Interaction of Nitric Oxide with Catalase: Structural and Kinetic Analysis

Abstract

We present the structures of bovine catalase in its native form and complexed with ammonia and nitric oxide, obtained by X-ray crystallography. Using the NO generator 1-(N,N-diethylamino)diazen-1-ium-1,2-diolate, we were able to generate sufficiently high NO concentrations within the catalase crystals that substantial occupation was observed despite a high dissociation rate. Nitric oxide seems to be slightly bent from the heme normal that may indicate some iron(II) character in the formally ferric catalase. Microspectrophotometric investigations inline with the synchrotron X-ray beam reveal photoreduction of the central heme iron. In the cases of the native and ammonia-complexed catalase, reduction is accompanied by a relaxation phase. This is likely not the case for the catalase NO complex. The kinetics of binding of NO to catalase were investigated using NO photolyzed from N,N′-bis(carboxymethyl)-N,N′-dinitroso-p-phenylenediamine using an assay that combines catalase with myoglobin binding kinetics. The off rate is 1.5 s^–1. Implications for catalase function are discussed.

Nitric oxide is known for its role in bioregulatory processes like vasodilatation, macrophage-induced cytostatis, cytotoxicity, and neurotransmission.¹⁻⁷ It also behaves as a ubiquitous signaling molecule operating through nitrosylation.(8) Further S-nitrosylation can regulate protein–protein interactions and may have an impact on cellular signaling networks.(9) Transition metal nitrosyl complexes within metalloenzymes play central roles in the biological chemistry of NO.(10) In our laboratories, we are investigating a number of metalloenzymes involved in NO trafficking. Because many of the nitrosylated metalloenzyme intermediates of interest to us are metastable, we have had to develop a number of methods for detecting such transient species (see, for example, refs (11−13)). Herein, we report on a methodology that allowed us to obtain the crystal structure of nitrosylated catalase, which is kinetically labile. This methodology should be readily applicable to the crystallographic analysis of other labile nitrosylated metalloenzymes.

Catalase (EC 1.11.1.6) is a heme protein, which is found in many bacteria and almost all plants and animals.(14) It catalyzes the disproportionation of toxic hydrogen peroxide into oxygen and water (eq 1)

a reaction of great importance to all aerobically respiring organisms. Reactive oxygen species (ROS) like H₂O₂ and HO₂ (superoxide) are inevitable byproducts of aerobic respiration, necessitating specialized enzymes for their elimination.¹⁵⁻¹⁷ Typical examples for mammalian catalases are bovine liver and human erythrocyte catalase.(18) Bovine liver catalase was crystallized for the first time in 1937(19) and bovine erythrocyte catalase in 1969.(20)

The structure of bovine liver catalase (Cat) was determined initially in 1981,(14) and refinements were subsequently reported by Reid et al. in 1981(21) and Ko et al. in 1999.(22) This catalase is a tetramer.^23,24 Each monomer (∼60 kDa) consists of 507 amino acid residues and contains a heme group with the iron in the ferric state.²⁵⁻²⁷ Structures of catalases containing a variety of ligands bound at the active site have also been reported: azide,(28) formate,(29) oxo (compounds I and II),^30,31 and cyanide and 3-amino-1,2,4-Triazol.(32) Such active site adducts were prepared primarily to provide structural information about intermediates important to the catalytic mechanism. Herein, we report the structure of Cat complexed with nitric oxide, along with a microspectrophotometric analysis of the crystals during X-ray exposure. NO has been shown to be a competitive inhibitor of Cat, which could be physiologically significant.(33) The structure could also potentially yield insights about the Cat-H₂O₂ adduct, which is not sufficiently long-lived to be structurally characterized within the active enzyme. The microspectrophotometric analysis reveals the oxidation state of the active site that is determined crystallographically. Cat-NO complexes had not been previously characterized by X-ray crystallography because NO tends to dissociate rapidly from the active site, complicating data collection. We also report new values for the rate constant (k_off) governing dissociation of NO from Cat, and the associated thermodynamic Cat-NO binding constant (K_Cat-NO).

Materials and Methods

Materials

Bovine liver catalase (Cat) was obtained from Sigma (C3155), as was horse skeletal muscle myoglobin (Mb). As obtained from Sigma, Mb is in the aquomet (Fe^III) form, but in our experiments, it was reduced as described below. Other than reduction, Mb was used without further manipulation. Cat on the other hand was purified using a Superdex 200 (GE) column. Fractions with a 405 nm to 276 nm absorption ratio of >7 were pooled. The Cat concentration was determined using the extinction coefficient of 420 cm^–1 L mmol^–1 at 405 nm (for the tetramer).^21,34 To the concentrated Cat solution used in the crystallographic experiments was added 1–5 wt % NH₄OH (30% in water). Methyl viologen dichloride (MV_ox) was obtained from Acros and sodium dithionite from Aldrich. The nitric oxide generators 1-(N,N-diethylamino)diazen-1-ium-1,2-diolate [DEANO (Scheme 1)] and N,N′-bis(carboxymethyl)-N,N′-dinitroso-p-phenylenediamine [BNN (Scheme 2)] were prepared using methods previously described in refs (13) and (35). All other reagents were obtained from Fisher.

Scheme 1. Molecular Structure of DEANO and Its Decomposition into Two NOs and Diethylamine.

Scheme 2. Molecular Structure of BNN and Its Photochemical Release of NO under Reducing Conditions.

