Interactions of azole-based inhibitors with human heme oxygenase

Abstract

Human heme oxygenase-1 (hHO-1) plays a crucial role in human physiology because of its ability to metabolize free heme. The heme degradation products, biliverdin and bilirubin, were shown to have protective antioxidant properties in cells. In the context of cancer, hHO-1 function grants cancer cells defense from standard chemotherapy treatments, leading to the development of azole-based inhibitors that target hHO-1 for potential anticancer therapy. This work reports experimental and theoretical characterization of interactions between three azole-based inhibitors and the active site of hHO-1. It was found that all three compounds have K_d values within the μM order. The electronic absorption and resonance Raman (rR) spectra indicated that they bind to the ferric heme and coordinate through a nitrogen atom. rR measurements revealed varying effects of inhibitors on the geometry of heme vinyl groups in the ferric form of hHO-1. Changes in peripheral group orientation are known to affect heme redox potential, and consequently can reflect the inhibitory properties of studied azoles. The subsequent docking studies showed that inhibitors with lower K_d values are located close to two vinyl groups, while the compound with higher K_d is situated near only one, consistent with the rR studies. Finally, the rR studies of the CO adducts showed that the inhibitors bind to the heme in a reversible manner. Altogether, the combination of ligand binding studies, UV-Vis and rR spectroscopies, as well as computational approach revealed an importance of the steric hindrance imposed by the inhibitor’s side chain.

Graphical Abstract

Besides its crucial role in physiology as heme degradation protein, human heme oxygenase-1 (hHO-1) was shown to protect cancer cells from standard chemotherapy treatments. This work reports experimental, spectroscopic and theoretical characterization of interactions between three azole-based inhibitors and the active site of hHO-1.

1. Introduction

Human heme oxygenase is an enzyme that is overexpressed in response to heat shock, oxidative stress, presence of free heme, heavy metals, and other stimuli and pathways.¹ It is typically recognized for its crucial role in cell physiology with respect to its role in free heme degradation.^2–5 The enzyme degrades heme to biliverdin with release of iron and carbon monoxide, and the biliverdin is subsequently converted to bilirubin by biliverdin reductase.^{2, 6–9} The products of the degradation reaction, biliverdin and bilirubin are reported to have antioxidant properties against hydrogen peroxide, peroxyl and hydroxyl radicals as well as anti-inflammatory properties.^{2, 10} The final product, bilirubin, decreases oxidative injury caused by hydrogen peroxide,¹¹ can protect vitamin A and linoleic acid from degradation¹² and acts against reactive oxygen species which have been linked to cancer, aging, and inflammation progression.¹² While it is well established that human heme oxygenase has a role in fighting oxidative stresses and protecting the cells from oxidative damages, ^{1, 2, 12} its cytoprotective nature has been utilized by cancer cells for self-protection against chemotherapy, giving tumors a resistance to standard treatments.¹ Hence, the enzyme has been linked to cancer progression and therapeutic resistance.^{1, 13, 14}

Accordingly, hHO-1 presents a promising target for cancer treatments. Various approaches, mainly genetic or pharmacological, can be employed for inhibition of the enzyme.¹⁵ Genetic inhibition entails use of CRISPR/CAS9 or RNAi to specifically target the hHO-1 gene. Genetic techniques have the advantage of targeting the gene or mRNA directly, albeit they could increase the chances of off-target effects.¹⁵ The alternative pharmacological approaches entail the use of metalloporphyrins like zinc protoporphyrin IX or azole-based drug inhibitors. Although metalloporphyrins have strong activity for inhibition, their potential lower selectivity poses some risk factors.¹⁵ On the other hand, azole-based compounds prove to be strong and more selective inhibitors, however, their drug metabolism and pharmacokinetics profiles (DMPK) need to be further explored.¹⁵

Comprehensive reviews on the development of azole-based inhibitors for hHO-1 inhibition have previously been presented.¹⁵ Development of the inhibitors initially started from a lead structure of azalanstat, a non-porphyrin inhibitor of 14 α-demethylase cytochrome P450 (CYP51).⁹ These types of inhibitors confer a unique advantage of having non-competitive inhibition unlike metalloporphyrins which are competitive with respect to the heme, presenting potential side reactivity and off-target selectivity issues, particularly with several physiologically relevant heme containing proteins like cytochromes P450s, globins and other heme enzymes.¹ The development of these inhibitors finally led to azole drugs with a unique pharmacophore previously described by Rahman et al.¹ which is based on the “initial” azalanstat structure and the previously defined regions of interest comprised of a central region containing a dioxolane group, a bulky northeastern group, an imidazole that binds heme in the eastern region, and a western clorophenyl group.¹

