REDOX PROPERTIES OF HUMAN ENDOTHELIAL NITRIC OXIDE SYNTHASE OXYGENASE AND REDUCTASE DOMAINS PURIFIED FROM YEAST EXPRESSION SYSTEM

Characterization of the redox properties of endothelial nitric oxide synthase, eNOS, is fundamental to understanding the complicated reaction mechanism of this important enzyme participating in cardiovascular function. Yeast overexpression of both the oxygenase and reductase domains of human eNOS, i.e. eNOS ox and eNOS red , has been established to accomplish this goal. UV-Vis and EPR spectral characterization for the resting eNOS ox and its complexes with various ligands indicated a standard NOS heme structure as a thiolate hemeprotein. Two low-spin imidazole heme complexes but not the isolated eNOS ox were resolved by EPR indicating slight difference in heme geometry of the dimeric eNOS ox domain. Stoichiometric titration of eNOS ox demonstrated that the heme has a capacity for 1-1.5 reducing equivalent. Additional 1.5 –2.5 reducing equivalents were consumed before heme reduction occurred indicating the presence of other unknown high-potential redox centers. There is no indication for additional metal center that could explain this extra electron capacity of eNOS ox . Ferrous eNOS ox , in the presence of L-arginine, is fully functional in forming BH 4 radical upon mixing with oxygen as demonstrated by rapid-freeze EPR measurements. CaM binds eNOS red at 1:1 stoichiometry and high affinity. Stoichiometric titration and computer simulation enabled the determination for three redox potential separations between the four half reactions of FMN and FAD. The extinction coefficient could also be resolved for each flavin for its semiquinone, oxidized and reduced forms at multiple wavelengths. This first redox characterization on both eNOS domains by stoichiometric titration and the generation of high-quality EPR spectrum for the BH 4 radical intermediate illustrated the usefulness of these tools in future detailed investigation into the reaction mechanism of eNOS.

7 forward primers, the EcoRI site and His 6 tag were added, and in each backward primer an XbaI site was added. The correct sequences of the PCR products were confirmed by primer extension sequencing. Both PCR products were double-digested with EcoRI and XbaI and subcloned separately into the corresponding sites of an AOX promoter-driven expression vector pPICZB to obtain a 1.5 kb insert of eNOS ox and a 2.1 kb insert of eNOS red . The constructs were linearized with PmeI, transformed into yeast P. pastoris GS115, and selected by growing on the YPDS/Zeocin plates containing 1% yeast extract, 2% peptone, 2% dextrose, 1M sorbitol and 100 µ g/ml Zeocin. The colony that grew fastest was inoculated into 25 ml buffered minimal glycerol medium (100 mM potassium phosphate, pH 6.0, 1 % yeast extract, 2 % peptone, 1.34 % yeast nitrogen base with ammonium sulfate without amino acids, 4 x 10 -5 % biotin and 1 % glycerol) and cultured at 30 o C overnight. This culture was then transferred to a 250 ml buffered minimal glycerol medium and grown at 30 o C overnight to A 600 = 7~10. Cells were harvested and resuspended in 250 ml buffered minimal methanol medium (100 mM potassium phosphate, pH 6.0, 1 % yeast extract, 2 % peptone, 1.34 % yeast nitrogen base with ammonium sulfate without amino acids, 4 x 10 -5 % biotin and 0.5 % methanol) with 4 mg/liter hemin chloride added and cultured for 72 hrs at 30 o C to induce protein expression. inhibitors. An equal volume of glass beads (425-600 microns) was added to the suspension. Cells were broken by 10 cycles of 30s vortexing and brief chilling on ice. Cell debris and glass beads were removed by centrifugation at 3400 r.p.m. The supernatant obtained after another by guest on March 24, 2020 http://www.jbc.org/ Downloaded from eNOS red activity assay -Cytochrome c reductase activity was measured as the absorbance increase at 550 nm using ∆ ε = 21 mM -1 cm -1 as described previously (17). Ferricyanide or 2,6dichlorophenol indolphenol oxidation assay was determined by ∆ ε = 1 mM -1 cm -1 at 400 nm and ∆ε = 21 mM -1 cm -1 at 600 nm, respectively (14). eNOS ox Activity in generating biopterin radical. This activity measurement essentially followed previous published procedure for iNOS ox (26,27). High concentration of BH 4reconstituted eNOS ox was reduced anaerobically in a tonometer by dithionite titration. The ferrous eNOS ox was then reacted with oxygenated buffer using rapid-freeze/EPR technique as we previously published (28). The rapid-freeze apparatus, System 1000, Update Instrument (Madison, WI), was placed inside an anaerobic chamber (Coy Laboratory, MI). The oxygen level was lower than 5 ppm during the whole experiment procedure monitored by an oxygen/hydrogen analyzer (Model 10, Coy Laboratory). One or two push programs were used to obtain samples freeze trapped at different reaction time.
