Structural basis for human NADPH-cytochrome P450 oxidoreductase deficiency

Abstract

NADPH-cytochrome P450 oxidoreductase (CYPOR) is essential for electron donation to microsomal cytochrome P450-mediated monooxygenation in such diverse physiological processes as drug metabolism (approximately 85–90% of therapeutic drugs), steroid biosynthesis, and bioactive metabolite production (vitamin D and retinoic acid metabolites). Expressed by a single gene, CYPOR’s role with these multiple redox partners renders it a model for understanding protein–protein interactions at the structural level. Polymorphisms in human CYPOR have been shown to lead to defects in bone development and steroidogenesis, resulting in sexual dimorphisms, the severity of which differs significantly depending on the degree of CYPOR impairment. The atomic structure of human CYPOR is presented, with structures of two naturally occurring missense mutations, V492E and R457H. The overall structures of these CYPOR variants are similar to wild type. However, in both variants, local disruption of H bonding and salt bridging, involving the FAD pyrophosphate moiety, leads to weaker FAD binding, unstable protein, and loss of catalytic activity, which can be rescued by cofactor addition. The modes of polypeptide unfolding in these two variants differ significantly, as revealed by limited trypsin digestion: V492E is less stable but unfolds locally and gradually, whereas R457H is more stable but unfolds globally. FAD addition to either variant prevents trypsin digestion, supporting the role of the cofactor in conferring stability to CYPOR structure. Thus, CYPOR dysfunction in patients harboring these particular mutations may possibly be prevented by riboflavin therapy in utero, if predicted prenatally, or rescued postnatally in less severe cases.

The role of human NADPH∶cytochrome P450 oxidoreductase (CYPOR) as the obligate electron donor to each of the many known microsomal cytochromes P450 that catalyze steroidogenesis and xenobiotic metabolism, as well as other monooxygenase activities, such as heme and squalene oxygenation (1, 2), identifies this enzyme as a protein of great interest in biology. Twenty-six missense mutations/polymorphisms of the POR gene, encoding CYPOR, have been described by Miller’s laboratory in patients with Antley–Bixler syndrome (ABS) and/or disordered steroidogenesis (3, 4). Others have also described such mutations (5–7). ABS (8) is a disorder characterized by severe midface hypoplasia, humeroradial synostoses, bowing and fracture of femora, and other malformations. Miller’s group (3) discovered a correlation between six allelic variants of the POR gene (encoding CYPOR, EC 1.6.2.4) and analyzed patients with disordered steroidogenesis, with and without ABS, identifying five missense and one nonsense mutations in the POR gene. Of the five missense mutations, three corresponded to the ABS phenotype: A287P, R457H, and V492E. The authors concluded that deleterious POR mutations could account for decreased lanosterol demethylase (CYP51) activity previously observed in ABS patients and were sufficient to cause the ABS phenotype in the absence of fibroblast growth factor receptor type 2 mutations (3, 4, 9). Retardation of somite and limb bud formation, observed previously in embryonically lethal CYPOR knockout mice (10, 11), supports this hypothesis. Subsequent work by these investigators (4, 12, 13) and our laboratories has supported this hypothesis by demonstrating the impaired activities of these and other mutant CYPOR preparations in purified, reconstituted systems (14) and engineered bacterial membrane systems (15, 16).

To address the biophysical and structural basis of the observed loss of CYPOR function in ABS patients, the soluble diflavin-containing catalytic domains of human CYPOR (WT, V492E and R457H) lacking 66 N-terminal amino acids were expressed in Escherichia coli. This construct was chosen based on the structure of the soluble domain of the rat enzyme (17), with which it bears 92.5% sequence identity, but human CYPOR contains three additional N-terminal residues and is missing H621 (rat numbering). It was not certain, a priori, whether the detailed structures, especially of the NADP(H)-binding domain and the residue(s) that might interact with the FAD isoalloxazine ring, would be the same, due to the deletion (H621) in the middle of helix R in the human enzyme compared to the rat enzyme (see Fig. 1, Inset, for the location of H621). Although this deletion site is not directly involved in FAD binding, the possibility exists that this mutation could influence the C terminus, in which W679 (the penultimate residue that stacks with the isoalloxine ring) is crucial for controlling hydride transfer from the nicotinamide ring of NADPH to FAD. In addition, because the original rat CYPOR structure was determined only at 2.6-Å resolution, it is extremely desirable to have a higher resolution structure.