Since the enzymatic inhibition can have profound physiological consequences, structure function correlation studies of the protein in the presence of an inhibitor are of great importance. Resonance Raman (rR) spectroscopy is one of the most powerful tools for efficient studies of the active site structure of heme proteins. rR spectroscopy has previously been successfully utilized to investigate the interactions between inhibitors with cytochromes P450; e.g., studies of CYP3A4 in the presence of N-carbene inhibitor showed its effects on heme planarity, ruffling, and interplay with the enzymes active site residues^{16, 17} CYP102 in the presence of a nitrogen-coordinating inhibitor revealed alteration of heme modes.¹⁸ Furthermore, rR studies of CYP2B4 in the presence of acetylenic inhibitor 4-(tertbutyl) phenylacetylene (BPA) and BPA-modified CYP2B4 showed remarkable potential of rR spectroscopy to effectively monitor structural changes associated with inhibition, not only in ferric state but also in the ferrous CO-ligated form. ^{18, 19} In this work we provide detailed spectroscopic insight into the interactions of three azole based inhibitors with human heme oxygenase. All inhibitors studied here, that is compounds E, X and OB-24 (Figure 1), contain imidazole moiety which was previously shown to enable inhibition of hHO-1.¹ The previous work with the OB-24 showed improved immunotherapy efficiency in treating tumors cells.¹³ X also exhibited inhibitory function and compound E was chosen based on the predicted function owing to its structural similarity to OB-24.¹ The equilibrium binding studies and rR measurement of hHO-1 in its ferric, ferrous and ligated states are complemented by molecular docking calculations.

Figure 1.

Structures of compounds E, X, OB-24, and a cognate ligand of 3K4F.

2. Experimental

2.1. Protein Expression and Purification

Human heme oxygenase (hHO-1) was prepared as previously published.¹⁸ Briefly, the enzyme was expressed in BL21 DE3 cells from New England Biolabs (Ipswich, MA) and protein expression was done in the presence of 0.1 % glucose and induced by addition of IPTG. The cells were grown at 30 °C for 18 hours and stored at −80 °C. The cells were lysed via sonication, followed by purification by binding to a chitin resin and subsequently being cleaved from the resin by the addition of β-mercaptoethanol.^20–23 Heme reconstitution was done by adding two-fold excess of heme followed by removing excess heme on a Bio-Rad P6 resin from Bio-Rad Laboratories, Life Science Group (Hercules, California). All samples were stored in 100 mM potassium phosphate buffer pH 7.4.

2.2. Equilibrium Binding Titrations

UV-Vis measurements were done using a Lambda 465 spectrophotometer equipped with a PDA array detector. Successive 1 μL aliquots of inhibitor were added to a 1.4 μM protein sample. In the initial experiment, excess inhibitor was added to the protein to estimate the time needed to achieve equilibrium. It was established that no further spectral change occurred after 4 minutes for all inhibitors. Thus, equilibrium for all titrations was accomplished by incubating samples for 5 minutes before measuring absorbance. The final concentration of inhibitor ranged from 1 μM to 60 μM and the final dilution volume was less than 1%.

The K_d values and their associated uncertainties were obtained using Origin software from the regression fitting curve using a hyperbolic model with the equation y = B_max· x/ (x+K_d) where y is the magnitude of absorbance of the Soret band in the difference spectra of the sample with inhibitor minus sample without inhibitor. Free ligand concentration x is obtained from the equation that utilizes total ligand concentration [L]_tot − (ΔA[_L] / ΔA_saturation)[P]_tot, where [L]_tot is the concentration of inhibitor added, ΔA[_L] is change in absorbance, ΔA_saturation is change in absorbance at inhibitor saturation, and [P]_tot is the protein concentration. K_d and B_max are obtained from the fitted curve where K_d is the binding dissociation constant and B_max is the maximum amplitude of titration, which is the plateau of the y axis at high concentrations of inhibitor. The final K_d values for each inhibitor were calculated from averaging K_d values obtained from 3 independent experiments and are reported with their corresponding propagated error. GraphPad Prism software was used to plot the graphs using the 1-site binding hyperbolic model described above.

2.3. Preparation of Samples for rR Measurements

All hHO-1 samples for rR measurements were ~50–100 μM protein concentration in 100 mM potassium phosphate, pH 7.5. To prepare inhibitor-bound ferric samples, 20-fold excess of a given inhibitor was added to ferric enzyme. 100 μL of each sample was transferred to a Wilmad 5 mm Economy NMR tube for rR measurements. To prepare ferrous samples, NMR tubes containing ferric inhibitor-free or inhibitor-bound hHO-1 were closed by a septum. Each sample was degassed under flow of argon for 25–30 minutes to ensure an anaerobic environment. The samples were then reduced by addition of ~3-fold molar excess of anaerobically prepared sodium dithionite using a gas-tight syringe. To prepare the ferrous CO adducts, the reduced sample was then exposed to appropriate isotopes of CO gas. Resonance Raman spectra were measured immediately after formation of ferrous or ferrous-CO adducts. It is noted that the same ferrous CO spectra were obtained when inhibitor-bound ferric hHO-1 was first saturated with CO gas then reduced by addition of sodium dithionite solution.

