Redox Partner Adrenodoxin Alters Cytochrome P450 11B1 Ligand Binding and Inhibition

Abstract

Human cytochrome P450 11B1 (CYP11B1) generation of the major glucocorticoid cortisol requires two electrons delivered sequentially by the iron-sulfur protein adrenodoxin. While the expected adrenodoxin binding site is on the opposite side of the heme and 15–20 Å away, evidence is provided that adrenodoxin allosterically impacts CYP11B1 ligand binding and catalysis. The presence of adrenodoxin both decreases the dissociation constant (K_d) for substrate binding and increases the proportion of substrate that is bound at saturation. Adrenodoxin additionally decreases the Michaelis-Menten constant for the native substrate. Similar studies with several inhibitors also demonstrate the ability of adrenodoxin to modulate inhibition (IC₅₀ values). Somewhat similar allosterism has recently been observed for the closely related CYP11B2/aldosterone synthase, but there are several marked differences in adrenodoxin effects on the two CYP11B enzymes. Comparison of the sequences and structures of these two CYP11B enzymes helps identify regions likely responsible for the functional differences. The allosteric effects of adrenodoxin on CYP11B enzymes underscore the importance of considering P450/redox partner interactions when evaluating new inhibitors.

INTRODUCTION

Cytochrome P450 (CYP) 11B1 performs the terminal steroid 11β-hydroxylation reaction in cortisol biosynthesis (Figure 1A). Cortisol is the major glucocorticoid hormone in the human body responsible for the stress response. Excess cortisol leads to Cushing’s disease, which is characterized by weight gain, immune suppression, hypertension, and osteoporosis (1). Treatment options for Cushing’s disease via CYP11B1 inhibition are limited, in part because of the difficulty in achieving drug selectivity for CYP11B1 vs. CYP11B2. CYP11B2, the enzyme responsible for aldosterone synthesis, is 93% sequence identical to CYP11B1. Until recently structures of the human CYP11B enzymes were not available to guide drug design. Now that structures are available (2,3), they reinforce the difficulty of selective inhibitor design in that the first shell active site residues are identical in composition and only moderately repositioned.

Figure 1. CYP11B1 substrate and inhibitors:

A) The major substrate of CYP11B1 is 11-deoxycortisol. CYP11B1 catalyzes an 11β hydroxylation to produce cortisol, the major glucocorticoid responsible for the stress response. B) LCI699 is an FDA-approved Cushing’s disease treatment that is an analog of the breast cancer drug fadrozole. LCI699, (R)- and (S)-fadrozole are all potent inhibitors of CYP11B1.

However, functional studies have led to the development of a moderately selective inhibitor. The racemic breast cancer drug fadrozole targets CYP19A1/aromatase but was reported to have some selectivity for CYP11B1 over CYP11B2 (2). Medicinal chemistry approaches led to the development of LCI699, an analog of (R)-fadrozole (Figure 1B). Originally developed by Novartis as a CYP11B2 inhibitor (4), LCI699 was eventually found more therapeutically useful as a CYP11B1 inhibitor (5). LCI699 (Osilodrostat or Isturisa^®) was approved by the FDA to treat Cushing’s disease in March 2020. Despite the approval of LCI699, it is clear this compound is only moderately selective between the two human CYP11B enzymes. In fact, when CYP11B1 and CYP11B2 were co-crystallized with racemic fadrozole, CYP11B1 selectively bound (S)-fadrozole, while CYP11B2 selectively bound (R)-fadrozole, which has the same stereochemistry as LCI699.

Interactions of CYP11B1 with other elements of the catalytic system are also poorly understood. Catalysis by P450 enzymes requires two electrons originating from NADPH and delivered via a redox chain. For mitochondrial P450 enzymes like the CYP11B enzymes, this redox chain consists of adrenodoxin reductase and adrenodoxin. First, membrane-bound adrenodoxin reductase binds NADPH and accepts two electrons via its FAD. Subsequently one electron at a time is transferred to the small, soluble iron-sulfur protein adrenodoxin (also known as ferrrodoxin-1). Adrenodoxin transiently associates with mitochondrial P450 enzymes to provide the two one-electron reductions to enable P450 catalysis. The seven mitochondrial P450 enzymes that adrenodoxin interacts with are involved in vitamin D metabolism and steroidogenesis, including CYP11B1. In addition to this role as electron donor, there are emerging lines of evidence that adrenodoxin binding to the proximal surface of these P450 enzymes may also allosterically alter P450-ligand interactions in the P450 active site, which is located on the opposite, distal side of the heme more than 15 Å away. First, binding to the proximal surface of the cholesterol side-chain cleavage enzyme CYP11A1 has been proposed to modulate the position of its intermediate substrate 22R-hydroxycholesterol 18 Å away in the CYP11A1 active site to allow a subsequent round of catalysis (6). Second, several recent studies with steroidogenic CYP11B2 indicates that adrenodoxin allosterically modulates CYP11B2 substrate binding (7,8), catalysis, and inhibitor binding (8). Finally, studies with vitamin D3-metabolizing CYP24A1 suggest that adrenodoxin modulates CYP24A1 substrate binding, potentially by conformational changes on the opposite site of the P450 involved with substrate access (9). However, studies with CYP27C1 suggest that adrenodoxin does not influence substrate binding saturation or affinity (10), so this allosteric effect may differ among mitochondrial P450 enzymes. Little is known about the effects of adrenodoxin on the remaining human P450 enzymes that interact with adrenodoxin, including CYP11B1, CYP27A1, and CYP27B1. Additionally, since only recently has recombinant, purified human CYP11B1 become available (11), much of the CYP11B1 work has been done with the bovine or rat forms and species differences are apparent (12,13).

The current study fills this gap by elucidating the effects of human adrenodoxin on the function of human CYP11B1. CYP11B1 was selected for further examination because of the high sequence identity with CYP11B2 and the ability to compare with recently published detailed adrenodoxin studies (8). Herein adrenodoxin binding, which is thought to occur on the proximal CYP11B1 surface, is indeed shown to alter the binding of substrates and inhibitors in the distant CYP11B1 buried active site, as well as catalytic action, inhibitor binding, and inhibition. This study provides further evidence that adrenodoxin is an allosteric modulator of mitochondrial P450 enzymes. Comparison of the functional effects of adrenodoxin on CYP11B1 and the closely related CYP11B2 combined with the sequence and structural differences between the two CYP11B enzymes may identify functional differences useful in future drug design.