Fig. 1.

Structure of human CYPOR and interactions between FAD and polypeptide. (A) Overall structure. The molecule consists of three domains, the FMN domain (blue), the connecting domain (gray), and the FAD/NADP(H) domain (green). Cofactors (FMN, blue; FAD, brown; and NADP⁺, red) are shown as stick models and mutation sites are marked with pink balls. Residue H621 (pink arrow) in rat CYPOR is missing in the human enzyme. (B) Interactions between FAD and the polypeptide observed in human CYPOR structure. Hydrogen bonds and salt bridges are shown with dotted lines and eye lashes indicate hydrophobic interactions. The FAD ring is sandwiched between W679 at the re-face and Y459 at the si-face. The dimethyl benzene ring of the isoalloxazine ring is in close contact with the FMN, and the pyrimidine side of the ring makes a tight hydrogen-bonding network with the polypeptide. The pyrophosphate group of FAD makes salt bridges with R457 and hydrogen bonds with the main-chain amide nitrogen of the peptide segment including V492–T494.

Previous studies showed that CYPOR activity measurements with V492E Δ66, as purified, lacked enzyme-bound FAD and retained a small fraction of the catalytic efficiency of the wild-type enzyme in NADPH-dependent cytochrome c reductase assays (14). V492E Δ66 was fully reconstituted upon addition of FAD, after purification, as shown by flavin analysis, cytochrome c reductase activity, and circular dichroism measurements. Prostaglandin E1 (PGE₁) ω-hydroxylation activity, supported by the full-length (holo) constructs with CYP4A4, was also “rescued” for both Y459H (also lacking FAD) and V492E by the addition of FAD (14). The potential of using riboflavin therapy to correct CYPOR deficiencies due to flavin-binding defects in certain human CYPOR polymorphisms was suggested in these studies. In addition, Kranendonk et al. (15) demonstrated in E. coli, engineered with relative amounts of CYPOR and CYP1A2 to approximate those observed in mammalian microsomal systems (CYPOR∶P450 = 1∶5–10), that O-dealkylation reactions and the mutagenic capacities of three different mutagens supported by V492E compared to wild-type CYPOR, were impaired, demonstrating the inability of this CYPOR variant to support drug-metabolizing P450s.

Here we report the crystal structures of the entire soluble portion (Δ66) of human wild-type CYPOR and two variants, V492E and R457H. Although the structure of rat CYPOR was determined some time ago (17, 18) and a modeled structure of the human enzyme has been used for the analyses of human mutations (19, 20), our structures represent experimentally determined structures of human CYPOR proteins. The global folds of the mutant proteins are the same as that of the wild type (Fig. 1); however, each of these variant structures reveals the structural basis for protein instability and/or abnormality in cofactor binding, as well as for the phenotypic expression of CYPOR deficiency. The implications of these findings in understanding the intricacies of protein folding/unfolding (present study), the possibility of riboflavin therapy for treatment of patients with these enzyme deficiencies (14–16), and the importance of understanding the actual structural aberrations resulting in catalytic and, thus, phenotypic outcomes has become even more critical.

Results and Discussion

Kinetic Properties of Mutants.

