Mitochondrial 3β-Hydroxysteroid Dehydrogenase Enzyme Activity Requires Reversible pH-dependent Conformational Change at the Intermembrane Space*

Background: Mitochondrial intermembrane proton gradient is essential for dual functionality of 3β-hydroxysteroid dehydrogenase 2 (3βHSD2) enzyme. Results: 3βHSD2 is more active in a partially unfolded conformation. Conclusion: Reversible conformation of 3βHSD2 is a mandatory requirement for the biological activity at the intermembrane space. Significance: Mitochondrial proton gradient activates 3βHSD2 in a molten globule conformation. The inner mitochondrial membrane protein 3β-hydroxysteroid dehydrogenase 2 (3βHSD2) synthesizes progesterone and androstenedione through its dehydrogenase and isomerase activities. This bifunctionality requires 3βHSD2 to undergo a conformational change. Given its proximity to the proton pump, we hypothesized that pH influences 3βHSD2 conformation and thus activity. Circular dichroism (CD) showed that between pH 7.4 and 4.5, 3βHSD2 retained its primarily α-helical character with a decrease in α-helical content at lower pH values, whereas the β-sheet content remained unchanged throughout. Titrating the pH back to 7.4 restored the original conformation within 25 min. Metabolic conversion assays indicated peak 3βHSD2 activity at pH 4.5 with ∼2-fold more progesterone synthesized at pH 4.5 than at pH 3.5 and 7.4. Increasing the 3βHSD2 concentration from 1 to 40 μg resulted in a 7-fold increase in progesterone at pH 4.5, but no change at pH 7.4. Incubation with guanidinum hydrochloride (GdmHCl) showed a three-step cooperative unfolding of 3βHSD2 from pH 7.4 to 4.5, possibly due to the native state unfolding to the intermediate ion core state. With further decreases in pH, increasing concentrations of GdmHCl led to rapid two-step unfolding that may represent complete loss of structure. Between pH 4 and 5, the two intermediate states appeared stable. Stopped-flow kinetics showed slower unfolding at around pH 4, where the protein is in a pseudostable state. Based on our data, we conclude that at pH 4–5, 3βHSD2 takes on a molten globule conformation that promotes the dual functionality of the enzyme.

The inner mitochondrial membrane protein 3␤-hydroxysteroid dehydrogenase 2 (3␤HSD2) synthesizes progesterone and androstenedione through its dehydrogenase and isomerase activities. This bifunctionality requires 3␤HSD2 to undergo a conformational change. Given its proximity to the proton pump, we hypothesized that pH influences 3␤HSD2 conformation and thus activity. Circular dichroism (CD) showed that between pH 7.4 and 4.5, 3␤HSD2 retained its primarily ␣-helical character with a decrease in ␣-helical content at lower pH values, whereas the ␤-sheet content remained unchanged throughout. Titrating the pH back to 7.4 restored the original conformation within 25 min. Metabolic conversion assays indicated peak 3␤HSD2 activity at pH 4.5 with ϳ2-fold more progesterone synthesized at pH 4.5 than at pH 3.5 and 7.4. Increasing the 3␤HSD2 concentration from 1 to 40 g resulted in a 7-fold increase in progesterone at pH 4.5, but no change at pH 7.4. Incubation with guanidinum hydrochloride (GdmHCl) showed a three-step cooperative unfolding of 3␤HSD2 from pH 7.4 to 4.5, possibly due to the native state unfolding to the intermediate ion core state. With further decreases in pH, increasing concentrations of GdmHCl led to rapid two-step unfolding that may represent complete loss of structure. Between pH 4 and 5, the two intermediate states appeared stable. Stopped-flow kinetics showed slower unfolding at around pH 4, where the protein is in a pseudostable state. Based on our data, we conclude that at pH 4 -5, 3␤HSD2 takes on a molten globule conformation that promotes the dual functionality of the enzyme.
Human 3␤-hydroxysteroid dehydrogenase (3␤HSD) 2 plays a central role in the synthesis of several steroid hormones and is thus essential for the survival of all species. Steroidogenic tissues do not store steroids, so synthesis occurs in the mitochondria on an as needed basis. Enzymes involved in the steroidogenic pathway are present in all steroidogenic tissues and some nonsteroidogenic tissues, including kidney and skin. 3␤HSD is encoded by two distinct genes and expressed in a tissue-specific pattern (1,2), with 3␤HSD2 expressed exclusively in the adrenal gland, ovary, and testes. 3␤HSD2, which lacks a heme group and requires NAD ϩ as a cofactor for activity, catalyzes the conversion of multiple steroids in the pathway: pregnenolone to progesterone, 17␣-hydroxypregnenolone to 17␣-hydroxyprogesterone, and dehydroepiandrosterone to androstenedione (4-androstene-3,17-dione). Synthesis of these steroids requires that 3␤HSD2 functions as both a dehydrogenase and isomerase, which it does as a single polypeptide.
The compartment-specific location of 3␤HSD2 in the mitochondria had remained elusive since the time of its discovery; however, we recently reported that the protein localizes to the inter membrane space (IMS) without insertion into the inner mitochondrial membrane (IMM) (3). Specifically, we demonstrated that 3␤HSD2 associates with several translocase complex proteins in the mitochondria. Translocase complexes, responsible for the import and sorting of mitochondrial proteins, can be found at either the outer mitochondrial membrane (OMM) or the IMM. We discovered that 3␤HSD2 associates closely with the IMM protein Tim50, allowing further and transient associations with Tim23, and Tom22 (3), an OMM protein whose C terminus faces the IMS (3). 3␤HSD2 has to adopt a flexible conformation to coordinate the continuous interaction with these proteins. Any change in the arrangement of the 3␤HSD2 stoichiometry with Tim50, Tim23, or Tom22 ablates steroidogenesis (3).
Most matrix proteins carry a positively charged N-terminal presequence that is proteolytically removed after import. After translocations through the TOM complex at the OMM, the preproteins are directed to the membrane potential (⌬⌿)driven presequence translocase of the IMM (4,5). Some inner membrane proteins carry a hydrophobic sorting sequence behind the cleavable matrix-targeting signal and thus arrested in the IMM (6). After passing through the TOM channel, precursors do not engage the presequence translocase, but use separate machinery for transport to and insertion into the IMM.
