Binding of cytochrome P450 27C1, a retinoid desaturase, to its accessory protein adrenodoxin

Abstract

Of the 57 human cytochrome P450 (P450) enzymes, seven are mitochondrial: 11A1, 11B1, 11B2, 24A1, 27A1, 27B1, and 27C1. Mitochondrial P450s utilize an electron transport system with adrenodoxin (Adx) and NADPH-adrenodoxin reductase (AdR). AdR reduces Adx, which then transfers electrons to the P450. The interactions between proteins in the mitochondrial P450 system are largely driven by electrostatic interactions, though the specifics vary depending on the P450. Unlike other mitochondrial P450s, the interaction between P450 27C1, a retinoid 3,4-desaturase expressed in the skin, and Adx remains largely uncharacterized. In this work, we utilized an Alexa Fluor 488 C₅ maleimide-labeled Adx to measure binding affinities between Adx and P450 27C1 or AdR. Both proteins bound Adx tightly, with K_d values < 100 nM, and binding affinities decreased with increasing ionic strength, supporting the role of electrostatic interactions in mediating these interactions. Cross-linking mass spectrometry and computational modeling were performed to identify interactions between P450 27C1 and Adx. While the residues of Adx identified in interactions were consistent with studies of other mitochondrial P450s, the binding interface of P450 27C1 was quite large and supported multiple Adx binding positions, including ones outside of the canonical Adx binding site. Additionally, Adx did not appear to be an allosteric effector of P450 27C1 substrate binding, in contrast to some other mitochondrial P450s. Overall, we conclude that P450-Adx interactions are P450-specific.

1. Introduction

Cytochrome P450s (CYPs or P450s) constitute a family of enzymes involved in the oxidation of a variety of endogenous and exogenous small molecules, including steroids, eicosanoids, vitamins, fatty acids, and xenobiotics [1]. P450 oxidation reactions require the donation of two electrons. There are two systems of proteins that can provide these electrons to human P450s, depending on the subcellular localization: the microsomal NADPH-cytochrome P450 reductase (POR) and cytochrome b₅ (b₅) system and the mitochondrial NADPH-adrenodoxin reductase (AdR) and adrenodoxin (Adx) system. AdR is a flavoprotein that transfers electrons from NADPH to Adx, a small, ~12 kDa, [2Fe-2S] cluster protein which can then subsequently transfer electrons to the mitochondrial P450s. Of the 57 human P450 enzymes, seven (P450s 11A1, 11B1, 11B2, 24A1, 27A1, 27B1, and 27C1) are intrinsically mitochondrial and utilize the AdR and Adx proteins for electron transfer [2]. The mitochondrial P450s are involved in the metabolism of endogenous substrates throughout the body: P450 11A1 converts cholesterol to pregnenolone, P450 11B1 and 11B2 generate cortisol and aldosterone, P450 24A1 and 27B1 hydroxylate vitamin D₃, P450 27A1 is involved in the biosynthesis of bile acids, and P450 27C1 desaturates retinoids (vitamin A) [1–3].

The interaction between mitochondrial P450s and Adx is dominated by electrostatic interactions [4]. Adx is generally thought to bind via its F-helix to the proximal surface of P450s, specifically to positively charged residues of the K-helix. Both of the existing P450-Adx fusion structures (P450s 11A1 [5] and 11B2 [6]), have Adx in this position. Outside of the generally conserved interaction interface, there is variability among the P450s with some having additional identified interactions. Due to the contribution of electrostatic forces to Adx interactions, the interactions between Adx and AdR and mitochondrial P450s are ionic strength dependent [7–9].

There are multiple models for the mechanism of electron transfer between the three proteins. The predominant model is the shuttle mechanism, where oxidized Adx binds to AdR, dissociates after being reduced, and then binds to the P450 and transfers the electron [8,10]. There is also a modified shuttle mechanism, where an Adx dimer is involved instead of a monomer [11]. The ability of Adx to form functional dimers has been proposed [12,13]. A ternary complex has also been proposed, with Adx interacting with both AdR and the P450 [14], as well as a quaternary complex involving two Adx proteins and one AdR and P450 [15]. Work with P450 11B1 and 11B2 has described the interaction of the P450 with a dimer of Adx [9]. An Adx dimer has also been proposed to interact with P450 24A1 based on the large binding interface detected by NMR [16]. On the other hand, cross-linking studies with P450 11A1 and Adx suggest interaction with a monomer [17]. Mechanisms may vary depending on the P450 involved.

P450 redox partners can be allosteric effectors of P450 activity in addition to their roles as electron donors. Perhaps the most well studied example of this is P450 17A1 and b₅, where it has been proposed that b₅ can stimulate the 17,20-lyase reaction by facilitating a conformational change of the P450 [18–20]. Allosteric stimulation by b₅ has also been observed with many other P450s (reviewed in [21]) and there is also some evidence for allosteric modulation with POR [22,23]. In the past decade, there have been multiple reports characterizing the potential allosteric role of Adx with mitochondrial P450s. P450 11B2 substrate binding was promoted by Adx [6,24]. In contrast, Adx binding to P450 24A1 reduced substrate binding affinity while stabilizing the enzyme-substrate complex and altering substrate positioning within the active site [16,25,26] and the P450 11A1-Adx fusion protein did not exhibit higher binding affinity for its substrate, unlike P450 11B2-Adx [5,6]. Adx has also been proposed to be an effector for oxygen transfer to P450 27B1 [27]. If Adx binding to the P450 affects substrate binding, substrate binding should also affect Adx binding to the P450, as the sum of these binding steps should be energetically equivalent based on the thermodynamic box principle [28]. Yablokov et al. have reported that the presence of substrate can also modulate the P450-Adx interaction in the cases of P450 11A1, 11B1, and 11B2 and that this effect this is P450- and substrate-specific [29]. Mutually facilitated Adx and substrate binding has been observed with P450 11A1 and P450 27B1 [30,31]. P450 24A1 ligand interactions can also affect Adx recognition [32].

