug-free enzyme remained after drug binding; these components have the same g-values as the drug-free spectrum and are identified as “resting state” in binding assignments.

2.3. Pulsed EPR measurements and analysis

CW EPR spectra were simulated for the HYSCORE frequency of 9.76 GHz and plotted as the absorbance rather than the first derivative to aid in data analysis, as done in previous studies [5, 11]. The g_z axis is taken to be the closest to the z axis of the heme, perpendicular to its molecular plane, using the convention where g_z > g_y > g_x. Pulsed EPR measurements were made at a nominal EPR frequency of 9.76 GHz. HYSCORE and ENDOR pulsed EPR measurements were made at 10–20 K using an ELEXSYS E680 EPR spectrometer (Bruker-Biospin, Billerica, MA) equipped with a Bruker Flexline ER 4118 CF cryostat and an ER 4118X-MD4 ENDOR resonator. HYSCORE measurements used a four-pulse sequence, π/2−τ−π/2−t1−π−t2−π/2−τ−echo. This sequence was repeated at a rate of 2 kHz with values of 16 ns and 32 ns for the π/2 and π pulses, respectively. The times t1 and t2 were varied independently, and the delay time τ was set to 240 ns to give the best signal in the proton region of the HYSCORE spectrum. ENDOR measurements used the Mims ENDOR pulse sequence, π/2−τ−π/ 2−T−π/2−τ−echo, with a 9 μs rf π pulse applied during the delay time T [18]. The delay time τ was set at 240 ns for all measurements because it resulted in the best resolution in the ¹H region of the spectrum. Mims ENDOR spectra were collected at magnetic fields spanning the g_z region for all samples.

HYSCORE and ENDOR spectra were processed using custom scripts in MATLAB (Mathworks, 2018a) and scripts written in the Python programming language. HYSCORE spectra were simulated using the EasySpin toolbox in MATLAB [17].

2.4. Analysis of hyperfine parameters

HYSCORE spectra from several different fields across the g_z region were used to determine the hyperfine interaction for the axial water protons. Eight points were selected from the HYSCORE spectrum at each magnetic field, along the ridge forming each cross-peak on the contour plot. The points yielded a curved arc, which were mapped onto a smooth curve by scaling all the frequencies to a common nuclear Zeeman frequency (v_i) of 12.5 MHz [19]. This frequency corresponds to the proton frequency at a magnetic field of 293.5 mT, near the g_z maximum in the samples. The frequency-normalized points were fit by the equation.

$${v_{\alpha (\beta )}}^2=Q_{\alpha (\beta )}{v}_{\beta (\alpha )}^2+G_{\alpha (\beta )}$$ Eq. 1

where Q_α(β) and G_α(β) can be described as:

$$Q_{\alpha (\beta )}=\left(T+2a-(+)4v_i\right)/\left(T+2a+(-)4v_i\right)$$ Eq. 2

$$G_{\alpha (\beta )}=+(-)2v_i\left(4v_i^2-a^2+2T^2-aT\right)/\left(T+2a+(-)4v_i\right)$$ Eq. 3

The v_α and v_β terms correspond to ENDOR frequencies taken from the HYSCORE arcs of protons in the α and β electron spin manifolds, respectively. [19–21]. This approach determines the hyperfine tensor from second-order shifts in the hyperfine spectra. The isotropic (a) and anisotropic (T) components of the hyperfine interaction were determined by fitting the curve in OriginLab using Eqs. 1–3. The fits closely match the frequency-normalized points for each drug combination. The anisotropic contribution of the hyperfine interaction, T, represents the through-space interaction between the magnetic moment of the surrounding nucleus and the unpaired electron. In the case of low-spin ferric heme, this interaction behaves like a point-dipole interaction because the unpaired electron density is largely localized in a spherical region around the iron and does not overlap with the water or drug protons [22, 23]. The point-dipole interaction (Eq. 4) has been used to calculate proton-heme distances in other CYPs [5, 22–25]. From the anisotropic contribution to the hyperfine interaction, T, we calculate the distance between the iron of the heme and the protons on the axial water as:

$$T=\left(\frac{{\mu }_0}{4\pi }\right)\frac{g_e{g}_n{\beta }_e{\beta }_n}{h{r}^3},$$ Eq. 4

where μ₀ is the permittivity of free space, g_e is the electronic g-value, g_n is the nuclear g-value, β_e is the electronic Bohr magneton, β_n is the nuclear Bohr magneton, h is Planck’s constant, and r is the distance in nm from the proton to the electron of the heme. This is a good approximation to the more exact procedure reviewed by Hutchison and Makinen and Mustafi [26, 27].