Crystallographic Analysis

Crystallization

Three different Cat crystal forms were prepared. Catalase was initially crystallized by the hanging drop vapor diffusion method, using conditions previously reported by Ko et al.(22) Four microliters of the concentrated protein (12–13 mg/mL, containing NH₄OH as described above) was mixed with an equal volume of the reservoir solution consisting of 45–60 mM magnesium formate. After 2–3 weeks, crystals with a typical size of 100 μm × 100 μm × 200 μm were obtained. These crystals are henceforth termed the Cat-NH₃ form. We obtained a second crystal preparation by soaking the Cat-NH₃ crystals overnight in an 80 mM magnesium formate solution (pH 6.7) and then washing them two more times with the same solution. We refer to the crystals thus obtained as Cat-5 crystals. Finally, we obtained crystals of the Cat-NO adduct by soaking the Cat-5 crystals for 5 min in pH 6.5 solutions containing 80 mM magnesium formate and 150 mM Bis-Tris as buffers, and DEANO at concentrations varying from 10 to 200 mM. To minimize the escape of NO into the atmosphere during the crystal nitrosylation process, we used the special apparatus shown in Figure 1. This apparatus consists simply of two glass plates with the mating surfaces ground to provide a very tight seal. The lower glass plate also has a hemispherical depression with a capacity of ∼200 μL. In any given experiment, a Cat-5 crystal was placed in the depression, which was then filled completely with the buffer/DEANO mixture, and the depression quickly capped with the top plate to prevent NO from leaving the cavity. A heavy weight was placed on top of the top plate to maintain the high internal pressure.

Figure 1.

Schematic diagram of the apparatus used to generate NO from DEANO. Abbreviations: V, glass vessel containing a 200 μL cavity; L, glass lid; M, mother liquor; W, weight used to prevent NO from leaking from the glass vessel. A Cat crystal is colored green.

Crystal Freezing

The crystals obtained under each of the three conditions were soaked for ∼20 s in a cryobuffer containing 20% polyethylene glycol 4000, 20% sucrose, and 50 mM Tris (pH 8.5). The crystals were mounted in a loop and shock-frozen at 100 K using a Cryojet (Oxford Diffraction). For the Cat-NO crystals, it was imperative to move quickly between removing the crystals from the pressure apparatus and freezing them in the Cryojet. NO will begin to diffuse out of the crystal as soon as the pressure is released, and NO occupancy at the active site will drop to unusable levels within ∼10 min unless the crystal is frozen.

Data Collection

Data were collected at BioCARS 14-BMC (Advanced Photon Source, Argonne National Lab, Argonne, IL). For single crystals of Cat-NH₃, Cat-5, and Cat-NO, complete data sets were collected using 13.8 keV X-ray radiation (λ = 0.9 Å). Reflection intensities were recorded on a charge-coupled device area detector (ADSC Q315). For each data set, the crystals were rotated through 180° in steps of 0.5° to cover the reciprocal space.

Data Analysis and Refinement

Data were indexed and integrated with MOSFLM(36) and scaled with the CCP4 program SCALA.^37,38 Because our crystals were orthorhombic, we used model 4BLC from the Protein Data Bank (PDB)(39) for a molecular replacement. This model was also determined from an orthorhombic crystal form of Cat(22) and could be used directly as the starting model for a refinement using CNS.(40) To prepare an initial model, we removed the NADPH molecules as well as the water molecules from the starting model. For each crystal form, we used the same refinement protocol. First, a rigid body refinement was performed at 3.0 Å, followed by simulated annealing using a 2000 K protocol to full resolution. After that, positional conventional refinement was used until convergence was reached. Finally, the B factors were determined using 40-step restrained B factor refinement. For inspection of the Fourier maps, “Xfit” was used (xtalview package(41)).

Water Search

Water molecules were searched up to the 3σ level and added to the molecular structure. Special care was taken during the water search not to insert a water molecule at the sixth coordination site of the iron. After insertion of all the water molecules, the positions of all atoms were refined using conventional refinement. Inspection of sigmaA-weighted 2mF_obs – DF_calc and mF_obs – DF_calc maps revealed misorientations of several amino acid side chains, which were manually reoriented into the electron density maps using Xfit. Refinement was repeated and the model corrected until no obvious misorientations could be detected. To determine the presence and orientation of any putative sixth iron ligand, the 2F_obs – F_calc electron density maps and the F_obs – F_calc difference maps were examined in the vicinity of the heme iron. Once a positive feature had been identified, the F_obs – F_calc difference electron density was integrated to calculate the total number of electrons at this site using Probe.(42)

Insertion of NO and NH₃

For Cat-NH₃, a positive electron density at the sixth coordination site was interpreted as NH₃, whereas for Cat-NO, a NO molecule was inserted into the electron density (see Results). Both the NO and the NH₃ were patched with the nitrogen atom to the iron at a distance of 1.9 Å using weak bond length restraints (50 kcal mol^–1 Å^–2). The NO was oriented parallel to the heme normal with a very weak Fe–N–O angular restraint (7 kcal mol^–1 rad^–2). The models were refined conventionally against their respective X-ray structure.

B Factor Averaging

The occupancy and B factor are usually hard to refine at nonatomic resolution. Because the electron count for the NO in Cat-NO indicated that not all Cat molecules in the crystal have an equal amount of NO bound, occupational refinement becomes necessary in addition to B factor refinement. For this, we alternated B factor and occupational refinement. First, the B factors for the N and O atoms of the NO determined from the conventional refinement (plus that of the iron and the four heme nitrogens) were averaged and assigned to both N and O. Then, the occupancy was refined for each individual NO using the grouped occupational refinement option in CNS. After that, the B factors for the NO molecules were refined again. This cycle was repeated until convergence was reached. In Cat-NH₃, we propose that ammonia is bound to the iron (see Results). The occcupancy and B factor for this crystal form were refined using a procedure similar to that just described for Cat-NO.

Geometry

To determine the orientation of the sixth axial ligand at the active site of Cat-NO, the tilt (τ) angle, the bend angle (ϕ), and the tilt and bend angle (α) were calculated with respect to the normal of the heme plane(43) (Figure 2). To compare the geometries of all four hemes in the three Cat structures, we determined the iron out of plane distance using Planefit.(43)

Figure 2.

Tilt (τ), bend (ϕ), and tilt and bend (α) angles for the NO bound to the heme iron in Cat-NO. The heme plane is defined by the four porphyrin nitrogens. A double-headed arrow marks potential iron out of plane displacements.