Overall, work surrounding the nature of interactions between Adx and mitochondrial P450s has pointed to numerous differences depending on the P450 in question. In comparison to other mitochondrial P450s, very little is known about the nature of the interaction between P450 27C1 and Adx. P450 27C1 is an all-trans retinoid desaturase expressed in the skin [3,33] and is the only human mitochondrial P450 for which there is no structural information describing the interaction with Adx. Our group has previously measured the rate of P450 27C1 reduction by Adx (3.6 min⁻¹ with substrate present) and the estimated binding affinity between the two proteins (220 nM) utilizing microscale thermophoresis (MST) [33,34], but additional details about the complex are not known. In this work, we characterized some aspects of the nature of the P450 27C1-Adx complex and investigated potential allosteric effects of Adx binding to P450 27C1.

2. Materials and Methods

2.1. Chemicals

All-trans retinol was from Toronto Research Laboratories (Toronto, ON, CA) or Sigma-Aldrich (St. Louis, MO). Retinoid stocks were prepared fresh in absolute ethanol and stored in amber glass. Stock concentrations of all-trans retinol were determined spectrophotometrically based on ε₃₂₅ = 52,770 M⁻¹ cm⁻¹ [35]. 2× Laemmli sample buffer was from Bio-Rad (Hercules, CA). NuPAGE 4–12% and 10% Bis-Tris Gels, Simply Blue SafeStain, NuPAGE MES and MOPS Running Buffers, 6-acrylodan-2-dimethylaminonapthalene (acrylodan), 5-((((2-iodoacetyl)amino)ethyl)amino)naphthalene-1-sulfonic acid (IAEDANS), and Alexa Fluor 488 C₅ maleimide were from Invitrogen (now Thermo Fisher Scientific, Waltham, MA). 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC), N-hydroxysulfosuccinimide (Sulfo-NHS), and Zeba Spin Desalting Columns were from Thermo Scientific (now Thermo Fisher Scientific). Dimethylsulfoxide (DMSO), L-α-dilauroyl-sn-glycero-3-phosphocholine (DLPC), HPLC-grade solvents, and all other reagents were purchased from Millipore Sigma/Sigma Aldrich.

2.2. Recombinant proteins

Human P450 27C1 (N-terminus modified, residues 3–60 deleted with expression optimized sequence), human Adx, and bovine AdR were expressed in Escherichia coli and purified as described previously [34,36–38]. The human Adx expression plasmid (pLW01) was provided by Dr. Richard Auchus (University of Michigan) [9].

2.3. Fluorescently labeled Adx

Human Adx was labeled with three different cysteine-reactive fluorescent dyes. Stocks of the fluorescent dyes were made in CH₃CN (acrylodan) or DMSO (IAEDANS, Alexa Fluor 488 C₅ maleimide). Human Adx (100 μM) was incubated with 10-fold molar excess dye in 100 mM potassium phosphate (pH 7.4) for 20 h at 4 °C in a black microcentrifuge tube. The final concentration of organic solvent was < 10% (v/v). Zeba Spin Desalting Columns were used according to the manufacturer’s instructions to remove excess dye. Spectra of the final purified, labeled proteins were recorded using a NanoDrop spectrophotometer (Thermo Fisher Scientific) (Alexa Fluor 488 in Fig. 1, other data not shown). The concentration of labeled Adx was calculated by measuring the absorbance and using ε₄₉₃ = 72,000 M⁻¹ cm⁻¹ for Alexa Fluor 488. Fluorescence spectra of each labeled Adx preparation were recorded using an OLIS DM-45 spectrofluorometer (On-Line Instrument Systems, Athens, GA) (Alexa Fluor 488 in Fig. 1, other data not shown). The extent of labeling was calculated by dividing this concentration by the calculated concentration of Adx based on ε₄₁₄ = 9,800 M⁻¹ cm⁻¹ [39]. The A₄₁₄ value from Alexa Fluor 488 dye itself is only ~2% of the A₄₉₃, and this was not corrected for in calculation of the Adx concentration. The purity of Alexa Fluor 488-Adx and unlabeled Adx was assessed by SDS-PAGE (Supplementary Material Fig. S1).

Fig. 1.

Spectral properties of Alexa Flour 488-Adx. (A) Dye structure; (B) dye-Adx absorbance spectrum (87 μM Alexa Fluor 488-Adx, NanoDrop); (C) dye-Adx fluorescence spectrum (43.5 nM Alexa Fluor 488-Adx) in 100 mM potassium phosphate (pH 7.4) (ex: 493 nm). Labeling efficiency was calculated to be ≥ 93%.

3. Experimental Procedures

3.1. Equilibrium Adx binding titrations

P450 27C1 (0–290 nM) or AdR (0–502 nM) was titrated into a solution of Alexa Fluor 488-Adx (50 nM) in potassium phosphate buffer (50, 100, or 200 mM, pH 7.4). Titrations were performed in a semi-micro 1-cm quartz cell with clear windows (Starna Cells, Atascadero, CA, catalog # 29F-Q-10) in an OLIS DM-45 spectrofluorometer with 1.24 mm slits and an integration time of 0.1 s with 493 nm excitation and emission spectra recorded from 500–600 nm. To determine potential effects of P450 substrate binding on Adx binding affinity, titrations were also performed with P450 27C1 incubated with an equal molar concentration of all-trans retinol. Titrations were performed in triplicate. The fluorescence at the emission maximum was normalized and plotted against the concentration of P450 or AdR and fit with a quadratic binding equation, known commonly as the Morrison equation [40], in GraphPad Prism (San Diego, CA) to calculate the K_d value (Eq. 1). In this equation, ΔF_max is the extrapolated fluorescence difference at an infinite ligand (P450 27C1 or AdR) concentration, E_T is the concentration of Adx, X is the concentration of P450 27C1 or AdR, and K_d is the dissociation constant.

$$\Delta F=\Delta F_{max}\frac{\left(E_T+K_d+X\right)+\sqrt{{\left(E_T+K_d+X\right)}^2-4E_{T}X}}{2E_T}$$ Eq. 1

For normalization, the minimum emission value during the titration was subtracted from the emission value at each point, and then this value was divided by the range in emission values observed in the titration. Preliminary titrations were also performed with P450 27C1 and the acrylodan- (excitation 391 nm; emission 450–700 nm) and IAEDANS-Adx (excitation 337 nm; emission 350–600 nm).