2.4. Resonance Raman (rR) Spectroscopy Measurements

Resonance Raman spectra of ferric and ferrous forms as well as ferrous CO adducts were measured using a 413.1 nm excitation line from an Innova 302C Kr⁺ laser (Coherent Inc.). Ferric and ferrous spectra were recorded using 10 mW of laser power while the ferrous CO were measured with 3 mW to prevent photodissociation. The laser beam was focused onto the sample using a cylindrical lens to form a line image. Measurements were done using a 180° backscattering geometry, and the sample tubes were spun to avoid local heating. Scattered radiation was collected using a 1250M spectrometer (Horiba), equipped with a PyLoN:400B charge coupled device (CCD) detector (Princeton Instruments). The slit width was set at 150 μm, and a 1200 grooves/mm grating was used. Spectra were calibrated with fenchone and acetone-D6, and Grams/32 AI software was used for spectral processing. All measurements were done at room temperature.

2.5. Molecular Docking

The Schrödinger Maestro 12.9 Platform was used for preparation, running, and analysis of docking simulations.²⁴ The structure of human heme oxygenase 1 (PDB: 3K4F) was imported into the Maestro workspace using the Protein Preparation Wizard. The protein structure was preprocessed to assign bond orders, add hydrogen atoms, create zero-order bonds to metals, create disulfide bonds, and to delete water molecules beyond 5 Å from hetero groups. Chain A was selected for further refinement and artifacts of crystallization such as chloride ions and hexanediol were removed from the structure to simplify calculations. These crystallization artifacts are not relevant to the binding of inhibitors or present in the physiological conditions of the protein. The heme and active site waters were included in the final structure. The protein structure was refined via automated H-bond assignment and restrained minimization using OPLS4 force field by converging heavy atoms to 0.3 Å RMSD.

Ligand structures were prepared with LigPrep in the Maestro Suite using the OPLS4 force field. Possible ionization states were generated at a target pH of 7.0 ± 2.0 using Epik for pK_a predictions to generate tautomers. Possible chiralities were determined from the 3D structures of the ligands to generate potential stereoisomers in Maestro output formats.

Docking simulations were prepared and performed using Glide, the docking application. A receptor grid for Glide was generated from the refined structure of 3K4F. The grid was set to a (30 Å)³ box centered on the cognate ligand. Dockings were performed individually using the desired receptor grid and the prepared ligand files of each ligand. Van der Waals radius was scaled by decreasing the default value of scaling factor to 0.8 to soften the potential for nonpolar parts of the receptor. The initial docking results on cognate ligand were analyzed and compared to previously reported crystal structures to evaluate binding poses. An output of 15 poses for each inhibitor was analyzed, taking into consideration the UV-Vis and rR spectroscopic data.

3. Results and Discussion

3.1. Inhibitor Binding Affinity Determination

The structures of compounds E, X and OB-24 (Figure 1 A–C) are similar to previously reported inhibitors of hHO-1, however instead of a triazole group, they contain an imidazole moiety.¹ To determine binding affinities, UV-Vis spectra of ferric hHO-1 titrated with compounds E, X and OB-24 were taken between 300 and 700 nm (Figure 2 A–C, left side). In all three cases, the Soret band gradually shifts from 405 nm (inhibitor free sample) to 413 nm in presence of 40-fold excess of each ligand. Location of the Soret band reflects different types of heme coordination, thus, the UV-Vis measurements can be employed to distinguish between the nature of the distal side ligation; i.e., whether the inhibitor binds through the imidazole nitrogen atom or oxygen atom of its central region.

Figure 2.

The UV-Vis spectra of ferric hHO-1 titrated with E, X and OB-24 (A-C, left side). The corresponding difference traces for K_d determination (D-F, center). The difference traces were obtained by subtracting the inhibitor-free spectrum from the inhibitor-bound spectrum for each titrimetric addition. Titration binding curves of E, X, and OB-24 inhibitors to hHO-1 (G-I, right side) with the associated standard error of the mean shown as error bars. The y-axis represents points derived from the difference traces (Figure 2 D–F), that is the total magnitude in y value between the absorption of the inhibitor-free hHO-1 represented as a trough at 405 nm and the 413 nm peak of the inhibitor-bound hHO-1s.

As shown previously, the bis-histidine complexes of heme proteins, that is, proteins in which heme is coordinated to histidine residues on both proximal and distal sides (N_His-Fe(III)-N_His), have Soret bands between 413 and 414 nm and Q(0,0) and Q(0,1) bands at around 565 nm and 535 nm, respectively.^{25, 26} On the other hand, heme proteins with His-Tyr coordination (O_Tyr-Fe(III)-N_His), exhibit Soret peaks around 406 nm.^{25, 26} The UV-Vis spectra of inhibitor bound hHO-1 indicate that all three inhibitors coordinate to the heme iron via the nitrogen of the imidazole group. The middle column in Figure 2 D–F shows the difference traces obtained by subtracting the inhibitor-free spectra from the inhibitor-bound spectrum in their ferric state at each titration step. The difference traces have peaks at 401 nm and 417 nm, resulting from the strong overlap of the two Soret peaks, and correspond to the samples without and with inhibitor, respectively.