In vitro, most proteins spontaneously fold into a unique, thermodynamically stable, three-dimensional conformation. Although the physical-chemical principles of folding processes are not fully understood, theoretical studies have shown that the free energy landscapes for protein folding are most often biased like a funnel toward the structurally well defined native state (7)(8)(9)(10). This implies that the transition state structure within a given folding resembles the native state of the protein.
Mitochondrial specific proteins that undergo membrane insertion form "molten-disc" intermediates that have a partial secondary structure with the ␤-strands sitting flat on the membrane surface (11). The fact that we observed a correlation between 3␤HSD2 activity and association with components of the inner and outer mitochondrial translocases strongly suggests that 3␤HSD2 undergoes a conformational change during catalysis. Moreover, protein conformation and folding are often sensitive to pH. With decreasing pH, most native proteins first lose tertiary and then secondary structure, so that under strongly acidic conditions, they revert to an unstructured random coil. As an inner mitochondrial resident protein (12), 3␤HSD2 faces the proton gradient generated by the proton pump and thus may undergo the conformational change required for its function. Because 3␤HSD2 is not integrated into the IMM, it likely faces a continuous change in proton gradient, which may result in an equilibrium of transient conformational changes that can be described as a molten globule state. Using different biochemical and biophysical tools, here we show that the two different functional activities of 3␤HSD2 at the inner mitochondrial membrane is possible due to a partially open conformation and remains in equilibrium at two different states with a minor change in free energy (⌬G).

Expression, Purification, and Characterization of 3␤HSD2
Protein-3␤HSD2 was expressed as a recombinant protein in the baculovirus system (13,14). Briefly, recombinant baculovirus containing 3␤HSD2 cDNA was added to 1.5 ϫ 10 9 Sf9 cells (1 liter) at a multiplicity of infection of 10. To identify recombinant 3␤HSD2, proteins from Sf9 cells were separated by SDSpolyacrylamide (12%) gel electrophoresis, probed with our anti-3␤HSD2 polyclonal antibody, and detected using the West Pico Western blotting system (Pierce). The enzyme was purified from the 100,000 ϫ g pellet of Sf9 cells (13,14) in the presence of the low-critical micelle concentration (CMC) detergent, Igepal CO 720 (Rhodia, Inc., Cranbury, NJ), which is necessary to elute the enzyme from the DEAE-ion exchange column (13,14). Using hydroxyapatite chromatography, we replaced Igepal with Cymal-5 (Anatrace, Inc., Maumee, OH), an UV invisible high-CMC detergent. The 3␤HSD2 fraction pool from the DEAE column was applied to the hydroxyapatite column (1 mg of protein/ml of packed gel); washed with 3.5 column volumes of 0.025 M potassium phosphate, pH 7.5, 20% glycerol, 0.1 mM EDTA, 0.01 M NAD, 1.8 mM Cymal-5, and eluted with 0.30 M potassium phosphate, pH 7.5, 20% glycerol, 0.1 mM EDTA, 0.01 M NAD, 1.8 mM Cymal-5. The peak of 3␤HSD2 activity was pooled and found to be free of Igepal CO-720 based on absorbance at 280 nm. Protein concentrations were determined by the guanidinum hydrochloride denaturation method (15) to avoid errors in extinction coefficients, but the total mitochondrial protein was estimated by (Bio-Rad dye reagents, Hercules, CA) using bovine serum albumin as the standard. During enzyme purification, the isomerase activity of 3␤HSD2 was measured by the initial absorbance increase at 241 nm (due to 4-androstene-3,17-dione, androstenedione formation from the intermediate substrate, 5-androstene-3,17-dione) as a function of time. Blank assays (zero-enzyme, zero-substrate) assured that specific isomerase activity was measured as opposed to nonenzymatic, "spontaneous" isomerization (13). Changes in absorbance were measured with a Varian (Sugar Land, TX) Cary 300 recording spectrophotometer.
Anti-3␤HSD2 Antiserum-We raised antiserum against recombinant, wild-type 3␤HSD2 protein expressed in baculovirus. Purified 3␤HSD2 protein was dissolved in Freund's adjuvant and 0.2 mg was injected into each of 2 New Zealand White rabbits, followed by a booster injection 6 weeks later (Lampire Biologicals, Pipersville, PA). Titers were assessed by radioimmunoassay (RIA). One rabbit generated titers above 1:1,000; following a third injection of antigen, the rabbit yielded a highaffinity antiserum that was used at a 1:10,000 dilution.
Cell Culture, Isolation, and Purification of Mitochondria-Mouse Leydig (MA-10) cells were incubated in 15% horse serum, 5% fetal bovine serum, and 1 ϫ gentamycin at 37°C in a humidified incubator under 5% CO 2 . Cells were removed from tissue culture plates by gentle scrapping using PBS. Cells were incubated in the hypotonic buffer 10 mM HEPES at pH 7.4 for 40 min prior to Dounce homogenization with 10 gentle up and down strokes. Cell debris was removed by spinning at 3,500 ϫ g for 10 min. The supernatant containing the mitochondrial fraction was spun down at 10,000 ϫ g for 10 min, and then the pellet was washed with the mitochondria isolation buffer. The washing step was repeated twice and the pellet was resuspended in energy regeneration buffer (125 mM sucrose, 80 mM KCl, 5 mM MgCl 2 , 10 mM NaH 2 PO 4 , 10 mM sodium succinate, 1.0 mM ATP, 1.0 mM NADP, 0.1 mM ADP, 30 mM creatine phosphate, 1 mM creatine kinase and 25 mM HEPES, pH 7.4).
Biological Activity of Purified Proteins-Mitochondria were isolated from steroidogenic MA-10 cells following our established procedure as described (16,17). For the metabolic conversion assays, mitochondria (300 mg) were incubated in potassium phosphate buffer along with 3 million cpm of [ 3 H]pregnenolone and then chased with 30 g of cold progesterone. Metabolic activity was initiated by the addition of 0.01 M NAD and allowed to proceed for 4 h at 37°C in a shaking water bath. To determine the effect of electrochemical gradient we incubated the mitochondria isolated from steroidogenic MA-10 cells with valinomycin (2 g/ml), oligomycin (50 g/ml), and carboxyatractyloside (50 g/ml) for 3 h. To evaluate the effect of pH-mediated partial unfolding on metabolic conversion, we incubated 3␤HSD2 at a pH of 3.0, 3.5, 4.5, 5.5, and 7.4 for 1 h at room temperature prior to adding the mitochondrial incubation mixture. Similarly, we induced 3␤HSD2 unfolding by incubating the protein for 30 min at room temperature with various concentrations of guanidinum hydrochloride (GdmHCl): 10 mM, 20 mM, 50 mM, 100 mM, 200 mM, 500 mM, 1 M, and 2 M. At 200 mM, GdmHCl induces partial unfolding, whereas at 1.75 M, unfolding reaches a plateau; thus, we assumed that 200 and 500 mM GdmHCl would induce an intermediate state of unfolding. After addition of all of the assay components, the reaction was incubated for 3 h at 37°C in a shaking water bath. Steroids were then extracted in ether/acetone, 9:1. Signal intensity was determined by a phosphorimager. We characterized all the steroids by gas chromatography/mass spectrometry.