Wild-type and mutant proteins were assayed for cytochrome c reduction activity without and with exogenous FAD supplement (Table 1). Whereas in the previous study, V492E catalyzed only approximately 9% of the wild-type activity and exhibited a comparable (14) in the absence of FAD, R457H exhibits approximately 32% of WT cytochrome c reductase activity (Table 1) and a similar affinity for NADPH as wild-type CYPOR (wild-type ; R457H ) in the present experiments. These values are in sharp contrast to the previously reported values for both mutants (4), which stated that V492E exhibited 7% of wild-type cytochrome c reduction with a of > 50 μM and that R457H has only 8% of the wild-type V_max for cytochrome c reductions with a . Whether these discrepancies are due to the different cloning constructs for heterologous expression is unknown. The fact that the R457H mutant has a similar affinity for NADPH as wild type in our studies suggests that the NADPH binding site in the mutant protein remains almost the same as wild type. These catalytic results will be confirmed in the structural studies to follow. Cytochrome P450 4A4-mediated ω-hydroxylation of PGE₁ reconstituted with the V492E mutant holoenzyme was likewise compromised to a level of approximately 9% that of the wild-type holoenzyme, and this monooxygenation activity could be reconstituted by the addition of FAD to approximately 58% of wild-type activity (14). The mutant form of CYPOR, V492E, was reported by Huang et al. (4) to be deficient in 17α-hydroxylation (ca. 7% of wild type), as was the R457H mutant at approximately 5%. R457H Δ66 was also found to be deficient, as was the full-length enzyme expressed in engineered E. coli (15) bearing both reductase (mutant 17% of wild-type cytochrome c reduction) and a specific cytochrome P450 (mutant 19% of wild-type methoxyresorufin O-deethylation catalyzed by CYP1A2). In addition, the flavin contents of the wild-type, R457H, and V492E Δ66 CYPOR preparations were determined (Table 1), showing that the protein∶FAD∶FMN contents were approximately 1∶0.3∶0.9 for the R457H variant compared to 1∶0.04∶0.74 for the V492E mutant. The enzymatic activities could be restored to varying extents by the addition of FAD, but not FMN, as would be expected.

Table 1.

Flavin contents and cytochrome c reductase activity

	Flavin content		Turnover, min-1		, μM
	[FAD]/[protein]	[FMN]/[protein]	−FAD	+FAD
Wild type	0.86 ± 0.26	0.8 ± 0.23	2385 ± 101	2,622 ± 165	0.0009
R457H	0.3 ± 0.12	0.9 ± 0.35	748 ± 83	1,811 ± 72	0.007 ± 0.001
V492E	0.04	0.74*	13.3 ± 0.8	2,323 ± 87	0.39 ± 0.09

Flavin contents were measured as described in Methods and were means of five measurements from three independent protein purifications ± SD. Rate of cytochrome c reduction for each protein was determined with and without FAD ([FAD] = 5x[protein]). (where K_app act is the apparent activation constant) was calculated by fitting the rates of cytochrome c reduction at different FAD concentrations to a one-site binding equation.

^*This experiment was performed on a different preparation to confirm our published observations (14); n = 1.

Both V492E and R457H Mutants Lack FAD and Are Unstable in Solution.

As previously described, although V492E contains a stoichiometric amount of FMN, it lacks FAD (< 5% of WT) (14). The R457H mutant also lacks FAD but contains approximately 35% of the wild-type value (Table 1). Furthermore, both mutants are relatively unstable. In order to assess the stability in solution, a limited trypsin digestion was performed. Fig. 2 shows trypsin digestion patterns of V492E and R457H in the absence and presence of exogenous FAD. Although the wild-type protein and V492E in the presence of excess FAD were stable even after 4 h of trypsin treatment, the V492E protein without FAD was cleaved to two major fragments after only 15 min of trypsinolysis, and most of the intact protein was degraded by 60 min. Interestingly, two major bands corresponding to molecular masses of 48 kDa (band a, Fig. 2) and 22 kDa (band b, Fig. 2) were obtained, and their N-terminal sequence analyses were shown to be GSHMV-67... (corresponding to the N terminus of the protein; the first four residues, GSHM, are from the cloning of the thrombin site) and 488-INKGEAT…, respectively. This indicates the first major cleavage point to be at R487, which is located near the tip of the β-turn-β region (β-flap) preceding E492 (Fig. 3). This result is consistent with the structural analysis that the β-flap containing R487 is flexible in V492E, making the Arg residue accessible to trypsin. In contrast, a large portion of the R457H protein was unstable and degraded to multiple fragments within 15 min of the trypsin treatment. However, about one-third of the mutant protein was stable even after 4 h of the trypsin treatment. This result is consistent with the FAD content in the R457H mutant—i.e., approximately 35% of the R457H protein contained endogenous FAD—and that this population of the mutant protein was stable and resistant to the limited trypsin digestion. As in the case of V492E, R457H was also stable in the presence of excess FAD, indicating that the R457H protein can be reconstituted with exogenous FAD to form a stable protein. Furthermore, the fact that the R457H mutant protein was degraded at multiple sites to smaller fragments suggested that the mutant protein without bound FAD exists in more open conformation(s) compared to the V492E variant.

