Solution NMR Characterization of an Unusual Distal H-bond Network in the Active Site of the Cyanide-inhibited, Human Heme Oxygenase Complex of the Symmetric Substrate, 2,4-Dimethyldeuterohemin* 210

The presence of variable static hemin orientational disorder about the α-γ-meso axis in the substrate complexes of mammalian heme oxygenase, together with the incomplete averaging of a second, dynamic disorder, for each hemin orientation, has led to NMR spectra with severe spectral overlap and loss of key two-dimensional correlations that seriously interfere with structural characterization in solution. We demonstrate that the symmetric substrate, 2,4-dimethyldeuterohemin, yields a single solution species for which the dynamic disorder is sufficiently rapid to allow effective and informative 1H NMR structural characterization. A much more extensive, effective, and definitive NMR characterization of the cyanide-inhibited, symmetric heme complex of human heme oxygenase shows that the active site structure, with some minor differences, is essentially the same as that for the native protohemin in solution and crystal. A unique distal network that involves particularly strong hydrogen bonds, as well as inter-aromatic contacts, is described that is proposed to stabilize the position of the catalytically critical distal helix Asp-140 carboxylate (Liu, Y., Koenigs Lightning, L., Huang, H., Moënne-Loccoz, P., Schuller, D. J., Poulos, T. L., Loehr, T. M., and Ortiz de Montellano, P. R. (2000) J. Biol. Chem. 275, 34501–34507). The potential role of this network in placing a water molecule to stabilize the hydroperoxy species and as a template for the condensation of the distal helix upon substrate binding are discussed.

and substrate, catalyzes the regiospecific conversion of heme to ␣-biliverdin, iron, and CO (1). The physiological roles of HO are heme catabolism (HO-1) (2)(3)(4) and the generation of CO as a putative neural messenger (HO-2) (5,6). Detailed mechanistic (7)(8)(9)(10)(11)(12)(13) and spectroscopic (13)(14)(15)(16)(17) studies of the fully active recombinant, soluble 265-residue portion of HO-1 have shown that, in contrast to heme peroxidase and cytochrome P450, HO does not act through a ferryl intermediate. Recent crystal structures (18,19) of the substrate-bound, water-ligated complexes of a more truncated 233-residue human HO, hHO (20), and the complete rat HO (18), rHO, have revealed a largely helical enzyme that confirms the binding of heme by His-25 and locates a highly bent distal helix that is sufficiently close to the heme to sterically block all but the ␣-meso position (see Fig.  1) for attack. Detailed studies have established that the reactive intermediate in HO catalysis is the hydroperoxy-heme (11) rather than the ferryl heme common to peroxidases and cytochromes P450 (21). Although the present structural information provides a ready interpretation of the stereoselectivity of the heme cleavage in terms of steric blocking of all but the ␣-meso position by the distal helix, the environmental influences that stabilize the Fe-O-OH rather than FeϭO species are incompletely understood.
Recent studies with both hHO (22) and rHO (23) mutants have shown that elimination of the carboxylate of Asp-140 on the distal helix completely abolishes HO activity and leads to formation of a catalytically incompetent ferryl species. Asp-140 is not oriented toward the heme iron and cannot directly interact with a heme ligand. However, the crystal structure (20) of hHO⅐PH⅐H 2 O (PH ϭ protohemin, with R ϭ vinyls in Fig. 1) identified a localized distal water molecule that bridges the ligated water and Asp-140 carboxylate and led to the proposals (22, 23) that such a localized water molecule may play the crucial role in stabilizing the Fe-O-OH species. It is reasonable that such an important role for a distal water is based on an extensive network of interactions that maintain the optimal Asp-140 side chain orientation. The characterization of such a distal network of hydrogen bonds would identify additional sites for more subtly modulating HO activity by mutagenesis.
Solution 1 H NMR investigations of the hHO and rHO have contributed to our understanding of the enzyme properties (24 -26). By far the most informative NMR spectra were observed (24 -26) for the low spin ferric, substrate-bound, cyanide-inhibited complex, HO⅐PH⅐CN, which serves as a model for HO-heme-O 2 , except that CN Ϫ , in contrast to the bent Fe-O-O unit, prefers to bind normal to the heme (27). The paramagnetism of the iron leads to significant hyperfine shifts (27,28) for the iron ligands and nearby residues that both improve resolution and provide a wealth of molecular and electronic structural details not readily available through other physical means. The informative dipolar shift, ␦ dip , is given by Equation 1 (27)(28)(29)(30), [⌬ ax (3cos 2 ЈϪ1)R Ϫ3 ϩ 3 2 ⌬ rh sin 2 Јcos2⍀ЈR Ϫ3 ]⌫(␣,␤,␦), (Eq. 1) where the position of a nucleus is given by R, Ј, ⍀Ј (xЈ, yЈ, zЈ) in an iron-centered reference coordinate system, ⌬ ax and ⌬ rh are the axial and rhombic anisotropies of the paramagnetic susceptibility tensor, , and ⌫(␣,␤,␦) is the Euler rotation that converts the reference coordinate system, xЈ, yЈ, zЈ, into the magnetic coordinate system, x, y, z, of interest. A number of active site residues in hHO⅐PH⅐CN have been assigned so that the orientation of the magnetic axes (⌫(␣,␤,␥) in Equation 1), can be determined (26). The major magnetic axis, which correlates with the Fe-CN tilt from the heme normal (27), was shown to reflect a large ϳ20°tilt of the ligand in the general direction of the ␣-meso position (26). Thus, two distinct steric influences contribute to the regioselectivity of the heme cleavage: the placement of the distal helix so close to the heme as to block access to all but the ␣-meso positions (20), and the steric tilt of the diatomic ligand toward the ␣-meso positions (25,26). A steric tilt of the Fe-O 2 unit had been proposed earlier on the basis of resonance Raman spectroscopy (31), but the direction of the tilt could not be determined. The NMR studies confirmed an active site structure for hHO⅐PH⅐CN in solution that is very similar to that of hHO⅐PH⅐H 2 O in the crystal (26), with two exceptions. The major (ϳ75%) species in solution has the heme rotated by 180°a bout the ␣-␥-meso axis relative (25,26) to that in the crystal (20), and a distal aromatic cluster appeared closer to the heme in solution (26) than in the crystal. Finally, we observed (26) a series of strongly low field shifted (9 -17 ppm) labile protons that are too weakly relaxed to exhibit any significant hyperfine shifts, and hence their low field bias likely represents unusually strong, and by implication, important, hydrogen bonds (32). These H-bond signals could not be assigned, although the furthest low field shifted labile proton was shown to reside near the distal aromatic cluster (26).