Spectrometry -UV-Vis spectra were measured on a HP8453 diode array spectrophotometer with 1 nm spectral bandwidth. EPR was recorded at liquid helium or liquid nitrogen temperature on a Bruker EMX. For liquid helium system, a GFS600 transfer line and an ITC503 temperature controller were used to maintain the temperature. An Oxford ESR900 cryostat was used to accommodate the sample. For liquid nitrogen transfer, a silver-coated double jacketed glass transfer line and a BVT3000 temperature controller was used. Data analysis was conducted using WinEPR and spectral simulation was done using SimFonia programs provided by Bruker. Flavin fluorescence was measured using SLM SPF-500C spectrofluorometer at ratio mode. About 2 µ M eNOS red in a 1-cm quartz cuvette was excited at 450 nm (5 nm spectral band width) and the emission spectrum between 450 -650 nm (7.5 nm spectral band width) was collected at 24 o C.
Stoichiometric titration-The redox capacity of eNOS red and eNOS ox were determined by anaerobic stoichiometric titration using sodium dithionite. Stock solution of sodium dithionite was freshly prepared by dissolving powdered reagent in 50 mM, pH 8.2 pyrophosphate buffer presaturated with pure nitrogen gas. The concentration of sodium dithionite was standardized by titration against a fixed amount of lumiflavin-3-acetic acid (ε 444 = 1.08 x 10 4 M -1 cm -1 ) anaerobically before and after individual real sample titration (29). The average concentration was used to calculate the number of reducing equivalents consumed in the titrations. Each protein sample was placed in an anaerobic titrator and made anaerobic by 5 cycles of evacuation (30 sec.) and argon replacement (5 min). Standardized dithionite solution contained in a gas-tight syringe engaged to the side arm of the titrator was quantitatively delivered and mixed with the protein sample under argon atmosphere. Electronic spectrum was recorded on an HP 8452 diode spectrophotometer and to confirm the system was equilibrated after each addition of dithionite reflected by a static absorbance.
Miscellaneous methods---Protein content was determined by BCA method (30). Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was performed on 10 % Ready-Gels in a Bio-Rad mini-gel apparatus. The gel filtration chromatography was performed on a Sephacryl 200 HR column (1.5 x 5.0cm). The kit for molecular weight 12,000-200,000 (product code: MW-GF-200) was used as the gel filtration marker.

Computer modeling
The SCoP program (Simulation Resources Inc., Redlands, CA) was used for simulating the data obtained from stoichiometric titration, mainly the eNOS red similar to the method used by by guest on March 24, 2020 http://www.jbc.org/ Downloaded from Iyanagi et al. (31). The absorbance changes at different monitoring wavelength during titration were simulated against accumulated reducing equivalents added.