Fig. 2.

SDS-PAGE analysis of the limited trypsin digestions of V492E, R457H, and wild type of CYPOR. The protein: trypsin ratio by weight was 100∶1, except for lanes with asterisks performed at a ratio of 1,000∶1. At indicated times, aliquots were taken for SDS-PAGE analysis. Molecular weight markers in V492E digestion are 1, 100 kDa; 2, 75 kDa; 3, 50 kDa; 4, 37 kDa; 5, 25 kDa; 6, 20 kDa; and in R457H or wild-type CYPOR digestion are 1, 70 kDa; 2, 50 kDa; 3, 40 kDa; 4, 30 kDa; 5, 25 kDa. N-terminal sequence analyses of V492E trypsin digestion products yielded GSGMV67… for band a; and 488INKGEAT… for band b, indicating that the first major cleavage site in V492E was at R487.

Fig. 3.

The β-flap in molecule B of V492E is disordered. Electron densities in the vicinity of FAD and E492 in molecule A (A) and molecule B (B) of the V492E structure. The omit (F_o - F_c) map electron density was contoured at 2.5σ level. The conformations of E492 are different in the two molecules. The densities for the ribityl-adenine portion of FAD and the β-loop-β region (V479–G491; shown in green ribbon) are well defined in molecule A, whereas those in molecule B are barely traceable.

Near UV-Visible CD Studies.

The R457H and WT preparations were subjected to UV-visible (UV-vis) CD spectroscopy (Fig. 4), performed previously with the V492E variant. For comparison, the normalized spectra of WT and V492E (14), and of Y181D (16), are shown. The contribution of bound FMN (represented by the spectrum of V492E) was subtracted from WT and R457H spectra (see Inset) in order to show the diminished FAD content of R457H, indicated by the decreased molar ellipticity at approximately 375 nm. Here, the FAD content can be estimated to be approximately 50%. The FAD content of the R457H mutant varies somewhat depending on enzyme preparations, approximately 30–50% of wild type, consistent with the FAD amount measured by the HPLC method (Table 1).

Fig. 4.

UV-vis circular dichroism spectroscopy of CYPOR variants. The CD spectrum of R457H ([protein] = 60 μM in MOPS buffer) was collected in the near UV-vis range. For comparison, the normalized spectra of WT and V492E (14), and of Y181D (16), are shown. Y181D is included to demonstrate the general spectral characteristics of CYPOR-bound FAD, because Y181D contains negligible FMN levels. The sum of Y181D + V492E spectra is also shown for comparison to WT. The contribution of bound FMN (represented by the spectrum of V492E) was subtracted from WT and R457H spectra (see Inset) in order to demonstrate the diminished FAD content of R457H, as indicated by decreased CD at approximately 375 nm.

Structure of Wild-Type CYPOR.

The wild-type structure used for comparison in the subsequent studies contains two innocuous polymorphisms that were in the original clone sent from the American Type Culture Collection (ATCC). Although the structure is identical to that obtained with the corrected polymorphisms (described in SI Methods), the degree of resolution is substantially improved (P228L/A503V vs. corrected wild type is 1.75 vs. 2.30 Å). The overall fold of wild-type CYPOR is essentially the same as that of the rat enzyme (Fig. 1A), except that H621 in rat CYPOR is absent in the human enzyme, making the tail end of helix Q in the human enzyme one residue shorter than the corresponding region in rat CYPOR. The rmsd between the rat and human structures is 0.78 Å for the 612 visible C^α atoms. The rmsd between the two molecules in the asymmetric unit of the human structure is 0.85 Å, whereas those of the individual FMN and FAD/NADP⁺ domains between the two molecules are 0.23 and 0.51 Å, respectively. Furthermore, when the two molecules in the asymmetric unit are overlaid by superimposing their FAD domains only, the rmsd value for the two FMN domains is 1.5 Å (Fig. S1). This result suggests that the relative orientations of the two domains in the two structures in the asymmetric unit are slightly different, consistent with the notion that the FMN domain is flexible relative to the FAD/NADP⁺ domain, as observed in various rat CYPOR structures (18, 21). As observed in the rat CYPOR structure, the ribityl-nicotinamide moiety of the bound NADP⁺ is disordered in both molecules in the asymmetric unit of the human CYPOR structure. Other disordered regions are the loop containing residues 504–508 and part of the hinge region (240–245) that connects the FMN and FAD domains in the A molecule; the hinge is also disordered in the B molecule. The resolution of the wild-type human CYPOR structure is the highest (1.75 Å) for all CYPOR structures thus far determined, revealing more detailed structure, including tightly bound water molecules that are involved in an intricate hydrogen-bonding network in cofactor binding and the hydride transfer reaction (e.g., Wat2 and Wat3 in Fig. 1), and possibly in interactions with its partner molecules. Thus, this structure will serve as the basis for comparisons of future human mutant CYPOR structures.