Microspectrophotometric Analysis of Cat Crystals during X-ray Irradiation

A microspectrophotometer (4DX) equipped with a deuterium/tungsten light source (DH2000, Ocean Optics) and a USB mini-spectrometer (USB 2000, Ocean Optics) available at BioCARS was mounted inline with the X-ray beam at BioCARS 14-BMC. During X-ray exposure (monochromatic flux of ∼5 × 10¹¹ photons/s), the absorption spectra of the three different catalase species were recorded at 100 K for up to 10 min in intervals of 10 s. Difference spectra were generated by subtracting the time-dependent spectra from the reference spectrum recorded before X-ray exposure. The absolute differences were integrated from 400 to 800 nm and plotted as a function of exposure time. Data were fitted empirically by functions that consist either of a single exponential, a sum of two exponentials, or an exponential and a linear phase.

Kinetic Analysis of Dissociation of NO from Cat-NO

Sample Preparation

All solutions were prepared and manipulated in a nitrogen-filled glovebox. Stock solutions were prepared daily and stored in a refrigerator at 4 °C until needed. All experiments were performed in solutions buffered with phosphate (μ = 50 mM, pH 7.4). Samples for any given experiment were prepared by mixing stock solutions of buffer, MV_ox, dithionite, BNN, Cat, and Mb in a 1.5 mL microcentrifuge tube, to give a total volume of 150 μL. The total concentration of MV in each sample was always 63 μM, and enough dithionite was added to reduce 25–45 μM of this to the monocation radical form (MV_red). The added dithionite also reduced aquometMb to the Fe^II form. The concentrations of the remaining reagents were 370–700 μM for BNN, 8.5–21 μM for Mb, and 11–15 μM for Cat. Once thoroughly mixed, the reaction mixtures were transferred to a 3 mm × 3 mm quartz fluorescence cuvette, fitted with a glass gastight stopper. The exact concentrations of all the reagents in the cuvette were determined by UV–vis spectroscopy using a previously described methodology.(44)

Data Collection and Instrumentation

Routine UV–vis spectra were recorded using a CARY 50 spectrophotometer (Varian) that was installed in the glovebox. Nitric oxide was photogenerated in situ from BNN using a methodology previously described^11−13,45 and summarized in Scheme 2. Briefly, photochemical fragmentation of BNN was initiated with a 10 ns, 308 nm, 6.0 mJ pulse from a XeCl excimer laser (TUI, Existar 200). This generated 1 equiv of free NO, which then reacted with Cat and Mb as described in Results. An OLIS RSM-1000 spectrophotometer was used to monitor the absorbance changes induced by the laser pulse. The configuration of the laser and spectrophotometric equipment has been described in general terms elsewhere.^11,13 All data were collected with the OLIS RSM-1000 in rapid-scanning mode (monochromator entrance slit width of 0.6 mm, scanning slit width of 0.2 mm, and exit slit width of 0.12 mm), which allows complete spectra to be obtained in 1 ms.

Data Analysis

Data were analyzed using the commercially available software packages Microcal Origin version 6.0 (Microcal Software, Inc.) and Mathcad 13 (Mathsoft Engineering and Education, Inc.). Changes in the complete spectra obtained using the Olis RSM as a function of time were analyzed using a previously described methodology.^11,44

Results

Crystal Structures

All the crystal forms obtained proved to be orthorhombic in space group P2₁2₁2₁. The unit cell dimensions vary slightly depending on the crystal form, in the following ranges: a = 83–86 Å, b = 139–140 Å, and c = 228–229 Å (Table 1). Multiple data sets were collected for each crystal form (see below), and the best one was picked on the basis of resolution, a low degree of mosaicity, and low R_sym values. Data typically extended to 1.9 Å and better and were more than 90% complete with an R_sym of <10%.

Table 1. Data Collection and Refinement Statistics.

	CAT-NH3	CAT-5	CAT-NO
soaking	directly from hanging drop	14 h in NH3-free buffer	5 min in 100 mM DEANO
a (Å)	83.48	83.22	86.11
b (Å)	140.72	140.92	139.94
c (Å)	229.52	229.37	228.02
α = β = γ (deg)	90	90	90
space group	P212121	P212121	P212121
volume (Å3)	2696241.6	2689905.1	2747694.22
resolution (Å)	1.99	1.90	1.88
Rsyma (last shell)	9.4 (30.3)	8.9 (34.3)	9.1 (32.9)
completeness (last shell) (%)	92.8 (92.8)	97.9 (97.1)	92.2 (89.9)
redundancy (last shell)	5.7 (5.5)	3.6 (3.3)	5.1 (4.3)
I/σI (last shell)	5.0 (2.5)	5.1 (2.5)	5.6 (2.1)
Rcrys/Rfree without water (%)	23.68/27.05	24.1/26.9	24.05/27.05
no. of water molecules	2126	1277	2013
Rcrys/Rfree with water (%)	16.83/20.69	22.16/25.82	18.56/22.05
rmsd for bond lengthsb (Å)	0.006	0.007	0.006
rmsd for bond angles (deg)	1.3	1.3	1.3
no. of Ramachandran outliers	53 (2.9%)	62 (3.3%)	53 (2.9%)
occupancy from electron count (%)
heme A	70	6.0	51
heme B	71	3.7	40
heme C	82	1.4	35
heme D	95	1.7	74
	80 ± 10	3.2 ± 1.8	50 ± 11
refined occupancy (%)
heme A	82		55
heme B	100		52
heme C	77		45
heme D	101		61
	90 ± 12.3		53.3 ± 5.8
FE out of plane distance (Å)
heme A	0.063	0.014	0.016
heme B	–0.054	0.123	–0.035
heme C	–0.044	0.132	–0.020
heme D	–0.052	–0.005	0.019
	–0.02 ± 0.06	0.066 ± 0.07	–0.020 ± 0.027
tilt angle (τ)/bend angle (ϕ)/tilt and bend angle (α) (deg)
heme A	7.98		9.14/19.6/28.74
heme B	6.01		8.17/5.03/3.13
heme C	0.51		12.28/11.78/24.06
heme D	5.33		8.23/9.72/17.96
	5.0 ± 2.75		9.46 ± 1.68/11.53 ± 5.6/18.47 ± 9.65
rms coordinate error (Å)	0.181	0.262	0.214
iron–nitrogen distance (Å)
heme A	2.10		1.92
heme B	2.11		1.88
heme C	2.11		1.85
heme D	2.10		1.83
	2.105 ± 0.01		1.87 ± 0.034
nitrogen (NH3)–water oxygen (NO)–water distance (Å)
heme A	3.05		2.42
heme B	3.05		2.69
heme C	2.96		2.58
heme D	2.76		2.52
	3.0 ± 0.13		2.55 ± 0.11
iron–Tyr distance (Å)
heme A	2.01	2.00	2.02
heme B	2.00	1.92	1.99
heme C	1.96	1.89	1.96
heme D	2.04	2.01	1.96
	2.0 ± 0.03	2.0 ± 0.06	1.98 ± 0.025
iron–water distancec (Å)
heme A	4.37	4.00	4.27
heme B	4.41		4.20
heme C	4.11		4.25
heme D	4.10	3.98	4.21
	4.25 ± 0.17	3.99 ± 0.01	4.23 ± 0.029