2.5. HYSCORE simulations

HYSCORE spectra were simulated using EasySpin, a toolbox in MATLAB (MathWorks, R2018b) [17]. Simulations include the g-values, the magnetic field, the microwave frequency, the delay time τ, the proton hyperfine coupling tensor, and the Euler angles α and β, which correspond to the angles φ and θ, relating the water protons to the x and y plane and the z axis of the heme, respectively. The proton hyperfine coupling tensor, A (Eq. 5), was calculated as the sum of the isotropic and anisotropic components of the hyperfine interaction derived from HYSCORE arc analysis according to equation

$$A=[a-T,a-T,a+(2*T\left)\right]$$ Eq. 5

where A is the hyperfine tensor, a is the isotropic hyperfine interaction, and T is the anisotropic contribution of the hyperfine interaction. Simulations at different magnetic fields spanning the experimental range were required to determine φ and θ.

2.6. ENDOR spectroscopy

Mims ENDOR spectra at various fields spanning g_z were processed using custom scripts in the Python programming language and Origin (OriginPro, 2018b). The drug-free spectrum at each magnetic field was subtracted from the drug-bound spectrum at the same field, and the resulting spectra were normalized to and centered at the nuclear Zeeman frequency for protons at each magnetic field. Spectra were symmetrized around the nuclear Zeeman frequency by plotting the data as the difference in absolute ENDOR effect versus the nuclear Zeeman frequency and the reflection of the difference spectrum about the nuclear Zeeman frequency versus the nuclear Zeeman frequency. In the case of weakly-coupled nuclei, such as the protons in the CYP active site, ENDOR peaks appear at

$$v_{ENDOR}=v_I\pm \frac{A}{2}$$ Eq. 6

where v_I is the nuclear Zeeman frequency and A is the hyperfine coupling. The hyperfine coupling is composed of both isotropic, a, and anisotropic, T, contributions. The isotropic component, a, arises from the delocalization of unpaired electron spin density from the heme onto nuclei of other molecules and is negligible for molecules not directly-coordinated to the heme iron carrying the unpaired spin [27]. For protons in water or drugs that are not ligands of the heme, a ≈ 0 and the hyperfine coupling, A, can be approximated using only the anisotropic component, T [27]. The point-dipole approximation (Eq. 4) can be used to estimate distance constraints for these proton ENDOR peaks. An ENDOR peak can only have a shift from v_I that is ≤ 2T_⊥ or $\le \frac{79.064kHzn{m}^3}{r^3}$, meaning that $r\le {\left(\frac{79.064kHz}{2\ast \left|v_{ENDOR}-v_{I(kHz)}\right|}\right)}^{\frac{1}{3}}nm$. In this way, we are able to set a limit on the distance from the heme iron to a proton that appears or disappears from the ENDOR spectrum upon drug binding. A proton that merely moves in the active site will have a negative and a positive lobe in the ENDOR spectrum corresponding to its original and its final positions. The distances were compared with published crystal structures using Chimera 1.12 (UCSF, San Francisco, CA) [28].

3. Results

3.1. CW EPR

CW EPR spectra of the low-spin, ferric heme in CYPs contain three characteristic peaks, which are denoted g_x, g_y, and g_z and correspond to the physical axes of the heme. Shifts in the g_z peak occur when the drug binds directly to the heme or through a water-bridge [11]. Figure 2 shows the CW EPR spectra of CYP51B1 with no drug and with 1,2,3-TRZ, with 17-click, or PPT added. Table 1 shows the best set of spectral fit parameters for each component [17]. The fit parameters include g-values, g-strains, (which are related to the line widths), and the relative weights of each component. The CW EPR spectrum of CYP51B1 with no drug added can be fit as a single spectral component with g-values similar to those of other resting state P450 isoforms [11].

Figure 2:

CW EPR Spectra of CYP51B1 with no drug added (solid), and with PPT (dashes), 17-click (short dots), and 1,2,3-TRZ (dots) added. The g_x, g_y, and g_z regions of the spectra are labeled.

Table 1:

CW EPR simulation parameters and binding mode assignments for CYP51B1 with no drug, PPT, 17-click, and 1,2,3-TRZ added.