FAD-Binding Site.

The binding of cofactors, FAD, FMN, and NADP⁺, is also very similar to that observed in the wild-type rat CYPOR, except for water molecules (Wat2 and Wat3 in Fig. 1B) that were not visible in the rat structure at a lower resolution. Fig. 1B shows the detailed interactions between FAD and the polypeptide. The isoalloxazine rings of the two flavins are juxtaposed to each other at their C7- and C8-methyl groups. The isoalloxazine ring of FAD is sandwiched between W679 at its re-face and Y459 at its si-face. The pyrimidine ring of FAD is interacting with the main-chain carbonyl groups of the residues in the loop containing I474–V477. The 2′-OH of the ribityl group is within hydrogen-bonding distance to the carbonyl oxygen of Y458, while the amide nitrogen of Y458 makes hydrogen bonds with both the ribityl 2′- and 3′-OH groups. The pyrophosphate group of FAD forms a salt bridge with the guanidinium group of R457 and hydrogen bonds with the main-chain atoms of the β-strand containing V492-T494, thus demonstrating the relationship between R457 and V492. The adenine ring of FAD is stacked with the Y481 phenolic ring. Thus, imbalance of any of these intricate interactions would be predicted to diminish the FAD binding to the polypeptide, resulting in an unstable and inactive enzyme.

Except for the mutant rat CYPOR structure with an engineered disulfide bond (22), all known mammalian CYPOR structures that have been reported so far have a bound nucleotide (either NADP⁺ or 2′-AMP). We attempted to crystallize the apo form of the human enzyme—i.e., without any pyridine nucleotide bound. Although the protein has not been exposed to exogenous nucleotide [neither NADP(H) nor 2′-AMP] during purification and/or crystallization, the resulting structure revealed the ADP-PP_i moiety (presumably of NADP⁺), suggesting that the protein was bound to endogenous NADP(H) from the E. coli cell and was purified as the CYPOR-NADP⁺ complex. Because the intracellular concentration of NADPH in E. coli is > 100 μM (23) and K_m (K_d) for NADPH is < 5 μM, it is not surprising to have bound NADP(H) in CYPOR proteins purified from E. coli cells. As was the case in the wild-type rat CYPOR structure, however, the nicotinamide moiety of NADP(H) is disordered.

Structure of V492E.

The initial crystals of V492E mutant were obtained without addition of exogenous FAD in the crystallization medium, resulting in small crystals under a heavy layer of precipitation in the crystallization dips. The crystals diffracted to approximately 2 Å resolution and the resulting structure contained a near full occupancy of FAD. Apparently, the majority of the protein molecules that did not contain FAD had precipitated and the small subpopulation (< 10% of total by the FAD content measurement; Table 1) that contained a nearly full complement of cofactors was crystallized. Therefore, in order to obtain better crystals, we reconstituted the V492E mutant protein with FAD and crystallized it under the same conditions. The crystallization dips had less precipitation and the resulting crystals grew larger and diffracted to a higher resolution (1.80 Å). Both sets of crystals have the same space group (P2₁2₁2₁) with the same cell dimensions and the same structure (Table S1). Thus, only the higher resolution structure is discussed here.

The β-Flap Consisting of Residues V479-E492 Is Flexible.