RESULTS

Adrenodoxin increases CYP11B1 substrate binding

CYP11B1 ligand binding studies were performed to assess the impact of adrenodoxin on binding of the CYP11B1 native substrate 11-deoxycortisol. While purified CYP11B1 has a heme Soret peak at 420 nm, 11-deoxycortisol binding in the active site causes a blue shift. In difference mode this is observed as an increase in absorbance at 389 nm and decrease in absorbance at 424 nm called a Type I shift (Figure S1). This absorbance change upon substrate titration can be used to determine the ligand dissociation constant (K_d) or affinity and the maximal change in absorbance (A_max) or relative amount of ligand-bound enzyme at saturation. Such titrations were used to evaluate CYP11B1 binding its substrate 11-deoxycortisol in the presence of varying amounts (Figure 2) of adrenodoxin. The range of adrenodoxin concentrations examined herein (up to 40x over the P450 concentration) were selected based on the range used in in vitro experiments in the literature 8x to 60x (13)) because the in vivo concentrations are not well established.

Figure 2. Influence of adrenodoxin on CYP11B1 substrate binding.

A) CYP11B1 binding its 11-deoxycortisol substrate was evaluated in the absence and presence of increasing amounts of adrenodoxin. These experiments used 1 μM CYP11B1 and adrenodoxin at 0 μM (black), 1 μM (red), 10 μM (blue), and 40 μM (green). B) Similarly, 11-deoxycortisol binding was evaluated for the adrenodoxin-CYP11B1 fusion enzyme, which has an inherent ratio of 1:1 for the two domains. All binding experiments were done in triplicate and all data is shown. C) The data in A and B were fit to the one-site specific binding equation to determine the dissociation constants (K_d) and maximal change in absorbance (A_max). R² values were 0.97–0.98.

Immediately obvious from such titrations was that adrenodoxin had systematic effects on two aspects of CYP11B1 substrate binding. First, the maximal change in CYP11B1 absorbance due to substrate binding systematically increased with increasing amounts of adrenodoxin (Figure 2A). Without adrenodoxin, the CYP11B1 A_max was 0.043, and this increased to 0.065, 0.072, and 0.080 in the presence of equimolar, 10-fold, and 40-fold excess adrenodoxin (Figure 2C). The overall increase in absorbance was two-fold, consistent with a commensurate increase in substrate binding. Second, increasing amounts of adrenodoxin substantially altered the CYP11B1 affinity for 11-deoxycortisol. With no adrenodoxin present, the dissociation constant K_d was 17.0 μM, but was 20.8, 12.2, and 7.3 μM with equimolar, 10-fold, and 40-fold excess adrenodoxin respectively (Figure 2C). While the 95% confidence intervals for no adrenodoxin and 1-fold adrenodoxin overlap substantially, the remaining experiments revealed a trend of decreasing K_d/increasing substrate affinity with increasing adrenodoxin concentrations. The increases in total substrate binding and substrate affinity are correlated.

As an orthogonal control experiment, spectral changes were also recorded while the amount of 11-deoxycortisol was held constant and the adrenodoxin was titrated. CYP11B1 was initially equilibrated with an amount of 11-deoxycortisol yielding about 50% of the maximal spectral shift. Addition of adrenodoxin, while keeping the total substrate concentration constant, resulted in further spectral changes even though the ligand concentration did not change (Figure S2). This result is also consistent with adrenodoxin interaction promoting ligand binding.

Adrenodoxin normally interacts with CYP11B1 transiently, and even a 40-fold excess may not capture the full effect of adrenodoxin/P450 interaction. While several human mitochondrial P450 enzymes including CYP11B enzymes are reported to have nanomolar affinity for adrenodoxin (13,14), a complex could not be isolated. Thus, we also designed an artificial fusion of adrenodoxin with CYP11B1. N-terminal adrenodoxin was connected to C-terminal CYP11B1 via the linker TDGTS, which was designed based on literature analysis (15). The idea was that colocalizing CYP11B1 and adrenodoxin might favor interaction between the two. Indeed, this adrenodoxin-CYP11B1 fusion enzyme followed similar trends seen above for the individual proteins (Figure 2B). The A_max was 0.074, most similar to the 10-fold adrenodoxin experiment above, while the K_d of 3.8 μM was almost two-fold lower than even the 40-fold adrenodoxin experiment above (Figure 2C). These results suggest that the artificial fusion both 1) encourages interaction of the two fused domains and 2) reinforces that adrenodoxin increases CYP11B1 substrate binding and affinity.

Adrenodoxin modulates CYP11B1 catalysis

To further determine the effects of adrenodoxin on CYP11B1 function, steady-state 11β-hydroxylation of 11-deoxycortisol to cortisol was also evaluated in the presence of increasing concentrations of adrenodoxin. While some adrenodoxin is required for CYP11B1 catalysis, the ratios of CYP11B1:adrenodoxin were varied from 1:1 to 1:10 to 1:40 (Figure 3A), with the latter representing the range frequently used in biochemical experiments (13).

Figure 3: Adrenodoxin influence on CYP11B1 catalytic activity.

A) CYP11B1-mediated conversion of 11-deoxycortisol to cortisol was determined in the presence of 1-fold (blue), 10-fold (red), and 40-fold (black) higher adrenodoxin concentrations. B) Conversion of 11-deoxycortisol to cortisol was also determined for the adrenodoxin-CYP11B1 fusion enzyme in the presence of 40-fold adrenodoxin. All reactions were completed in triplicate and all data are shown. C) These data were fit to the Michaelis-Menten equation to determine the k_cat and K_m values. R² values were 0.94–0.97.

As expected, increasing amounts of adrenodoxin increased the k_cat. The observed rates were 3.1, 3.8, and 13 min⁻¹ for the 1:1, 1:10, and 1:40 CYP11B1:adrenodoxin ratios, respectively (Figure 3C). Interestingly, however, the presence of adrenodoxin ratio also substantially decreased the K_m. As the adrenodoxin:CYP11B1 ratio increased from 1:1, 1:10, and 1:40, the K_m decreased from 72.8 to 43.4 to 27.3 μM (Figure 3C). While the K_m and K_d values are not the same (unless ES->EP is very, very slow), the K_m trend observed is related to the K_d observed above: increasing adrenodoxin makes CYP11B1 half saturated at lower and lower substrate concentrations.