Gas Chromatography-Mass Spectrometry-The spots extracted from TLC plates were subjected to GC-MS analysis on an Agilent 7890 GC with 5975C mass spectrometer. The column used was an Agilent HP-5 with dimensions 30 m ϫ 0.25-mm inner diameter with a 0.25-m film thickness. Samples were dissolved in 50 l of dichloromethane and 1 l was injected onto the column using a pulsed splitless injection. Helium was used as the carrier gas at a flow rate 1 ml/min. The temperature program is as follows: ramp at 10°C/min from 70 to 310°C and hold for 6 min. Spectra were collected in full scan mode with 70 Ev ionization over the mass range m/z 50 to 500 to facilitate comparison of MS spectra with the NIST/EPA/NIH NIST08 mass spectral library.
Circular Dichroism (CD)-CD experiments in the far-UV region (195-250 nm) were carried out using a 1.0-mm pathlength quartz cuvette at 20°C in a Jasco J-815 spectropolarimeter equipped with a Peltier temperature-controlled cell holder and a four-syringe stopped flow (New Biologic System). The instrument was purged with a continuous flow of nitrogen of 20 liter/min to reduce the maximum signal to noise ratio. The CD spectra obtained in the near-UV (250 -375 nm) were recorded in a 1-cm path length cuvette with a protein concentration of 2.0 mg/ml. Spectra obtained in the far-UV are presented without mathematical smoothing. Purified wild-type 3␤HSD2 was equilibrated in NaH 2 PO 4 buffers ranging in pH from 2.5 to 7.4, and the CD spectra were recorded as a function of pH. The mean residue molar ellipticity [⍜] at 208 and 222 nm was plotted versus pH. Secondary structural analysis was carried out using the CD-Pro (18 -21) to determine the relative proportions of ␣-helix and ␤-sheet as a function of pH. In a separate set of experiments, purified wild-type 3␤HSD2 was equilibrated in 10 mM sodium phosphate at pH 3.0, 3.5, 4.0, 4.5, and 7.5, prior to any experiments, and aggregation was removed by centrifugation at 100,000 ϫ g for 1 h and recording the optical density before and after centrifugation. Temperature-dependent unfolding was measured at 222 nm where the temperature was increased from 4 to 80°C. For pH-mediated or GdmHCl-mediated unfolding, the appropriate buffer blank was subtracted from each spectrum, and ⍜ at 222 nm was plotted with respect to the pH of solution or GdmHCl concentration.
Measurement of ANS-3␤HSD2 Binding-Steady state fluorescence experiments were performed using a Hitachi (Hitachi, Model F-7000) spectrofluorometer using a 10-mm path length quartz cuvette. Fluorescence spectra for 1-anilinonaphthalene-8-sulfonic acid (ANS) were recorded at an excitation of 357 nm with 5-nm band pass slits for both excitation and emission channels. All experimental values were corrected for those obtained for buffer 10 mM NaH 2 PO 4 at pH values of 3.0, 4.5, 5.5, and 7.5, and free ANS run under identical conditions. The sample temperature was maintained at 22°C for all experiments unless otherwise indicated. Stock solutions of 400 M ANS in Me 2 SO were diluted into water, and the concentration was determined using a molar absorption coefficient (⑀) of 4990 M Ϫ1 cm Ϫ1 (22). Protein-ANS dissociation constants were determined by nonlinear curve fitting of the binding isotherm based on the following equilibrium: L ϩ P ϫ L Ϫ P, where L and P are ANS and protein, respectively. ANS binding experiments were performed using 14 g of protein, to which increasing amounts of ANS were added, and the emission intensity was measured at 470 nm to monitor protein-bound ANS in 15 different data points. The data points obtained from fluorometric titrations were analyzed by fitting to Equations 1 and 2. A total of 15 data points were recorded with a fixed increase in ANS concentration of 1 M while the maintaining protein concentrations remained unchanged. The dissociation constants of ANS binding to 3␤HSD2 (K d ) were determined by nonlinear curve fitting analysis using Equation 2 (22). All of the experimental points for the binding isotherm were fitted by least square analysis using a Microcal Origin software package (version 6.0, Microcal Software Inc., Northampton, MA). The binding stoichiometries of the ANS-protein complex were estimated from the intercept of two straight lines of the nonlinear fitted plot of ⌬F/⌬F max against the mole fraction of ANS.
In both equations, ⌬F is the change in fluorescence emission intensity at 470 nm ( ex ϭ 360 nm) for each point on the titration curve, and ⌬F max change in fluorescence emission intensity when the protein is completely bound to ANS. C p is the concentration of the ANS, and C 0 is the initial concentration of protein. The double reciprocal plot was used to determine ⌬F max following Equation 3, and determined the apparent binding constant (K app ) (23).
The linear double reciprocal plot of 1/⌬F against 1/(C p Ϫ C 0 ) is extrapolated to the ordinate to obtain the value of ⌬F max from the intercept. The apparent binding constant, K app , is the inverse of the dissociation constant (K d ). The approach is based on the change in emission intensity of ANS is proportional to the concentration of the protein. C p and C 0 represents protein (14 g) and ANS concentrations. Thus, the apparent binding rates reflect the ANS effect at the molten globule core as described from the changes with the incremental addition of ANS.
Stabilization of 3␤HSD2 with Specific Detergent-Igepal CO 720, a polyoxyethylene phenyl ether nonionic detergent that has phenyl groups, eluted the enzyme from a DEAE ion exchange column with a very sharp peak during purification. To determine whether the Igepal CO 720 detergent can promote stabilization, we added different concentrations of this detergent to a fixed concentration of 3␤HSD2 and then determined the stoichiometry of binding by recording the equilibrium CD at 190 -250 nm. The detergent itself has strong absorption at 205 nm, and thus the signal to noise ratio is higher in this region. To circumvent this problem, we determined binding at 208 and 222 nm.