Expression
The purified oxygenase domain also contained endogenous biopterin at a stoichiometry lower than 0.3/monomer. As our sample buffers did not include DTT, most of these biopterin molecules were present as dihydrobiopterin, BH 2 , as analyzed by our HPLC method (data not shown). The functional form of biopterin, BH 4 , can be reconstituted back to the purified eNOS ox according to the anaerobic procedure similar to that described by Rusche and Marletta (22). The reconstituted eNOS ox has biopterin content as high as 0.72 per monomer (Table 1) and are present as fully reduced form, BH 4 , as analyzed by our HPLC method (data not shown).
The content of both FAD and FMN in purified eNOS red is essentially stoichiometric based on our HPLC determination against authentic FAD and FMN standards ( Table 1). The ratio of FAD and FMN to that of eNOS red monomer is 1.0 and 1.14, respectively, thus further reconstitution of flavins is unnecessary.
Spectroscopic characteristics of eNOS ox . UV-Vis spectrum from 250 -700 nm of purified eNOS ox showed a Soret peak at 400 -404 nm, 81 mM -1 cm -1 , a broad α / β band at 518 nm, 15.4 mM -1 cm -1 , and a charge-transfer band at 645 nm, 5.8 mM -1 cm -1 . Treatment with L-arginine shifted the Soret peak to 396 nm with comparable amplitude, 82 mM -1 cm -1 , and only slight changes in the visible region (Fig. 2). When eNOS ox was reduced by dithionite, the Soret band is red-shifted to 413 nm with sizable decrease in intensity, 66.7 mM -1 cm -1 . The α / β band also shifted to 552 nm, 13.0 mM -1 cm -1 , and the charge-transfer band at about 650 nm is abolished as the lower lying three metal dorbital are completely filled. Further addition of CO resulted in the hallmark 444 nm Soret band for P450 hemeproteins with an extinction of 91.3 mM -1 cm -1 with the features at visible region very similar to that of ferrous eNOS ox . These spectral behaviors are very similar to our bacterialexpressed eNOS ox and other NOS oxygenase domains (9,13,17). The ratio of 280 nm to the Soret peak to either the resting or L-arginine-treated eNOS ox was ~1.5 which is an index of the purity of the hemeprotein and is a good number compared with other NOS preparations (9,13,17).
Liquid helium temperature EPR of the resting eNOS ox showed mixture of high-spin and low-spin heme structures (Fig. 3, (32,33). Addition of excess amount of L-arginine essentially wiped out the low-spin heme signals and substantially increased the high-spin heme signals (Fig. 3,  imidazole low-spin heme is even better than that found for whole eNOS (33).
Stoichiometric titration of the eNOS ox . To determine the redox capacity of the purified eNOS ox , a stoichiometric titration was conducted using standardized dithionite solution. L-arginine was added to make the heme homogeneously high-spin. The spectral conversion from ferric to ferrous heme during the course of titration appeared to involve only one simple redox reaction as evidenced by the isosbestic points at 410, 489, 532, 615, and 678 nm for both the absolute and difference spectra relative to the resting eNOS ox spectrum. An isosbestic point at 338 nm was slightly perturbed by dithionite whose absorption peaks at 314 nm ( Fig. 4A and Inset). However, the optical amplitude changes at 444 -388 nm showed a long lag for 1.5-2.5 (in three separate titrations) reducing equivalents before a sharp rise. 1-1.5 reducing equivalents were needed to completely reduce eNOS ox (Fig. 4B). The additional 1.5-2.5 reducing equivalents consumed during titration were not due to oxygen contamination as assessed by lumiflavin titration using the same titration vessel and conditions. Furthermore, similar stoichiometric titrations performed on eNOS red did not show an initial long lag (see below). The extra reducing equivalents used to titrate eNOS ox could be due to oxidized biopterin or free sulfhydryl group at the protein surface. The former may be a consequence of autooxidation of BH 4 and the latter could be due to the loss of the zinc cluster, which coordinate with four cysteine ligands with two cysteines from each monomer (2, 5, 6).
However, the samples used in these titrations are not BH 4 -replenished. The content of biopterin was as low as 0.2-0.3/heme and was preset as BH 2 , which is not reducible by dithionite. This left the zinc loss as the most possible cause of the additional consumption of dithionite in the titration.