These initial NMR studies (24 -26), moreover, revealed some serious obstacles to extracting the maximum information from the 1 H NMR spectral parameters of the native PH complex. On the one hand, it was shown that the protohemin orientation is equilibrium (static) disordered about the ␣-␥-meso axis in a 55:45 ratio in rHO (24), and in a ϳ3:1 ratio in hHO (25), with the minor form in solution for each case possessing the heme orientation found in the crystal structures (18,20). This disorder has only marginal relevance for HO function, because the location of the ␣-meso position in the protein matrix is conserved. However, the alternate heme orientations induce sufficiently large differences in hyperfine shifts for a large number of the active site residue protons so as to severely complicate the resolution of one-and two-dimensional NMR spectra (24 -26). Moreover, even the individual species with a given heme orientation exhibited a second, dynamic, equilibrium heterogeneity resulting from interconversion of two (or more) species (26). This latter interconversion is fast on the NMR chemical shift time scale but results in severe broadening or complete loss of the signals for numerous residues.
Our interests here are to structurally characterize by NMR a unique HO complex free of heme disorder, which also exhibits significantly reduced line broadening from the dynamic equilibrium heterogeneity. Such a complex can serve as the reference for future solution NMR structural studies of hHO mutants, apo-hHO, and the ligand-free hHO-substrate complex, as well as for other natural genetic variants such as the bacterial HOs (33)(34)(35). The symmetric hemin, 2,4-dimethyldeuterohemin (36), DMDH ( Fig. 1 with R ϭ CH 3 ), provides a single species with narrower resonances than with protohemin, and thereby allows significantly more extensive and definitive assignments as well as structural characterization of residues in the substrate-binding cavity. The more extensive and definitive assignments achievable with DMDH than PH, moreover, serve as a test of the predicted values of ␦ dip generated from the magnetic axes on native hHO⅐PH⅐CN (26) and allow us to unambiguously identify, in large part, the members of the residues involved in the distal hydrogen-bonding network and aromatic cluster.

EXPERIMENTAL PROCEDURES
Protein Sample-The 265-residue soluble portion of hHO was expressed and purified as described previously (7). 2,4-Dimethyldeuteroporphyrin was purchased from Mid-Century Chemicals and the iron incorporated to yield DMDH by standard procedures (36). DMDH was titrated into apo-hHO to a 1:1 stoichiometry in the presence of a 10-fold molar excess of KCN in a 90% 1 H 2 O, 10% 2 H 2 O solution buffered at pH 8.0 with 100 mM phosphate. The final hHO⅐DMDH⅐CN complex was ϳ1.5 mM.
NMR Spectroscopy-1 H NMR data were collected on a Bruker AVANCE 600 spectrometer operating at 600 MHz. Reference spectra were collected in both 1 H 2 O and 2 H 2 O over the temperature range 10 -40°C at a repetition rate of 1 s Ϫ1 using a standard one-pulse sequence with saturation of the water solvent signal. Chemical shifts are referenced to 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) through the water resonance calibrated at each temperature. Non-selective T 1 values were determined in both 1 H 2 O and 2 H 2 O at 20, 25, and 30°C from the initial magnetization recovery of a standard inversion-recovery pulse sequence. The distance of proton H i from the iron, R H i , was estimated from the relation R H i ϭ R* Fe [T 1 */T 1 i ] 1/6 , using the heme for the ␣-meso-H for H* (R* Fe ϭ 4.6Å and T 1 * ϭ 50 ms) as reference. NOESY (37) spectra (mixing time, 40 ms; 10 -40°C) and Clean-TOCSY (38) spectra (25, 35, 40°C, spin lock of 15 and 30 ms) using MLEV-17 (39) were recorded over a bandwidth of 14 (or 28) KHz (NOESY) and 14 KHz (TOCSY) with recycle times of 1 s (or 0.33 s), using 512 t1 blocks of 128 and 250 scans each consisting of 2048 t2 points. Two-dimensional data sets were processed using Bruker XWIN software on a Silicon Graphics Indigo work station and consisted of 30°sine-squared-bellapodization in both dimensions, and zero-filling to 2048 ϫ 2048 data points prior to Fourier transformation.
Magnetic Axes-The magnetic axes (see Fig. 1) were determined by a least-squares search for the minimum in the error function (27,30,40) (Equation 2).
␦ DSS (obs) is the chemical shifts, in ppm, referenced to DSS, for the paramagnetic hHO⅐DMDH⅐CN complex, and ␦ DSS (dia) may be estimated from the available molecular structure for the presumed isostructural hHO⅐PH⅐H 2 O complex (26).