Like the binding experiments above, steady-state kinetics were also evaluated for the adrenodoxin-CYP11B1 fusion protein. In the absence of added free adrenodoxin, little to no turnover was observed. This would be consistent with the fused adrenodoxin primarily interacting with the CYP11B1 and poorly available to interact with adrenodoxin reductase. However, when the same steady-state kinetics experiments were performed with the adrenodoxin-CYP11B1 fusion protein plus 40-fold excess free adrenodoxin, some product formation (k_cat of 1.1 min⁻¹) was observed. First, this suggests that the CYP11B1 domain is catalytically active in the context of the fusion protein. Second, this suggests that if one adds high enough concentrations of free adrenodoxin, that free adrenodoxin is sometimes able to outcompete the fused adrenodoxin, delivering electrons to support catalysis. The k_cat for the adrenodoxin-CYP11B1 fusion protein cannot be directly compared to the k_cat of the isolated proteins because low signal required a higher protein concentration for the fusion protein experiments, but the trend in the K_m can be compared. For the fusion protein with 40-fold excess adrenodoxin the K_m was 16.1 μM, almost half the non-fused proteins at the same ratio (27.3 μM).

Altogether the decreasing trends for K_m observed in separate and fused proteins with increasing adrenodoxin suggest that adrenodoxin promotes saturation of CYP11B1 with 11-deoxycortisol, in addition to its established role in electron transfer.

Adrenodoxin also modulates CYP11B1 binding of inhibitors

While the substrates discussed above bind in the CYP11B1 active site in such a way that they displace water from the active site heme iron, most of the inhibitors bind differently. Inhibitors including the FDA-approved drug LCI699 (also called osilodrostat or Isturisa^®) and its precursors (R)- and (S)-fadrozole have an imidazole that replaces this water, with the nitrogen lone pair of electrons forming a coordinate covalent bond with the iron. This results in a red shift in the Soret peak wavelength. Usually viewed in difference mode, for CYP11B1 this occurs as an increase in absorbance at 428 nm and a decrease in absorbance at 410 nm called a Type II shift (Figure S3). When CYP11B1 was titrated with these three type II inhibitors, the K_d could be determined in the presence of increasing adrenodoxin, as described above for substrates.

Consistent with prior evidence (16), our initial titration experiments indicated that all three compounds bound very tightly (nM K_d values). To evaluate such high affinity compounds, the CYP11B1 concentration was minimized as much as signal would allow and the path length increased to 5 cm to offset the reduction in signal. The lowest reasonable CYP11B1 concentrations for the LCI699, (S)-fadrozole, and (R)-fadrozole binding studies were 0.20, 0.15, and 0.10 μM respectively. The dissociation constants for LCI699, (S)-fadrozole, and (R)-fadrozole were 14.7, 14.1, and 11.5 nM respectively. These values are 9- to 14-fold lower than the protein concentrations used to determine them, so they are estimates at best. It is likely these inhibitors bind so tightly that the K_d value is below what can be accurately quantified. Because these experiments were optimized to get the best estimates of the K_d values, the maximal change in absorbance (A_max) could not be compared because of the different protein concentrations used.

To make direct comparisons of the A_max values between inhibitors and upon addition of adrenodoxin, the binding assays were then repeated at a constant CYP11B1 concentration of 0.2 μM. Among the three inhibitors evaluated, both fadrozole inhibitors showed higher saturation of CYP11B1 than LCI699 (Figure 4). Saturation of CYP11B1 by all three inhibitors was substantially increased when even an equimolar amount of adrenodoxin was present. For LCI699, the A_max increased 2.5-fold from 0.006 for CYP11B1 alone to 0.015 using 1:1 CYP11B1 to adrenodoxin. For (R)-fadrozole, the A_max increased 1.5-fold from 0.014 to 0.021 upon addition of equimolar adrenodoxin, and for (S)-fadrozole, the A_max increased 2.3-fold from 0.010 to 0.023 with adrenodoxin present.

Figure 4: Influence of adrenodoxin on CYP11B1 inhibitor binding.

The effect of adding equimolar adrenodoxin (blue) vs. no adrenodoxin (red) was examined for CYP11B1 binding of three different active-site-directed inhibitors: A) LCI699, B) (R)-fadrozole, and C) (S)-fadrozole. Each binding assay was performed using 0.2 uM CYP11B1 in 5 cm cuvettes and the resulting data was fit to the Morrison equation. In all cases, the binding affinity was much lower than the protein concentration, which limits the accurate quantification of K_d values. However, the A_max nearly doubled for each inhibitor, demonstrating that adrenodoxin increases ligand saturation. Each binding study was completed in triplicate and all data are shown.

Adrenodoxin modulates CYP11B1 inhibition efficiency

Since the information above suggested that the presence of adrenodoxin promotes inhibitor binding to CYP11B1, studies were performed to determine how varying adrenodoxin impacts the inhibition of CYP11B1 enzymatic capacity by measuring IC₅₀ values. As in the Michaelis-Menten kinetic studies, ratios of 1:40, 1:10, and 1:1 CYP11B1 to adrenodoxin were used for each compound, and the 11-deoxycortisol concentration was held constant at the K_m value for each experimental condition (Figure 5).

Figure 5: Adrenodoxin influence on CYP11B1 inhibition.

Inhibition of CYP11B1 cortisol production was evaluated in the presence of 1-fold, 10-fold, and 40-fold higher adrenodoxin over the CYP11B1 concentration for inhibitors A) LCI699, B) (R)-fadrozole, and C) (S)-fadrozole. D) The resulting data was fit to the four-parameter dose-response equation to determine IC₅₀ values. R² values were 0.94–0.98.

For all three compounds, increasing amounts of adrenodoxin decreased the value of the IC₅₀ significantly. For LCI699, the IC₅₀ was decreased 10-fold, from 429 nM to 161 and 43 nM for the 1:1, 1:10, and 1:40 excess adrenodoxin conditions, respectively. For (R)-fadrozole, the IC₅₀ decreased 12-fold, from 209 nM to 144 and 18 nM for the 1:1, 1:10, and 1:40 excess adrenodoxin conditions, respectively. For (S)-fadrozole, the IC₅₀ decreased 10-fold, from 422 nM to 287 and 43 nM for the 1:1, 1:10, and 1:40 excess adrenodoxin conditions, respectively. In all three cases, there is approximately a 10-fold decrease in the IC₅₀ value between the 1:1 to 1:40 CYP11B1:adrenodoxin ratios. These results suggest that the presence of excess adrenodoxin promotes inhibition.

