Home

Abstract (Summary)

DISCUSSION

In this article, we present novel methodologies for modeling cytochrome P450s, which were applied to model CYP27A1, an important mitochondrial enzyme involved in bile acid metabolism and the 25-hydroxylation of vitamin D. We used this model to guide the selection of nine site-directed mutations directed toward seven putative substrate contact residues affecting the metabolism of the vitamin D prodrug, 1α-OH-D^sub 3^. Changes in activity and the regioselectivity of side-chain hydroxylation were interpreted to result from conformational changes affecting cavity shape, steric conflicts with bound substrate, or disruption of a secondary heme-binding interaction, though we cannot rule out that small differences in total enzyme activity might be due to differences in enzyme expression.

The model was also used to explore, in silico, broader aspects of substrate recognition in CYP27A1 and the location of potential channels governing substrate access and metabolite egress. In the remainder of the discussion, we have highlighted the important aspects revealed by our study, namely: a), the contribution of selected residues to the active site structure and catalytic activity of CYP27A1 toward IaOH-D3, and b), a modeling strategy based on the systematic analysis of conserved hydrogen bonding patterns and structural motifs in 11 CYP crystallographic templates.

The major objective of our metabolism studies was to structurally probe putative substrate contacts in the binding cavity of CYP27A1. In our opinion, one of the most interesting findings in this study was that subtle changes in bound orientation directly affect the ratio of 25- and 27-hydroxylated products formed. The hydroxylation at c25 implied that the substrate is bound deeper into the pocket. Mutations targeted to the β-3a strand, β-5 hairpin, and F-helix changed the structure predictably and altered CYP activity from positions that have not been previously observed in patients with CTX, VDDR-I, or any other P450 deficiency.

Mutations affecting the hydrophobic binding cavity surface at the β-5 hairpin

The effect of the Phe-248-Met mutation on CYP27A1 activity was unique and shifted the ratio of 1α-OH-D^sub 3^ metabolism toward 27-hydroxylation. Phe-248 was predicted to be well within the F-helix and was clearly modeled into the top of the substrate binding site. Whereas Phe-248 may contact the substrate directly, its presence in CYP27B1 and CYP11 isoforms suggests that it plays a more generic function, such as shaping the hydrophobic surface of the binding cavity near Leu-516 in the â-5 hairpin. Phe-248 may deflect Leu-516 away from the substrate so that the mutation to methionine allows Leu-516 to have increased contact with c27 to prevent deep penetration of substrate.

The small decreases in total enzyme activity associated with the Phe-248-Met and Leu-516-Val mutations may suggest that part of this hydrophobic surface is also critical to the gating function of Asp-338 in regulating proton access during the reaction cycle via the adjacent solvent access channel. Interestingly, the Phe-248-Lys mutation reported in a previous study altered the regioselectivity of bile acid precursor hydroxylation by inducing a novel demethylase activity and the formation of non-c27-oxidizcd and/or unsaturatcd products (53). Consumption of protons during the reaction cycle could make lysine a proton acceptor capable of directing a 27-hydroxylated product toward demethylation. This could also disrupt bound substrate orientations and/or Asp-338-regulated proton transport resulting in altered product formation and product release.

All of the β-5 hairpin mutations transformed CYP27A1 into a more efficient 25-hydroxylase, shifting the ratio of hydroxylation products toward 1α,25-(OH)^sub 2^D^sub 3^. In Leu-516-Val, the smaller side chain had reduced contact with the substrate, enlarged the binding cavity, allowed the deeper penetration of substrate leading to 25-hydroxylation, and may have decreased overall activity by disrupting Asp-338-regulated proton transport. In our model, the side chain of Val-515 made deep penetration of the substrate somewhat unfavorable through steric contact with the c18 D-ring methyl.

A longer side chain could adopt rotomer conformations that reduce this contact, and this was seen in the increased activity and deeper penetration of substrate in Val-515-Ile and Val-515-Leu. The effect of the Ile-514-Phe mutation is more complicated because the large loss of activity can be interpreted as a large and unfavorable impact on the presumed substrate access channel. Resulting in a preferential loss of 27-hydroxylated product, the residual 25-hydroxylase activity of the Ile-514-Phe mutation implied that what little substrate was bound was deflected deep into the active site. Whether this was caused by conformational changes in the F-G loop, a productive interaction with the vitamin D cw-triene, or a deflection of the adjacent Val-515 is not known. Overall, the mutations in the β-5 hairpin study suggest that this structure forms a contoured hydrophobic surface that affects substrate orientation.