Sample	gx	gy	gz	gx strain	gy strain	gz strain	weight (%)	binding assignment
No drug complex 1	1.913	2.258	2.435	0.023	0.021	0.053	100	resting state
PPT complex 1	1.913	2.258	2.435	0.020	0.033	0.050	38.0	resting state
PPT complex 2	1.924	2.252	2.407	0.015	0.017	0.031	44.4	WB
PPT complex 3	1.889	2.261	2.493	0.046	0.014	0.088	17.6	DC
17-click complex 1	1.913	2.258	2.435	0.020	0.017	0.045	31.4	resting state
17-click complex 2	1.881	2.270	2.493	0.039	0.023	0.066	19.9	DC
17-click complex 3	1.924	2.250	2.407	0.015	0.018	0.030	48.7	WB
1,2,3-TRZ complex 1	1.911	2.256	2.430	0.019	0.019	0.043	22.6	WB
1,2,3-TRZ complex 2	1.887	2.268	2.467	0.037	0.035	0.070	37.9	DC
1,2,3-TRZ complex 3	1.867	2.273	2.509	0.051	0.021	0.103	39.5	DC

The addition of each drug changes the spectrum significantly. Each of the three drug-bound spectra can only be fit well by a mixture of two or more different spectral components, each representing a different active site conformation. The spectra of CYP51B1 bound with PPT or 17-click are similar. In both cases, one spectral component has g-values identical to the resting state enzyme. This spectral component, labeled complex 1 for each sample in Table 1, is the residual resting state enzyme with no drug perturbing the active site. For PPT, the residual resting state component comprises 38.0 % of the overall CW EPR signal. The other two components in each sample, labeled complexes 2 and 3 in Table 1, represent drug-bound complexes with significant g-value shifts. The second PPT-CYP51B1 component has a smaller g_z value, which is indicative of a complex where the axial water has not been replaced, suggesting a water-bridged complex representing 44.4 % of the overall CW EPR signal of the PPT-bound enzyme. The third simulated PPT-CYP51B1 spectral component has an increased g_z value that is indicative of the axial water being replaced, resulting in a directly-coordinated complex that contributes 17.6 % of the overall CW EPR signal.

The spectrum of 17-click bound to CYP51B1 is similarly heterogeneous: a mixture of residual resting state enzyme (31.4 % of the overall signal), a spectral component with a decreased g_z value (48.7 %), and a third spectral component with an increased g_z value (19.9 %). Overall, the 17-click-CYP51B1 complex exhibits an increased percentage of directly-ligated complex, which is concomitant with a decrease in both water-bridged and ligand free enzyme structures relative to the equivalent CYP51B1-PPT ensemble at concentrations 10 × K_D. Despite these notable differences between the resultant PPT and 17-click CYP51B1 complexes, both samples highlight an important phenomenon: both are a mixture of directly-coordinated and water-bridged CYP51B1 complexes based on shifts in g_z values.

The spectrum of CYP51B1 with 1,2,3-TRZ is also comprised of three spectral components. The first component makes up 22.6 % of the spectrum and has g-values that are barely different from those of the resting state enzyme. To distinguish whether it is resting enzyme or drug-bound enzyme, the hyperfine coupling of the water protons in the drug-bound complex was compared with that of the resting state. The hyperfine coupling should remain unchanged if the enzyme remains in the resting state.

The other two complexes have larger g_z values, indicating that both are complexes where the axial water has been replaced by drug. While a directly-coordinated nitrogen heterocycle is typically considered inhibitory, the presence of two distinct directly-coordinated components is interesting. One component makes up 37.9 % of the spectrum, while the other makes up 39.5 %. These two components could arise from two different orientations of the drug in the active site or two different conformations of the protein.

3.2. HYSCORE measurements and binding assignments

Shifts in g_z reveal whether a drug is directly-coordinated or water-bridged, but HYSCORE spectra can independently confirm whether the axial water is absent [11]. HYSCORE measurements are ideal because they directly observe the ¹H of the axial water. HYSCORE spectra are roughly symmetric around the diagonal from the lower left to the upper right corner of the plots. Most protons, such as those on the heme and nearby residues, appear along the HYSCORE diagonal in the “matrix” proton peak at the ¹H ENDOR frequency (small peak on the diagonal at ~12.6 MHz in Figure 3). Water peaks, however, are resolved from all other protons by first and second-order shifts caused by the large, anisotropic hyperfine interactions between the water protons and the heme [25]. These water peaks are ideal monitors of CYP-drug binding; if a drug displaces the water, then the water peaks disappear, whereas if a drug binds via a water-bridge, the peaks remain but shift slightly. HYSCORE can also resolve alternate ligand orientations in low-spin complexes based on perturbations of the axial water proton signals. [12, 22, 29].