RESULTS
The 600-MHz 1 H NMR spectra of hHO⅐PH⅐CN immediately after preparation ( Fig. 2A) and at equilibrium (Fig. 2B) are compared with that of hHO⅐DMDH⅐CN (Fig. 2C), illustrating the significant improvement in spectral resolution with respect to both the number of resonances and the linewidths when the symmetric DMDH is utilized. The previously assigned (26)  The target residues for assignment are all those with significant hyperfine shift and/or paramagnetic relaxation and for which scalar and dipolar connectivities can be uniquely confirmed over a range of temperatures despite the large size of the complex. Also emphasized are the residues that exhibit unusually large low field peptide NH shifts indicative of strong H-bonding (32), as well as aromatic residues and their aliphatic residue contacts near the active site. We will pursue the assignments, to the degree possible, by standard sequential assignments (44) of NOESY backbone connections among TOCSY-detected spin systems that do not require the crystal coordinates. The large number of new TOCSY connectivities for hHO⅐DMDH⅐CN not observable in hHO⅐PH⅐CN (26) are detailed below, but data are shown only in Supplemental Data (available in the on-line version of this article only). Because numerous target residues (the proximal helix and a portion of the distal helix) had been assigned previously for hHO⅐PH⅐CN (26), we will illustrate 1 H NOESY spectra only for the important new assignments for strongly H-bonded or aromatic residues.
Heme Assignment-The assignment of the prosthetic group in hHO⅐DMDH⅐CN follows standard procedures in low spin ferric hemins/hemoproteins containing native protohemin (27). Thus, NOESY spectra allow the detection of all expected dipolar contacts among TOCSY-connected pyrrole substituents and the meso-Hs in a circular fashion (data not shown) as reported for the metMbCN complex with the same symmetric heme (45). The substrate shifts at 25°C are listed in Table I, where they can be compared with similar data for the same complex of native PH. It is important to note that the symmetric DMDH, unlike PH, does not have a unique numbering system. Hence, we will label the heme methyls in DMDH on the basis of the position of the similarly situated methyl in the major isomer hHO⅐PH⅐CN complex in solution. Because the PH orientation of the major isomer of hHO⅐PH⅐CN (26) in solution is rotated 180°a bout the ␣-␥-meso axis of PH in the crystal (20) of hHO⅐PH⅐H 2 O, our present numbering requires the comparison of the solution 138 substituent environments with the crystallographic 831 positions, respectively. Sequential Assignments-Six fragments, labeled I-VI, could be uniquely identified by sequential backbone NOESY connectivities (see Figs. 3 and 4) among TOCSY-detected (data not shown; see Supplemental Data) spin systems, five of which , and/or ␣ i -␤ iϩ3 connectivities indicative of portions of helices (44).
The Proximal Helix-The eight-residue helical fragment I, , contains the uniquely low field contact shifted axial His-25 backbone as iϩ3 (data not shown; see Supplemental Data). This proximal helix contains the strongly up-field dipolar-shifted side residues Ala-28 (iϩ6) and Glu-29 (iϩ7) that make the expected NOESY contacts to 2-CH 3 and 3-CH 3 . The improved resolution allowed the detection of the complete Val-24, and at least the C ␣ H of Gln-27 in the DMDH, but not in the PH complex (26). The observed NOESY cross-peak pattern is summarized in Fig. 5. The dipolar shifted complete proximal Phe-207 (only ring seen in native complex (Ref. 26)), the Phe-214 ring, and a complete Ile-211 (not resolved in native complex (Ref. 26)) spin systems were detected and assigned on the basis of their expected NOESY contacts to the heme and proximal helix. The chemical shifts for the newly assigned residues are listed in Table II, and the data for the remainder of the proximal helices are provided (and compared with the data for hHO⅐PH⅐CN, in Supplemental Data).
The Distal Helix-The 10-member helical fragment II, is unique to the fragment Tyr-134 -Gly-143, and contains the strongly relaxed and dipolar-shifted signals of Leu-138 -Gly-143, of which the Leu-138 and Gly-139 NHs exhibit the expected dipolar contacts to the 8-CH 3 (data not shown; see Supplemental Data). Two TOCSY-detected two-spin aromatic rings show NOESY cross-peaks to the backbone of residues i and iϩ3, confirming the assignment of Tyr-134 (with ring contacts to 8-CH 3 ) and Tyr-137. The complete Asp-140 with the expected dipolar shifts is identified (see Fig. 4 (A and C)). It is noted that the TOCSY connectivities of neither the backbone of Tyr-134, Tyr-137, nor the side chain of Asp-140 could be observed in hHO⅐PH⅐CN (26). It was not possible to detect TOCSY cross-peaks to the strongly low field shifted and severely relaxed (T 1 ϳ50 ms) Gly-143 NH. However, strong NOESY crosspeaks of N iϩ9 to a TOCSY-detected (not seen in hHO⅐PH⅐CN) pair of strongly low field shifted non-labile protons locate the Gly-143 C ␣ Hs. It was not possible to observe either TOCSY or NOESY cross-peaks from the Gly-139 NH to the C ␣ Hs. Moreover, although it was possible to locate a candidate for the Gly-144 NH by the its expected NOESY cross-peak to the Note that the heme is rotated 180°about the ␣-␥-meso axis in the PH complex in solution relative (35) to that in the crystal. The numbering of DMDH is not unique, but we will adopt a pyrrole numbering system for the methyls in the complex that corresponds to the numbering position of PH bound to HO in solution. The reference coordinate system xЈ, yЈ, zЈ, with zЈ normal to the heme, is shown. The magnetic coordinate system, x, y, z, where the paramagnetic susceptibility tensor is diagonal, is related to xЈ, yЈ, zЈ by the Euler rotation ⌫(␣,␤,␥) by [x, y, z] ϭ [xЈ, yЈ, zЈ]⌫(␣,␤,␥). The angles ␤ and ␣ represent the tilt of the major magnetic axis, zЈ, from the heme normal, and ␣, referenced to xЈ, gives the direction of tilt; the rhombic axes, x and y, are rotated to the reference xЈ, yЈ axis by ϳ ␣ ϩ ␥. The angles ␣ and ␤ can be related to the tilt direction and magnitude for the Fe-CN unit, and can be related to the angle between the axial His-25 imidazole plane and the xЈ axis.