The overall structure of the V492E mutant is very similar to that of wild type with an rmsd of 0.5 Å. However, upon closer examination, there are subtle, but significant, differences in the mutant structure that are reflected in the activity and cofactor content measurements (14) and also reproduced in the present studies. Fig. 3 shows electron densities in the vicinity of the adenine-pyrophosphate binding site of the V492E structure. In one of the two molecules in the asymmetric unit (Mol A), the electron densities show that the adenine-pyrophosphate moiety of FAD and the β-turn-β (β-flap) of the polypeptide, containing residues V479–E492 (V492 in WT), are well defined as in the wild-type structure. The two β-strands form an antiparallel β-finger, and residues K484 and R487 at the tip of the β-flap make salt bridges with E425 and D436, respectively, of the connecting domain (gray ribbon in Fig. 3A), resulting in a well-defined, firmly packed region and making it possible for the phenol ring of Y481 to stack onto the adenine ring of FAD (Fig. 3A). Furthermore, the side chain of E492 is also in the same position as V492 of the wild-type structure, but is now within hydrogen-bonding distance of the amide nitrogen of N329 (Fig. S2). Thus, the protein-FAD interactions are completely conserved as in wild type (see above and Fig. 1, vicinity of FAD). In contrast, electron densities for the corresponding regions in the other molecule in the asymmetric unit (Mol B) are disordered (Fig. 3B)—i.e., residues from E482 through K490 are not traceable—indicating that the β-flap is mobile, which in turn results in weaker interactions between the entire β-flap and the helix-turn-helix region (E425–D436) of the connecting domain. Furthermore, the flap also contains Y481, whose phenol ring stacks onto the adenine ring of FAD, and T494, within hydrogen-bonding distance of the pyrophosphate moiety of the bound flavin (Fig. 1). Most significantly, the side chain of E492 is flipped inward from the V492 position and makes a salt bridge with R457, weakening the interactions between R457 and the pyrophosphate of FAD and resulting in weaker binding of FAD to the polypeptide (Fig. 3B and Fig. S2B). In addition, the disordered β-flap, which includes Y481, will weaken the binding of the adenine portion of FAD. Taken together, the V492 mutant protein in solution is loosely packed, binds FAD less tightly, and is less stable. This observation is entirely consistent with the results of the limited tryptic digestion (Fig. 2), in which the first cleavage site is R487, in the middle of the flexible β-flap and with the lower activities reported previously by Fluck et al. (3) and in Marohnic et al. (14), as well as in the present studies.

The facts that better crystals of V492E were obtained upon the addition of FAD with less precipitation and reconstitution of V492E variant activity is possible in solution indicate the population of enzyme in these preparations was mixed and selective crystallization of the enzyme containing FAD had occurred. This fact further indicates that when FAD is added to this FAD-deficient population, the binding of flavin occurs, and catalytic function is restored to a large degree. Because Mol A in the asymmetric unit adapts the WT structure (Fig. 3A) in which the β-flap is well defined, it is assumed that this is the structure of the mutant enzyme when it has been fully reconstituted with excess FAD. On the other hand, in the absence of the excess FAD, the mutant molecule adopts a conformation similar to the one observed in Mol B (without bound FAD) with the β-flap “flapping.” Mutation of one residue at the base of the β-flap causes the instability of the entire β-flap structure, which in turn is sufficient to alter the binding of FAD to this human CYPOR variant. This causes disruption of the monooxygenation activities dependent upon CYPOR as an electron donor. Thus, the observed structures, together with the limited trypsin digestion results, indicate that the mutant exists in at least two different conformers in dynamic equilibrium in solution.

Structure of R457H.