^aR_sym = ∑_i∑_j(I_i,j – ⟨I⟩)/(I_i,j).

^bRoot-mean-square deviation (rmsd) from ideal geometry. The rmsd values of dihedrals and impropers were 23° and 1°, respectively, for all models.

^cDistance from the heme iron to the closest water molecule near His74.

Ammonia-Bound Catalase (Cat-NH₃)

The initial model prepared from PDB entry 4BLC (see above) was refined to 1.99 Å with an R value of 23.7%. The structure is shown in Figure 3A. Earlier studies have shown that Cat typically incorporates one tightly bound NADPH per subunit,^34,46 and an earlier structure obtained under the same conditions as ours revealed a bound NADPH group. However, our structure showed no electron density for the NADPH in the same region (Figure 3B), from which we conclude that the NADPH was lost during purification by size exclusion. After addition of all water molecules to the structure, the R value decreased to 16.8%. A strong positive electron density feature was present in the difference map (mF_obs – DF_calc) of Cat-NH₃, near the heme iron (Figure 4A). Because high concentrations of ammonium were present during the crystallization, we propose that the electron density arises from an ammonia molecule bound to each of the hemes. For each of the four hemes, the integrated electron count at the sixth coordination site is ∼9 e^–, which supports our proposal that NH₃ molecules are present there. NH₃ was placed into the electron density at a distance of 1.9 Å, and its position was refined. The occupancy of the NH₃ molecule refined to ∼1.0 for all four subunits. The iron out of plane distance (Figure 4A) was approximately ±0.05 Å for all hemes; hence, the iron is almost in plane. The model precision was 0.18 Å as estimated by a Luzzati plot (see, for example, ref (47)).

Figure 3.

Quaternary structure of beef liver catalase. (A) Cartoon representation, with subunits A–D colored differently; the central heme B is highlighted. The dotted red box highlights the location of an NADPH molecule found in other catalases. (B) Electron density observed in our experiments (green), in the area where other groups have observed NADPH. A putative NADPH molecule is overlaid, showing that the presence of NADPH is not supported. (C) Water structure in the asymmetric unit of CAT-5.

Figure 4.

Structure of the heme pocket in the three different Cat forms. Water molecules within 8 Å of the heme iron are shown (red spheres). A sigmaA-weighted mF_obs – DF_calc simulated annealing difference map is colored green (3.7σ): (A) Cat-NH₃, (B) Cat-5, and (C) Cat-NO. Important heme pocket residues are marked in panel B.

Five-Coordinate Cat-5

The final model of Cat-NH₃ was used as a starting model to refine Cat-5. After the ammonia molecule at the sixth coordination site of the iron had been removed, as well as the water molecules, the model was refined to 1.9 Å. Inclusion of water lowered the R value to 22.1%. No positive density appeared at the heme active site (Figure 4B). Via integration of the faint electron density features around the sixth coordination site, the counted number of electrons was found to be <1.0 on average (Table 1). This demonstrates that NH₃ was successfully removed from this crystal form, leaving a vacant sixth coordination site at the heme.

Cat-NO Structure

The amount of NO needed to observe significant binding to the catalase crystals was optimized by testing different concentrations of DEANO, from 10 to 200 mM. DEANO concentrations of >100 mM gave satisfactory results, with more than 50% NO occupancy in the heme active site pocket. Data were collected to 1.88 Å (Table 1). The ammonia-free (Cat-5) structure without the water molecules was used as the initial model for the refinement. After water molecules had been inserted, the R value decreased to 18.6%. A positive electron density feature appeared at the sixth coordination site of the heme iron (Figure 4C), which was absent in Cat-5. This positive electron density was identified as a NO molecule. The refined NO occupancy is ∼55% on average (Table 1). Interestingly, a single water molecule was observed in the vicinity of the NO for all four hemes in this case. This seems to be important in defining the binding geometry of the NO with respect to the heme plane (see Discussion). The NO is slightly bent with respect to the heme (Table 1). For the different subunits, the bending angle varies between 5° and 20°, with an average of 12°. Note that the bending angle depends strongly on the restraints employed. If the bending angle were allowed to vary in an unrestrained fashion, the bending angle would increase to ∼70° (data not shown). The distance from the iron to the nitrogen of NO refined to 1.85 Å, and the distance from the oxygen of NO to the heme pocket water is 2.5 Å. This indicates a strong hydrogen bond (Table 1 and Figure 4C).

Microspectrophotometric Analysis

Figure 5a shows the spectral changes that are observed when a crystal of Cat-5 is exposed to X-rays; Figure 5b shows the same data as difference spectra, and Figure 5c shows how the integrated difference spectra change over time. Over a period of ∼100 s, one sees changes in the visible spectrum characteristic of reduction of heme from the Fe(III) (met) form to the Fe(II) (deoxy) form.(48) Over longer time scales, the area of the integrated difference spectrum decreases, in a process that may indicate crystal degradation. When the experiment was repeated with other crystals, the rate of the initial increase in the difference spectrum remained fairly invariant, whereas the rate of the subsequent decay varied substantially from one experiment to the next (data not shown).