A complete TOCSY-detected Val exhibited the NOESY crosspeak to the heme 5-CH 3 as expected for Val-146. Both N i -N iϩ1 and ␤ i -N iϩ1 cross-peak locate the NH of Leu-147, for which TOCSY encompasses the majority of the predicted, very strongly down-field shifted, side  chain protons. The failure to resolve an additional NOESY cross-peak from a NH to that of Val-146 precluded the location of the Glu-145 NH. The characteristic backbone NOESY crosspeak patterns are summarized in Fig. 5, and shifts for the newly assigned distal helix residues are included in Table II. Less than half of the residues could be sequence-specifically assigned in hHO⅐PH⅐CN (26). The shifts for the previously assigned residues are listed (and compared with those for hHO⅐PH⅐CN) in Supplemental Data.
H-bonding Fragments-The signals of ϳ15 labile protons with insignificant paramagnetic relaxation resonate to the low field of 9 ppm. TOCSY spectra reveal that 2 are Trp ring N ⑀ H and 13 are peptide NHs; 3 fail to exhibit any TOCSY connectivities (data not shown; see Supplemental Data). Several of these strongly low field shifted labile protons are retained briefly upon changing the 1 H 2 O solution to 2 H 2 O (data not shown) confirming their role in H-bonds. The majority of these labile protons will be shown to arise from four fragments, designated III-VI (of which three, III-V, correspond to portions of helices), pertinent to the distal cavity (residues 92-98, 163-167, 57-62, and 84 -87). The chemical shifts of the residues are listed in Table II, and the pattern of dipolar contacts among residues in the four sequential fragments and to the heme and/or distal helix is illustrated schematically in Fig. 5.
The labile protons at 11.71 (iϩ4), 9.31 (iϩ5), and 9.13 ppm (iϩ2) are part of the helical fragment III connectivities, Fig. 3 (A-D), where the backbones of residues iϩ3 to iϩ5 each make contact to a TOCSY-detected aromatic ring. This unique pattern identifies the segment Asp-92-Gly-98 that includes the aromatic rings of Phe-95, Trp-96 (with its N ⑀1 H at 11.71 ppm), and Tyr-97. The inconsequentially hyperfine-shifted segment III members do not, and are not expected to, exhibit dipolar contacts to either heme or the distal helix. A five-member helical segment IV involves labile protons at 12.05 (jϩ1) and 11.58 (jϩ2) with sequential connectivities described by: Leu j -Ala jϩ1 -AMX jϩ2 -AMX jϩ3 -N jϩ4 , and illustrated in Fig. 3 (D-G), with an obvious missing ␣ (under solvent) for jϩ2. The residues exhibit small-to-moderate low field hyperfine shifts but mini-mal paramagnetic relaxation (see Table II). This fragment is uniquely identified as Leu-164 -Phe-167 and the Gly-163 NH, and the backbone of jϩ2 and jϩ3 make the contact to two three-spin aromatic rings of Phe-166 (Fig. 3F) and Phe-167 (Fig. 3C). It is important to note that Phe-166 was initially proposed to be a Tyr because only one ring TOCSY peak could be detected in hHO⅐PH⅐CN (26) Two shorter fragments encompass the low field peptide labile protons 9.06 (kϩ1) of (fragment V) and 13.03 (mϩ1) and 9.46 (mϩ2) (fragment VI). The helical fragment V shows the TOCSY/NOESY connectivity AMX k -Val kϩ1 -Ala kϩ2 -Val kϩ3 -(N) kϩ4 , where AMX k makes contact to a two-spin aromatic ring (see Fig. 4B). This portion is unique to Tyr-58 -Glu-62. The C ⑀ Hs of Tyr-58 exhibit strong NOESY cross-peaks to the lowest field (16.7 ppm), but paramagnetically inconsequentially influenced, labile proton ( Fig. 4F) with no TOCSY peaks that identifies the Tyr-58 ring OH. The Tyr-58 ring exhibits NOESY cross-peaks to both the ring of Phe-166 on fragment IV (Fig.  4B) and the NH of Arg-85 (on fragment VI ( Fig. 3H; see below)) (as expected from the crystal structure) and the Tyr-58 OH displays the expected NOESY cross-peak to Tyr-137 and Asp-140 on the distal helix (fragment II; see Supplemental Data). NOESY cross-peaks from both the Ile-57 C ␦ H 3 and Tyr-58 NH to a two-spin aromatic ring locate the Tyr-114 ring. TOCSY/ NOESY characterize a non-helical fragment VI as AMX m -Z mϩ1 -Z mϩ2 -Ala mϩ3 (Z ϭ long chain), with N-N connectivities shown in Fig. 3 (D, F, and H), where AMX m makes contact to an aromatic ring proton with no TOCSY connectivity; the connectivities are summarized in Fig. 5. Although the fragment is not unique in the sequence of hHO, NOESY cross-peaks between the NH of residues mϩ1 and the distal helix (fragment II) Tyr-137 ring uniquely identify VI as arising from His-84 -Ala-87. The His-84 and Arg-85 NH make the expected dipolar contact to the ring and Tyr-137 on the distal helix (fragment II (Fig. 3, G and H)).