3.2. Catalytic assays

P450 27C1 catalytic assays were performed similarly as previously described [3,33,38]. Reactions contained 0.02 μM human P450 27C1, 0–10 μM human Adx or Alexa Fluor 488-Adx, 0.2 μM bovine AdR, 16 μM DLPC, 0.5 μM all-trans retinol, and 1 mM NADPH in 50 mM potassium phosphate buffer (pH 7.4). The final reaction volume was 500 μL (< 1% ethanol, final, v/v) and reactions were done in amber vials. Samples were preincubated at 37 °C for 5 min in a shaking water bath before initiation of the reaction with NADPH. Reactions were quenched after 60 s by vortex mixing with 1.0 mL tert-butyl methyl ether containing 20 μM butylated hydroxytoluene. An aliquot of the upper layer (0.7 mL) was removed, dried under N₂, and resuspended in 50% (v/v) ethanol for analysis. Chromatography (UPLC-UV/Vis) was performed as previously described [38]. Relative product formation was calculated by dividing the product peak area by the sum of the substrate and product peak areas.

3.3. Equilibrium substrate binding titrations

All-trans retinol (0–1.04 μM) was titrated into P450 27C1 (90 nM) in 200 mM potassium phosphate buffer (pH 7.4) in the presence of varying concentrations of Adx (0–1.8 μM). To enable absorbance measurements at this low P450 27C1 concentration, titrations were performed in a 10-cm cell (25 mL solution) (Starna Cells, catalog # 34-Q-100). The final volume of ethanol in the titration was <0.05 % (v/v). Spectra were recorded with an OLIS Cary-14 spectrophotometer from 350–500 nm. The reference spectrum (with no substrate) was subtracted from each titration spectrum. ΔA₃₉₀–A₄₂₀ was calculated at each concentration, with the substrate-free value corrected to 0. The plot of absorbance difference versus substrate concentration was fit in Prism software (GraphPad, San Diego, CA) using the quadratic equation (same as shown in Eq. 1, but with ΔA), where E_T is the concentration of P450 27C1 and X is the concentration of all-trans retinol.

3.4. Crosslinking

EDC (2 mM) and Sulfo-NHS (5 mM) was added to Adx or Alexa Fluor 488-Adx (40 μM) in 100 mM potassium phosphate buffer (pH 7.4) and incubated at 23 °C with shaking for 15 min. EDC and Sulfo-NHS solutions were made in water immediately before addition. P450 27C1 (2 μM) was added to the reaction (final volume 20 μL) and the samples were incubated for 2 h at 23 °C with shaking before quenching with an equal volume of 2× Laemmeli buffer (with β-mercaptoethanol). The samples were heated at 90 °C for 10 min and then loaded onto a NuPAGE 4–12% Bis-Tris SDS-PAGE gel with MOPS running buffer. The gel was stained with SimplyBlue SafeStain.

3.5. Proteomic analysis

The excised gel pieces were destained, reduced, and alkylated as previously described [41]. Briefly, regions of interest were excised from SDS-PAGE and cut into small, ~1 mm cubes and incubated with 100 mM ammonium bicarbonate, 30% CH₃CN for 30 min with shaking. This solution was removed, and then samples were reduced with 10 mM dithiothreitol in 100 mM ammonium bicarbonate for 30 min at 55 °C. Samples were then alkylated with 55 mM iodoacetamide in 100 mM ammonium bicarbonate for 30 min at room-temperature in the dark. Gel pieces were shrunk by shaking with CH₃CN between treatments and for storage at −20 °C until digestion. Digestion was performed with trypsin (0.2 μg) at 37 °C for 16 h. Extractions (two with 0.1% CF₃CO₂H, 60% CH₃CN (v/v), and one with 0.1% CF₃CO₂H, 80% CH₃CN (v/v)) were done to remove peptides from the gel pieces. This solution was dried and resuspended in 0.2% HCO₂H. To determine the site of Alexa Fluor 488-labeling, samples were worked up similarly.

Peptides were loaded onto a 21.5 cm capillary (360 μm outer diameter, 100 μm inner diameter) octadecylsilane (C18) reversed phase (Jupiter, 3 μm beads, 300 Å, Phenomenex) analytical column using a Dionex Ultimate 3000 nanoLC. The mobile phase solvents consisted of: solvent A: 0.1% HCO₂H in H₂O; solvent B: 0.1% HCO₂H in CH₃CN (all v/v). Peptides were eluted with gradient of: 0–2 min, 2% B; 2–73 min, 2–40% B; 73–78 min, 40–95% B; 78–79 min, 95% B; 79–80 min, 95–2% B; 80–90 min, 2% B (all v/v) at a flow rate of 0.35 μL min⁻¹. A Q-Exactive Plus mass spectrometer was used in positive ion mode for full MS/data-dependent MS2 (Top 15) analysis. The full MS settings were as follows: microscans, 1; resolution, 70,000; AGC target, 3e6; maximum IT, 60 ms; scan range, 375 to 1800 m/z. The dd-ms2 settings were: microscans, 1; resolution, 17,500; AGC target, 1e5; maximum IT, 100 ms; loop count, 15; MSX count, 1; TopN, 15; isolation window, 2.0 m/z; scan range, 200 to 2000 m/z; (N)CE, 26.