Figure 5.

Spectral changes of Cat crystals after X-ray exposure (from 500 to 700 nm). The Soret band region (∼410 nm) was excluded, because the high level of absorption of the crystals leads to saturation effects. Panels a, d, and g show spectra for the Cat-5, Cat-NH₃, and Cat-NO crystals, respectively; the thick black lines represent the reference spectra, and the thick blue lines are the spectra after exposure to X-rays for 10 min. Note the disappearance of the met peak around 630 nm and the appearance of deoxy features around 560 nm (dotted arrows) for the Cat-5 and Cat-NH₃ species. Panels b, e, and h show the corresponding difference spectra obtained by subtracting the reference spectra; the thick black lines represent the difference spectra directly after opening the X-ray shutter, and the blue difference spectra were collected after 10 min. The met and deoxy peak positions are again indicated by dotted arrows. Panels c, f, and i show the temporal progression of the changes observed in the integrated difference spectra. Data in panel c were empirically fitted by an exponential and a linear decay. Data in panel f were fitted by two exponentials. Data in panel i were fitted by a single exponential. The characteristic times are indicated.

Panels d–i of Figure 5 are analogous to panels a–c, respectively, except that the observed changes are now for the Cat-NH₃ and Cat-NO crystals, respectively. In the case of Cat-NH₃, the observed spectral changes are somewhat more complicated than those for Cat-5, but they still exhibit features characteristic of heme reduction.(48) The visible spectra of ferrous and ferric heme nitrosyl complexes tend to vary depending on the protein, but a broadening of the signals as seen in Figure 5g is consistent with what is observed when nitrosylated myoglobin(49) or nitrophorin(50) is reduced. Note that the decay in the integrated difference spectrum amplitude that is observed on longer times cales for both the Cat-5 and Cat-NH₃ crystals (Figure 5c,f) is not apparent in the Cat-NO difference spectra (Figure 5i).

The microspectrophotometric data for the three crystal types were obtained over the span of ∼10 min, while 120 min was typically needed to obtain sufficient data for the three crystal structures. Because in each case the spectrophotometric data show that Cat is reduced after 10 min in the X-ray beam, we conclude that in all the crystal structures we obtained the iron centers must be in the +2 oxidation state. However, the cryogenic temperatures that we used would have prevented any major structural changes to the protein in response to the change in metal oxidation state. We are thus confident that the structures reflect the original form of the enzyme, even if the metal was reduced during the experiment.

Kinetic Analysis of Dissociation of NO from Cat-NO

Figure 6a shows the spectral changes observed after a solution initially containing 350 μM BNN, 32 μM MV_red, 10 μM Mb, and 15.2 μM Cat had been irradiated with a 308 nm, 10 ns laser pulse. Analysis of the difference spectra by singular-value decomposition (SVD) revealed three spectral components. Subsequent modeling of the spectral changes using the three known extinction coefficient difference spectra, Δε_Cat-NO (ε_Cat-NO – ε_Cat), Δε_Mb-NO (ε_Mb-NO – ε_Mb), and Δε_MV (ε_MVred – ε_MVox),^11,44 allowed us to explain the changes as follows.

Figure 6.

(a) Spectral changes (blue traces) observed after a solution initially containing 350 μM BNN, 32 μM MV_red, 10 μM Mb, and 15.2 μM Cat had been irradiated with a 308 nm, 10 ns laser pulse. Calculated difference spectra (red traces) obtained from least-squares fitting of the experimental spectra using the known difference spectra Δε_Mb-NO, Δε_Cat-NO, and Δε_MV. (b) Changes in [Cat-NO] and [Mb-NO] as a function of time, calculated from the data fits shown in panel a (green and blue traces, respectively). Least-squares fits (red traces) using exponential functions. The theoretical trace for Mb-NO has a linear component in addition to the exponential one, which becomes noticeable only over longer time scales (data not shown).

The laser pulse produced ∼4.2 μM NO and resulted in an equivalent amount of MV_red, as shown in Scheme 2.a Within the first few milliseconds of the laser pulse, the free [NO] was reduced to a small, steady-state concentration by reaction with either Cat or Mb. The rate constant for binding of NO to Cat [k_on1 = (1.3 ± 0.1) × 10⁷ M^–1 s^–1 (Supporting Information)] is comparable to that for binding of NO to Mb (k_on2 = 1.7 × 10⁷ M^–1 s^–1),(49) so comparable amounts of Mb-NO and Cat-NO are generated in this initial phase. From our analysis, we estimate that the initial concentrations of Mb-NO and Cat-NO (the concentrations of each species present when the first spectrum was collected 19 ms after the laser pulse) were 1.9 μM ([Mb-NO]₀) and 2.3 μM ([Cat-NO]₀), respectively.

The binding constant for Mb-NO formation (K_Mb-NO) is extremely high (∼10¹¹)(49), so binding of NO to Mb can be considered an irreversible process within the time scale of our experiments. Cat also binds NO tightly, but much less so than Mb. Consequently, Cat-NO will release NO within the time scale of our experiments. With these points in mind, the major spectral changes observed in Figure 6a after the first few milliseconds can be attributed to a decrease in [Cat-NO] and a corresponding increase in [Mb-NO], due to the reactions depicted in Scheme 3. These spectral changes were complete within ∼10 s and could be empirically fitted with single exponentials (eq 2 and Figure 6b).

where C_0,1 and C_0,2 are constant offsets, A_g is a global amplitude, and k_obs is the observed (apparent) rate of the reaction.

Scheme 3.