Equilibrium binding titration curves for hHO-1 binding to the inhibitors (Figure 2 G–I, right side) were constructed by taking the difference in absorbance at 413 nm and 405 nm and plotting the value against inhibitor concentration. The K_d values were calculated using the binding equation described in the experimental section. Compound E was found to have a K_d value of 2.92 ± 0.37 μM, virtually identical to that of OB-24 (4.46 ± 0.80 μM), while X binding was the weakest with a K_d of 8.91 ± 1.29 μM. It is noted that all three compounds have K_d values within the same order of magnitude. Inasmuch as the uncertainties associated with the K_d values of E and OB-24 inhibitors, they can be considered to be the same; however, X is definitely the most weakly bound.

Although the relative strength of inhibitor binding is the highest for E and OB-24 and the lowest for X, they are of the same μM order as compared to K_d’s of other analogous compounds. A similar class of inhibitors, such as SLV-8289 and SLV-11199, which also bind to heme iron through an imidazole group, were previously used for further profiling and screening of novel HO inhibitors for use as cancer treatments.²⁷ The K_d values for SLV-8289 and SLV-11199 were estimated using microscale thermophoresis and are reported to be 3.42 μM and 2.31 μM, respectively, similar to the K_d values determined spectroscopically in this work. While the direct comparison of the K_d numerical values should be avoided when obtained under different experimental conditions, it is recognized that they all are of the same order of magnitude.

The structures of inhibitors studies here differ by the presence of the dioxolane moiety, lengths of the central linker region, presence of the heteroatoms in the linker as well as presence of polar groups such as 3-bromophenyl or hydroxyl groups. The small differences in the K_d values associated with all three inhibitors indicate that the changes in the H-bonding interaction with the distal ordered water molecules or the amino acid residues, as well as hydrophobic and aromatic stacking interactions with the active site residues do not impact the binding affinities, or if they do, these effects must be offset by additional factors, such as a difference in steric energy cost of distorting the heme or its peripheral groups (vide infra). The common structural motif in all three inhibitors is the presence of the imidazole group which coordinates to the heme iron and this feature is most reasonably deterministic for these inhibitors’ affinities.

3.2. Resonance Raman (rR) Measurements

The high frequency (HF) region of resonance Raman spectrum of heme proteins provides information on coordination, oxidation and spin state of the heme iron.¹⁶ The resting state ferric hHO-1 exists as a mixture of dominant six-coordinate high-spin form (6cHS) with a minor six-coordinate low-spin (6cLS) component as indicated by the position of the ν₃ spin state marker modes seen at 1481 cm⁻¹ and 1504 cm⁻¹, respectively (Figure 3). The mixed spin state was previously interpreted as indicative of a heme iron bound to a proximal histidine residue and a water molecule/hydroxide anion on the distal heme side.²⁸ Upon binding the inhibitors there is a clear change in spin state distribution, i.e., a new dominant 6cLS species arises with ν₃ at 1506 cm⁻¹, with no detectable high spin component. This change is consistent with binding of the nitrogen atom of the azole groups of the inhibitors to the heme iron on the distal heme site. More specifically, the ν₃, ν₂, and ν₁₀ spin state markers seen around 1506 cm⁻¹, 1584 cm⁻¹, and 1640 cm⁻¹, respectively; resemble the frequencies of previously documented for bis-histidine N_His-Fe(III)-N_His complexes.^{25, 29} Consequently, these HF rR data are consistent with the UV-Vis studies, further solidifying the conclusion that the mode of inhibition involves binding to the heme through the N atom of the imidazole base of the inhibitors.

Figure 3.

High frequency resonance Raman spectra of ferric hHO-1 without inhibitor (A), in the presence of compound E (B), in the presence of compound X (C) and in the presence of compound OB-24 (D). Spectra are normalized to ν₄ mode at 1376 cm⁻¹ and measured using 413.1 nm excitation line_.

The low-frequency (LF) rR region provides information on deformation of heme macrocycle, disposition of heme peripheral groups, and the nature of heme axial ligands. The spectrum of ferric hHO-1 is dominated by strong ν₇ mode at 678 cm⁻¹, and exhibits ν₈ mode at 347 cm⁻¹, δ(C_βC_cC_d) propionate bending mode at 380 cm⁻¹ and two δ(C_βC_aC_b) vinyl bending modes at 410 cm⁻¹ and 422 cm⁻¹ (Figure 4). While addition of inhibitors does not alter geometry of propionate groups or deformation of heme macrocycle, it does induce changes in the disposition of vinyl groups. The vinyl bending modes observed at lower frequency are associated with the in-plane (ip) geometry of heme vinyl group and these observed at higher frequency represent the out-of-plane (oop) deformations of heme vinyl groups, with respect to heme plane.³⁰ The inhibitor E with the lowest K_d value induces a noticeable shift of both δ(C_βC_aC_b) vinyl bending mode, shifting them to 414 cm⁻¹ and 424 cm⁻¹, respectively, implying more out-of-plane geometry of both vinyl groups with respect to their orientation the inhibitor-free samples. Interestingly, OB-24, with K_d comparable to E, also upshifts the in-plane vinyl bending mode by 4 cm⁻¹; however it also upshifts the out-of-plane vinyl bending mode by 5 cm⁻¹ with a significant concomitant increase of its intensity, implying even stronger oop deformation of the vinyl groups than that caused by E. The compound with the weakest K_d, X, has a δ(C_βC_aC_b) bending modes at 410 cm⁻¹ and 425 cm⁻¹, implying that only one vinyl group might be affected by the presence of this inhibitor.