A labile proton at 14.7 ppm exhibits strong NOESY crosspeak to two non-coupled protons in the aromatic spectral window, which identify the C ␦ H and C ⑀ H of the likely His-132 ring (25). Left unassigned in the 9.5-15 ppm spectral window are two labile protons at 9.6 and 9.9 ppm, which exhibit TOCSY peaks to identify peptides (one an Ala), but do not provide contacts to any currently assigned residues and hence remain unassigned at this time.
Aromatic Residues⁄Clusters-The rings of six residues on the H-bonding fragments participate in a cluster with inter-ring dipolar contacts that include two additional interacting Phe rings, the proximity of which to Phe-167, Trp-96, and Leu-164 for one (Fig. 4B) (26), or as deduced here by comparison to hHO⅐DMDH⅐CN. e Protons not stereospecifically assigned, so ␦ dip (obs) and ␦ dip (calc) are determined for both possibilities. f NHC ␣ HC ␤ H 2 fragment detected by TOCSY. The C ␥ HC ␦ H 2 fragment was assigned by its NOESY cross-peak to the Arg-136 N H assigned on the basis of the crystal structure (see text). g TOCSY cross-peak connectivity, NH to C ␤ Hs, but not to C ␣ H or the expected relaxed C ␦ H 3 s were observed. The C ␣ H and the C ␦ H 3S , however, could be assigned by their unique predicted low-field dipolar shifts and NOESY cross-peaks to the NH and C ␤ Hs.
h Severe spectral overlap in both TOCSY and NOESY spectra in this spectral window dictates that these assignments are only tentative. i The C ␣ H signal is under the solvent signal.
with Phe-37 (see Fig. 4B). Both Leu-93 and Ala-94 exhibit NOESY cross-peaks to a complete Trp (N ⑀1 H at 10.8 ppm) which is the only other in hHO, Trp-101, as expected (20). A very narrow obvious methyl group with no TOCSY cross-peaks exhibits the expected cross-peaks to C ␣ H of Leu-164 on fragment III, and the above assigned Phe-47 ring identifies the C ⑀ H 3 of Met-51 (data not shown; see Table II). A weakly upfield shifted complete Val exhibits NOESY cross-peaks to both the Phe-47 and Phe-37 ring (data not shown), and is assigned to Val-42. Both the Tyr-58 ring and OH make strong dipolar contacts with a labile proton (Fig. 4E) with no TOCSY connectivity, which the crystal structure (20) suggests arises from the guanidyl group N ⑀ H of Arg-136 (Table II). The methyls of distal helix Leu-138 make contact with the adjacent Tyr-134 ring and an additional two-spin ring that the crystal structure identifies as Tyr-182 (data not shown). Tyr-182, in turn, exhibits the expected cross-peaks to the rings of two Phe that the crystal structure indicates must be the interacting rings Phe-79 and Phe-178. The rings of Phe-79, Phe-178, and Tyr-182 exhibit strong NOESY cross-peaks to a methyl that must arise from Leu-83. Another very narrow methyl peak with no TOCSY connectivity exhibits the expected cross-peak to the rings of Tyr-134 on fragment II and Tyr-182, and must arise from C ⑀ H 3 of Met-186. The dipolar contacts are shown schematically in Fig. 6, and the chemical shifts for these residues are provided in the Supplemental Data.
Magnetic Axes Determination-The orientation of the magnetic axes in hHO⅐DMDH⅐CN were determined using the conserved magnetic anisotropies of isoelectronic low spin met-MbCN (30), the coordinates of molecule A in the crystal structure of the 233 fragment hHO⅐PH⅐H 2 O complex (20), and the ␦ dip(obs) for hHO⅐DMDH⅐CN. Comparison of the chemical shifts for the two complexes for residues that could be assigned in both reveals the same pattern of shifts for the two complexes (Table II and Supplemental Data). Not surprisingly, the optimized orientation for hHO⅐DMDH⅐CN exhibits a tilt of ϳ20°(␤) and tilt direction ␣ (-z or FeCN vector toward ␣-meso-H), similar to that deduced earlier for hHO⅐PH⅐CN (26). The shifts predicted (see Table II and Supplemental Data Fig. 6S), and observed ␦ dip for distal residues agree reasonably for residues 134 -142, but deviate significantly for all protons for Gly-143 (as observed before (Ref. 26)). For the majority of the residues on sequential fragments III-VI, the dipolar shifts are small (Ͻ ͉0.5͉ ppm) to negligible (see Table II). Thus, the extreme low field positions of labile protons in particular is not related to the paramagnetism of the heme iron, but must arise from strong H-bonds (32).
There are insufficient data available at this time to adequately model the distal helix position by its dipolar shifts. For this, distance estimates from rise curves must be obtained which are impractical at this time. However, some insight into the potential movement of the distal helix is obtained by considering the Gly-143 N p H, for which ␦ dip (obs) ϳ10 ppm is considerably less than the predicted 5 ppm (Table II), although its relaxation (T 1 ϳ 50 ms) is well predicted (26). A ϳ0.5-Å movement by Gly-143 N p H toward the ␣-meso position leads to a negligible change in R Fe that is consistent with the observed T 1 , but to an increase in ␦ dip to ϳ10 ppm, which is consistent with ␦ dip (obs).
Extension of Assignments to hHO⅐PH⅐CN-The definitive assignments achievable for hHO⅐DMDH⅐CN, could, in part, be extended to hHO⅐PH⅐CN, primarily on the basis of conserved NOESY patterns among similarly shifted protons, as listed in Table II and Supplemental Data. This resulted in the reassignment (26) of the aromatic rings of Tyr-58, Tyr-137, and Phe-166 and the identification of a residue as Leu (Leu-164) rather than the original Ala (see "Discussion"). As is apparent in Table II, the chemical shifts for a given residue in the two complexes are very similar for the hyperfine shifted residues, and essentially the same for residues only weakly influenced by paramagnetism.