Figure 3:

HYSCORE spectrum of resting state CYP51B1 with no drug added at 295.5 mT.

Figure 3 shows the HYSCORE spectrum of the drug-free enzyme at 295.5 mT. The spectrum contains water proton peaks seen as an intense pair of ridges or arcs on the upper right with coordinates near (11, 15) and (15, 11) MHz. A set of smaller ridges shifted slightly to the lower left near (12, 13) and (13, 12) MHz correspond to β-protons on the proximal cysteine ligand. These cysteine proton peaks act as an internal intensity standard for the HYSCORE spectra. When the cysteine peaks are seen, the water protons should also be observable, because they have comparable intensities. If the water peaks disappear when a drug binds but the cysteine proton peaks remain, then the drug binding mode is direct coordination. If the drug does not displace the water then both sets of peaks remain, and the drug binding mode is water-bridged.

Figure 4 shows how the binding modes of multiple complexes in the same sample can be determined for CYP51B1 bound to PPT. In brief, simulated CW spectra are plotted as the absorbance spectra rather than as first-derivative spectra to best view spectral component overlap. HYSCORE measurements take advantage of a phenomenon called orientation selection where paramagnetic centers with different g-values are observed at different magnetic fields. Therefore, HYSCORE measurements made at a magnetic field where only one spectral component contributes to the overall spectrum will only see the binding of that component. For example, with PPT bound to CYP51B1, the HYSCORE measurement at 280.5 mT, Figure 5, only sees component 3. Here, no water peaks are seen, but the cysteine peaks are clear. Therefore, this spectral component is a directly-coordinated complex. HYSCORE measurements at higher magnetic fields also include contributions from components 1 and 2, and the axial water proton peaks appear in the spectra. Component 1 in this case has the same g-values as the resting state enzyme, but component 2 does not, yet the spectrum still has intense water proton peaks. This component is therefore assigned as a water-bridged complex. The binding modes of all complexes could be assigned in this manner and are listed in Table 1. The spectra used to assign each binding mode can be found in section S3.

Figure 4:

CW EPR spectrum (solid line) and simulation (dotted line) of CYP51B1 bound to PPT. Bottom: the first-derivative spectrum and simulation consisting of three spectral components: component 1 (dashes), component 2 (short dots), and component 3 (short dashes). Top: The simulated spectral components plotted as absorbance rather than first-derivative to better visualize their overlap.

Figure 5:

HYSCORE spectra of CYP51B1 with PPT at 280.5 mT (left) showing only the cysteine ¹H and 297.0 mT (right) showing the cysteine ¹H and the axial water ¹H.

3.3. HYSCORE simulations

The water-bridged complexes make up a significant portion of the total drug-bound complexes, so we used pulsed EPR to determine whether the drugs perturb the position of the axial water ¹H relative to the heme. Changes in the distance or orientation of the water protons upon drug binding would support a hydrogen-bonding network, whereas no change might indicate that the drug does not interact directly with the axial water and might be located elsewhere in the active site or at an allosteric site. HYSCORE measurements were taken at magnetic fields near the maximum of g_z, where the water proton peaks are best resolved. Except for CYP51B1 with no drug, all samples contain at least one directly-coordinated complex where the water proton peaks disappear from the HYSCORE spectra (see all spectra in section S3). These directly-coordinated complexes appear at a lower magnetic field in the CW spectrum. The top of the axial water ridges in the HYSCORE spectra at different magnetic fields were measured and plotted as described in the Materials and Methods. Figure 6 shows the fit (solid line, Eqs. 1–3) of points chosen across the HYSCORE arcs from all the HYSCORE spectra, and Table 2 shows the isotropic and anisotropic contributions to the hyperfine interaction and the distance from the water protons to the heme iron.