Figure 4.

Low frequency resonance Raman spectra of ferric hHO-1 without inhibitor (A), in the presence of compound E (B), in the presence of compound X (C) and in the presence of compound OB-24 (D). Spectra are normalized to ν₇ mode at 678 cm⁻¹ and measured using 413.1 nm excitation line. The insert shows the structure of heme b to illustrate the notation of local coordinates.

The interaction of E and OB-24 with vinyl groups is further confirmed by observation of the 4–7 cm⁻¹ upshifts of the ν(C=C) stretching modes with concomitant sharpening of the ν(C=C)/ν₁₀ envelope caused by the overlap of the vinyl modes and the ν₁₀ mode (Figure 3). Such behavior further reflects an alteration of the vinyl groups geometries to adopt more out-of-plane configurations with respect to the heme plane. Interestingly, the HF spectrum of X-bound hHO-1 in this region shows that the ν(C=C) stretching mode is at the same position as in the inhibitor-free sample and the presence of now clearly defined 6cLS ν₁₀ mode at 1640 cm⁻¹ (as opposed to the 6cHS inhibitor-free hHO1 where the ν₁₀ mode is hardly noticeable) results in an apparent broadening of this ν(C=C)/ν₁₀ envelope. The docking studies allowed an additional insight into the inhibitor – vinyl group interactions, as discussed below.

The detection of the altered geometries of the heme vinyl groups conveys important potential functional consequences since the orientation of vinyl group with respect to the planes of adjacent pyrrole rings were previously shown to be an important factor that can affect the heme iron redox potential, and as such are important to clarify.³¹ The small, but detectable relative differences between stronger and weaker binding compounds might arise from the steric interaction of their central linker region and associated functional groups with the heme peripheral group. Such interaction can be responsible for the selectivity, specificity, and potency of these inhibitors. Furthermore, based on the careful analysis of the low-frequency rR spectra it can be concluded that the presence of inhibitors does not cause excessive or differing deformation of the heme plane macrocycle, i.e., no significant enhancement of the out-of-plane mode is observed in the spectra of inhibitor-bound samples.

The electronic absorption spectrum of ferrous hHO-1 (Figure 5, insert) has Soret band at 430 nm and a single Q band at 556 nm, as reported previously. ³² The addition of E, X, and OB-24 compounds results in identical UV-Vis spectral response with the Soret band being 5 nm blue shifted and the appearance of two Q bands at 528 nm and 558 nm. Such spectral pattern is characteristic of the ferrous heme iron containing two nitrogen atoms coordinated on the proximal and distal heme side; e.g., identical UV-Vis spectra were reported for HupZ protein (424 nm, 530 nm ands 559 nm),²⁹ cytochrome b5 and the K⁺-APX mutant (424 nm, 527 nm and 557 nm),³² all of them having bis-imidazole (histidine) ligated heme.

Figure 5.

Low and high frequency resonance Raman spectra of ferrous hHO-1 without inhibitor (A), in the presence of compound E (B), in the presence of compound X (C) and in the presence of compound OB-24 (D). Spectra are normalized to ν₇ mode at 678 cm⁻¹ and ν₄ mode at 1354 cm⁻¹, respectively, and measured using 413.1 nm excitation line. The insert shows electronic absorption spectra of ferrous hHO-1 without and in the presence of inhibitors.

To further probe the heme environment, the rR spectra of ferrous hHO-1 samples were measured. The high-frequency rR spectra of ferrous hHO-1 without the inhibitors are shown in Figure 5 A, right and are consistent with 5cHS state with the ν₄, ν₃ and ν₂ modes at 1354 cm⁻¹, 1469 cm⁻¹ and 1562 cm⁻¹, respectively. The ν(Fe-N_His) stretching mode is observed at 216 cm⁻¹, indicative of the neutral proximal histidine, consistent with the previously published data. ³³ The rR spectra of ferrous samples containing E, X, and OB-24 inhibitors are indistinguishable (Figure 5 B–D). Their oxidation and spin state markers are at 1354 cm⁻¹ (ν₄), 1491 cm⁻¹ (ν₃) and 1562 cm⁻¹ (ν₂), similar to that of the bis-His ligated heme proteins, such as human neuroglobin,²⁵ cytochrome b₅, ³⁴ and NAD(P)H cytochrome b₅ oxidoreductases (Ncb5or). ³⁵ Consistent with this picture is the lack of the enhancement of the ν(Fe-N_His) stretching mode, which was not observed for the other bis-His ligated proteins. Hence, the electronic absorption of rR studies of ferrous hHO-1 in the presence of all three compounds clearly indicates that these inhibitors remain coordinated to the heme iron in the ferrous state.