DISCUSSION
Advantages of the Symmetric DMDH Substrate-The observation in hHO⅐DMDH⅐CN of the same pattern of NOESY crosspeaks between the hemin substituents and both proximal and distal helix residues as in hHO⅐PH⅐CN (26)  modes of substrate binding for native PH and DMDH. Hence, we conclude that information on the molecular and electronic structures, as well as magnetic properties of the hemin deduced for hHO⅐DMDH⅐CN, is also valid for hHO⅐PH⅐CN. However, the 1 H NMR spectra for the former complex are much more informative than for the latter complex for several reasons.
Vastly improved spectral resolution, and hence confidence in and extent of assignments, result upon substituting DMDH for PH. On the one hand, the ϳ25% "minor" heme orientation is abolished by the symmetric heme. However, the more dramatic resolution afforded by DMDH arises from the fact that the intrinsic dynamic heterogeneity of the hHO-substrate complex is more effectively averaged (i.e. more rapid motion), lending to significant line narrowing, and in turn, the detection of numerous TOCSY cross-peaks crucial to the unambiguous, sequencespecific, residue assignment (see Supplemental Data). Thus, the backbone TOCSY connection of distal helix residues Tyr-134, Thr-135, Tyr-137, Asp-140, and Gly-143 is observed in hHO⅐DMDH⅐CN, and, although strong candidates for those signals are now available based on conserved NOESY cross-peaks in hHO⅐PH⅐CN, the crucial TOCSY cross-peaks in this complex still cannot be observed (26). Notable limitations of the two-dimensional NMR spectra of hHO⅐PH⅐CN involve the presently unambiguous assignment of Leu-164 and Phe-166. The former exhibited a single TOCSY C ␣ H-CH 3 cross-peak consistent with an Ala, and the ring exhibited a single TOCSY cross-peak indicative of a Tyr (26). Both residues exhibit additional TOCSY cross-peaks for hHO⅐DMDH⅐CN that trace out a complete Leu and Phe.
On the proximal side, essentially the complete, moderately hyperfine shifted, but only inconsequentially paramagnetically relaxed, Ile-211 (Table II) side chain is observed by TOCSY for the DMDH complex (data not shown, see Supplemental Data), with NOESY peaks to 3-CH 3 and Ala-28 (data not shown) as predicted by the crystal structure (20). For the PH complex, these TOCSY peaks are all undetectable, although the conserved dipolar shifted and NOESY cross-peaks to Ala-28 C ␤ H 3 allow the assignment of a few Ile-211 signals (Table II). Because only selective residues/protons lose their TOCSY connectivity in the PH relative to DMDH complex, we must ascribe the loss of these TOCSY peaks to increased linewidths caused by slower interconversion of the dynamic heterogeneity in the PH than DMDH complex. Fortunately, the resolution for the labile protons (and in large part, backbone and side chain protons) involved in the unique H-bonding network, as well as the ring protons from the aromatic cluster, on the distal side can be identified (although less thoroughly, see Table II) for hHO⅐PH⅐CN on the basis of the present definitive assignments for hHO⅐DMDH⅐CN.
Thus, despite the extensive and unambiguous assignments of the target residues in hHO⅐DMDH⅐CN, as well as the dem- Ile-57, and C ␦ H 3 of Leu-83), which could be uniquely assigned on the basis of the crystal structure are also shown. The dashed lines represent the pairs of residues that exhibit the expected NOESY contact; the asterisk on the dotted line between Phe-167 and Gly-143 indicates an observed weak NOESY cross-peak for which the crystal separation is Ͼ8 Å. onstrated conservation of both hyperfine shifts and inter-residue and heme-residue dipolar contacts between the two substrate complexes, it is still not possible to characterize by 1 H NMR the active site environment as thoroughly for the native substrate as for the symmetric 2,4-dimethyldeuterohemin. We conclude, therefore, that future 1 H NMR characterization of both other natural genetic variants (i.e. bacterial HO and mutants of mammalian HO) will be much more effective and informative using the symmetric hemin DMDH, rather than native protohemin, as substrate. The symmetric hemin, moreover, may also benefit crystallographic studies because it removes one of the contributions (heme orientation) to static disorder in data interpretation (18,20).
However, although the averaging of the dynamic heterogeneity is more rapid for DMDH than PH, making DMDH the superior candidate for NMR study, several residues in hHO⅐DMDH⅐CN remain to be assigned and likely still suffer residual line broadening from the dynamic heterogeneity that precludes their detection. The effect of this dynamic line broadening is observed solely for one heme methyl in hHO⅐DMDH⅐CN, where the 3-CH 3 peak in Fig. 2C is broader than the 2-CH 3 peak; their difference is suppressed at higher temperature. Two residues that must exhibit similar line broadening are Gly-139 and Gly-144. Paramagnetic relaxation is not expected to interfere with TOCSY detection, so the loss of one C ␣ H for Gly-139 and both C ␣ Hs of Gly-144 must be attributed to line broadening. Their location must await planned NMR studies with 15 N/ 13 C-labeled hHO.