Figure 6:

HYSCORE peak arc fits for each sample. Each plot represents frequency coordinates of points chosen along water proton HYSCORE arcs at several magnetic fields up to 300.0 mT. The points at each field are shifted to where they would appear at a common magnetic field of 293.5 mT and are fit using Eqs. 1–3 (solid line) to determine the hyperfine coupling parameters and the distance from the water protons to the heme.

Table 2:

Hyperfine interaction parameters, distance, and orientation of the axial water protons for the resting state and for the water-bridged complexes.

Sample	aiso (MHz)	T (MHz)	r (Å)	θ (°)	φ (°)
No drug (resting state)	−1.46 ± .12	5.16 ± .03	2.483 ± .005	24	49
PPT complex 2	−1.83 ± .07	5.17 ± .02	2.482 ± .003	21	42
17-click complex 3	−1.60 ± .10	5.21 ± .02	2.475 ± .003	21	41
1,2,3-TRZ complex 1	−0.72 ± .17	5.04 ± .04	2.503 ± .006	24	38

Using this information, the spectra were simulated in EasySpin to determine the orientation of water protons relative to the heme from the Euler angles α and β, in the x and y plane and the z axis, respectively (Figure 7).

Figure 7:

The angles φ and θ with respect to the active site heme.

Figures 8, 9, 10, and 11 show HYSCORE spectra and simulations of CYP51B1 with no drug, 17-click, PPT, and 1,2,3-TRZ, respectively. The changes in anisotropic hyperfine coupling and the distances from the heme to the axial water protons between CYP51B1 with no drug added and CYP51B1 with either PPT or 17-click are not large. The orientation of the axial water protons and the isotropic hyperfine coupling, however, are perturbed slightly. In both the PPT and 17-click-bound complexes, the angle θ between the axial water proton and the g_z axis of the heme decreases slightly. The changes are similar to those observed for the pyridine-based drug, LP10, with CYP125A1, another Mtb CYP isoform [25]. The angle φ indicates how the water is rotated around the Fe-O coordinated bond relative to the heme, and it changes by ~7 ° upon the binding of PPT or 17-click. These water-bridged complexes have significantly different g-values than the resting state, and small changes in the orientation of the axial water. This is similar to what has been observed with water-bridged complexes in CYP2C9d and CYP125A1. In those complexes the axial water was only slightly perturbed upon drug binding, but binding of ¹⁵N labeled PPT to CYP2C9d showed that the drug was in the active site with the nitrogen label only 4.44 Å from the heme iron [5].

Figure 8:

HYSCORE simulations of resting state CYP51B1 with no drug added at 288.0 mT, 291.0 mT, 295.5 mT, and 300.0 mT. HYSCORE peaks are shown in color and simulations are shown as black contour lines.

Figure 9:

HYSCORE spectra and simulations of CYP51B1 in complex with PPT at 219.0 mT, 294.0 mT, 297.0 mT, and 300.0 mT. HYSCORE peaks are shown in color and simulations are shown as black contour lines.

Figure 10:

HYSCORE spectra and simulations of CYP51B1 with 17-click at 291.0 mT, 296.5 mT, 298.0 mT, and 300.0 mT. HYSCORE peaks are shown in color and simulations are shown as black contour lines.

Figure 11:

HYSCORE simulations of CYP51B1 with 1,2,3-TRZ at 290.0 mT, 294.0 mT, 297.0 mT, and 300.0 mT. HYSCORE peaks are shown in color and simulations are shown as black contour lines.

The wate-bridged complex of CYP51B1 and 1,2,3-TRZ has g-values that are very similar to the resting state enzyme; however, the hyperfine interaction between the heme and the axial water protons is significantly different. The isotropic and anisotropic contributions to the hyperfine interaction are both different from the drug-free enzyme, and the distance from the heme to the water protons is slightly longer. The angle θ does not change from that of the drug-free enzyme, indicating that the H-Fe-O angle remains constant, but φ decreases, indicating a rotation of the water about the Fe-O coordination bond relative to the heme. Most of the complexes formed when 1,2,3-TRZ binds to CYP51B1 are directly-coordinated complexes with the water replaced by the drug, but this minor complex retains the water and has a hyperfine coupling distinct from that of the drug-free enzyme. This altered hyperfine coupling could be due to hydrogen-bonding between a water proton and the drug. Alternatively, the drug might occupy the active site but not be close enough to the heme to replace the water, yet it still perturbs the water. In either case, having three distinct complexes is unusual and unexpected for a small nitrogen heterocycle.