Over a longer time scale (60 s), the [Mb-NO] continued to increase at a slow, linear rate, in a process that also consumed MV_red (data not shown). We attribute the slower linear process to generation of NO from BNN by the spectrophotometer’s lamp. Note that in Figure 6b the calculated concentration of Cat-NO toward the end of the reaction appears to become negative. We believe this artifact arises mainly because of the uncertainties associated with the extinction coefficient spectra that we used. In particular, we know that there is substantial variation in the published literature values for ε_Cat. We generated our ε_Cat spectrum from the published value for ε_Cat of 1.2 × 10⁵ M^–1 cm^–1 at 403 nm (for the monomer), reported by Vlasits et al.(51) However, lower ε_Cat values ranging from 1.05 × 10⁵ M^–1 cm^–121,34 to 1.07 × 10⁵ M^–1 cm^–1(52) can be found in earlier literature reports. Despite these uncertainties, the increase in [Mb-NO] roughly matches the decrease in [Cat-NO] on the same time scale. Equally importantly, the k_obs rate constants obtained by fitting the two concentration versus time traces in Figure 6b agreed within 3% of each other.

Theoretical modeling of the reactions depicted in Scheme 3 shows that the transfer of NO from Cat-NO to Mb-NO is only truly exponential when [Cat] and [Mb] are both much greater than [Cat-NO] and [Mb-NO] (Supporting Information). Nevertheless, the empirical exponential fits of Figure 6b can be used to accurately determine the initial rate for the process (V₀ = A_gk_obs), which will have the theoretical interpretation shown in eq 3

under all reaction conditions (Supporting Information). k_on1 and k_on2 are the on rates of catalase and myoglobin, respecitively, and [Cat-NO]₀ and [Mb]₀ are the corresponding concentrations directly after NO is generated (see the Supporting Information). All of the concentrations and parameters in eq 3 except k_off are known or were determined experimentally from analysis of the data in Figure 6a. Substituting all the known values into eq 3 and solving for k_off yields a value of 1.6 ± 0.3 s^–1 for the desired parameter.

The experiment depicted in Figure 6 was repeated using a range of Mb and Cat concentrations, and in each case, the value of k_off was obtained as described above. As expected, k_off exhibited no dependence on the [Mb]_tot/[Cat]_tot ratio (Figure 7). Averaging all of the k_off values shown in Figure 7 gives a best estimate for k_off of 1.5 ± 0.1 s^–1. From the calculated k_off and the value of k_on1 for binding of NO to Cat (Supporting Information), we estimate the binding constant K_Cat-NO for Cat-NO formation to be (8.7 ± 1.2) × 10⁶. This value is almost 50 times greater than that previously reported by Hoshino et al. (1.8 × 10⁵),(52) but in good agreement with the value of 5.6 × 10⁶ that can be calculated from the reported K_i, for inhibition by NO of catalase’s H₂O₂ disproportionation activity.(33) Hoshino et al. obtained their value of K_Cat-NO from NO manometric measurements,(52) which are very difficult to perform at the low pressures required to measure binding constants greater than ∼10⁵. We thus feel that our value and that obtained from Brown’s K_i value(33) are the more reliable ones. Finally, it should be noted that Hoshino et al. also measured the rate constant k_on1 for binding of NO to Cat, by photolyzing Cat-NO with a laser pulse and then measuring the rate of NO rebinding in the presence of varying free NO concentrations. Our value for k_on1 [(1.3 ± 0.1) × 10⁷ M^–1 s^–1] agrees reasonably well with theirs (7.5 × 10⁶ M^–1 s^–1 per monomer).(52)

Figure 7.

Rate constant values obtained for dissociation of NO from Cat (k_off) at different [Mb]_tot/[Cat]_tot ratios. k_off is seen to be independent of [Mb]_tot/[Cat]_tot, which is expected if k_off is truly being measured.

Discussion

Nitric Oxide Binding to Catalase

DEANO is a universal NO generator in powder form (Scheme ). It can be easily handled and, because of its high water solubility, can be used to generate large amounts of NO in aqueous solution. With DEANO, the use of the inconvenient, poisonous NO gas tapped from pressurized tanks can be completely avoided, and the amount of NO in solution can be exactly controlled (see also refs (53) and (54)). High concentrations of NO dissolved in aqueous solution can be achieved without the vexing pH changes common to gas usage.

In our study of Cat-NO, the use of DEANO and the extremely simple apparatus illustrated in Figure 1 allowed us to expose our crystals to NO concentrations on the order of 100 mM. At 1 bar pressure and 25 °C, the solubility of NO in aqueous solution is 1.93 mM.(55) A 100 mM NO concentration corresponds to a partial pressure of ∼50 bar (5 MPa). The NO concentration reached is on the same order of magnitude as the Cat concentration within the crystals. By immersing the catalase crystal in the tightly closed cavity that is completely filled with stabilization buffer, we literally “pressurized” the crystals with NO. It was crucial to start with such high NO concentrations because, though Cat has a relatively large binding constant, it also has a relatively high dissociation rate (k_off) of 1.5 ± 0.1 s^–1 (Table 2). This means that as soon as the crystals are removed from the NO-containing solution they will quickly begin to lose their bound NO by diffusion. Note that we had to immerse the crystals in NO-free cryobuffer before freezing them in the ice-cold nitrogen gas stream. This procedure took a few seconds, time enough for some NO to diffuse out of the crystal. By initially pressurizing the crystals with NO concentrations much higher than 1 bar, we were still able to detect substantial occupation of NO in the catalase crystals.

Table 2. On and Off Rates and Binding Constants for Several Protein-NO (or CO) Complexes.

	kon(M–1 s–1)	koff (s–1)	KCat-NO = kon/koff (M–1)
Mb(Fe2+)COa	5.1 × 105	1.9 × 10–2	2.7 × 107
Mb(Fe2+)NOa	1.7 × 107	1.2 × 10–4	1.4 × 1011
Mb(Fe3+)NOb	4.8 × 104	42	1.14 × 103
Lb(Fe2+)NOa	1.2 × 108	2.0 × 10–5	5.9 × 1012
NP1(Fe3+)NOc	1.5 × 106 (pH 5)	0.18 (pH 5)	8.3 × 106 (pH 5)
	1.5 × 106 (pH 8)	1.25 (pH 8)	1.2 × 106 (pH 8)
NP4(Fe3+)NOc	2.1 × 106 (pH 5)	0.11 (pH 5)	2.0 × 107 (pH 5)
	2.3 × 106 (pH 8)	1.24 (pH 8)	1.9 × 106 (pH 8)
Cat(Fe3+)NOd	1.3 × 107	1.5	8.7 × 106

^aFrom ref (60). Lb is leghemoglobin.