3.6. Peptide data analysis

pLink 2 software (version 2.3.9) [42] was used to identify cross-linked peptides. Software settings were as follows: cross-linker: EDC-DE; enzyme: trypsin; number of missed cleavages: 2; peptide mass: 400–6,000; peptide length: 4–60; precursor tolerance: 20 ppm; fragment tolerance: 20 ppm; fixed modifications: carbamidomethyl[C]; variable modifications: oxidation[M]; filter tolerance: 10 ppm; false discovery rate (FDR): separate FDR < 5% at PSM label. The E. coli proteome with the sequences for recombinant human Adx and P450 27C1 was used as a database (with contaminant proteins added by pLink 2). MS/MS spectra were visualized in pLabel version 2.4.1.

For identification of the site of the Alexa Fluor 488 label, raw data files were analyzed using MyriMatch software version 2.2.140 [43] against the sequence for recombinant human Adx. Software settings were: enzyme: trypsin; precursor ion resolution: high; fragmentation mass resolution: high; modifications: methionine oxidation (15.9949 Da, dynamic), cysteine carbmidomethylation (57.0215 Da, dynamic), Alexa Fluor 488-cysteine (698.0988 Da, dynamic). This was the predominant protonation state of the Alexa Fluor 488 modification, though others were observed (data not shown). Spectra were visualized in IDPicker software [44] and the false discovery rate was set to 5%.

3.7. Structural modeling

The structure of P450 27C1 has not been experimentally determined. The human P450 27C1 structure was downloaded from the AlphaFold Protein Structure Database [45]. The structure in the database was for the full-length protein sequence and the PDB file was modified to remove the first 63 residues (disordered) to align the sequence with the recombinant protein used in this study. The X-ray crystal structure for human Adx was obtained from the Protein Data Bank (PDB ID: 3P1M [46]). Only chain A was used.

PDB files for each protein were used as inputs to HADDOCK 2.4 [47,48] and the amino acids identified in cross-linking mass spectrometry studies were selected as active residues in the interaction. Given the likely conformational flexibility of Adx Lys-127, this residue and cross-links with it were not listed as active residues in the interaction. Other parameters were kept at the default settings. HADDOCK scoring weighs intermolecular van der Waals energy and empirical desolvation energy (both 1.0 multiplier) more highly than intermolecular electrostatic energy (0.2 multiplier) and interaction restraints (0.1 multiplier). Given that electrostatic interactions between P450s and ferredoxin proteins are generally considered more important than van der Waals forces [4], and the cross-linking data was available, preference was given to models with low restraint and electrostatic energies, as opposed to selecting the lowest HADDOCK score model. Heme and the [2Fe-2S] cluster were added into best model from cluster 3 and 4 (prosthetic groups and co-factors are not compatible with docking software, no heme in AlphaFold structure) for measuring distances. All structures were visualized in PyMOL software [49].

4. Results

4.1. Selection of Alexa Fluor 488 for fluorescent-labeling of Adx

Measuring mitochondrial P450-Adx binding affinity has previously been accomplished through assessing the perturbation of the P450 spectrum upon Adx binding [14]. Our previous work has shown that this spin state change does not occur with P450 27C1 [33]. We have previously utilized a fluorescence-based labeling in another technique (microscale thermophoresis) for measuring the interaction, but the labeling lacked specificity (see additional details in Discussion) [34]. For this reason, three cysteine-reactive dyes—acrylodan, IAEDANS (dansyl), and Alexa Fluor 488 C₅ maleimide—were used to label Adx. The absorbance and fluorescence properties of each dye-Adx conjugate were assessed. All three dye-Adx conjugates displayed the expected fluorescence spectrum based on known properties of the dye, indicating successful labeling (Alexa Fluor 488-Adx in Fig. 1, other data not shown). Only the Alexa Fluor 488-Adx resulted in clear changes in the emission spectra upon P450 27C1 binding (Fig. 2, other data not shown), so this dye conjugate was utilized for further studies. Coupling of Alexa Fluor 488 with Adx was also efficient, with ≥ 93% labeling.

Fig. 2.

Representative spectra from titration of P450 27C1 with Alexa Fluor 488-Adx. Excitation of Alexa Fluor 488-Adx was at 493 nm. The fluorescence at the emission maximum increased with increasing concentrations of P450 27C1. The titration was performed in 100 mM potassium phosphate (pH 7.4) with 50 nM of Alexa Fluor 488-Adx and 0–0.29 μM (0, 0.0086, 0.017, 0.026, 0.034, 0.052, 0.069, 0.086, 0.12, 0.15, 0.19, 0.22, 0.29 μM) P450 27C1.

The fluorescence of Alexa Fluor 488-Adx was linear over a large range (Supplementary Material Fig. S2) and labeling Adx with Alexa Fluor 488 did not negatively affect the electron transfer capabilities of Adx to P450 27C1 (Supplementary Material Fig. S3). While Adx has five cysteine residues, the labeling calculation and presence of a single Alexa Fluor 488 modification by proteomic analysis suggests only one cysteine is labeled (Supplementary Material Fig. S4).

4.2. Alexa Fluor 488-Adx binds tightly to P450 27C1 and AdR

Fluorescence spectroscopy titrations were used to measure the dissociation constant of P450 27C1 and AdR with Adx. Varying concentrations of potassium phosphate were used to assess the ionic strength dependence of each interaction. At lower concentrations of potassium phosphate, P450 precipitated from solution causing Rayleigh scattering which prevented the measurement of K_d values (see spectra in Supplementary Material Fig. S5). The K_d of P450 27C1 with Adx ranged from 12–22 nM and the K_d of AdR with Adx ranged from 19–90 nM (Fig. 3, Table 1). While a larger effect was observed with AdR, the binding affinity for Adx with both proteins decreased with increasing ionic strength. Even at higher concentrations of potassium phosphate, these values indicate tight binding between Adx and P450 27C1 or AdR. The presence of all-trans retinol did not affect the binding affinity of P450 27C1 for Adx (Supplementary Material Fig. S6).