The rR spectra of ferrous CO adducts provide information on the nature of the proximal ligand as well as the active site environment such as the distal side hydrogen bond network and steric hindrance. Hence, carbon monoxide is a useful probe of the changes in the hHO-1 active site upon binding various inhibitor. The oxidation state marker band, ν₄ mode, of all samples without and with inhibitors, is observed in a typical range for ferrous CO species at 1372 cm⁻¹, and the spin state marker bands, ν₃, ν₂ and ν₁₀ modes, are at 1497 cm⁻¹, 1583 cm⁻¹, and 1628 cm⁻¹, respectively, indicative of a 6cLS species, as expected (data not shown).

Figure 6 shows the low-frequency region of the ferrous CO adducts of wild type hHO-1 without inhibitor and the protein in the presence of E, X, and OB-24 compounds. The ν(Fe-C) stretching and the δ(Fe-C-O) bending modes are seen at 503 cm⁻¹ and 575 cm⁻¹, respectively, for the CO natural-abundance WT hHO-1 (solid lines). The identity of these modes is confirmed by their expected isotopic shift upon substitution with ¹³C¹⁸O analogues; e.g., the ν(Fe-C) mode shifts to 493 cm⁻¹ and the δ(FeCO) mode shifts to 558 cm⁻¹ (dotted line). Interestingly, the frequencies and relative intensities of these isotope sensitive modes do not change upon addition of any of the inhibitors studied here. The inset of Figure 6 shows the ¹²C¹⁶O - ¹³C¹⁸O difference traces, revealing the ν(C-O) stretching mode located around 1956 cm⁻¹ for the inhibitor containing samples shifting to the 1864 cm⁻¹ upon ¹³C¹⁸O substitution and exhibiting expected 92 cm⁻¹ downshift. The frequencies of the ν(Fe-C) at 503 cm⁻¹, δ(Fe-C-O) at 575 cm⁻¹ and ν(C-O) at 1956 cm⁻¹ modes of hHO-1 observed here for inhibitor-free sample are almost identical (within 1–2 cm⁻¹) to those previously published for wild-type rat HO.³⁶

Figure 6.

Low frequency resonance Raman spectra of ferrous CO adducts of hHO-1 without inhibitor (A), in the presence of compound E (B), in the presence of compound X (C) and in the presence of compound OB-24 (D). Spectra are normalized to ν₇ mode at 675 cm⁻¹ and measured using 413.1 nm excitation line. The solid lines represent spectra of samples with natural-abundance CO and the dotted lines represent samples containing ¹³C¹⁸O isotope (95 atom % ¹⁸O, 99 atom % ¹³C). Inset shows the ¹²C¹⁶O – ¹³C¹⁸O difference traces in the ν(CO) region for hHO-1 in the presence of E (i), X (ii) and OB-24 (iii) inhibitors.

It is noted that the heme modes in the LF region of these CO adducts remain unchanged for the samples with inhibitors, as opposed to what was seen in the ferric state where the geometry of vinyl groups of E-, X- and OB-24-bound samples were affected. Since these rR spectra of hHO-1 in the absences and presence of all three inhibitors are virtually identical, it is reasonable to conclude that binding of the exogenous CO ligand to the reduced inhibitor-bound hHO-1 results in expulsion of the inhibitors from the active site of the protein; i.e., the CO molecule has a higher affinity for ferrous heme iron than the imidazole nitrogen atom of the inhibitor. Consequently, the absence of the inhibitors in the active site of the hHO-1 in its ferrous CO state indicates a reversible binding of imidazole-type inhibitors to human HO-1, meaning that the enzyme is inactivated via noncovalent, reversible interactions. These studies provide the first spectroscopic evidence for the reversible inhibition by these compounds.

3.3. Molecular Docking of hHO-1 Inhibitors

Molecular modeling is typically implemented to help understand drug-target interactions by predicting the preferred orientations of ligands against a target protein, receptor, or enzyme that forms a stable complex. Here, we perform molecular docking calculations of cognate ligand as well as E, X, and OB-24 inhibitors using the Schrodinger Maestro 12.9 suite.²⁴ The receptor grid was defined by docking the cognate ligand of 3K4F (PDB: 3K4F) crystal structure to evaluate the appropriate settings later used in the docking procedure (Figure 7). The cognate ligand has a triazole moiety that, according to X-ray crystallographic data, coordinates to the ferric iron of heme through its 4’- triazole nitrogen atom. In our studies we used rR and UV-Vis data to choose the correct poses; e.g., the spectroscopic data clearly showed that all inhibitors bind to the heme iron via nitrogen atom of the imidazole, so consequently the poses with this type of interactions were chosen.