Active Site Electronic Structure and Magnetic Properties-The heme contact shift pattern for DMDH is consistent with the spin delocalization into a filled e MO that places larger and comparable spin density at pyrrole positions 2, 3, 6, and 7 than at positions 1, 4, 5, and 8, as expected for an axial His ring orientation essentially collinear with the ␣-␥-meso axis (27,46). This conclusion had been reached earlier for native PH (25,26), but the removal of intrinsic asymmetry of the free substrate (i.e. replacement of vinyls by methyls) makes the 2-fold symmetry much more obvious. The conserved ␦ dip shifts for the active site residues, in particular those of the proximal helix (see Supplemental Data), lead to magnetic axes for hHO⅐DMDH⅐CN, and hence tilt of the Fe-CN unit, by ϳ20°in the direction of the major magnetic or -z-axis (and hence, Fe-CN unit) toward the ␣-meso position as found for hHO⅐PH⅐CN (26). It is noted that the ␦ dip (obs) values for the newly assigned Asp-140, Val-146, and Leu-147 (Table II) are very well predicted by the magnetic axes that were determined on the basis of the ␦ dip (obs) for the proximal residues. The position of Gly-143 relative to the heme in the hHO⅐PH⅐H 2 O crystal structure allows a normal Fe-CϵN bond perpendicular to the heme (Fig. 7A). The proposed ϳ0.5-Å movement of Gly-143 to match the observed dipolar shift (and at least portions of the distal helix) toward the iron results in a position for Gly-143 that sterically tilts the Fe-CϵN unit toward the ␣-meso position, as shown in Fig. 7 (B and C). Similar steric tilt of the Fe-O-OH would contribute significantly to the stereoselectivity of heme cleavage at the ␣-meso position (31). The determination of a more definitive position of the distal helix over the heme must await the determination of NOESY rise curves and the assignment of more of the dipolar contacts to the ligated helix.
Active Site Molecular Structure-The pattern of the intraresidue NOESY connections among the five distal residue frag-   ments II-VI deduced completely independently of the crystal structure, as depicted schematically in Fig. 6, as well as the few residues with connections to two or more of the fragments or to the heme (Phe-47, Val-42, Phe-37, Trp-101) the assignment of which was affected on the basis of the crystal structure, confirm an active site environment that is largely consistent with the same structure of the cyanide complex in solution as for the aquo complex in the crystal (20), with three exceptions. First, the orientation of PH differs in solution and crystal, as discussed previously (26), but this point is moot in the DMDH complex. Second, the distal helix Gly-143 NH and, by inference, at least portions of the distal helix are closer to the iron in solution than crystal for hHO (20), with the solution position more closely resembling that in the crystal of rHO (18). However, the finding of the distal helix closer to the iron in solution than crystal is likely not functionally relevant for hHO, because a similar difference in positions of the distal helix is found in the crystals of hHO (20) and rHO (18), for which activities are essentially the same (7,47). Instead, different positions of the distal helix more likely reflect the intrinsic flexibility of at least portions of the helix for which the exact position is easily perturbed by small environmental influences. The crystal structure of hHO⅐PH⅐H 2 O yielded two molecules in the unit cell with slightly different positions of the distal helix and with part of the distal helix backbone exhibiting disorder (20).
The third difference involves the proximity of the Phe-167 ring, and possibly the Tyr-58 ring (see below), to the distal helix, as reflected in weak NOESY cross-peaks (ϳ5-6 Å) between the ring C ⑀ H and C ␦ H to the Gly-143 NH observed in both hHO⅐PH⅐CN (26) and hHO⅐DMDH⅐CN (data not shown), where the crystal coordinates predict distance Ͼ8 Å, and the very strong H-bond between the Tyr OH and carboxylate of Asp-140 (see below). The other contact between fragment IV and the distal helix (Phe-166 ring and Asp-140) is expected (ϳ4 Å). The likely movement of the distal helix closer to the iron by ϳ0.5 Å would not shorten the Phe-166 ring to Gly-143 distance significantly. A possible role of the distal ligand as an H-bond acceptor in modulating the interaction between the distal helix and the aromatic cluster/H-bonding network will be considered below.
Distal H-bonding Network and Aromatic Interactions-The critical role of Asp-140 in HO activity (13,22,23), although connected to the heme ligand in the hHO⅐PH⅐H 2 O crystal structure solely by a H-bonded water (18,20), emphasizes the potential important role of H-bonding interactions in facilitating the formation of the Fe-O-O-H rather than FeϭO activated species. Numerous peptide NHs and side chain-labile protons exhibit unusually low field shifts (Tables II and III). Because the magnetic axes predict small (Ͻ0.5 ppm) to negligible dipolar shifts for the assigned labile protons, their strong low field bias must be interpreted as reflecting strong hydrogen bonds (32). The presently characterized strong H-bonding network among residues in fragments II-VI, and between fragments IV, V, and VI and the distal helix (II), and the fact that the spatial arrangements of these fragments qualitatively conform to the prediction of the crystal structure (20) (see Fig. 6) allows the use of the crystal structure to identify the candidates for Hbond acceptors to the low field assigned labile protons.
The structural constraints for strong H-bonds are inter-heteroatomic distances N-H⅐⅐⅐O of ϳ2.5 Å and a near linear N-H⅐⅐⅐O configuration (32,48). The crystal structure (20) does not reveal such sites, but for each labile proton, it identifies a unique candidate, as listed in Table III. The H-bonding distances and angles in Table III are far from ideal for a strong H-bond (32,48). However, the available resolution of the structure allows for significant uncertainties in the positions of the relevant atoms, and it is very likely that there is some structural rearrangement in the geometries of at least the side chains of residues on the distal fragments II-VI (see above) upon replacing a ligated water H-bond donor in the crystal with diatomic ligated H-bond acceptor such as cyanide (ore molecular oxygen). The fact is that the strong H-bonds exist, as reflected in the strong low field bias to the labile proton chemical shift (32) such that more "ideal" H-bond geometries must exist in both hHO⅐DMDH⅐CN and hHO⅐PH⅐CN. Moreover, as observed in the weak Phe-167-Gly-143 dipolar contact, it is likely that the distal fragments III-VI may all move slightly with respect to the distal helix in the cyanide complex in solution relative to the aquo complex in the crystal.