^bFrom ref (67).

^cFrom ref (64). The nitrophorins adopt closed and open conformations at pH 5 and 8, respectively.

^dFrom this work.

Another advantage of our methodology is that with 100 mM DEANO the peak NO concentration reached is ∼2 orders of magnitude larger than that of oxygen (O₂ solubility at 25 °C and 1 atm is 1.28 mM).(55) NO normally has to be handled under anaerobic conditions because it reacts very rapidly with O₂. However, in the presence of a large excess of NO, we were able to scrub out the O₂ using only a fraction of the available NO and leaving plenty of NO for the reaction with Cat. Consequently, there was no need to use rigorous anaerobic handling techniques, which greatly complicate crystallographic experiments. Note though that the reaction of O₂ with NO generates acidic products,(56) so even with our methodology the crystals must be buffered with a suitable buffer at a concentration of at least 50 mM to avoid pH changes. The results presented herein will pave the way for easy crystallographic analyses of other proteins that weakly bind NO.

Kinetics of NO Binding

In higher organisms such as mammals, NO is a cell hormone, and its main effect is to increase blood flow by vasodilation.^5,8 NO is strongly bound by (ferrous) deoxyhemoglobin (Hb) located in the red blood cells(57) and rapidly oxidized to NO₃^– by oxyhemoglobin (HbO₂),^58,59 processes that normally keep the concentration of NO low in the bloodstream. In tissues, myoglobin takes over the role of hemoglobin in keeping the NO concentration low. Myoglobin, in both its ferric (metMb) and its physiologically more abundant ferrous (Mb) forms, also binds NO.^60,61 The ferrous Mb-NO is remarkably stable with a binding constant K_Mb-NO on the order of 10¹¹ M (Table 2),(49) and like HbO₂, the oxymyoglobin (MbO₂) can oxidize NO to NO₃^–. Thus, Hb and Mb normally keep the NO concentrations in the blood and tissues low, and poisoning of the terminal respiratory cytochrome c oxidase with NO is prevented (see ref (62) for the copper B-NO complex of cytochrome c oxidase). Uptake of NO through the red blood cells, however, is reportedly(63) slowed by diffusional barriers such as the cell membrane. This can allow transient accumulations of relatively high NO concentrations in the bloodstream, though the impact of such spikes is restricted to a relatively small area. In an example of an event that can transiently increase the NO concentration, blood-sucking insects like the “kissing bug”, Rhodnius prolixus, inject NO into the host to locally enhance blood flow. To achieve this, the bug uses the heme protein nitrophorin (NP) that contains a ferric central iron to which the NO binds.⁶⁴⁻⁶⁶ In some nitrophorins, the proximal iron ligand is a cysteine, which may detach from the heme and itself become nitrosylated. This leaves the iron in its ferrous form, which binds NO tightly.(50)

In general, ferrous iron-containing heme proteins such as Mb and Hb bind NO extremely tightly, with small dissociation rate coefficients [k_off (Table 2)]. NO also binds to ferric iron in proteins such as nitrophorin, metMb, and Cat; however, k_off values for the ferric compounds are comparably larger, leading to smaller binding constants. In the case of metMb, nitric oxide binds only weakly; the off rate for the ferric form differs from that of the ferrous form by many orders of magnitude.

Like metMb-NO, Cat-NO is comparatively labile; however, K_Cat-NO is substantially higher than K_metMb-NO, because of both a higher k_on and a lower k_off in the case of Cat (Table 2). Our structural investigations reveal potential reasons for these differences. First, metMb has a water molecule bound at the sixth heme coordination site, whereas our data show that Cat-5, the immediate precursor to Cat-NO, has a free sixth position. In metMb nitrosylation, the water molecule must be displaced before NO can enter,(67) which most likely accounts for the higher value of k_on for Cat (1.3 × 10⁷ M^–1 s^–1) versus that of metMb (4.8 × 10⁴ M^–1 s^–1).

Our crystallographic results seem to suggest that the NO in Cat-NO is bent away from the heme normal with angles of ∼20° (Table 3). Caution is necessary here, because at the resolution and NO occupancy reached in our experiments the true geometry can be only roughly determined. Nevertheless, slightly bent NO geometries were also found in ferric nitrophorin but not in pure heme-NO adducts where the NO is found to be aligned with the heme normal. In general, ferrous NO complexes tend to have more bent Fe-NO bonds, whereas ferric Fe-NO bonds tend to be more linear.(68) Therefore, our crystallographic data seem to be consistent with the heme having somewhat reduced character. This in turn is consistent with our microspectrophotometric data, which show that Cat is photoreduced in the X-ray beam; however, at the low temperature at which the data were collected (100 K), very little atomic rearrangement can occur within a molecule, so we suspect that the Cat-NO structure represents a metastable ferrous form in which the Fe-NO bond is “frozen in the ferric” arrangement. If so, then, in a catalase that has never been exposed to any X-rays, the Fe-NO bending angle would suggest that the heme iron in Cat has some ferrous character, despite being formally Fe(III). Such behavior has been observed in model systems.(68) Depending on the amount of extra density supplied, the iron gains Fe(II) character. The bend increases the degree of orbital overlap of NO with the molecular orbitals of the iron/protoporphyrin system,(68) which in turn decreases the dissociation rate relative to that of metMb. A final factor that could stabilize the NO in Cat relative to metMb-NO is the water molecule that sits nearby and may form a hydrogen bond to the oxygen of NO, bending and stabilizing it further (Figure 4C).