Figure 7.

Overlay of the PDB: 3K4F (ligand in blue)⁴ and the docked cognate ligand (green) in the binding site of hHO-1.

The docked cognate ligand aligns similarly with the ligand located in the crystal structure with a calculated Fe-N_inhibitor distance of 2.63 Å as compared to the 2.27 Å given by the X-ray data. The 0.36 Å discrepancy in the Fe-N distance results mostly from the 0.33 Å translation of the heme macrocycle during protein preparation, a movement of the heme b parallel to a heme normal toward the proximal side of the heme pocket, as can be seen in Figure 7. As such, it is expected that the translation of the heme upon protein preparation will result in extended ligand to heme iron distance of a comparable magnitude; i.e., 0.36 Å difference due to 0.33 Å heme translational movement. Taking this into account, the Fe-N_inhibitor distance 2.63 Å - 0.33 Å = 2.30 Å, is very close to that seen in the crystal structure. More importantly all key functional groups of the docked cognate ligand that are capable of interactions with the active site overlay with corresponding groups in the crystal structure. In both cases, the ketone group orients in similar direction and maintains H-bonding to the water molecule 585, although in the docked pose its oxygen atom is shifted slightly closer (~ 0.26 Å) to an active site solvent molecule (water) than that observed in the crystalized protein. The phenyl of the cognate ligand is slightly shifted with respect to that in X-ray structure, however, the π-π interactions with Phe37 are still preserved, as shown in Figure 7. The active site water molecules, water 585 and 605, were included in the final prepared structure. Since the receptor structure is rigid in the docking procedure, the waters stay stationary during docking of each compound. The docking score for the cognate ligand was - 5.185 (kcal/mol). Based on this data, the assigned receptor grid was selected for further use in inhibitor docking using E, X, and OB-24 compounds.

Compounds E, X, and OB-24 were docked to hHO-1 according to procedure described above with a glide score of −4.802 (kcal/mol), −4.215 (kcal/mol) and −5.019 (kcal/mol), respectively. Corresponding docked poses were generated for E, X, and OB-24 bound to hHO-1 and are shown in Figure 8. For E, the imidazole nitrogen coordinates to the ferric iron of heme with a bond length of 2.52 Å (Figure 8 A). The hydroxyl group forms a direct hydrogen bond with active site water 585, however the π-π interactions with Phe37 are lost. For X, the Fe-N bond length was calculated to be 2.59 Å and an aromatic H-bonding interaction with the Asp140 was observed (Figure 8 B). For OB-24, the Fe-N distance is 2.54 Å while the ligand’s bromine and Phe37 experience bad contacts and the active site water 585 interacts with inhibitor’s dioxolane moiety (Figure 8 C). In summary, in all cases, the distance between heme iron and the imidazole nitrogen is around 2.5 Å and the orientations of the inhibitors within the active site are similar to that observed for cognate ligand.

Figure 8.

Docked binding poses of the examined compounds against hHO-1 (PDB: 3K4F). hHO-1 with E (A), hHO-1 with X (B) and hHO-1 with OB-24 (C).

The rR studies of the ferric hHO-1 samples containing E, X and OB-24 revealed that these inhibitors alter the disposition of a heme vinyl groups (vide supra). Such conformational changes can affect the redox potential of the heme, and as such modify the reactivity of the enzyme and consequently impact the inhibitory properties of studied compounds. Analysis of the docked poses showed several interesting trends, summarized in Table 1. The distances between all atoms of a given inhibitor and the C_a and C_b atoms of 2-vinyl and 4-vinyl groups were measured. It was determined that if the inhibitors are within 5.1 Å of any heme vinyl group, a change in the rR spectral pattern is observed, reflecting a change in the disposition of the vinyl group. The hydroxyl group of the E or dioxolane moiety of OB-24 are located within approximately 5 Å or less of both heme vinyl groups, and the frequencies of both vinyl bending modes are changed in the spectra of E- and OB-24-bound samples compared to inhibitor-free hHO-1 (Figure 4). On the other hand, the carbon atom of X linker chain is within 4.12 Å of the vinyl-2 and the oxygen atom of the linker is almost 6 Å away for the vinyl-4 groups and in the rR spectra only the higher frequency vinyl bending mode (at 425 cm⁻¹) is affected while the lower frequency one (410 cm⁻¹) stays the same as in sample without inhibitor (Table 1). Taking together the rR spectroscopic and computational docking results, it is now possible to assign and distinguish the vinyl bending modes in the rR spectra; e.g.; the lower frequency mode represent the vibrational mode of the 4-vinyl group and the higher frequency is associated with the 2-vinyl group.

Table 1.

Comparison of the rR frequency of the vinyl bending modes and the distances between vinyl modes and inhibitors as determined by docking studies. In parenthesis are observed upshifts of the vinyl groups.