3.4. ENDOR difference spectra

The changes in the orientation of the axial water in drug-free versus drug-bound spectra suggest that water-bridged drug complexes do interact with the axial water, which would be similar to observations in CYP121 with the substrate, cYY [9]. Yet, CYPs can have allosteric sites where binding might affect the heme and the axial water. We use electron-nuclear double resonance (ENDOR) spectroscopy to try to verify if the drugs are in the active site. ENDOR is a double resonance technique that can detect the hyperfine-shifted NMR spectrum of an individual proton near the heme [25, 30]. Subtraction of ENDOR spectra reveals changes in the position of nuclei near the paramagnetic center. In this work, the proton ENDOR spectrum of the drug-free complex is subtracted from the spectrum of the drug-bound complex, revealing positive peaks when protons are introduced into the active site by the drug. If the peaks are negative, they represent protons that were displaced by drug. A proton that has shifted from its original position but not been displaced would give an adjacent positive and negative peak.

Figures 12, 13, and 14 show Mims ENDOR difference spectra, i.e. drug-bound spectra minus drug-free spectra at various magnetic fields. Spectra are relative to the proton Zeeman frequency, v_I, and stacked according to magnetic field. The spectra are also symmetrized by flipping them around the proton Zeeman frequency and overlaying them. As described in the experimental section, the splitting between ENDOR peaks provide a limit on the distance from the iron for protons that are gained and lost upon drug binding. Distance constraints were calculated for peaks labeled in Figures 12, 13, and 14.

Figure 12:

Mims difference ENDOR spectra of CYP51B1 with 1,2,3-TRZ. Peak splittings used to calulate distances are labeled.

Figure 13:

Mims difference ENDOR spectra of CYP51B1 with 17-click. Peak splittings used to calulate distances are labeled.

Figure 14:

Mims difference ENDOR spectra of CYP51B1 with PPT. Peak splittings used to calulate distances are labeled.

The difference ENDOR spectra of the 1,2,3-TRZ bound complex (Figure 12) reveal several sets of positive peaks that become most intense near g_y. The set of peaks with the smallest hyperfine splitting, labeled a in Figure 12, come from a proton within 4.75 Å from the heme, whereas the positive peaks with the largest splitting, labeled b in Figure 12, are from a proton within 3.10 Å. These are sharp peaks, which implies that they are highly ordered. Crystal structures of other Mtb CYPs with inhibitors indicate protons at similar distances. For example, the CYP121-fluconazole structure (PDB file: 2IJ7), which has six CYP121 molecules per asymmetric unit, has both directly-coordinated and water-bridged complexes [8]. In the directly-coordinated complexes, protons on C3 and C5 of the 1,2,4-TRZ ring are between ~2.97 Å and ~3.76 Å away from the heme iron. This distance increases in the water-bridged fluconazole complexes where protons on C3 and C5 are between ~4.60 Å and ~5.92 Å away from the heme iron. A similar CYP-ligand proton distance was determined with HYSCORE for protons on an imidazole ring directly-coordinated to human CYP2C9 that were ~3.27 Å away from the heme [22]. The positive proton peaks from protons within 4.75 Å correspond well with drug proton positions of directly-coordinated and water-bridged CYP121-fluconazole complexes and directly-coordinated CYP2C9-imidazole complexes. These peaks are likely from protons on a highly ordered 1,2,3-TRZ ring in the active site very close to the heme.

There are also several sets of negative peaks in the difference spectrum for CYP51B1 with 1,2,3-TRZ. The smallest coupling for negative peaks, labeled c in Figure 12, corresponds to a proton within 4.95 Å. These peaks could represent protons from active site residues or from a water molecule that is replaced by drug binding. The peaks with the largest coupling appear at 288.0 mT, and perhaps 292.4 mT, and are labeled d in Figure 12. These peaks have a hyperfine coupling constant of ~6 MHz and are within 2.44 Å of the heme iron. These can only be the axial water protons [23, 25]. We also verified this assignment with ENDOR simulations of the axial water protons based on our HYSCORE analysis (Table 2). The simulated axial water proton spectra match the experimental ENDOR peaks (See section S4 for simulation); therefore, these negative peaks indicate the replacement of the axial water at this field. This confirms that 1,2,3-TRZ displaces the axial water in the directly-coordinated complexes. At higher magnetic fields, these ENDOR peaks broaden and disappear (see section S3 for all CW spectra).