The locations of the resulting H-bonds among the distal fragments II-VI are displayed schematically in Fig. 8. First and foremost, the Tyr-58 side chain OH serves as a donor to the carboxylate of the catalytically critical Asp-140. Although the geometry of the H-bond is not very favorable in the crystallographic orientation of the carboxylate, a ϳ45°rotation of the carboxylate (or a small movement of the Tyr ring) leads to a much more favorable H-bond interaction (Table III) without interfering with the simultaneous H-bond between the Arg-136 N ⑀ H and Asp-140 carboxylate (Table III). Second, the other identified strong H-bonds occur exclusively between donors and acceptor of different fragments II-VI (see Fig. 8), indicating that one of the prime roles of the H-bond network is to stabilize the relative positions of these fragments. The "center" of this network appears to be fragment V containing Tyr-58. The helix backbone exhibits the greatest dynamic stability (49) in that the peptide NHs of both Tyr-58 and Val-59 exhibit their respective NOESY cross-peak with undiminished intensity 6 months after dissolution of the complex in 2 H 2 O (data not shown).
The different protein segments for which relative positions are stabilized by strong H-bonds may also be stabilized by interaction of aromatic rings. Thus, fragments IV and V interact via Tyr-58 and Phe-166 ring contacts, whereas the Phe-47 ring interacts with rings on fragment III (Trp-96) and fragment IV (Phe-167), as shown in Fig. 8. The mutation of Asp-140 has already been shown to adversely affect the stability of the Fe-O-OH species (13,22,23). The present NMR analysis makes the proposal that the mutation of the side chains of Glu-62, Glu-92, and/or Tyr-58, candidates not obvious from the crystal structure alone, would similarly, but less severely, affect HO activity by disrupting the distal H-bonding network.
It is noteworthy that our present assignments include not only the donor residues, but also the key acceptor residues (i.e. Glu-62, Asp-92, Asp-140) in this network, setting the stage for future NMR studies of the effect of mutation on each of these H-bonds. The acceptors to the four strongest peptide NH Hbonds, Glu-62 and Asp-92, are conserved among mammalian HO and are retained in some (33), but not other (34), bacterial HO. If the water molecule H-bonded to Asp-140 indeed serves as a stabilizing influence for the Fe-O-OH in the activated form of the enzyme, then the network of strong H-bonds can be viewed as a scaffold for facilitating the correct orientation of the Asp-140 side chain for optimal catalysis.
The presently characterized H-bonding network, quite aside from being potentially relevant to HO activity, exhibits some unique properties. Thus, the strong low field bias for the chemical shift of the NH and OH indicates strong H-bonding (32) that is consistent with its presence in both the crystal and solution. Moreover, several of the NHs, including the side chain of Trp-96 and a His, as well as several peptide NHs, exchange sufficiently slowly to be detected in 2 H 2 O (data not shown). The strength of the H-bonding, as reflected both in the chemical shift (32) and their distinctly reduced lability, is all the more remarkable because the crystal structure indicates that these residues are located close to the protein surface. Finally, although the 1 H NMR spectra and the two-dimensional maps differ strongly between the apo-HO and the substrate-bound complexes (25), the H-bonding network and aromatic cluster, as reflected in the strong low field shifts, and the aromatic cluster as described herein, appear largely conserved. 2 The portion of the hHO⅐PH⅐H 2 O crystal structure (20) that encompasses the presently described H-bond network/aromatic cluster in human heme oxygenase is shown in the stereo view of Fig. 9. The strongly interacting and highly stabilized H-bonding/aromatic cluster network serves to form a wall that interacts with the distal helix, and this wall may be retained in the apo-HO to serve as a template for the distal helix to condense upon the binding of the substrate. Preliminary studies of a complex 3 of a bacterial HO⅐PH⅐CN complex (33,50) suggest that the distal H-bond and part of the aromatic cluster network are, in large part, present and may have general role in HO activity.
Mechanistic Implications-As argued previously, partitioning of the ferric hydroperoxide intermediate between hydroxylation of the ␣-meso carbon of the porphyrin ring and formation of a ferryl (Fe IV ϭO) species represent a critical branchpoint in the function of the family of catalytic hemoproteins (11)(12)(13). In the case of peroxidases and cytochrome P450 enzymes (21), for example, the distal active site environment provides residues that hydrogen bond, directly or through intervening water molecules, to the oxygen of the peroxo ligand that is distal to the iron. This hydrogen-bonding pattern facilitates scission of the oxygen-oxygen bond to produce the critical ferryl intermediate. In contrast, the role of the ferric hydroperoxide as the actual oxidizing agent in the case of heme oxygenase appears to require an alternative hydrogen-bonding scheme in which the hydrogen bond to the distal oxygen, if any, is relatively weak. Indeed, it is likely that proton delivery to the oxygen coordinated to the iron facilitates electrophilic porphyrin hydroxylation by making the iron-bound oxygen a better leaving group. The stringent hydrogen-bonding pattern observed in the present NMR experiments is likely to be critical in the fine control of proton delivery that helps to channel the reaction toward the heme oxygenase outcome rather than the formation of a ferryl species. This inference is consistent with the finding that perturbation of the distal hydrogen-bonding pattern by mutation of Asp-140 (22, 23), or of the flexibility of positioning of the distal helix by mutation of Gly-139 or Gly-143 (51), leads to formation of a ferryl species and the acquisition of peroxidase activity at the expense of heme oxygenase activity.