Stopped-flow Kinetics-The kinetics of protein unfolding were measured in a JASCO-815 spectropolarimeter interfaced with an SF4 stopped-flow module (Biologic, France). A stock concentration of 3␤HSD2 (2.0 mg/ml) in the syringe was changed to 0.1 mg/ml in the cell. For acid-mediated unfolding, 3␤HSD2 (25 l) and varying amounts of buffer at different pH values were mixed with a "Delta" mixer so that the protein concentration remained unchanged, but the pH decreased to 3.0, 3.5, 4.0, 5.5, and 7.4. For acid gradient unfolding, varying amounts of 10.0 mM NaH 2 PO 4 , pH 3.0, in a second syringe and water in the third syringe were mixed in such a fashion that the pH of the solution changed without significant interruption of the ionic concentration of the protein. The syringes were operated 10.0 ml/s with a 6.8-ms dead time. Fluorescence was measured by excitation at 280 nm with emission intensity using a 320-nm cutoff filter. Data were collected in four segments. First, 5 data points were collected to monitor the basal signal of the solution. Second, the solution was mixed for 60 ms. Third, 100 data points were collected over 100 ms to monitor initial rapid kinetics, then 200 data points were acquired over a 2-s time period to monitor slow kinetics (24).

3␤HSD2 Protein Does Not
Aggregate at Higher Concentrations-To characterize 3␤HSD2, we expressed the protein in the baculovirus system and then purified it from Sf9 cells in the presence of the low-CMC detergent, Igepal CO 720, using a DEAE ion exchange column. Coomassie Blue staining of fractions separated by SDS-PAGE (12%) revealed a single band (42.0 kDa) that co-migrated with the crude 3␤HSD2 extract (Fig. 1A). To facilitate analysis of protein folding, the Igepal detergent was replaced with the high-CMC detergent Cymal-5 using hydroxyapatite column chromatography. Based on enzymatic assay, we determined that the peak binding capacity occurred in fractions 20 -26, indicating active 3␤HSD2 (Fig.  1B). To confirm the mitochondrial localization, we stained fractionated mitochondria from steroidogenic MA-10 cells with our human 3␤HSD2 antibody. 3␤HSD2 was present only in the IMM fraction and was resistant to proteolysis. Probing the same membrane with antibodies directed against OMMassociated voltage-dependent anion channel 2 (VDAC2) and IMM-associated Tim50 confirmed the association of 3␤HSD2 with the IMM (Fig. 1C).
Many proteins change their conformation and/or form aggregates at high concentrations. The signal intensities of CD spectra obtained in far-UV for this preparation of 3␤HSD2 at pH 7.5 depended solely on concentrations over a 50-fold range, from 2 to 100 g (Fig. 1D). The CD data gave a linear correlation between concentration and ⍜ at both the -* transition at 208 nm and the n-* transition at 222 nm (Fig. 1D). Thus, the bio- logically active 3␤HSD2 protein employed in this study did not aggregate or change its conformation at this pH. Computational analysis with increasing 3␤HSD2 concentrations showed ( Fig. 1E) a minor increase in ␣-helical content at or above 40 g. At the same time, ␤-sheet content decreased from 15 to 10% when the protein concentration increased from 40 to 60 g, and then remained unchanged with further increases in concentration. The observation that the protein did not aggregate at 100 g supports the notion that, due to the intramolecular interaction, the change in ␤-sheet content occurs at the expense of maintaining the entropy of the protein.
Effect of Electrochemical Gradient on 3␤HSD2 Activity-To investigate the role of electrochemical gradient at the IMS, we have determined activity with the mitochondria isolated from the MA-10 cells incubated with various electrochemical gradient inhibitors: oligomycin inhibits the H ϩ -translocating ATP; valinomycin collapses electrochemical potential across the inner membrane; and carboxyatractyloside inhibits access of ATP to the matrix (25)(26)(27). Oligomycin and valinomycin, but not carboxyatractyloside, inhibited conversion from pregnenolone to progesterone ( Fig. 2A). Because of the inhibition of proton pump at the IMS no metabolic conversion is possible. A quantitative estimation of the intensity of bands (Fig. 2B), as measured by the PhosphorImager, shows that incubation with carboxyatractyloside retained 45% activity, whereas incubation with valinomycin and oligomycin retained ϳ22%. As a control we also performed Western blot analysis with different antibodies on the mitochondria after incubation with the different inhibitors. For OMM-associated expression, we probed with the VDAC2 antibody and matrix proteins with the COX IV antibody. The Western blot shows (Fig. 2C) that the expression of VDAC2 and COX IV is unchanged suggesting that the addition of these inhibitors did not reduce mitochondrial architecture.
Effect of pH on 3␤HSD2 Conformation-CD spectroscopy can distinguish secondary structural components by the presence of minima near 198 nm for random coil, at 208 and 222 nm for the ␣-helices, and 218 nm for ␤-sheets (28,29) and so changes in the CD spectra were used to monitor pH-induced unfolding of 3␤HSD2 (Fig. 3A). We found that the CD spectrum generally remained unchanged from pH 7.4 down to pH 4.5, indicating that the protein retained its predominately ␣-helical character. However, at pH 4.5, there was a minor reduction in ellipticity at the 208 nm -* position. With a further decreases in pH, the ellipticity remained almost unchanged at the 222 nm n-p* position, but changed significantly at the 208 nm -* position. At pH 3.0 or 2.5, the protein conformation changed to a random coil. Secondary structural analysis (18 -21) clearly shows the effect of pH on 3␤HSD2 conformation (Fig. 3B): decreasing the pH to 3.25 resulted in a decrease in the ␣-helical conformation without a significant change in the ␤-sheet conformation. When the pH was titrated back from pH 3.25 to 7.4, the conformation was restored (Fig.  3C), suggesting that conformational changes were reversible. However, titrating back the pH from 3.0 to 7.4 did not restore the initial conformation, and thus the complete denaturation to random coil may have destroyed protein flexibility. Based on these observations, pH 3.25 appears to represent the threshold for 3␤HSD2 to maintain a reversible, flexible conformation (Fig. 3D). This range in pH resembles the mitochondrial proton gradient, which supports the possibility that mitochondrial resident 3␤HSD2 undergoes conformational changes associated with the proton gradient.