These experiments additionally allowed comparison of the inhibitory potencies of three structurally similar compounds. In all three CYP11B1 to adrenodoxin ratios, the IC₅₀ was the lowest for (R)-fadrozole while the IC₅₀ values were generally similar for (S)-fadrozole and LCI699. Using a 1:40 CYP11B1:adrenodoxin ratio, the IC₅₀ values were 18, 43, and 43 nM for (R)-fadrozole, (S)-fadrozole, and LCI699. That is, the IC₅₀ for (R)-fadrozole was ~2-fold lower than the other two compounds. Using a 1:10 CYP11B1 to adrenodoxin ratio, the IC₅₀ values were 144, 287, and 162 nM for (R)-fadrozole, (S)-fadrozole, and LCI699. That is, the (R)-fadrozole IC₅₀ was still 2-fold lower than that of (S)-fadrozole IC₅₀, but about the same as for LCI699. At a 1:1 CYP11B1:adrenodoxin ratio, the IC₅₀ values were 209, 422, and 429 nM for (R)-fadrozole, (S)-fadrozole, and LCI699 respectively. Under these conditions, the (R)-fadrozole IC₅₀ was again ~2-fold lower than the IC₅₀ values for the other two compounds. These inhibition studies support that (R)-fadrozole is a more potent CYP11B1 inhibitor than (S)-fadrozole or LCI699.

DISCUSSION

Adrenodoxin modulates CYP11B1 binding to substrate 11-deoxycortisol

The human mitochondrial CYP enzymes, including CYP11B1, require electron transfer from adrenodoxin for catalysis. Several previous studies have evaluated communication between what happens in a P450 active site and their interactions with adrenodoxin for select mitochondrial P450 enzymes. The identity of the ligand in the active site of vitamin D metabolizing CYP24A1 altered the specificity of the CYP24A1 interaction with adrenodoxin as measured by NMR (17). Steroidal substrate binding in the active site of CYP11A1 was reported to increase the affinity of CYP11A1 for adrenodoxin by promoting their association, but decrease CYP11B1 and CYP11B2 interactions with adrenodoxin as measured by surface plasmon resonance (18). Based on this information it appears that there is communication between mitochondrial P450 active sites and the adrenodoxin binding site, which is different for different P450 enzymes. Herein, we examine the opposite directionality, the effects of the protein-protein interaction on what occurs in the P450 active site, in terms of ligand binding, metabolism, and inhibition.

CYP11B1 binding studies with its native substrate 11-deoxycortisol indicate that increasing amounts of adrenodoxin monotonically decrease the K_d and increase the maximal change in absorbance (A_max), with the trends also consistent for the adrenodoxin-CYP11B1 fusion protein (Figure 2). These changes are significant in scale, with the K_d values decreasing as much as 4-fold (for the fusion) and the maximal saturation nearly doubling (for the fusion and 40-fold excess adrenodoxin). These results establish that even the oxidized adrenodoxin used in these studies, which cannot transfer an electron, allosterically promotes substrate binding in the active site. Additionally, the fusion of adrenodoxin and CYP11B1 likely stabilizes the P450/adrenodoxin complex because the K_d is ~5-fold higher than the same 1:1 ratio when the two proteins are not fused and most similar to the K_d for the 40-fold excess adrenodoxin situation (Figure 2A-C). Thus, the most likely interpretation is that fusion of the adrenodoxin increases the adrenodoxin concentration in the vicinity of its proximal binding site on CYP11B1, magnifying the effects on substrate binding affinity.

Comparison with corresponding results for CYP11B2 is instructive. Increasing free adrenodoxin concentrations or fusing adrenodoxin also resulted in CYP11B2 binding its substrate 11-deoxycorticosterone with decreasing K_d values and increasing saturation, but the concentration regimes of these effects were quite different for the two enzymes. While it took 40-fold excess adrenodoxin or fusion to yield the effects on CYP11B1 described above, it took only equimolar adrenodoxin to yield a 2.5-fold increase in CYP11B2 substrate affinity (8), with subsequent additions of adrenodoxin having relatively little effect on CYP11B2 K_d. Conversely, the effects of increasing adrenodoxin on maximal saturation were much more modest for CYP11B2 (8) than the ~2-fold increase seen for CYP11B1 over the same range of adrenodoxin concentrations (Figure 2C) This suggests that, while CYP11B1 and CYP11B2 are 93% identical, there are functional differences in their sensitivity to adrenodoxin allostery. One way to compare adrenodoxin effects on P450 binding is to calculate the A_max/K_d or binding efficiency, which corresponds to k_cat/K_m for catalytic efficiency. The A_max/K_d for CYP11B1 increases more than 6-fold over the 40-fold range of isolated adrenodoxin concentrations evaluated and more than 11-fold when adrenodoxin was fused. By comparison, the A_max/K_d for CYP11B2 increased more than 3-fold over the same 40-fold range of excess adrenodoxin and more than 6-fold when adrenodoxin was fused. Thus, in summary, the overall magnitude of the adrenodoxin allosteric effect is greater for CYP11B1, but the effects on CYP11B2 are observed at lower adrenodoxin concentrations than for CYP11B1.

Comparisons with the mitochondrial adrenodoxin-supported CYP11A1 are more difficult because less information is available. Isolated CYP11A1 and an artificial adrenodoxin-CYP11A1 fusion protein demonstrated that the presence of adrenodoxin does not impact CYP11A1 affinity for the primary cholesterol substrate and the affinity of the subsequent substrate 22R-hydroxycholesterol is too tight to discern any increases in affinity that adrenodoxin might cause (19). CY11A1 is only 39% identical to CYP11B1 so there are many more differences that could contribute to functional differences.