The overall structure of R457H is essentially the same as wild type, except for the mutation at the 457 position. The rmsd between the mutant and wild-type structure are 0.33 Å for the A molecule and 0.49 Å for the B molecule of the two molecules in the asymmetric unit. The slightly higher rmsd values for the B molecule are consistent with the fact that the relative position of the two flavin domains of the B molecule is more variable than the A molecule, as seen in other structures of CYPOR (18). R457 is located on a long β-strand of the β-sheet that links the FAD-binding domain and the connecting domain and cradles the isoalloxazine-ribityl-PP_i moiety of FAD (Fig. 1). In the wild-type structure, R457 makes salt bridges with the pyrophosphate moiety of FAD (2.9 and 3.0 Å between the guanidinium group of R457 and PP_i; Fig. S2A), whereas H457 makes hydrogen bonds with the pyrophosphate (3.2 and 3.5 Å; Fig. S3). Thus, the mutant with two hydrogen bonds has weaker interaction with FAD compared to the wild type having two salt bridges. Studies of a rat CYPOR mutant of the corresponding residue R454E showed < 5% of FAD content and only 0.3% of the wild-type activity in cytochrome c reduction (24). Because the Glu substitution in the rat enzyme is more drastic compared to a His substitution in the human enzyme, it is not surprising to observe a milder phenotype in the R457H human mutant.

Despite the fact that V492E has a lower affinity for FAD than R457H (Table 1), V492E can be more fully reconstituted, indicating another difference in the degree of disruption. R457H retains its FAD more tightly than V492E, but a portion of the molecules appear to be resistant to FAD reconstitution. Examination of the catalytic data shows that V492E is reconstituted fully by the addition of FAD (6% of WT minus FAD; 89% of WT plus FAD; Table 1), whereas R457H is reconstituted less fully (31% of WT minus FAD and 69% of WT plus FAD). Because we were not able to crystallize the form of R457H protein that is deficient in FAD, the R457H and WT preparations were subjected to UV-vis circular dichroism spectroscopy, as performed previously with the V492E variant (14). These data (Fig. 4), together with the rescued activity and the structure of the R457H mutant, clearly show that the overall polypeptide fold in the absence of FAD is intact.

The flavin contents of three preparations of R457H were determined and, consistent with the distinctly higher affinity for FAD than V492E demonstrated in the catalytic experiments, the isolated, purified enzyme consistently showed between 30–50% FAD content compared to much higher (ca. 90%) FMN content. These data correspond well to the 30–50% activity compared to WT enzyme determined for the unsupplemented R457H preparations (Table 1). These combined results beg the question of why the R457H mutant in the cellular milieu is not activated to its fullest capacity. The mutation itself has not destroyed and has left largely unaffected the ability to reconstitute enzyme activity with excess FAD, so its catalytic deficiency could be due to lower than optimal cellular concentrations of FAD. The FAD affinity is lower than wild type and only approximately 30% of the mutant enzyme is bound to FAD. It is possible that, in mammalian cells, the FAD concentration could be much lower than that in E. coli, resulting in much less than 30% flavinated proteins and decreased CYPOR activity. It should be stated that the FAD concentrations required for activation in the titration experiments were determined catalytically, so they do not represent actual binding constants but rather “activation constants.”

With the R457H polymorphism, the work of Fukami et al. shows the importance of knowing the genetic makeup of affected individuals. It is apparent that human patients with this mutation exhibit milder defects than those of other single mutations and that only in combination with other mutations in the opposite allele does this particular human polymorphism exhibit the strongest phenotypes (6). This is in agreement with our data, which indicate that the R457H mutation is not as detrimental to enzyme function as V492E, for example. Examining a total of 35 patients, Fukami et al. observed a group of 14 with a homoallelic form of the POR gene defect. However, because the phenotypes differ, even in the case of identical POR alleles, it is difficult to draw conclusions with respect to genotype/phenotype evaluation until a sufficient number of patients with identical mutations in both alleles are well characterized. Other genetic mutations in P450 genes or possible epigenetic factors can also influence the phenotype observed.

Both R457H and V492E mutations disrupt the interactions between the FAD pyrophosphate and the polypeptide. However, there are subtle, but distinct, differences in destabilization/unfolding of the enzyme molecule between these two mutations. R457, shown in Fig. 1A, lies on one of the two long β-strands that connect the FAD/NADP(H) domain to the connecting domain. Whereas V492 lies at the base of the β-flap and is located at one side of the pyrophosphate moiety of FAD, R457 is located on the other side of the FAD pyrophosphate. As discussed above, the V492E mutation destabilizes the β-flap, leading to a somewhat localized unfolding of the CYPOR molecule as evidenced by the discreet bands obtained upon limited trypsin digestion. In contrast, the R457H mutation results in less unstable but more global destabilization of the polypeptide, leading to multiple, simultaneous cleavages upon trypsin digestion. There appear to be three different populations of the R457H protein in solution: (i) one that is active with a full complement of FAD, (ii) another that can be reconstituted with FAD to form an active conformer, and (iii) a remaining population intractable to FAD reconstitution (Table 1). This is in contrast to our observations with V492E with which FAD reconstitution is virtually complete (Table 1), indicating only two populations of conformers. The nature of polypeptide folding state(s) of the third population of R457H (i.e., not convertible to an active form) remains unknown.