Fig. 3.

Fluorometric titration data for P450 27C1 and AdR equilibrium binding to Adx. Binding titrations were completed with: (A) P450 27C1; (B) AdR. Ionic strength was varied by changing the concentration of potassium phosphate buffer: 50 mM (▲), 100 mM (▼), 200 mM (●). At lower ionic strengths with P450 27C1, Raleigh scattering prevented binding measurements. Experiments were completed in triplicate and points are shown as means ± SD. Representative spectra are shown in Supplementary Material Fig. S5.

Table 1.

Dissociation constants for P450 27C1 and AdR with human Adx at varying ionic strengths.

[KPhos] (mM)	P450 27C1 Kd (nM)	AdR Kd (nM)
50	a	19 ± 5
100	12 ± 3	40 ± 10
200	22 ± 6	90 ± 30

Values are derived from fitting in Fig. 3 from experiments completed in triplicate.

^aRaleigh scattering prevented binding measurement.

4.3. Effect of Adx on substrate binding by P450 27C1

With some other mitochondrial P450s, Adx has been shown to affect P450 substrate binding [6,26]. To investigate this possibility with P450 27C1, substrate binding affinity was assessed with varying concentrations of Adx present (0 μM, equal molar, 10-fold excess, and 20-fold excess). P450 27C1 displays a Type I spectral change when binding all-trans retinol ([3], Fig. 4A). Dissociation constants were calculated from quadratic fits of this absorbance change plotted versus the substrate concentration (Fig. 4B, Table 2). The binding affinity for all-trans retinol did not notably vary in the presence of Adx; K_d values ranged from 10 ± 7 nM to 30 ± 20 nM. Variation in ΔA_max values have previously been associated with changes in the amount of the P450 in solution that is able to bind substrate [6]. No changes in A_max were observed across titrations with the addition of Adx, suggesting that the proportion of P450 27C1 that binds all-trans retinol remains the same. At equilibrium, Adx did not appear to alter P450 27C1 substrate binding.

Fig. 4.

Lack of effect of Adx on the equilibrium binding of P450 27C1 to all-trans retinol. (A) Representative all-trans retinol substrate binding spectra from titration with equal concentrations of P450 27C1 and Adx. With subsequent additions (indicated with different colors), the absorbance increased at 390 nm and the A₄₂₀ decreased. Spectra were adjusted to 0 at the isobestic point (407 nm). (B) Quadratic fits for calculation of K_d and A_max values for all-trans retinol binding to P450 27C1 in the presence of no Adx (●), equal molar Adx (■), 10-fold Adx (▲), or 20-fold Adx (▼).

Table 2.

Equilibrium binding constants for P450 27C1 and all-trans retinol at varying concentrations of Adx.

Adx	Kd (nM)	ΔAmax (ΔA390–A420)
No Adx	30 ± 10	0.027 ± 0.002
+ Equimolar Adx	10 ± 7	0.027 ± 0.002
+ 10-fold Adx	30 ± 20	0.030 ± 0.003
+ 20-fold Adx	30 ± 10	0.030 ± 0.002

Values are derived from fitting in Fig. 4 from single titrations.

4.4. Characterization of the P450 27C1-Adx binding interface

No structural information is currently available describing the interaction between P450 27C1 and Adx. To identify the protein-protein interaction interface that exists between P450 27C1 and Adx, chemical cross-linking with EDC was utilized. Both unlabeled Adx and Alexa Fluor 488-Adx were used in reactions. In the absence of cross-linking, Adx and P450 27C1 displayed molecular weights of approximately 13 kDa and 50 kDa, respectively (Fig. 5). Incubation of P450 27C1 and Adx resulted in the appearance of two new bands, at approximately 29 kDa and 62 kDa (labeled by arrows in Fig. 5). The sizes of the cross-links are consistent with a P450 27C1-Adx (monomer) complex (62 kDa) and an Adx dimer (29 kDa). The presence of the Alexa Fluor 488 label did not negatively affect crosslinking with P450 27C1 (Fig. 5, lane 7 vs. 9). Bands of interest were excised and in-gel trypsin digestion was performed. Peptides were analyzed by high-resolution mass spectrometry (HRMS) to identify proteins and cross-linked peptides within the bands of interest.

Fig. 5.

SDS-PAGE of P450 27C1 and Adx cross-links. Proteins in crosslinking reactions mixtures were separated on a 4–12% Bis-Tris gel with MOPS running buffer. SimplyBlue SafeStain was used for total protein staining. Lanes 1 and 10: SeeBlue Plus2 pre-stained protein ladder; lanes 2–4: Adx (2), Alexa Fluor 488-Adx (3), and P450 27C1 (4) proteins (no crosslinking); lanes 5: intentionally left blank; lanes 6 and 7: Adx and P450 27C1 (without (6) and with crosslinking (7)); lanes 8 and 9: Alexa Fluor 488-Adx and P450 27C1 (without (8) and with crosslinking (9)). Cross-linked proteins indicated by arrows were excised for proteomic analysis.

Analysis of the 62 kDa band showed peptides corresponding to both P450 27C1 and Adx. Many cross-linked peptides between P450 27C1 and Adx were identified by LC-MS/MS (Supplementary Material Table S2, Fig. 6, Fig. 7, Supplementary Material Fig. S7). The Lys-117, Glu-127, Glu-147, Lys-318, Lys-403, and Lys-478 residues of P450 27C1 and Asp-77, Glu-78, Asp-84, and Lys-127 residues of Adx were involved in cross-links in the identified peptides (Table 3, numbering according to recombinant protein sequences in Supplementary Material Table S1). The interacting residues on both proteins are localized to a single surface of each protein (Fig. 8A–B). The 62 kDa band also contains three Adx-Adx cross-links, but all of these cross-links involved Lys-127, part of the flexible C-terminal part of the protein, and not other Adx-Adx cross-links identified in the 29 kDa band (Supplementary Material Table S3). These results further support the designation of the 62 kDa cross-link as a P450 27C1-Adx monomer complex as opposed to a P450 27C1-Adx dimer complex.