Difference spectra of 17-click and PPT bound to CYP51B1 (Figure 13) also contain a few sets of positive ENDOR peaks. As in the case of CYP51B1 with 1,2,3-TRZ, these spectra are from three spectral components that represent a mixture of bound complexes. However, the directly-coordinated complexes fall at lower magnetic fields (see section S2 for CW EPR spectra) and do not contribute much to the overall signal in the ENDOR difference spectra. For CYP51B1 with 17-click, there are two sets of weak, positive ENDOR peaks, labeled a and b in Figure 13. Peaks a correspond to a distance ≤ 4.07 Å, and peaks b correspond to a distance ≤ 3.39 Å. CYP51B1 with PPT (Figure 14) has positive ENDOR peaks labeled a at 301.2 mT that correspond to a distance of ≤ 3.42 Å from the protons to the heme iron, and positive peaks labeled b, most apparent at 292.4 mT, correspond to a distance of ≤ 3.03 Å. Both drugs primarily form water-bridged complexes, so these peaks are likely from protons on the drugs. The EPR spectrum at these fields is dominated by water-bridged complexes and by residual resting state enzyme, the latter of which does not produce any peaks in the difference spectra.

There are also negative ENDOR peaks for both 17-click and PPT with CYP51B1. The peaks in the 17-click spectra (labeled c in Figure 13) correspond to a distance of ≤ 4.00 Å. For PPT, the peaks (labeled c in Figure 14), correspond to a distance ≤ 4.00 Å from the heme iron and dominate the ENDOR difference spectra from 292.4 mT to 301.2 mT. In both cases, these negative peaks could correspond to protons that are displaced by drug binding, either from a water molecule or from active site residues.

The ENDOR difference spectra for all CYP51B1-drug combinations reveal drug protons that are within 5 Å of the heme. Although specific protons cannot be assigned because the difference spectra represent a mixture of bound complexes, the distance constraints conclusively show large changes in the active site rather than subtle changes in conformation from binding of drug at an allosteric site that is far from the heme. This further supports our assignment of changes in the orientation of the axial water in water-bridged complexes to the presence of a drug in the active site rather than an allosteric interaction. This is in line with published crystal structures of water-bridged complexes as well as our previous work [5, 8, 9].

4. Discussion

4.1. Mixture of binding modes

First and foremost, this work shows that each CYP51B1-drug combination gives a mixture of bound complexes. Each mixture contains both directly-coordinated and water-bridged complexes. The complexity of CYP-drug binding revealed by EPR and associated spectroscopy methods demonstrates that more sensitive techniques are needed to resolve these species than is routinely provided by ensemble measurements like optical difference spectra, which are typically interpreted in terms of a single bound complex. A mixture of binding modes in frozen solution implies a similar mixture of binding modes under physiological conditions. Consequently, these mixtures, all of which contain water-bridged complexes, have significant implications for drug design, particularly in the design of inhibitors for Mtb CYPs. The metabolic activity of water-bridged drugs is not well understood, but enzymatically active water-bridged complexes are known [9, 12].

4.2. Water-bridged complexes in Mtb CYPs

Because water-bridged complexes appear in each CYP51B1-drug combination, it is possible that the complexes have a slightly modified resting state heme rather than a true drug-bound complex. Such would be the case if the drugs occupied an allosteric site far from the heme rather than the active site. However, the CW EPR parameters and hyperfine interactions between the axial water protons and the heme change with drug binding, which demonstrates a difference in both the electronic environment of the heme and in the orientation of the axial water. In addition, ENDOR difference spectra of water-bridged complexes place drug protons in the active site just a few angstroms from the heme. The drugs are positioned close enough to directly interact with the axial water, which argues strongly that they are not occupying an allosteric site.

There are a few examples in the crystallographic database of Mtb CYP121 complexes with substrates, substrate analogs, or inhibitors that do not replace the axial water and instead occupy the space above it in the active site. These complexes all have substrates or analogs that are between ~5–9.5 Å from the heme iron, which leaves plenty of room for the axial water and for additional solvent water molecules (PDB files: 3G5H, 1N4G, 4IQ7, and 5IBI) [10, 13–15]. In contrast, all drug-bound crystal structures of CYP51B1 have the inhibitor replacing the axial water to form a six-coordinate, low-spin heme. (PDF files: 1E9X and 2W09) [31, 32]. These results suggest that drug binding in CYP51B1 is similar to that of CYP121 and that the crystallographic database does not capture the full range of CYP51B1-drug interactions.