Effect of pH on Progesterone Synthesis-To evaluate if the change in pH alters 3␤HSD2 activity, we measured the conversion of pregnenolone to progesterone. 3␤HSD2 protein (20 g) was equilibrated with the appropriate pH buffer and then incubated with isolated mitochondria from steroidogenic MA-10 cells. We normalized progesterone levels to the amount produced in the absence of the cofactor, NAD (0.01 M). At pH 4.5, 3␤HSD2 synthesized 25% progesterone, the highest level, whereas at pH 7.4 3␤HSD2 only synthesized 11%, similar to the amount at pH 3.5 (Fig. 4A). This experiment clearly indicates that pH 4.5 supports the maximum level of 3␤HSD2 activity.
We next addressed whether the 3␤HSD2 protein concentration affected the level of metabolic activity seen at the different pH values. Fig. 4B demonstrates that 3␤HSD2 at pH 7.4 increased progesterone synthesis with an increase in protein concentration. However, 3␤HSD2 activity at other pH values exhibited varying degrees of sensitivity to protein concentration. We observed that 3␤HSD2 (40 g) produced 35% relative progesterone at pH 4.5, 28% at pH 5.5, and 10% at pH 3.5. At a concentration of 100 g of 3␤HSD2, the relative amount of progesterone produced decreased slightly at pH 5.5, remained similar at pH 7.4, and substantially increased at pH 4.5. These observations corroborate our previous finding that pH 4.5 supported the most active state of the protein. At the other pH values, more than 40 g of 3␤HSD2 had no significant effect on the catalytic conversion, suggesting that the large amount of protein may have blocked the mitochondrial metabolic transport by partially blocking the import channel and thus become ineffective (Fig. 4, B and C). This did not appear to be the case for  Rather, at this pH, the limiting factor appeared to be the catalyst.
We next measured the kinetics of progesterone synthesis at either pH 7.4 or 4.5 using a fixed amount of 3␤HSD2 (40 g). Synthesis at either pH started at about 30 min, but the amount of progesterone produced at pH 4.5 was 2-fold more than that at pH 7.4 (Fig. 4, D and E). Within 2 h, synthesis reached its saturation point at pH 7.4. At pH 4.5, synthesis at 1 h yielded 3-fold more progesterone than at pH 7.4. The synthesis at pH 4.5 did not reach saturation until 4 h, and the amount produced was more than 6-fold its initial synthesis (Fig. 4, D and E). The radioactive pregnenolone and its metabolite(s) (Fig. 4, B and D) were characterized by GC-MS. The samples scraped from the TLC plate were dissolved in dichloromethane and analyzed on the GC-MS instrument (supplemental Fig. S1, A-D). Supplemental Fig. S1A shows the extracted ion chromatogram for the molecular ion of pregnenolone at m/z 316.2. Supplemental Fig.  S1B shows the full scan mass spectrum at a retention time (RT) of 21.02 min. This matched the NIST08 library and identified pregnenolone with a quality factor of 99/100. Supplemental Fig.  S1C shows the extracted ion chromatogram for the molecular ion of progesterone at m/z 314.2 and supplemental Fig. S1D shows the full scan mass spectrum at a RT of 21.90 min. This spectrum matched the NIST08 library with a match for progesterone with a quality factor of 96/100. The spectral match to the NIST08 library confirms that spots from the TLC plate contained the expected steroid. In all, these observations provide further support that the unfolded state of 3␤HSD2 has higher activity. To further confirm this, we performed a metabolic conversion assay at pH 7.4 in the presence of different concentrations of the chaotropic agent, GdmHCl (Fig. 5A) in the presence of mitochondria isolated from MA-10 cells. Concentrated GdmHCl disrupts the higher structure of water, decreasing hydrophobic effects and promoting protein unfolding and dissociation. Addition of 10 mM GdmHCl increased progesterone synthesis, and at 200 mM GdmHCl, relative progesterone levels were 4-fold greater than levels measured in the absence of GdmHCl (Fig. 5A). Peak synthesis of 7-fold occurred with 200 mM GdmHCl and then subsequently declined until there was a complete loss of activity at 2 M GdmHCl (Fig. 5B), likely due to the denaturation of 3␤HSD2. These results confirm that 3␤HSD2 exhibits peak activity when in a partially open conformation (Fig. 5C).
Cooperative Denaturation-To better understand why partial unfolding of 3␤HSD2 led to increased activity, we evaluated the effects of acidic and GdnHCl cooperative denaturation on the protein. Fig. 6A shows GdnHCl denaturation curves obtained at pH 3.5, 4.0, 4.5, 5.5, and 7.4. As compared with the melting curve from proteins at pH 7.4, all the curves exhibited a decrease in ellipiticity. This decrease may represent the protein transitioning from the native folded state to the intermediate ion core state due to the presence of long hydrophobic amino acids at the N terminus. In addition, pH 3.5 may represent the threshold pH before complete denaturation (Fig. 6A). The melting curves at pH 7.4, 5.5, and 4.5 show that the protein at those pH values underwent a three-step transition, suggesting that a small decrease in pH could not open the hydrophobic core of 3␤HSD2. Based on the melting curves, we calculated ⌬G and plotted it as a function of GdmHCl concentration (Fig. 6B). For 3␤HSD2 equilibrated to pH 7.4, addition of GdmHCl at concentrations greater than 0.2 M resulted in ⌬G decreasing from 1.5 to Ϫ5.4 kcal/mol. After the initial opening, the decrease in ⌬G Ϫ5.4 kcal/mol may be the result of the participation of very loose amino acids from the peripheral side of the domains involved in relatively tighter folding. Increasing con-centrations of GdmHCl, from 0.2 to 1.75 M, caused the protein to undergo a very slow, cooperative transition. The second cooperative change, beginning at 1.75 M GdmHCl, was faster than the first change. At pH 7.4, the unfolding was completed at 3.5 M GdmHCl. Thus, the protein needs to undergo a very slow, smooth cooperative unfolding for pH 7.4 and 5.5 prior to its complete denaturation. The three-step transition also exists for pH 4.5, but it is subtle because of the already attained partially

. Metabolic conversion of [ 3 H]pregnenolone to [ 3 H]progesterone after incubation of 3␤HSD2 pre-equilibrated at different pH values.