Three experiments were conducted to assess this hypothesis. Zinc analysis by ICP-mass was carried out using either eNOS ox or purified whole eNOS. Sufficient amount of zinc was determined by ICP-MS analysis (data not shown). Gel-filtration chromatography was conducted to determine the population of eNOS ox monomer and dimer. Molecular sieving using five molecular weight standards and purified eNOS ox indicated that the whole population is present as a dimer with a molecular weight greater than 100 kDa (Fig. 5). Titration of free thiol by 4,4'-dithiopyridine was also conducted on eNOS ox using free L-cysteine as positive control. Time-dependent modification of the thiol was monitored in parallel with urea-treated eNOS ox and a bovine eNOS ox predetermined to have zinc and present as a dimer (5)  Tetrahydrobiopterin radical formation of eNOS ox . BH 4 -reconstituted eNOS ox prepared at a concentration of ~300 µ M was premixed with excess L-arginine and reacted with oxygenated buffer anaerobically at room temperature and freeze trapped at several time points. EPR spectrum corresponding to 100 ms reaction time is given in Fig. 7. EPR recorded at 11K between 200 -4200 G revealed both the heme component and the radical component (Fig. 7A). EPR spectrum of a control L-arginine-treated eNOS ox was also recorded under the same EPR conditions. Two spectra are normalized to the same concentration of heme. Approximately 50% of the BH 4 is converted to BH 4 radical and other diamagnetic heme intermediates estimated from the decrease of the highspin heme signal amplitude. The radical signal observed at g = 2 region was remeasured at 120K Enzyme activities of eNOS red . Cytochrome c reductase activity, DCPIP and ferricyanide reduction activities were evaluated for eNOS red expressed in yeast (  (Table 2). Our data are compatible with literature data for eNOS red or full eNOS (Table 2). Although there are some variations of cytochrome c and ferricyanide reductase activities among different eNOS or eNOS red preparations, most of them are within the same order of magnitude (Table 2). Differences in assay temperature, expression system and buffer composition could be the reasons that resulted in these variations. Computer simulations for the reductive titration data. The data shown in Figure 9D at three different wavelengths were simulated by SCoP program according to Eqs. 1 -12 to obtain three redox potentials gaps, ∆ E 1 -∆ E 3 , between four half reactions of two flavins in eNOS red (31).
Computer simulation for the data at 456, 508 and 600 nm was successful as indicated by the close match of the simulations and the actual data except the initial < 0.3 reducing equivalents (Fig. 9D).
The initial short lag was probably due to residual amount of oxygen in the titrator. This simulation process was tested using any arbitrary absolute midpoint potential value for one of the four half reactions and to zoom in the values for ∆ E 1 -∆ E 3 . The variation for each of the redox potential gap is not significant as shown in Table 3. The optimal value for ∆ E 1 is the biggest, 180 mV, and a clean conversion from one oxidized flavin to the semiquinone form is expected in the first stage of titration. In contrast, ∆ E 2 is almost zero indicating that the second stage of reduction consisted of two almost parallel half reactions. The value of ∆ E 3 , 73 mV, is in the middle and could be used to estimate the cutoff point to obtain the absorbance contribution from only one specific halfreaction. The extinction coefficients at three different wavelengths were also converged by several simulation cycles and were shown in Table 4. In principal, we could conduct this simulation cycles on any wavelength between 300 and 700 nm and to reconstruct the spectrum for each of the six flavin redox species.

DISCUSSION
We have successfully prepared both eNOS ox and eNOS red using yeast expression system.