Why might the three human adrenal mitochondrial CYP11 enzymes have different adrenodoxin responses? All three CYP11 enzymes are expressed in the adrenal gland, but with different zonation. CYP11A1 is expressed in all three adrenal zones, CYP11B1 in the middle zona fasciculata, and CYP11B2 only in the outer zona glomerulosa. As a result, these three enzymes may experience both varying adrenodoxin levels and substrate levels. One possibility is that different expression levels of adrenodoxin in different zones could be a biological control mechanism modulating substrate binding to adjust corticosteroid concentrations in the body. An in vivo study with the myxobacterial CYP260A1 determined that the ratio of P450: adrenodoxin influences the product formation pattern, suggesting that the adrenodoxin concentration may act as a biological control mechanism for certain products (20). One limitation to this area of discussion is that adrenodoxin levels in the different zones of the human adrenal are poorly established. Studies attempting to quantify the ratio of adrenodoxin and P450 have come to many different answers depending on the system and tissue type (21,22) (genome.ucsc.edu). As a result, it is difficult to rationalize the current results in that context of adrenodoxin concentration as a potential control mechanism. More is known about the zonal abundance of substrates. Under basal conditions in adult adrenocortical cells the CYP11A1 major substrate cholesterol is highly abundant, whereas the CYP11B1 primary substrate 11-deoxycortisol is much less prevalent and the CYP11B2 primary substrate 11-deoxycorticosterone is even less abundant (23). The evidence (8) suggests that CYP11B2 is most sensitive to adrenodoxin levels, the evidence herein suggests that CYP11B1 is intermediately sensitive to adrenodoxin concentrations and the sparse evidence available (19) suggests that CYP11A1 may be insensitive to adrenodoxin modulation. Thus, one possibility is that adrenodoxin differentially promotes CYP11B substrate binding more than CYP11A1 because the CYP11B expression profile is narrower and substrate access is more limited.

Adrenodoxin modulates CYP11B1 catalysis

While ligand binding studies could employ oxidized CYP11B1 to separate allosteric and redox effects, probing potential adrenodoxin allosteric contributions to substrate catalysis is much more complex because adrenodoxin must cycle through oxidized and reduced states. Two separate electron transfers from adrenodoxin to CYP11B1 are required to complete a single catalytic cycle to generate cortisol. Previous kinetic studies determined that the rate of P450 reduction by adrenodoxin was constant with increasing the ratio of adrenodoxin:P450 up to 5:1 but this could vary for different P450 enzymes (24).

In this study, increasing adrenodoxin from 1:1 to 1:40 increased the k_cat ~4-fold (Figure 3). Increases in k_cat are expected if the two transient adrenodoxin/CYP11B1 interactions and electron transfers are rate-limiting components of the catalytic cycle. This is also consistent with other studies documenting that increasing amounts of adrenodoxin increase the catalytic rate (13,25). However, there is one report that increasing adrenodoxin ratios decreased CYP11B1 11-deoxycortisol metabolism at 1:8 and 1:16 P450:adrenodoxin concentrations. (2). Less expected were the effects on K_m. While the binding studies above describe increasing substrate affinity with increasing adrenodoxin concentrations, K_d (k_off/k_on) and K_m are only the same if k_cat is much smaller than k_off and herein k_cat is only increased by higher adrenodoxin concentrations. However, the data herein reveals that increasing adrenodoxin also decreases K_m, by ~2.7-fold from equimolar enzymes to 40-fold excess adrenodoxin (Figure 3C). The decreases in K_d combined with the increases in k_cat amount to a greater than 11-fold increase in catalytic efficiency (k_cat/K_m) from the 1:1 condition to the 1:40 excess adrenodoxin condition.

Results of substrate metabolism by the adrenodoxin-CYP11B1 fusion are a little more difficult to interpret than the binding results. The fusion protein alone is by definition 1:1 CYP11B1 to adrenodoxin and has almost no activity (a very small peak below the limits of detection). This would be consistent with the adrenodoxin iron-sulfur surface binding tightly enough to its fused CYP11B1 that this same surface is rarely available to interact with adrenodoxin reductase to receive electrons. However, the P450 domain of the fusion protein can perform catalysis under the right conditions. If 40-fold excess of free adrenodoxin is added to the fusion protein reactions, then the normal cortisol product is generated. The simplest explanation for this recovery is that at high enough concentrations, free (reduced) adrenodoxin can at least partially outcompete the fused adrenodoxin to deliver electrons to CYP11B1. While the fusion protein k_cat is low, the K_m is also quite low, even lower than the 1:40 ratio of the non-fused proteins (Figure 3C). A different fusion of all three proteins--CYP11B1, adrenodoxin, and adrenodoxin reductase--was capable of generating product (26), but also had lower activity than the isolated proteins. Overall, the effects observed are consistent with increasing adrenodoxin-mediated improvements in catalysis beyond electron delivery.

The experiments herein are most readily compared with corresponding investigations of the effects of variable adrenodoxin concentrations on CYP11B2 conversion of 11-deoxycorticosterone to corticosterone. As was observed for K_d values, increasing adrenodoxin up to 40-fold had more substantial effects on the K_m for CYP11B1 (2.6-fold) than for CYP11B2 (1.6-fold, (8)), with further ~1.6-fold decreases in the K_m garnered from fusing the two proteins (Figure 3C and (8)). Thus, both binding and catalytic studies demonstrate that while adrenodoxin allosterically regulates both CYP11B enzymes, this effect is more pronounced for CYP11B1.

Adrenodoxin modulates CYP11B1 inhibition

Often substrates and inhibitors bind differently in P450 active sites. While substrates typically displace water on the heme iron but leave the iron free to bind O₂, many inhibitors have a nitrogen lone pair which binds directly to the heme iron, thereby preventing O₂ binding to the heme iron as required for catalysis. Thus, the next set of experiments were designed to determine if the allosteric influence of adrenodoxin was restricted to substrates binding or also modulated inhibitor binding in the active site. Three structurally related inhibitors were evaluated: LCI699, which is an FDA-approved for CYP11B1 inhibition to treat Cushing’s, and both enantiomers of the compound it was derived from, the breast cancer drug fadrozole (Figure 1A).

All three inhibitors bind so tightly to CYP11B1 that spectral determination of the K_d values is limited by the assay sensitivity. The estimated values in the absence of adrenodoxin (11–15 nM) are substantially below the lowest CYP11B1 concentrations, reducing the accuracy of the values. This also precludes evaluation of the effects of adrenodoxin on these K_d values. However, changes in the maximal saturation are significant and in the range that can reasonably be evaluated. For all three inhibitors the addition of even an equimolar amount of adrenodoxin increased the saturation (1.5- to 2.5-fold), with (S)-fadrozole binding to the greatest extent (Figure 4).