Fig. 6.

Representative fragmentation of P450 27C1-Adx cross-linked peptides from proximal binding site (cluster 4). MS/MS spectra with fragment assignment for cross-links observed between: (A) Asp-77 of Adx and Lys-117 of P450 27C1; (B) Glu-78 of Adx and Lys-117 of P450 27C1. Full peptide information is shown in Supplementary Material Table S3.

Fig. 7.

Representative fragmentation of P450 27C1-Adx cross-linked peptides from distant binding site (cluster 3). (A) Asp-77 of Adx and Lys-478 of P450 27C1; (B) Asp-84 of Adx and Lys-318 Full peptide information is shown in Supplementary Material Table S3.

Table 3.

Cross-linked residues between P450 27C1 and Adx.

P450 27C1	Adx
Lys-117	Asp-77, Glu-78, Asp-84
Glu-127	Lys-127
Glu-147	Lys-127
Lys-318	Glu-78, Asp-84
Lys-403	Asp-77, Asp-84
Lys-478	Asp-77, Asp-84

Results are from proteomic analysis performed on the 62 kDa band from P450 27C1 cross-linking with Adx (unlabeled). Numbering is based on sequence of recombinant proteins shown in Supplementary Material Table S1. Representative peptide MS/MS information for identification is shown in Supplementary Material Table S2, Fig. 6, Fig. 7, and Supplementary Material Fig. S7.

Fig. 8.

Model of P450 27C1-Adx interaction. Localization of P450 27C1 (A) and Adx (B) residues identified in cross-links. Residues are shown as sticks (red for Adx, blue for P450 27C1) and the rest of the structure is shown as cartoon (light orange for Adx, pale cyan for P450 27C1). (C–D) Modeled P450 27C1-Adx complexes ((C), cluster 4; (D), cluster 3). Residues identified from cross-linking data at the interface are shown as sticks, the rest of the structure is shown as a transparent surface. The heme of the P450 and the [2Fe-2S] cluster of Adx are shown in black. The distance between the [2Fe-2S] cluster and the center of the heme was 26.1 Å in the cluster 3 model and 44.8 Å in the cluster 4 model (shown as dashed line).

HADDOCK software was used to generate P450 27C1-Adx complex models utilizing residues identified in cross-linking studies. HADDOCK clustered 89 structures into 11 clusters (44% of generated models). Energy values for clusters are shown in Supplementary Material Fig. S8. The best structures from the top 10 clusters were manually compared. The P450 27C1 molecule in each complex was aligned to visualize the range of Adx docking positions (Supplementary Material Fig. S9). The range of Adx orientations covers the range of P450 27C1 residues identified in cross-links. The best scoring structures from each cluster were aligned with two experimentally determined x-ray crystallography structures of mitochondrial P450-Adx fusion proteins (11A1, PDB: 3N9Y [5] and 11B2, PDB: 7M8I [6]). The models from cluster 4 had low energy values and were the most similar to the fusion protein structures (Fig. 8C, Supplementary Material Fig. S10). Models from cluster 3 also had low energy values and several experimentally observed cross-links were fit well (Fig. 7D). Interactions between amino acids determined by cross-linking are shown for cluster 4 and 3 (MS/MS spectra in Fig. 6 and 7, respectively). The distance between the [2Fe-2S] cluster and the heme varied depending on the model. Distances were 26.1 Å and 44.8 Å for cluster 4 and 3, respectively.

5. Discussion

With Adx and mitochondrial P450s, the primary methods reported to measure binding affinity are UV/Vis-difference titrations [14], surface plasmon resonance (SPR) [50], microscale thermophoresis (MST) [34], and titrations with fluorescently labeled proteins [51]. In our previous work with MST, we found that our K_d values were lower than those previously reported with SPR [34], suggesting that free diffusion of the proteins may be important for optimal binding. One concern with our previous MST work was that the labeling method utilized could interfere with the protein-protein interactions (a lysine reactive dye was used to label Adx, and there are many lysine residues). In this work we have pursued a method where perturbation of the interaction would be minimized. Adx has five cysteine residues. Four of these residues are involved with binding the [2Fe-2S] cluster, leaving only one available free cysteine for labeling [52]. Because of this, we elected to label Adx with Cys-reactive fluorescence dyes. IAEDANS has previously been utilized to specifically label Adx at an individual cysteine residue to assess interactions with AdR and P450 11A1 [51]. Specific labeling of Cys-95 (Cys-100 of our construct, Supplementary Material Table S1) of Adx was proposed in a study with 5-iodoacetamidofluorescein (5-IAF) [52], though binding affinities were not assessed in that work. Our group has recently utilized Alexa Fluor 488 C₅ maleimide (structure shown in Fig. 1A) for labeling b₅ (T70C mutant) to quantify interactions with a range of microsomal P450s [53,54]. Acrylodan was originally utilized to label a homolog of this b₅ mutant [55]. In the present study, we attempted to label Adx with acrylodan, IAEDANS, and Alexa Fluor 488 C₅ maleimide. While labeling with each dye was successful, we found that the Alexa Fluor 488 was the most effective for labeling Adx in detecting interactions with P450 27C1, in that it was the only dye that resulted in a strong fluorescent signal that was clearly altered by the addition of P450 27C1 (Fig. 2, other data not shown).