Table 3. Tilt and Bending Angles for NO Found in Different Heme-Iron Proteins with the Iron in Different Oxidation States (NP4, nitrophorin 4).

	tilt and bend angle (deg)		tilt and bend angle (deg)
Mb(Fe2+)NOa	60	NP4(Fe3+)NOc	24
Mb(Fe3+)NOa	33	NP4(Fe2+)NOd	40
NP4(Fe3+)NOb	70/3	catalase (Fe3+) (this study)	19

^aFrom ref (73).

^bFrom ref (74).

^cFrom ref (75).

^dFrom ref (76).

Possible Implications for Catalysis of H₂O₂ Disproportionation

H₂O₂ is both an oxidizing and a reducing agent. Catalase works by transferring a total of four electrons, in two different steps, at a rate of two electrons per step. First, two electrons are transferred to one molecule of H₂O₂, and then two electrons are accepted from a second H₂O₂ according to

The oxyferryl-porphyrin complex generated by eq 4 is called compound I. In the Fita–Rossman model,(34) the oxoferryl attacks another H₂O₂ molecule abstracting a hydride ion. We are left with an OH^– at the iron. The iron itself again becomes Fe(III), and one electron is transferred back to the porphyrin. Another proton is temporarily stored at the N_ε atom of the distal heme pocket His74 and finally transferred to the OH^– at the iron to form water. This water binds only weakly to the catalase iron and is quickly released. Other mechanisms without the involvement of His74 have also been proposed.(69) Reaction 4 requires that the porphyrin-Fe(III) complex easily donate electrons; hence, it acts as a reducing agent. This is not conceivable unless there is evidence that this complex has an excess of electron density. Our putatively bent NO suggests that the iron has iron(II) character, which is obtained from binding the negatively charged hydroxyl oxygen of the proximal ligand tyrosine (see the discussion above). This largely will facilitate the formation of compound I and contributes to the speed of the catalase.

Our NO catalase seems to be good model for mimicking the Fe(H₂O₂) adduct, with NO taking the place of H₂O₂. The water molecule that appears ∼4 Å from the Fe in all three structures, and is strongly H-bonded to the NO, could facilitate oxoferryl production in the H₂O₂ adduct by donating a proton to the incipient water being formed in reaction 4. The nearby water could also mark the spot where the second H₂O₂ sits to interact with both the distal His74 and the oxoferryl in reaction 5.

Note that the structural changes in catalase that accompany changes in coordination at the Fe distal site are minute. The iron out of plane distance is almost zero regardless of whether there is a distal ligand such as NO or NH₃ (Table 4). The positions of all distal and proximal residues remain essentially unchanged within the coordinate error of our structure analysis [∼0.18 Å (Table 1)]. Similar results are also observed upon ligand binding in a bacterial catalase.(30) Catalase’s rigidity through the catalytic cycle seems to be designed to make it an effective electron transfer protein where structural changes remain as small as possible, optimized for speed. In contrast, in myoglobin, the iron moves out the heme plane by ∼0.3 Å and the heme itself domes once the distal ligand is removed.^42,43 Large structural changes are triggered by this relaxation. Another possibly important feature of catalase is that it is internally wet. We identified ∼1280 water molecules; ∼100 (8%) are located in 140 cavities within the protein matrix (Table 5 and Figure 3C). In a protein similar in size, cytochrome c oxidase, whose structure has been determined to similar resolution,(62) 5% of the water can be found in 108 cavities. In terms of electron transfer and substrate diffusion, the terminal oxidase and catalase have similar functions. This seems to be reflected in the vast number of cavities. The importance of cavities for substrate diffusion and storage was outlined extensively for myoglobin.^43,70,71 Myoglobin, however, has only a few cavities in which <1% of the water can be found, even at atomic resolution.(72) Catalase can effectively guide the substrate through the network of cavities to its active center to cope with the high turnover rates.

Table 4. Structural Differences in Angstroms Relative to CAT-5 (no sixth iron ligand) Averaged over All Four Subunits.

	CAT-NH3	CAT-NO
heme Fe and N atoms	0.1 ± 0.01	0.1 ± 0.01
proximal side of the heme
Tyr357 Oη	0.11 ± 0.02	0.14 ± 0.04
distal side of the heme
His74 Nδ1/Nε2	0.12 ± 0.06	0.14 ± 0.06
Asn147 Oδ1/Nδ2	0.12 ± 0.07	0.16 ± 0.08
Phe160 Cζ	0.15 ± 0.03	0.13 ± 0.07
total rms differencea	0.34 ± 0.04	0.37 ± 0.05

^aThe structures were superimposed by a least-squares fit (Xfit(41)).

Table 5. Cavities in Selected Proteins^a.

	no. of cavities	H2O in cavity/total H2O	%	PDB entry
catalaseb	140	100/1277	8	this work
catalase (human, 1.5 Å)c	108	197/2298	9	1DGF
cytochrome c oxidased	95	85/1579	5	3ABK
L29W deoxyMbe	2	1/100	1	2BLI
deoxyMb (to 1.15 Å)f	2	0	0	1A6N

^aCalculations were conducted using “voidoo”(77) using a probe with a radius of 1.4 Å on a 0.5 Å grid.

^bCat-5, from this study.

^cFrom ref (32).

^dFrom ref (62).

^eFrom ref (43).

^fFrom ref (72).

Glossary

Abbreviations

BNN

N,N′-bis(carboxymethyl)-N,N′-dinitroso-p-phenylenediamine

Cat

catalase

DEANO

1-(N,N-diethylamino)diazen-1-ium-1,2-diolate

Mb

myoglobin

MV (MV_ox)

methyl viologen

MV_red

methyl viologen monocation radical

NADPH

nicotinamide adenine dinucleotide phosphate (reduced)

NO

nitric oxide.

^a Actually, as explained in ref (11), the amount of MV_red oxidized is slightly smaller than the amount of NO generated. The [MV_ox]/[NO] ratio is known and can be used to independently estimate the amount of NO generated by the laser pulse.

Supporting Information Available

Determination of k_on1 for the binding of NO to Cat and theoretical analysis of the experiments in which NO was photogenerated in the presence of both Cat and Mb. This material is available free of charge via the Internet at http://pubs.acs.org.