As an orthogonal way to profile allosteric effects on inhibitor interactions with CYP11B1, IC₅₀ values were determined in the presence of increasing amounts of adrenodoxin for all three inhibitors. For each compound, the IC₅₀ decreased by 1.5- to 2.6-fold upon increasing the CYP11B1: adrenodoxin ratio from 1:1 to 1:10. The IC₅₀ values for all three decreased ~10-fold when going from 1:1 to 1:40. Earlier experiments herein demonstrated that substrate affinity and saturation is increased in the presence of increasing adrenodoxin. If adrenodoxin increased the affinity/saturation by substrate and inhibitor equally, then the effects would offset each other in IC₅₀ experiments. The fact that inhibition is significantly more effective at high adrenodoxin concentrations is consistent with the idea that allosteric effects on inhibitor binding and saturation may be much more substantial than its effects on substrates. The marked differences in IC50 values depending on the adrenodoxin concentration employed suggests caution in comparing drugs in transformed cells and other systems where adrenodoxin levels may be unknown or poorly controlled and could lead to misleading compound ranking.

While increasing adrenodoxin concentrations also decreased the IC₅₀ for CYP11B2 inhibition by LCI699, the decrease was 4-fold for the 1:1 vs. 1:40 excess adrenodoxin condition. Thus, again the magnitude of allosteric effect was much larger in CYP11B1 than in CYP11B2.

Possible mechanisms

While the results herein establish that there is complex communication between the proximal surface of CYP11B1 where adrenodoxin binds and the CYP11B1 active site, the mechanism for this communication is unknown. Evaluation of the effects on catalysis are most complex because the adrenodoxin must cycle between oxidized and reduced forms and multiple intermediates are involved. A range of observations for other P450 enzymes are potentially relevant. Raman spectroscopy of CYP11A1 found that adrenodoxin binding modulated the bond strengths of the dioxygen species (27), thus influencing the P450 catalytic cycle. A study of the vitamin D metabolizing CYP27B1 found that coupling efficiency increases with higher amounts of adrenodoxin (28), which would increase k_cat. Crosslinking studies with CYP24A1 suggested that adrenodoxin binding promotes a conformational change in the F and G helices (9) that might facilitate catalysis. Hydrogen/deuterium exchange studies with the microsomal CYP46A1 demonstrated that adrenodoxin binding causes changes in the deuteration of several areas of the protein, including the active site (29). Any or all of these might also apply to CYP11B1 as well. However, it is likely that adrenodoxin interactions and effects vary between P450 enzymes. Kinetic studies revealed that adrenodoxin mutations differentially affected the apparent K_m for CYP11A1 and CYP11B1 (25) and catalytic rates depend differently on adrenodoxin concentration among P450 enzymes (13). Additionally, the effect of the adrenodoxin oligomeric state on the allosteric effect is unknown and potentially additionally confounding. While several studies report that adrenodoxin is a functional dimer (30,31), others report that adrenodoxin exists in an equilibrium between a monomer and dimer in the oxidized form, but is a monomer in the reduced form (32). Finally, effects of adrenodoxin concentration on IC₅₀ are an important alert in the evaluation of mitochondrial P450 inhibitors as potential drugs, in these experiments both inhibitor and substrate are present simultaneously and while adrenodoxin increases the binding of both ligands, the results may vary between substrate and inhibitor. Regardless of the mechanism, significant decreases are observed in K_m and IC₅₀ values with increasing adrenodoxin concentrations.

Narrowing down the focus to the effects of adrenodoxin on ligand binding simplifies matters by excluding multiple catalytic intermediates, adrenodoxin redox cycling and potential changes in oligomeric state, heme redox cycling, and competition between two ligands. In this case the substantial increases observed for CYP11B1 affinity for its 11-deoxycortisol substrate and for saturation by both type I substrate and type II inhibitors strongly support close communication between the proximal adrenodoxin-binding surface and CYP11B1 interactions with ligands on the opposite side of the heme in the relatively buried active site. Several underlying mechanisms are possible. First, there may be a specific series of residues between the two binding sites that transduce adrenodoxin interaction to active site residues dictating either direct, specific ligand interactions. Earlier studies with CYP11A1 and adrenodoxin established that binding is dominated by electrostatic interactions and predicted that the CYP11A1 residues K339 and K343, which are conserved in mitochondrial but not microsomal P450s, would be critical to adrenodoxin binding (33). A later structure of CYP11A1 with bound adrenodoxin did incorporate the predicted residues but did not show substantial structural changes compared to the structure of CYP11A1 alone (19). Second, it is also possible that adrenodoxin has larger conformational effects on active site opening and closing dynamics via residue differences in the substrate access channel. Binding of cytochrome b₅ on the proximal surface of steroidogenic CYP17A1 has been demonstrated to alter the positioning of residues involved in substrate access in the F/G region on the opposite side of the P450 protein (34). While closed active sites are most commonly observed for human P450 enzymes, open forms must exist and have been observed in a number of cases. However, comparison of the CYP11B2 structure and a structure of CYP11B2 with bound adrenodoxin did not reveal substantial changes in the substrate access channel or other aspects of the active site conformation. It is possible that crystallographic structures are not capturing relevant dynamic aspects of adrenodoxin allostery and other structural approaches will be more informative.

Some clues for these allosteric effects may be garnered by comparing the effects of adrenodoxin on different mitochondrial P450 enzymes. Comparison of CYP11B1 and CYP11B2 (8), which are 93% identical, reveal that while the trends are the same, CYP11B1 is more affected over the 40x range of adrenodoxin concentrations but CYP11B2 is more responsive at lower adrenodoxin concentrations. This variable responsiveness can be more reasonably assumed due to the 7% of the residues that differ between the two enzymes. Two possibilities appear most obvious. First, the differences in their allosteric responsiveness may be due to differences in the specific interactions that adrenodoxin makes with the two CYP11B enzymes. While we recently determined a structure of CYP11B2 with adrenodoxin (8) which identifies specific interactions, there is no corresponding structure of CYP11B1 with adrenodoxin to compare. However, there is a first structure of CYP11B1, co-crystallized with an inhibitor (3) and alignment of this structure with the CYP11B2/adrenodoxin complex revealed differences in the conformation of the meander region correlated with the only substitution in the interface (8). In CYP11B2, H439 interacts with an arginine in the meander region, while CYP11B1 has a tyrosine at this position that cannot make these contacts (3). The change in this residue correlates with a completely distinct conformation of the meander region, which could lead to the observed functional differences with adrenodoxin. A second possibility is that the adrenodoxin/CYP11B interactions are essentially conserved, but the effects are expressed differentially due to substitutions elsewhere in the proteins. While the first shell of active site amino acids is completely conserved in both CYP11B enzymes, there are somewhat differential positioning of these residues, which is likely due to second-shell substitutions. Alternatively, larger scale dynamics might be differentially influenced due to substitutions across the proteins, nowhere near the active site.