Like previous reports of Adx labeling with IAEDANS and 5-IAF [51,52], Alexa Fluor 488-Adx still efficiently supports P450 catalysis (Supplementary Material Fig. S3) and labeling presumably occurs at a single site (Supplementary Material Fig. S4). With the high fluorescence of the Alexa Fluor 488-Adx, very low concentrations of protein could be used (50 nM) which is important given the low K_d values between Adx and AdR/P450s. In the previous work with IAEDANS labeling, 0.5–1 μM of AEDANS-Adx was used and the K_d with AdR was reported to be 0.07 μM [51]. This K_d value is in the range of values of what we determined with AdR and the Alexa Fluor 488-Adx (~20–90 nM, depending on ionic strength, Table 1) but is subject to higher error due to the high concentration of AEDANS-Adx used. The K_d determination for the interaction of AdR and Adx in this work utilized bovine AdR and human Adx. A previous study has shown that human Adx binds to human AdR 5-fold less tightly than human Adx binds to bovine AdR [56]. While this work is focused predominantly on interactions with P450 27C1 and Adx (both human proteins), the studies with AdR are limited due to this apparent species difference. In comparison with MST results [34], the K_d values in this study were lower for P450 27C1 (12 nM versus 220 nM) but equal for AdR (both 40 nM) (Table 1). This points to some improvement in P450 27C1 binding to Adx with the single fluorescent label at Cys-100.

Interactions with Adx are driven by electrostatic interactions, generally making the complexes sensitive to changes in ionic strength. Consistent with previous reports, Adx binding affinity for AdR decreased with increasing ionic strength (Table 1) [8]. Differences in Adx affinity for P450 27C1 were also noted (Table 1), though the range of ionic strengths investigated was limited due to precipitation of P450 27C1 at low ionic strengths (Supplementary Material Fig. S5). The interaction between P450 27C1 and Adx is not nearly as dependent on low ionic strengths as P450 11B1 and 11B2, which did not form cross-links with Adx at potassium phosphate concentrations above 20 mM [9].

The P450 27C1-Adx interaction interface was modeled utilizing data from cross-linking mass spectrometry. While there are many residues of P450 27C1 identified in cross-linked peptides that cover a large surface across the protein (Fig. 8A), the interactions do appear specific, in that there are many other lysine residues on the surface of P450 27C1 that were not detected in cross-links. While previously identified large interfaces have been used to suggest the interaction of an Adx dimer [16], we do not believe that our crosslinking results are consistent with that model, given the apparent molecular weight on SDS-PAGE (Fig. 5) and the lack of Adx-Adx cross-linked peptides in the P450-Adx crosslink band (Supplementary Material Table S3).

The cross-links identified cannot be well explained by any single model of a P450 27C1-Adx complex. The presence of multiple P450-Adx protein conformations has been proposed before [57]. The first model illustrated involves interactions between Asp-77 and Glu-78 of Adx and Lys-117 of P450 27C1 (Table 3, Fig. 6, Fig. 8C). This complex has Adx positioned similarly to the previously published P450-Adx fusion structures (Supplementary Material Fig. S10). Typically, the interaction with Adx at this position involves two lysine residues of the K-helix. In P450 27C1, these two residues correspond to Arg-338 and Lys-342. Lys-342 was not found in any cross-linked peptides, which was surprising given its conserved role in Adx binding (Fig. 9). The distance between the [2Fe-2S] cluster and the heme in this model is 26.1 Å, further than in the two P450-Adx fusion structures (17.4 Å for 11A1, 17.8 Å for 11B2) [5,6], but still in line with distance constraints for productive electron transfer [58]. It is possible that the model we have generated from docking is representative of an initial recognition structure driven by electrostatic interactions, and that the Adx could reposition to shorten the distance between the [2Fe-2S] cluster and the heme as previously proposed by Strushkevich et al. [5]. We believe that could be the case, given that structural model generated from 11B2-Adx crosslink data also resulted in a longer [2Fe-2S] cluster to heme distance (24.4 Å) than what was later observed in the 11B2-Adx fusion protein structure (17.8 Å) [6,9]. Alternatively, if this is the structure associated with electron transfer, the increased distance may contribute to the lack of spectral perturbation of P450 27C1 by Adx, as observed with P450 11A1 [14,33] and the slower rate of reduction of P450 27C1 (0.35 s⁻¹) in comparison to 11A1 (2.0 s⁻¹) [33,59].

Fig. 9.

Comparison of mitochondrial P450 sequences. Sequences for the human mitochondrial P450s from UniProt [60] were aligned using the Clustal Omega program [61]. Gaps in alignments are shown as dashes (−) and portions of sequences are omitted by ellipses (…). Relevant numbering of the recombinant P450 27C1 sequence is shown below. The P450 27C1 residues identified in cross-links are bolded and highlighted in yellow. The corresponding aligned residues from other mitochondrial P450s are bolded. Residues that have been implicated in interactions with Adx in other P450s are bolded and shown in red [6,9,27,57,62–65]. Note that some of the cited work utilized species homologs of the P450s.

Many other model clusters are focused away from the traditional Adx binding site. The most favorable scoring model of this type of interaction involves Asp-77 and Asp-84 of Adx interacting with Lys-478 and Lys-318 of P450 27C1, respectively (Table 3, Fig. 7, Fig. 8D). Interactions with residues in these regions have not been identified with other mitochondrial P450s (Fig. 9). P450 27C1 is the only mitochondrial P450 with multiple positively charged residues here, and this may allow for P450 27C1 to bind Adx at a unique interface. Like P450 27A1, the K_d value for binding between P450 27C1 and Adx is very low. Pikuleva et al. demonstrated that this was due to Arg-418 of P450 27A1 also being involved with Adx binding [57]. It was proposed that there may be additional binding sites for Adx outside of the common site for P450s with high affinity for Adx. Prior to identification of P450 27C1, this residue was thought to be specific to P450 27A1 as it is not present other mitochondrial P450s, including P450 family member 27B1 [27]. P450 27C1 also has this arginine residue (Arg-402). Lys-403 is the subsequent amino acid in P450 27C1 and it was detected in cross-links with Adx (Table 3, Supplementary Material Table S2). Thus, as in P450 27A1, Arg-402 of P450 27C1 and the region surround it may serve as an additional binding site for Adx. Given the distance between the [2Fe-2S] cluster and the heme with these other binding sites, they are not likely involved in electron transfer. The functional relevance of these other binding sites is unknown.