Purified eNOS ox domains appear to have most of the heme characteristics in intact eNOS. The optical spectra of ferric and ferrous eNOS ox and its ferrous-CO complex and the spin-state change by L-arginine are all typical for eNOS and eNOS ox we observed previously (17,21,24). The purity index expressed as the ratio of A 280 /A 396 was as good as 1.3-1.6 compared to 1.5 -1.7 in our previous baculovirus expressed recombinant protein (17). This ratio is not as good as recent iNOS ox preparation expressed from E.coli (9) but our oxygenase domain does not have the Nterminal heterogeneity as observed for iNOS ox as only one single peptide band was observed on SDS-PAGE at ~50 kDa region (9).
EPR spectrum of the isolated eNOS ox showed dominant low-spin P450-type heme. Using the low-field g = 7.5 signal amplitude to estimate the spin-state population against the L-argininetreated eNOS ox , about 75 % of the heme was present as low-spin. This high proportion of low-spin heme is at odds with the room temperature optical data (Fig. 2 vs. Fig. 3 The extra 1.5 -2.5 reducing equivalents required to initiate the heme reduction in the stoichiometric titration is puzzling (Fig. 4). These additional equivalents are not originated from biopterin as the amount of pterin was too low to account for this amount of reducing equivalents and dithionite does not reduced BH 2 to BH 4 due to unfavorable redox thermodynamics (35). The zinc loss is not the reason as there is plenty of zinc present in isolated eNOS ox assessed by ICP mass analysis. The isolated eNOS ox is a perfect dimer as analyzed by gel-filtration. Monomeric form was not even detected. Furthermore, the titration for the surface exposed thiol function group showed identical modification kinetics as another bovine eNOS ox whose crystallographic data indicates the presence of zinc cluster. Thus the hypothesis that a zinc loss leading to surface exposed disulfide linkage as that found in the iNOS ox crystallographic data does not apply to our isolated eNOS ox (7). Furthermore, the possibility for propensity of zinc loss in recombinant eNOS ox but not in intact eNOS may be unfounded in our yeast expression system. We do not see additional metal redox centers such as heme, iron-sulfur cluster or copper center by optical or EPR spectroscopy, thus leave us with no explanation for the extra reducing equivalents which show much higher redox potential than the heme center. Reduction of the heme center appeared to consume more than one reducing equivalent (1-1.5 in three experiments). A similar case in iNOS ox was also observed recently (9). The sharp increment of absorbance change at the beginning of the titration and a curvature and even tailing approaching the end of the titration seems to indicate that the titrant may not have electron donating power strong enough to completely reduce all the heme molecules. Considering the very negative midpoint potential of thiolate-ligated heme, it is possible in the later portion of the titration that only part of the reducing equivalents from dithionite were donated to the heme center dictated by the midpoint potential difference between the heme and dithionite (36). We thus put more emphasis in using the initial linear sharp rise to estimate the end point of titration. By doing this, we get a stoichiometry closer to 1 rather than 1.5.
The biological activity of our eNOS ox was demonstrated by its capability in forming biopterin radical (Fig. 7). We chose this method as it is directly linked to the redox function of the protein and provide detailed information regarding the reaction mechanism of eNOS. In our study, the radical signal plateaued at ~100 ms at room temperature. The line shape, intensity and initial kinetics of biopterin radical formation appeared very similar to those published for iNOS ox (26,27). Computer simulation for the BH 4 radical indicates a minimal requirement for one nitrogen and two proton nuclei to match the EPR data. As N5 (or its 4a carbon) is para to the electron-releasing amino group at C2 and thus has high electronegativity, it is thus more favorable than N8 to give the first electron. N8 (or its 8a carbon) is meta to C2 and C4; thus electron withdrawal can only occur via conjugation with 4-oxo group. The pKa of the N5 proton is much higher than neutral and is not dissociated easily, thus a hyperfine interaction of this proton with the unpaired electron at N5 is expected. The second proton has to come from the C6 beta proton. Such initial trial of simulation appears fairly promising. Our observed biopterin radical is thus likely a BH 4 + · cation radical (37,38). Although N8 nuclei and its associated proton have been proposed to involve in the