Further work is required to determine the basic mechanism by which adrenodoxin binding on the proximal surface of some P450 enzymes like CYP11Bs has such a significant effect on ligands on the opposite side of the heme in the relatively buried active site. Determining whether the increase affinity is due to adrenodoxin modulation of the on vs. off rate might inform ideas about conformational changes. Additionally, investigation and quantitation of adrenodoxin allosteric effects on other mitochondrial P450 enzymes might yield important clues. For example, unlike the results for the CYP11B enzymes, fusion of adrenodoxin to CYP11A1 did not alter the CYP11A1 affinity for its initial substrate cholesterol (19).

CONCLUSIONS

CYP11B1, the primary human enzyme generating the stress hormone cortisol, is a drug target for the treatment of Cushing’s disease. The work herein reveals that CYP11B1 is allosterically modulated by interactions with its redox partner adrenodoxin, separate from its electron-delivery role. Herein it was revealed that oxidized adrenodoxin allosterically modulated CYP11B1 substrate and inhibitor binding, catalysis, and inhibition. Similar allosteric effects are seen in CYP11B2, but the magnitude of the allosteric effect is greater for CYP11B1 and appears to be absent for CYP11A1. These differences observed may arise from substitutions yielding disparate adrenodoxin interactions or from more distributed substitutions with either indirect effects on ligand/active site interactions or more global conformational changes. While this work reveals an important mechanistic aspect of cytochrome P450 enzymes that may be part of their complex biological control, more detailed structural and functional studies are required to elucidate the mechanism.

EXPERIMENTAL PROCEDURES

Materials

LCI699 (osilodrostat or Isturisa^®) (CAS 928134-65-0) was purchased from Selleckchem. Progesterone (CAS 57-83-0), cortisol (50-23-7), and 11-deoxycortisol (152-58-9) were obtained from Sigma-Aldrich (St. Louis, MO). (R)- and (S)-Fadrozole were a gift from Pfizer.

Protein expression and purification

Generation of CYP11B1, adrenodoxin, and adrenodoxin reductase

Expression and purification of human CYP11B1, human adrenodoxin, and human adrenodoxin reductase were as described previously (3).

Generation of an adrenodoxin-CYP11B1 fusion protein for functional studies.

An artificial adrenodoxin-CYP11B1 fusion enzyme was generated to study the normally transient interaction between the two proteins. The construct coding for this fusion was generated using overlap extension PCR. The N-terminal AB fragment coding adrenodoxin followed by the TDGTS linker was generated by using pLW_Adx (expression plasmid for human adrenodoxin; gift from Richard Auchus) as the template with the forward primer 5’-catgcatatgAGCAGCAGCGAGGATAAAATC-3’ and reverse primer 5’-gaacgctggtaccatcggtGCTGGTTTTACCCACATCAATG-3’. In all primers, the capitalized nucleotides code for adrenodoxin, underlined for the TDGTS linker, and bold for CYP11B1. The C-terminal CD fragment coding for the TDGTS linker followed by truncated CYP11B1 was generated by using pCW11B1 (3) as the template with the forward primer 5’-CCAGCaccgatggtaccagcgttcccgcacagttctc-3’ and reverse primer 5’-catgaagcttttaatgatgatgatggttaatc-3’ (italics=stop codon). The resulting 400 bp fragment expected for reaction AB containing the adrenodoxin+linker and the 1,450 bp fragment expected for reaction CD containing the linker+truncated 11B1 were run on an agarose gel and purified from it. Fragments AB and CD were then used as templates in a final round of PCR employing the forward primer of the AB reaction and the reverse primer of the CD reaction. An agarose gel revealed a fragment just under 2000 bp, the expected size for the sequence encoding the adrenodoxin-TDGTS-CYP11B1 fusion protein. This fragment was purified from the gel, digested with NdeI and HindIII and ligated into the similarly digested pCW11B2dH vector, replacing the 11B2 coding region with the adrenodoxin-TDGTS-11B1 coding region. This construct was validated by redigestion and DNA sequencing. The adrenodoxin-CYP11B1 fusion protein was expressed from this construct and purified as described previously for adrenodoxin-CYP11B2 (8). An SDS-PAGE of the purified adrenodoxin—CYP11B1 fusion protein resulted in a single band at 70 kDa. The ratio of absorbance at the Soret peak (419 nm) over the absorbance at 280 nm was 1.05.

Ligand binding assays: Constant adrenodoxin with ligand titration

Ligand binding studies were utilized to determine the CYP11B1 dissociation constant (K_d) and maximal absorbance (A_max) at saturation for substrate and inhibitors. For each binding titration, CYP11B1 and a specific molar ratio of adrenodoxin were incubated in assay buffer (50 mM potassium phosphate, pH 7.4, 20% glycerol, 0.5% CHAPS) for 10 minutes. Each experiment thus had adrenodoxin constant at 1:0, 1:1, 1:10, or 1:40 adrenodoxin:CYP11B1 ratios. For 11-deoxycortisol substrate binding experiments the CYP11B1 concentration was 1 μM and the mixture was transferred to quartz cuvettes with a path length of 1 cm. To evaluate tightly-binding inhibitors (LCI699, (R)-fadrozole, and (S)-fadrozole) the CYP11B1 concentrations were reduced as much as possible (0.1, 0.15, or 0.2 μM, respectively) and the mixture was transferred to quartz cuvettes with a 5 cm path length. Ligand stocks dissolved in DMSO were added to the sample cuvette, while an equal amount of DMSO was added to the reference cuvette. The cuvettes were incubated for 8 minutes, and difference spectra were recorded from 300–500 nm. Upon substrate binding a peak forms at 389 nm and a trough at 424 nm. Upon inhibitor binding, a peak forms at 428 nm and a trough forms at 410 nm. For each spectrum, the difference between the peak and trough was measured and plotted against the ligand concentration. The data was then fit to either the one-site binding equation (for substrate) or the tight-binding (Morrison) equation in GraphPad (La Jolla, CA) to account for ligand depletion, yielding the K_d and A_max.