Mutations affecting structure and substrate contact in the β-3 sheet

In our model, it was found that the β-3a strand in CYP27A1 connected the β-5 sheet and B-B' loop to an extended β-sheet structure, bound the heme A-ring propionate from two positions, and made significant contact with bound substrate. The Thr-402-Phe mutation severely impaired the activity of the enzyme. We hypothesize that this occurred through the loss of a structural link to the β-5 hairpin near Ile-514, which disrupted protein stability and the loss of a contact with the substrate 1α-hydroxyl. The Asn-403-Thr mutation appears to have ablated activity by disrupting a heme-binding interaction between the His-428-stabilized Asn-403 and the A-ring propionate that disrupts protein stability.

The loss of a proton donor function associated with the Ser-404-Ala mutation disrupted the contact with the substrate 3β-hydroxyl and the B-B' loop through the release of hydrogen bonding interactions, the latter centered on Asn-128 and Gln-423 side chains. The ensuing conformational change oriented bound substrate deeper into the active site, preferentially disrupting 27-hydroxylation. A second mutation at this position, Ser-404-Thr, did not influence bound substrate orientation toward 25- or 27-hydroxylation, but the methyl group sterically interfered with binding and caused a significant loss of activity.

Substrate recognition in the substrate access channel

A survey of proton donors capable of recognizing the hydroxyl groups on CYP27A1 substrates revealed a multifaceted hydrogen bonding network of conserved asparagines, glutamines, and histidines in the extended β-sheet and amino-terminal helices. It was found that these residues occupied discrete positions and have side-chain orientations that are independently determined by cytochrome P450 structure. Some of these residues may assist the partitioning of substrates down a substrate access channel into the active site through transient interactions with substrate hydroxyl groups and presumably provide a mechanism for substrate specificity by filtering or selecting against previously hydroxylated substrates.

Our model also revealed the existence of a previously unrecognized aromatic cluster over the binding cavity. A similar cluster in CYP3A4 formed a prominent hydrophobic core, which was highly ordered and composed of residues known to affect activity (31). In CYP27A1, this aromatic cluster appears to contribute a large surface at the junction of the binding cavity and a pw2a substrate access channel, and contains residues shown in previous studies to affect the kinetics of cholesterol hydroxylation (53,62).

Structural motif-based modeling methodology

In this study, we have used a new approach to homology modeling which was based upon a strategy which is broadly applicable to other cytochrome P450 proteins. We have used the observed secondary structure and a systematic analysis of hydrogen bonding patterns in 11 CYP crystal structures to refine the multisequence alignment and simplify the assignment of secondary structure to a new CYP sequence. This analysis revealed a number of residues and tertiary structural motifs that are functionally conserved in most of the available CYP templates and probably in many other CYPs as well.

Although the function of some conserved elements has been correctly deduced in other studies, for example the PGP motif (42), our analysis revealed a high degree of commonality in known CYP structures and drew attention to functionally equivalent interactions that may be easily overlooked. These structural motifs were used to guide the application of distance constraints during energy minimization and molecular dynamics. This strategy maximizes the likelihood of successfully propagating known structure into regions that are difficult to model due to template heterogeneity. Finally, we have used this approach to model CYP27A1 and explore the enigmatic structure and location of access channels.

The finding of such a functionally conserved and highly structured heme-binding region was unexpected and emphasizes the need for objective input during modeling. The hydrogen bonding network around the propionates helped position the B- and C-helices, B-B' and B'-C loops, and most of the β-sheet structure relative to the heme group. The extensive hydrogen bonding network surrounding the ERR-triad suggests an important structural core that stabilizes adjacent structures and may, for some, change how protein cofactor binding at the proximal heme face will be viewed. This is probably also true for the conserved links between the J- and K-helices and the K- and L-helices. The EH-gap motif was an interesting finding since it helps delineate and position the E-, G-, and H-helices and the E-F loop.

The incorporation of this motif in our model of CYP27A1 revealed a putative disulfide bond and raises the question of the potential for formation of this structure in other CYPs. The spatially conserved phenylalanine at the start of the F-helix is an important alignment and modeling landmark, as is the well-conserved aspartate at the start of the E-helix. The unique aspects of cytochrome P450 structure revealed in this modeling approach can now be used to predict and study the binding pocket in other vitamin D-related hydroxylases. The authors acknowledge the excellent technical assistance of Kerith R. Geh in the metabolism studies, the contributions of Dr. Donald F. Weaver, Kelly A. Dakin, and Oreola A. T. Donini in the preparation of a preliminary model of CYP27A1, the instruction of Dr. Michael J. Kuiper in the usage of the SYBYL software, and the assistance of Jimin Zheng and Robert Campbell. They also acknowledge data not shown, which was collected by Valarie Byford, Martin Kaufmann, Sonoko Masuda, and Shelly West.