unpaired spin system (37), it remains to be clarified by further spectroscopic studies using isotope replacement. Nonetheless, we presented here the first high quality EPR spectrum of eNOS biopterin radical and will pursue the mechanistic role of biopterin using rapid-freeze EPR approach in parallel with stopped-flow and other kinetic methods. eNOS red activity was assessed by three different assays. The cytochrome c reduction and the DCPIP reduction assays require the participation from both flavins and ferricyanide reduction activities were believed to involve only FAD (39). In all cases, CaM enhanced the activity and the enhancement for cytochrome c reduction and DCPIP reduction activity to a similar extent. Why we only see ~ 3 × activity increase for the cytochrome c reductase activity by CaM by our yeast protein and 10 × increase in our previous sf9 expressed eNOS red is unclear (Table 2). Nonetheless, CaM appeared to interact both between FMN and the heme as well as between FAD and FMN as initially observed in nNOS (40). There are many factors that enhance the reductase activity of eNOS including the CaM binding, the removal of the autoinhibitory peptide and the phosphorylation of the C-terminus of the reductase (1,39,41,42). Furthermore, the presence of DTT, EDTA and variation in ionic strength during different stage of the purification also affect the sensitivity of the reductase domain activity to Ca +2 / CaM (17,39,41,42). The only redox centers present in eNOS red are the two flavins as exactly four reducing equivalents were consumed in the stoichiometric titration (Fig. 9D). Optical changes of the flavins occurring almost immediately after dithionite addition contrasts with the data for eNOS ox which required 1.5 -2.5 additional reducing equivalents before reduction of heme and supports that additional redox centers were present in eNOS ox (Fig. 4B). In addition to quantifying the redox capacity of eNOS red , stoichiometric titration also enabled determination of the relative redox potentials between different half reactions as illustrated in this study ( Fig. 9D and Table 3).
Successful simulation in data of all three wavelengths using the same set of difference midpoint potential values attested to the utility of this approach. The redox potential gap between the first and second half-reaction was 170 -190 mV, thus a complete separation of the first half-reaction from the other three is expected. The optical spectrum at the point of addition of one reducing equivalent should contain one intact flavin (FAD) and one flavin semiquinone (FMNH·) (Fig. 9A, heavy line). The extinction coefficient for the FMNH· semiquinone could thus be unambiguously determined. The spectral change at 600 nm is completely due to semiquinone forms of the flavins as the fully reduced and fully oxidized sample are silent at this region. The titration data indicate that the second semiquinone was gradually reduced to its hydroquinone at the addition of the fourth reducing equivalent. However, ∆ E 3 is only 56 -90 mV and prohibits clean separation of the last half-reaction from the other three. As 10 (∆ E 3 /0.059) = (Ox 1 · Red 2 )/(Ox 2 · Red 1 ) for the last two overlapping half reactions: We expect that only 70 -80 % of the reaction after addition of the 4 th electron is only contributed by FADH· semiquinone reduction to FADH 2 . Simulation for the 600 nm data here is very useful to define the extinction coefficient for the second semiquinone species as the trapezoidal titration profile is very sensitive to the difference midpoint potential as well as the extinction coefficient (31) thus greatly assisted in converging the value of the ∆ E values and the extinction for the second flavin semiquinone by simulation. There is a 20 % difference in the extinction coefficient at 600 nm for these two flavin semiquinones and almost 3 × difference at 456 nm and ~50 % difference at 508 nm with the FMNH· having the higher values ( Table 4). The middle two halfreactions, corresponding to that after addition of the 2 nd and 3 rd reducing equivalents, attributable to the formation of FAD semiquinone and FMNH· transformation into the fully reduced form are almost equipotential and were titrated together. The absolute values of the midpoint potential for all four half-reactions will be determined by potentiometric titration. Once these values are available, they will be used to validate the ∆ E values obtained in this study and to refine the accuracy of the extinction coefficient derived in this study. The spectral contribution from each half reaction at every single wavelength can be reliably determined and will be useful in future mechanistic studies by stopped-flow measurements.