Ligand binding assays: Constant ligand with adrenodoxin titration

As an orthogonal control to the above binding experiments with constant adrenodoxin and increasing ligand concentrations, spectral changes were also investigated for constant ligand and increasing adrenodoxin concentrations. CYP11B1 was diluted to 1 μM in the same assay buffer above in a 1 cm quartz cuvette and baselined. Then 11-deoxycortisol was added to 10 μM, which yields a normal type I spectral shift but only partially saturates CYP11B1. Adrenodoxin was then titrated into the sample cuvette and reference cuvette, along with enough ligand to keep the ligand concentration constant at 10 μM. After each addition, the difference spectrum was again recorded. Spectra were multiplied by a dilution factor to account for protein dilution.

Determination of kinetic parameters

Purified CYP11B1, purified adrenodoxin reductase (3), and purified adrenodoxin (3) were incubated together for 20 minutes at room temperature, then added to the reaction buffer (50 mM potassium phosphate, pH 7.4, 20% glycerol, 0.5% w/v CHAPS) containing 1–200 μM of 11-deoxycortisol in DMSO. The total DMSO concentration in each reaction was kept constant at 0.5%. The final enzyme concentrations were varied as follows: 0.05/0.05/2 μM of CYP11B1/adrenodoxin reductase/adrenodoxin for 40-fold adrenodoxin, 0.05/0.05/0.5 μM for 10-fold adrenodoxin. Extensive efforts were made to also keep the CYP11B1 concentration constant at 0.05 μM for reactions investigating equimolar ratios of CYP11B1 and adrenodoxin as well as the adrenodoxin-CYP11B1 fusion enzyme, but production of the cortisol product was so low that the respective enzyme concentrations had to be increased to 0.2/0.2/0.2 μM for these conditions. After these reconstituted enzyme systems were assembled, they were incubated for 3 minutes in a 37 °C water bath, then initiated by 1 mM NADPH dissolved in reaction buffer, giving a final reaction volume of 500 μL. The equimolar and 10-fold excess adrenodoxin reactions were each run for 10 minutes, but the 40-fold excess adrenodoxin reactions were run for 30 seconds to prevent excess substrate depletion. Under these conditions substrate consumption was less than 2.0%, 6.6%, 3.0% for the 40-fold, 10-fold, and equimolar adrenodoxin reactions, respectively. At the stated time, each reaction was quenched by the addition of 500 μL chloroform. Progesterone in DMSO was then added to 40 μM to each reaction as an internal standard. Steroids were extracted by vortexing the aqueous buffer and organic chloroform layers, centrifugation to separate the layers, and by removal of the bottom organic layer by pipetting. This process was repeated a second time with another 500 μl chloroform to increase extraction efficiency. Residual chloroform was removed via evaporation overnight and the product was resuspended in 100 μL 20% acetonitrile for HPLC analysis.

Determination of the half-maximal inhibitory concentration (IC₅₀)

IC₅₀ assays were conducted as described immediately above, except the 11-deoxycortisol substrate concentration was held constant at the value of the K_m and increasing amounts of the inhibitors LCI699, (R)-fadrozole, and (S)-fadrozole dissolved in DMSO were added. The DMSO concentration was kept constant at 0.5% for all reactions.

Chromatography for steady state and IC₅₀ assays

To quantitate the steroid products for both steady state and IC₅₀ assays, 30 μL of each sample was injected onto a reverse-phase Phenomenex Luna^® 5 μm C18 column (150 × 4.6 mm) at 40 °C with a flow rate of 0.8 ml/min. The methods were run as a gradient from A (10% (v/v) acetonitrile, 90% (v/v) water) to B (100% acetonitrile). A 50-minute gradient as follows was sufficient for steady-state kinetic analysis: 0–6 minutes, 20% B; 6–32 minutes, 20–40% B; 32–43 min, 40–80% B; 43–45 min, 80% B; and 45–50 min, 20% B. Longer gradients were necessary to achieve optimal separations for IC₅₀ assays. For IC₅₀ assays using LCI699, a 70-minute gradient was used as follows: 0–6 min, 20% B; 6–52 min, 20–40% B; 52–63 min, 40–80% B; 63–65 min, 80% B; and 65–70 min, 20% B. For IC₅₀ assays using the fadrozole enantiomers, a 90-minute gradient was used as follows: 0–6 min, 20% B; 6–72 min, 20–40% B; 72–83 min, 40–80% B; 83–85 min, 80% B; and 85–90 min, 20% B.

Steroids were detected via absorbance at 240 nm. For the 50-minute gradient used for steady-state analysis, retention times for cortisol, 11-deoxycortisol, and progesterone were 10.5–11.4, 18.9–19.8, and 39.8–40.2 minutes, respectively. For the 70-minute gradient, retention times for cortisol, 11-deoxycortisol, and progesterone were 10.9–11.3, 21.3–21.9, and 57.5–57.7 minutes, respectively. For the 90-minute gradient, retention times for cortisol, 11-deoxycortisol, and progesterone were 11.2–12.3, 23–25, and 71.5–74 minutes, respectively. A standard curve was constructed to quantify the amount of cortisol produced from the ratio of the cortisol to progesterone peak areas. Samples containing varying amounts of cortisol and a constant progesterone concentration were run on the HPLC and the ratio of the peak areas was plotted against the known cortisol concentration to construct the standard curve.

For steady-state kinetic experiments, the amount of cortisol produced was used to calculate initial rates and these initial rates were plotted against the substrate concentration in GraphPad Prism (La Jolla, CA). The Michaelis-Menten equation was used to fit to the data and determine K_m and k_cat.

For IC₅₀ assays, the amount of cortisol produced with no inhibitor present was set as 100% maximal activity, and the amount of cortisol produced in inhibitor-containing samples was scaled to this value to calculate the percent maximal activity. These percent maximal activity values were plotted against the inhibitor concentration in GraphPad Prism (La Jolla, CA). The data was fit to the [inhibitor] vs. response, 4 parameter variable slope equation, with constraints to hold the Top=100 and the Bottom=0 and as Rout outlier analysis to determine the IC₅₀ value.

Data Availability Statement

All data are included in the manuscript and the supplement.