After pre-equilibration, 3␤HSD2 was further incubated with the isolated mitochondria (Mito) from MA-10 cells, [ 3 H]pregnenolone and NAD for an additional 3 h (unless otherwise indicated) and then metabolites were separated by thin-layer chromatography. Panel A, pregnenolone to progesterone conversion where 3␤HSD2 was previously equilibrated at the indicated pH values. The amount of conversion was measured by PhosphorImager, normalized to levels synthesized in the absence of NAD, and plotted against pH. The activity was maximum at pH 4.5. Panel B, thin layer chromatogram showing the measurement of progesterone conversion with different concentrations of 3␤HSD2 pre-equilibrated at the indicated pH. The activity was maximum at pH 4.5 and saturation was not evident at that pH. Panel C, quantitation of radioactive conversion observed in panel B. Panel D, thin layer chromatogram showing the conversion of (pregnenolone to progesterone) kinetics at pH 7.4 and 4.5. The result shows that synthesis began at 30 min for proteins equilibrated to either pH. Over time, 3␤HSD2 pre-equilibrated to pH 4.5 synthesized more progesterone than protein at pH 7.4: at 1 h 3␤HSD2 at pH 4.5 synthesized 2-fold more and at 3 h the amount increased to 7-fold more. Saturation was reached at 8 h for the 3␤HSD2 at pH 4.5. Panel E, quantitative analysis of the progesterone conversion from panel D. Data presented in panels A, C, and E is the mean Ϯ S.E. of three independent experiments.  open conformation at the lower pH. The change in ⌬G from 1.5 to Ϫ5.4 kcal suggests that the protein did not have tight shielding, from pH 7.4 to 4.5, prior to the start of the cooperative unfolding. At lower pH values, increasing GdmHCl concentrations resulted in a two-step, rapid unfolding, possibly due to the complete loss of structure.
These curves fit a multiple state model (Fig. 6D) in which the free energy differences between the native (N) and unfolded (U) conformations depend linearly on the concentration of the denaturant; thus ⌬G ϭ ⌬G H2O Ϫ m(GdmHCl), where ⌬G is the apparent free energy difference at any GdmHCl concentration, ⌬G H2O is the free energy difference in the absence of denaturant, and m is a parameter describing the cooperativity of the transition. Hence, at pH values between 3.25 and 4.5, ⌬G H2O ϭ 1.95 kcal/mol and m ϭ 0.9 kcal and higher values from pH 4.5 to 7.4. ⌬G changed from 1.67 to Ϫ5.4 kcal/mol, but the nature of the two-step cooperative denaturation remained unchanged (Fig. 6B). A plot of ellipticity with pH shows retention of increased structures with an increase in pH (Fig. 6C). The loss of the ellipticity may be due to the putative loosely folded conformation of the outer core of the protein. The conclusion that the inner core remained intact indicates that the tightly folded inner core is shielded by loosely folded domains of the inner core. Thus, loss of the outer sphere of the protein at pH 4.5 may allow interactions between the unfolded inner core and mitochondrial membrane to occur, promoting activity.
Unfolding States Analysis by ANS-To further confirm the notion that 3␤HSD2 at a lower pH (pH 4.5) is partially unfolded and thus exhibits higher activity than when at its native conformation, we performed ANS binding with 3␤HSD2 at different pH values (Fig. 7, A-I). ANS binds to hydrophobic regions of partially unfolded proteins exposed to solvents that induce a conformational change in the protein structure (30). In particular, ANS binds strongly to proteins in a molten globule structure (31). Binding of ANS to 3␤HSD2 at pH values ranging from 4.5 to 7.4 (Fig. 7, A-C) suggests more binding at pH 4.5 and 5.5. Upon excitation at 357 nm and in the absence of protein, ANS shows a very low quantum yield emission intensity maxima at ϳ525. However, binding of ANS in the acidic region led to an increase in the emission intensity as well as a concomitant shift in the emission maximum (Fig. 7, B and C), indicating the presence of exposed hydrophobic groups. The increasing concentration of ANS to a fixed concentration of 3␤HSD2 (14 g) increased fluorescence intensity. The difference between Fig. 7, A and B, shows the complete shift of the equilibrium as the isosbestic point for pH 7.4 (Fig. 7A) to 5.5 (Fig. 7B) or 4.5 (Fig.  7C) is changed from 460 to 490 nm because of the stronger binding of ANS. The binding efficiency of ANS at pH 4.5 was at least 2-fold more than at pH 7.4 (Fig. 7D), suggesting that protein at an acidic pH has a more open or more unfolded conformation, resulting in greater binding. This greater binding was possibly due to the loosening of the tertiary interactions that helps in exposing the hydrophobic groups. To further confirm the flexible conformation, we titrated ANS with 3␤HSD2 and estimated binding by determining the difference in fluorescence at each concentration of ANS with respect to the fluorescence maxima (⌬F/⌬F max ). Fig. 7E shows that the ANS binding to 3␤HSD2 was greater at pH 4.5 and 5.5 than at pH 7.4. Next, we determined the apparent binding constant for ANS binding to 3␤HSD2 at different pH values (Fig. 7F). The apparent binding constant (K app ) at pH 4.5 and 5.5 was 4.4 and 4.1 ϫ 10 5 M Ϫ1 , respectively. However, at pH 7.4 and 3.0, K app was 2.9 and 1.8 ϫ 10 5 M Ϫ1 , respectively. The higher apparent binding constant at acidic pH indicates a stronger interaction than what was seen at pH 7.4. Further analysis of the stoichiometry of 3␤HSD2 binding to ANS is shown by a Job's plot (Fig. 7, G-I). Fig. 7, G-I, shows that the protein to ANS binding ratio at pH 4.5 was 1:1.3, (Fig. 7I), and at pH 5.5 this ratio was 1:1.1 (Fig. 7H). The intersection point of the two limiting lines of a Job's plot will yield the binding stoichiometry of a ligand (32). For example, if the intersection point occurs at a mole fraction of 0.5, this would indicate 1:1 binding stoichiometry. However, at pH 7.4, this ratio was 1:0.7 (Fig. 7G), indicating greater binding of ANS at an acidic pH and supporting our notion that 3␤HSD2 at pH 4.5 is in a partially unfolded conformation.