	Kd [μM]	4-vinyl frequency [cm−1]	4-vinyl – inhibitor distance [Å]	2-vinyl frequency [cm−1]	2-vinyl – inhibitor distance [Å]
hHO-1		410		422
+ E	2.92 ± 0.37	414 (+4)	4.93	424 (+2)	5.06
+ X	8.91 ± 1.29	410	5.90	425 (+3)	4.12
+ OB-24	4.46 ± 0.80	414 (+4)	4.67	427 (+5)	4.11

Perhaps even more importantly, a clear correlation emerged between the K_d values and the relative distances of the inhibitors to the heme vinyl groups as well as the number of vinyl groups interacting with the particular inhibitor. E and OB-24 have the lowest K_d values and affect the geometry of both vinyl groups, as discussed above. X has the highest K_d value and interacts with only one vinyl group (vide supra). It is observed that in all cases when the inhibitor is positioned in a shorter distance to the vinyl groups than computationally derived distance of 5.1 Å, the vinyl modes in the rR spectra shift to the higher frequency and in some cases increase in their intensities. Such spectral behavior is interpreted as the heme vinyl groups adopting an out-of-plane geometry with respect to the heme plane. Such inhibitor-induced increase of out-of-plane orientation of the vinyl group can result in the lowering of their electron-withdrawing abilities and consequently a decrease of the reduction potential of the heme. ^{31, 37} Such decrease in heme reduction potential might have dramatic effects on the key first step of hHO-1 catalytic cycle which is reduction of the ferric heme that allows binding of the molecular oxygen. Consequently, it is reasonable to hypothesize that such steric interactions between inhibitor’s central region and the vinyl groups of heme prosthetic group are related to the strength of inhibitor binding.

4. Conclusions

This work focuses on structural elucidation of the protein-inhibitor interactions. The K_d values of E, X, and OB-24 were estimated to be at 2.92 ± 0.37 μM, 8.91± 1.29 μM, and 4.46 ± 0.80 μM, respectively. These spectroscopically determined values are of similar order of magnitude as compared to other novel hHO-1 inhibitors, such as SLV-8289 (K_d = 3.42 uM) and SLV-11199 (K_d = 2.31 μM) for which the K_d values were determines by microscale thermophoresis.²⁷ All of these ligands rely on an imidazole moiety as the iron binding group, which appears to be a deterministic for these inhibitors’ affinities. The UV-Vis and rR spectroscopic studies provided clear experimental evidence for binding of all three inhibitors to the heme iron via the nitrogen of the imidazole moiety. The rR spectra of the ferric samples revealed clear differences in the disposition of the heme vinyl group in the spectra of inhibitor-free and inhibitor-bound samples. Such changes can be responsible for the small, but detectable differences between the affinities of X and other, more strongly binding compounds. In other words, the steric interaction of inhibitors central linker regions and associated functional groups with the heme peripheral group might be responsible not only for the differences in the K_d values, but can also affect the selectivity, specificity, and potency of these inhibitors. The presence of inhibitors does not caused excessive or differing deformation of the heme plane macrocycle.

As revealed by the docking computations, the general orientation of the studied inhibitors within the heme active site is comparable to the orientation of similar compounds observed previously in X-ray studies. The docking data revealed that the positioning of a given inhibitor within 5.1 Å or less of the heme vinyl groups results in changes in rR spectral patterns, reflecting the interaction between the vinyl groups and inhibitors. Additionally, while equilibrium binding studies showed that the E and OB-24 have comparable K_d values, they are lower than that for X. Since E and OB-24 affected the geometry of both vinyl groups, as opposed to X interacting with only one vinyl, it is reasonable to hypothesize that such interactions might be related to the strength of inhibitor binding. Therefore, the inhibitors containing functional groups that are in greater steric contact with vinyl groups might exhibit stronger binding properties.

Additionally, the UV-Vis and rR studies showed that all three inhibitors are still bound to the heme iron in the ferrous state. The rR spectra of ferrous CO adducts revealed that the modes associated with the Fe-C-O fragment, as well as heme modes, in the inhibitor bound samples are identical to those observed in the spectra of inhibitor free hHO-1. The lack of such changes indicates that the hydrogen bonding network in both cases are the same; in other words, the water network is preserved after the expulsion of inhibitor from active site by the CO gas. Consequently, the rR experiments provide a strong evidence that binding of the inhibitors is reversible with the addition of CO, a mimic of dioxygen. Finally, these studies showed remarkable capabilities of rR spectroscopy, and its complementarity with docking methodologies, in revealing important structure-function relationships which might aid in the selection of better structure-based drugs.

Highlights.

• Inhibitors coordinate the heme iron of heme oxygenase via nitrogen atom

• K_d values are all in the μM range

• Resonance Raman studies reveal reversible inhibition

• There is correlation between K_d values and inhibitor – heme vinyl group distances

• Bulky functional groups in the central region of inhibitors improve binding