4.3. Role of the axial water

While most directly-coordinated nitrogen heterocycles are associated with CYP inhibition, water-bridged complexes are observed with both substrates and inhibitors [5, 9, 11, 12]. For example, numerous water-bridged complexes have been identified in a variety of human CYP-drug combinations [11]. In addition, catalytically active water-bridged complexes have been identified; namely, CYP3A4 with 17-click [12]. This suggests that the portion of CYP51B1–17 click complexes that are water-bridged might also retain some metabolic activity, which clouds the role of water-bridged complexes. One study with CYP121 suggests that the axial water serves a structural purpose and plays a role in positioning substrates for metabolism [9]. Such a role would help explain why water-bridged structures have been observed with CYP substrates, but it does not imply a similar conclusion for water-bridged inhibitors. These findings suggest that water-bridged complexes are just as prevalent with inhibitor-like ligands, at least in the Mtb proteome. This research warrants further study into the prevalence and activity of water-bridged inhibitors. Since Mtb CYPs are drug targets, the metabolic consequences of water-bridged complexes and mixtures of bound complexes need to be considered in inhibitor design.

4.4. Implications for drug design

Several CYP isoforms from various species are potential drug targets. A common design strategy to obtain CYP inhibitors includes incorporation of fragments, such as imidazoles or triazoles, into drugs, which are capable of heme ligation via formation of iron-nitrogen bonds. Data from the current results, and our previous work with several CYP isoforms, indicate that undecorated imidazoles or triazoles that lack additional functional groups are likely to form heterogeneous complexes where only a fraction of the bound complexes contain the ‘designed’ iron-nitrogen bond [5, 11, 12, 29]. The remainder of the population retains water as an axial ligand. In many cases, the water ligand hydrogen bonds to the imidazole or triazole fragment to form the water-bridged species we have characterized. More importantly, this heterogeneity is observed with drugs like 17-click and PPT that have substituents that provide additional interactions with active site residues and have increased affinity. Presumably, the heterogeneity observed in our EPR studies of frozen samples reflects rapidly equilibrating species at physiological temperatures. Therefore, even high affinity inhibitors that make multiple contacts with active site residues can adopt multiple orientations in the active site and yield a mixture of heme ligation states, including direct heme ligation, water-bridged nitrogen-heme interactions, or resting state, water-ligated heme with drug nearby. Although adding functional groups and hydrophobic bulk to azoles can result in higher affinity, this does not ensure homogenous ligation to the heme.

The details of this heterogeneous population are important for drug design because the functional properties of the water-bridged complexes are not yet known. In particular, the reduction potential of the water-bridged heme could allow for catalytic activity rather than complete inhibition. This is consistent with our previous observations with 17-click, where the drug formed a water-bridged complex with human CYP3A4 that was metabolized into a library of different metabolites [12]. Since catalytic turnover is an undesired property of inhibitory drugs, our results emphasize the importance of characterizing the population of low-spin complexes that are present with potential inhibitors that incorporate heme ligation as a design element.

Synopsis.

Drugs form a mixture of complexes with Mycobacterium tuberculosis (Mtb) cytochrome P450 (CYP) 51B1. Some of these complexes retain the axial water ligand present in drug-free enzyme, while others have drug that displaces and replaces the water. This mixture of binding modes should be considered in inhibitor design for Mtb.

Highlights.

• There is a mixture of binding modes for nitrogenous drug fragments

• Water-bridged complexes are present in each cytochrome P450 (CYP)-drug combination

• A mixture of binding modes impacts inhibitor design for Mycobacterium tuberculosis CYPs

Abbreviations

CYP

cytochrome P450

Mtb

Mycobacterium tuberculosis

TB

tuberculosis

CW

continuous wave

EPR

electron paramagnetic resonance

HYSCORE

hyperfine sublevel correlation

ENDOR

electron-nuclear double resonance

LP10

α-[(4-methylcyclohexyl)carbonyl amino]-N-4-pyridinyl-1H-indole-3-propanamide

PPT

4-(3-phenylpropyl)-1H-1,2,3-triazole; 17 click, 17α-(2H-2,3,4-triazolyl)-estradiol

1,2,3-TRZ

1,2,3-triazole