Stability by Thermal Denaturation-To better understand the process of 3␤HSD2 unfolding, we calculated the melting temperature (T m ) as a function of pH (Fig. 8). The protein at pH 7.4 started to unfold at 30°C and was completely unfolded at 70°C. For this pH, we calculated the T m as 42°C (Fig. 8B). Not surprisingly, reducing the pH to 5.5, which would contribute to unfolding, resulted in a T m of 37°C. A plot of T m values versus pH shows that proteins at pH 5.0 to 4.0 had T m values in close range, but with a steady decrease, indicating continuous unfolding (Fig. 8C). However, proteins in the pH range of 4.25 to 3.25 had almost identical T m values; this plateau may indicate a pseudo stable conformation in which most of the secondary structure is preserved due to slower unfolding. Slower unfolding may serve as a way for the protein to maintain equilibrium with the native state. We have summarized the results in the form of a schematic (Fig. 8D). The protein is in an unfolded (U) state at pH 3.0 and remains in a pseudo stable (I 1I ) conformation from pH 3.5 to 4.0. From pH 4.0 to 5.0, the protein arranges itself to the most open state conformation (I 1 ), prior to reaching its most stable native (N) state. Thus, N 3 I 1 N I II 3 U represents the change in 3␤HSD2 conformation from the native state to the completely unfolded state, where the interchange between the I 1 and I II states remain reversible. At pH 7.4 to 6, 3␤HSD2 was at its most stable, native (N) state. Decreasing the pH induced a slow unfolding and between pH 5.0 and pH 4.0, the protein adapts an open conformation (I 1 ) that has the highest enzymatic activity. A further decease in pH, from 4.0 to 3.5, results in a pseudo stable conformation (I II ). These intermediate I 1 and I II states appear reversible. Below pH 3.5, the protein becomes irreversibly unfolded (U).
Stopped-flow Kinetics and Real Time Unfolding of 3␤HSD2 Domains-To illuminate the mechanism of 3␤HSD2 activity, we measured protein unfolding using stopped-flow spectrophotometry (Fig. 9, A-C). Stopped-flow devices can monitor the unfolding and refolding of a protein in real time, thus revealing the different kinetic steps involved. The dead time or the mixing time of this instrument was set to 6.8 ms. The experiment requires rapid mixing of a protein in buffer with a high concentration of a denaturing agent such as urea or GdmHCl. Conformational changes in the protein are measured in milliseconds by fluorescence emission following excitation at 295 nm for Trp or at 280 nm for Trp and Tyr.
The unfolding kinetics appeared to be a single exponential process, and the values found by extrapolating back to time 0 agreed with those for the native state of the protein. When the buffers at different pH values are suddenly mixed, the protein starts to unfold, depending on the degree of denaturation. The rapid unfolding phase is defined as the "burst phase," which may or may not entail complete unfolding. The unfolding is completed when there is no change in fluorescence emission. The unfolding of 3␤HSD2 occurred after the burst phase, not within the dead time for all pH values, except pH 7.4, which shows only diffusion. As measured by fluorescence emission, the time of unfolding (t1 ⁄ 2 ) at pH 3.0 was 13 s with a relaxation time of 50 s (Fig. 9A), demonstrating that 3␤HSD2 unfolds by a two-step process. Under similar conditions, the unfolding at pH 3.5 was slower, with an initial relaxation of 20 s (Fig. 9B). At higher pH values, 3␤HSD2 might have been shielded by the flexible domains outside the core. Indeed, at pH 5.5 the protein gradually opened as shown by the t1 ⁄ 2 of 45 s, and then underwent diffusion (Fig. 9C). Thus, the burst phase is buried within the relaxation phase. As expected, faster unfolding was observed at pH 3.25 and 4.0 (supplemental Fig. S2, A  and B). However, at a further increase in pH to 7.4, no unfolding was observed, but only a slow diffusion (supplemental Fig. S2C). Thus, the slow unfolding at pH 5.5 suggests that there are no rearrangements during unfolding (33). The burst phase of unfolding shows no energy jump, confirming the rearrangement must have to open between pH 3.25 and 4.5, which remained in the molten globule state as indicated by the thermal denaturation (Fig. 8, C  and D).
We also examined the reversibility of the unfolding. CD kinetic measurements at both 208 and 222 nm indicate that 3␤HSD2 unfolds at pH 3.5 with increasing time (Fig. 9D). Also, a gradual decrease in negative ellipticity at 208 nm with increasing time shows the significant loss of ␣-helices at pH 3.5. However, the immediate development of negative and positive ellipticities around 208 and 222 nm, respectively, shows that the helical conformation was restored back after a pH jump from 3.5 to 7.4 (Fig. 9E) and the threshold limit for returning back to its original conformation was 25 min.

DISCUSSION
We had previously found that the steroidogenic enzyme, 3␤HSD2, does not get cleaved upon import into the mitochondria. Rather, the N-terminal mitochondrial leader sequence directs the N terminus to enter the mitochondrial membrane. 3␤HSD2 does so in an unfolded state. It then translocates to the IMM, where it associates with IMS proteins, but does not integrate into the membrane (3). This suggests that 3␤HSD2 may lack a receptor, which would allow greater flexibility of the protein. The proton pump can also be found at the IMM. The electrochemical gradient produced by the proton pump results in localized change in pH. Because 3␤HSD2 associates with the IMM, we posit that the protein faces the proton gradient and likely remains partially unfolded.
To test whether pH modulates the 3␤HSD2 conformation, we equilibrated purified protein at different pH values and evaluated conformations by CD. We found that changes in pH cor-responded to changes in conformation. This led us to conclude that pH causes a completely reversible equilibrium denaturation of 3␤HSD2. The pH-mediated changes in CD and fluorescence spectra of the protein may be described reasonably well by a simple three-state denaturation/renaturation model. However, the chaotropic agent GdmHCl induced further denaturation of the same protein, which indicates the presence of a stable folding intermediate that represents a significant portion of the equilibrium population. This intermediate retains a significant amount of secondary structure, amounting to ϳ30% of the native protein. The folding intermediate observed after incubation with GdmHCl was not due to differential denaturation of monomeric and dimeric 3␤HSD2, because the denaturation profile and CD spectra of 3␤HSD2 over a 50-fold range in concentration clearly showed that the protein is monomeric.