Conclusions.

NADPH-cytochrome P450 oxidoreductase, expressed by a single gene, interacts with over four dozen cytochromes P450 and other redox partners. Its role with this seraglio of partners represents a model for understanding protein–protein interactions at the structural level. In addition to reporting the X-ray structure of human CYPOR, the data reported herein present insights into the impact of mutations on its structure and resulting function required in understanding the numerous CYPOR deficiencies reported in human patients. The effects of these mutations, although dramatic in altering catalytic activity, are not global with respect to overall structural alterations in the molecule. However, mutation at one residue affects another part of the molecule—i.e., V492E at the base of the β-flap influences folding (packing) of the molecule between the FAD/NADPH domain and the connecting domain by weakening salt bridges that are some 20-Å away from the mutation site. Effects of these alterations in single residues can be rescued to varying extents by the addition of FAD to each of these mutant proteins, as previously reported in the case of V492E (14). Whereas R457 is located on a β-strand interacting with the pyrophosphate moiety of FAD, V492 is situated at the “base” of the β-flap, which covers the PP_i-ribityl-adenine half of FAD and its tip (the loop portion) is tightly bound to the connecting domain. Thus, although the residues in the β-strand and the β-flap contribute to FAD binding to the polypeptide, FAD, itself, also influences the conformation of these residues by binding to them, maintaining the folding of the entire molecule. This explains the stabilization of both the V492E and R457H mutants to trypsin digestion in the presence of FAD. Any imbalance of this intricate relationship (i.e., mutation of any of these residues that influence FAD binding) results in weaker FAD-binding affinity and instability of the enzyme molecule, albeit to varying extents. The varying abilities to restore activities represent the degrees to which the binding affinity for FAD has been altered in these preparations and could predict the ability to rescue these defects therapeutically.

Methods

Cloning, Mutagenesis, Protein Expression, and Purification.

Subcloning of the cDNA encoding the C-terminal 614 residues of human CYPOR (residues 67-VRESSFV through 675-SLDVWS) from Mammalian Gene Collection 9411 (ATCC) into the pET-28a vector (Novagen), yielding the expression plasmid, pETPORΔ66, was previously described (14) and detailed in SI Methods. Histidine-tagged Δ66 CYPOR proteins were expressed in the JM109(DE3) strain and purified as described previously (14) and described in SI Methods. FAD and FMN contents of each CYPOR protein were determined by extracting flavins from samples by boiling and subjecting the supernatant to HPLC as described in SI Methods.

Kinetic Analysis, Circular Dichroism Spectroscopic Analysis, and Limited Trypsin Digestion.

Detailed description of these experimental procedures is available in SI Methods.

Crystallization, Data Collection, and Structure Determination.

All crystals were grown using the hanging drop vapor diffusion method (25) at 19 °C, followed by the macroseeding technique (26), as detailed in SI Methods. All datasets were collected at the Structural Biology Center-CAT 19ID beamline at the Advanced Photon Source, Argonne National Laboratory, and were processed using HKL 2000 (27). All crystals of the Δ66 proteins belong to the orthorhombic space group, P2₁2₁2₁, with approximate unit cell dimensions of a = 70 Å, b = 118 Å, and c = 115 Å with two molecules in an asymmetric unit. In all cases, the initial structures were solved by the molecular replacement method using Molrep in the CCP4 program package (28) with the structure of the corresponding domains of wild-type rat CYPOR structure (17) (Protein Data Bank ID 1AMO). Subsequent refinements of the initial structures were carried out using the CNS program package (29). The final data collection and refinement statistics are summarized in Table S1.