For the mechanism of electron transfer, our data is most supportive of the shuttle mechanism [8,10]. The lack of Adx-Adx crosslinks suggests that an Adx dimer does not interact with P450 27C1, which would be required for the proposed quaternary structure [15]. Additionally, the interface of Adx that interacts with P450 27C1 is the same as what interacts with AdR [66], preventing the possibility of Adx to interact with both AdR and the P450 at the same time. One potential area of concern with the shuttle mechanism and the measured K_d values is that Adx appears to bind preferentially to the P450 over AdR (Table 1), and that binding of oxidized Adx to P450 27C1 may prevent the binding of reduced Adx. These studies utilized oxidized Adx, so binding to AdR and reduction is necessary before productive electron transfer to the P450. It is possible that reduced Adx would have a different binding affinity with P450 27C1, allowing it to dissociate any bound oxidized Adx. Adx structural changes upon reduction have been observed [11], so this is not outside of the realm of possibility. As far as preferential binding of Adx to the P450 over AdR, the respective concentrations of AdR and P450 27C1 within the system are also important to consider. P450 27C1 is expressed in the skin and has been quantified [33], but the concentrations of Adx and AdR within the skin are currently unknown to our knowledge. The difference in binding affinity may be important to facilitate P450 27C1 reduction if AdR is present at much higher concentrations than P450 27C1.

Given the presence of multiple Adx binding sites on P450 27C1, some of which do not seem conducive for electron transfer, Adx binding to P450 27C1 may play an allosteric role in the activity of the enzyme. A potential allosteric role of Adx on P450 substrate binding has been proposed for other mitochondrial P450s [6,30,31]. The potential for Adx binding to stimulate P450 27C1 substrate binding was assessed by equilibrium titrations (Fig. 4, Table 2), but no differences in equilibrium substrate binding were observed. Given the low K_d for all-trans retinol binding to P450 27C1, it may be possible that there are minor changes in binding affinity that are not detected due to the concentration of P450 necessary for these measurements. The physiological relevance of assessing these changes may be limited though, as retinoid metabolizers like P450 27C1 are thought to receive their substrates from cellular retinoid-binding proteins in vivo, and we have recently provided in vitro evidence for such interactions with these binding proteins [38]. There was also no difference in Adx binding to P450 27C1 when substrate was present (Supplementary Material Fig. S6). This work focused exclusively on effects to P450 27C1 substrate binding to assess allosteric effects of Adx binding. There are other potential allosteric effects of Adx binding that have been described that were not assessed in this study (e.g., substrate positioning within the active site), so Adx allosteric regulation with P450 27C1 cannot be completely ruled out.

This work provides a basis for understanding the interaction between P450 27C1 and Adx, although much remains unknown. Cross-linking mass spectrometry has some limitations, and complementary methods would help refine models for the P450 27C1-Adx interaction. The protein-protein interaction modeling was also performed with the P450 27C1 AlphaFold structure. Modeling the interaction could be improved with the availability of a P450 27C1 crystal structure. The validation and assessment of the respective contribution of each P450 27C1 residue/interface to the interaction with Adx could be addressed through site-directed mutagenesis. Utilizing EDC cross-linking to identify interactions between P450 27C1 and Adx limited our study to focus on interactions between primary amines and carboxyl groups, and residues may interact that were not detected in our study. Additionally, other components may alter the interaction between P450 27C1 and Adx. For example, phospholipids can sometimes alter P450-redox partner protein binding affinity [4], though we have previously shown that this P450 27C1 construct does not require lipids for catalysis [33]. In this work, we focused on characterizing interactions with oxidized, non-modified Adx. It is possible that reduced Adx may have a different affinity for P450 27C1 or interact with it differently, in light of conformational changes that can occur when Adx is reduced [67]. There are also reports of Adx post-translational modifications affecting binding with P450s [68].

6. Conclusions

The goal of this work was to characterize the P450 27C1-Adx complex and to investigate potential allosteric effects of Adx binding to P450 27C1. Adx binds to P450 27C1 with very high affinity even at higher ionic strengths. This tight binding may be due to the increased number of contacts formed between P450 27C1 and Adx identified by cross-linking mass spectrometry. No significant Adx allosteric effect on P450 27C1 all-trans retinol substrate binding was observed. Overall, the interaction between P450 27C1 and Adx appears to be different than what has been observed with other mitochondrial P450s, supporting the conclusion that interactions between P450s and Adx are P450-specific. Differences in P450-Adx binding may serve an important function in partitioning Adx-mediated electron transfer among multiple P450s and in specific modulation of P450 activity.

Highlights:

• Alexa Fluor 488-labeling of adrenodoxin was utilized for complex detection.

• Adrenodoxin binds to P450 27C1 with low nanomolar affinity.

• P450 27C1-Adx interactions were identified by cross-linking mass spectrometry.

Abbreviations:

5-IAF

5-iodoacetamidofluorescein

acrylodan

acrylodan-2-dimethylaminonapthalene

Adx

adrenodoxin

AdR

NADPH-adrenodoxin reductase

b₅

cytochrome b₅

DLPC

1,2-α-dilauroyl-sn-glycero-3-phosphocholine

DMSO

dimethylsulfoxide

EDC

1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride

IAEDANS

5-((((2-iodoacetyl)amino)ethyl)amino)naphthalene-1-sulfonic acid

MST

microscale thermophoresis

P450 or CYP

cytochrome P450

POR

NADPH-cytochrome P450 reductase

SPR

surface plasmon resonance

Sulfo-NHS

N-hydroxysulfosuccinimide.