Several scenarios could account for the presence of a stable intermediate after denaturation of 3␤HSD2. The simplest explanation is that GdmHCl only partially denatures the protein, with the first transition producing the observed metastable intermediate. Even after incubation in 2 M GdmHCl, there may be some residual secondary structure remaining at pH 7.4. When we decreased the pH, it appeared that this metastable state persisted until pH 5.5. With a further decrease in pH, although, the metastable state underwent a two-step cooperative unfolding due to the presence of additional hydrogen ions.
A second scenario is that GdmHCl stabilizes an intermediate, increasing its equilibrium concentration. If nondenaturing concentrations of GdmHCl stabilize a folding intermediate, relatively low concentrations of GdmHCl might also lead to accumulations of the intermediate similar to acid-induced denaturation. Incubating nondenaturing concentrations of GdmHCl (0.5-1.14 M) with protein at pH 5.5 resulted in a directly biphasic denaturation (Fig. 6A), reminiscent of what we observed with GdmHCl-induced denaturation. Based on the ellipticity at 222 nm, we calculated the free energy, and then plotted these values as a function of the GdmHCl concentra- tion. Through extrapolation, we determined that the ⌬G 0 value in the absence of denaturant equaled 1.9 kcal/mol and used this as an independent estimate of the free energy of the N N I transition. This value is similar to that observed in the absence of acid denaturation. The second transition, I 1I 3 U, is more difficult to quantitate because it occurs at higher denaturant concentrations and produces a somewhat smaller signal change; however, it appears that increasing the GdmHCl concentration increases the free energy difference between the intermediate and unfolded state. With increasing acid denaturation in the presence of GdmHCl, the sum of the free energy change for the N N I I and I 1I N U transitions is nearly constant at 1.3 kcal/mol, particularly at the intermediate concentrations of GdmHCl as these concentrations yield the most accurate ⌬G measurements. This suggests that the estimated free energy difference between native and unfolded states is independent of the concentration of GdmHCl and that the accumulation of the intermediates results from GdmHCl-mediated stabilization.
Thus, there appears to be an equilibrium between I I to I 1I .
Maintaining an equilibrium state would conserve energy, resulting in a stable protein that can revert back to its native conformation when the pH reaches 7.4.
The fact that denaturation by pH in the absence of GdmHCl is characterized by a ⌬G N 3 U of 0.78 kcal/mol rather than 1.9 kcal/mol and the observation that the residue ellipticity pH 3.5 is significantly higher than that at 6 M GdmHCl implies that denaturation in pH alone produces a less stable intermediate that is difficult to detect experimentally. If our model is correct and the ⌬G N 3 U is independent of the denaturant, then we would expect ⌬G from I I 3 I II would have a value of ϳ0.5 kcal. Thus, the inferred stability of the intermediate in the absence of GdmHCl is consistent with the inability to observe it initially.
Based on this data, we have concluded that 3␤HSD2 is in a molten globule state at the IMM, which promotes activity. Folding can (34) proceed through a variety of intermediate states with decreasing Gibbs free energies (35). The intermediates may comprise partially folded forms of individual domains or of the entire protein: intermediates that have attained a high degree of secondary structure but lack defined tertiary structure are often termed "molten globules" (36). Molten globules can be viewed as highly dynamic, collapsed, compact polypeptide chains maintained by hydrophobic interactions that promote helical conformations but that lack fixed, specific longrange interactions (37)(38)(39)(40). Studies with bacterial ␤-barrel proteins have shown that insertion into the mitochondrial membrane causes the proteins to form molten-disc intermediates that have a partial secondary structure with the ␤-strands sitting flat on the membrane surface (11). This likely occurs in eukaryotic mitochondrial proteins as the membrane insertion pathway is fairly conserved (41). Molten globules are incompletely folded proteins and may go through molten globule intermediate states during membrane insertion (42) and biological activity is highly questionable at the molten globule state. However, in the present study the 3␤HSD2 is 7-fold more active at pH 4.5 than the protein in its native state, pH 7.4. This observation is further supported by the higher efficiency of ANS binding to the molten globule conformation of 3␤HSD2 at around pH 4.5 (31).
We also did stopped-flow spectrophotometry to measure real time unfolding of 3␤HSD2, and found the results consistent with a molten globule conformation. The protein rate of unfolding was observed only when the outer hydrophobic core was opened at or below pH 3.25 followed by relaxation. Also, the rate of unfolding between pH 4 and 5 embedding of the burst phase with the relaxation time, which is the condition of partial opening and existence of the molten globule conformation.
3␤HSD2 has an overall very flexible conformation (43). We had previously shown that the protein has no defined proteaseresistant domain (3). The N-terminal mitochondrial leader sequence is the most hydrophobic portion of 3␤HSD2, and secondary structure predictions suggest that all regions of the protein are highly structured, but flexible due to the presence of the Rossmann-fold domain, with a ␤-␣-␤-␣-␤-␣-␤-␣-␤-␣-␤ folding pattern that binds cofactor and substrate (44). An electrostatic interaction could occur between the positively charged residues of 3␤HSD2 and the negatively charged cofactors, resulting in a molten-disc conformation (11). The additional ejected protons in the IMS are likely to be absorbed by the lipid vesicles (45). Alternatively, small structured domains of 3␤HSD2 might interact with an as yet unidentified receptor on the IMM or at the IMS. In either event, the physical data suggest that the helices in the soluble active 3␤HSD2 structure could be preserved in the IMM (42), suggesting that the membrane bound protein retains its flexible conformation.
Enzymes catalyzing proton transfer are arranged asymmetrically on the inner mitochondrial membrane so that protons are transferred across the two-mitochondrial membranes. This proton pump establishes an electrochemical gradient leading to a local reduction of pH inside the mitochondria. If the interaction of 3␤HSD2 with the inner mitochondrial membrane causes a reduction in the membrane potential (⌬⌿) (42), our data suggest that this would increase the flexibility of the N-terminal hydrophobic region by accepting or releasing the additional hydrogen ions. Thus partial unfolding induced by the mitochondrial electrochemical gradient and by inner mitochondrial compartmental location would result in the transition to a molten globule state (11,46). By preserving some secondary structure, the molten globule state would provide the best pathway for minimizing the energetic cost and also the most active form for the catalysis of pregnenolone to progesterone. This will allow 3␤HSD2 to interact directly with the outer and inner mitochondrial translocase at the inter membrane space (3), as this transition to a flexible conformation lowers the energy required to open the structure, possibly supporting the ability to participate in two different steps as described by our model.