Proc. Nati. Acad. Sci. USA Vol. 85, pp. 2979-2983, May 1988 Biophysics aggregation and interaction with cholesterol oxidase in supercritical carbon dioxide (electron paramagnetic resonance spectroscopy/Krafft behavior/ in nonaqueous solvents) T. W. RANDOLPH, D. S. CLARK, H. W. BLANCH*, AND J. M. PRAUSNITZ Department of Chemical Engineering, University of California, Berkeley, CA 94720 Contributed by J. M. Prausnitz, December 7, 1987

ABSTRACT High-pressure EPR spectroscopy indicates concerned only with cholesterol oxidase from G. chryso- that cholesterol forms aggregates in supercritical carbon diox- creas. As has been shown elsewhere (5), addition of small ide. In pure carbon dioxide, changes in cholesterol aggregate amounts of cosolvents to supercritical carbon dioxide may size or packing structure are observed with changing pressure. cause substantial increases in the rate of enzymatic reaction. Near the critical point of carbon dioxide, cholesterol solubility These changes are not easily explained in terms of choles- is too low to permit sigificant aggregation, and monomeric terol-solubility increases caused by cosolvent addition; al- cholesterol is observed. Addition of small amounts of dopants though methanol and acetone cause the largest increases in to supercritical carbon dioxide strongly affects cholesterol cholesterol solubility, these cosolvents yield smaller in- aggregation. Branched butanols (2-methyl-1-propanol and 2- creases in reaction rate than do 2-methyl-1-propanol (iso- methyl-2-propanol) and ethanol (to a lesser degree) promote butyl alcohol) and 2-methyl-2-propanol (tert-butyl alcohol), cholesterol aggregation, while methanol, acetone, and 1- which cause the smallest solubility increases. butanol do not. Cosolvents that promote aggregation also High-pressure EPR spectroscopy was used to study the increase the rate at which cholesterol oxidase from Gloeocysti- effect of cosolvent addition. Nitroxide-labeled cholesterol cum chrysocreas catalyzes the oxidation of cholesterol. In oxidase from G. chrysocreas was used in addition to a supercritical carbon dioxide solutions, the EPR spectroscopy nitroxide-labeled derivative of cholesterol, 3-doxyl-5-a cho- reveals little or no conformational change in cholesterol oxi- lestane, to study (i) the conformation of the as a dase as 2-methyl-2-propanol or methanol is added. Damp function of cosolvent addition, (ii) the self-association of cholesterol oxidase binds multiple cholesterol molecules; dry cholesterol in supercritical carbon dioxide and supercritical enzyme loses the ability to bind cholesterol. When molecular carbon dioxide/cosolvent mixtures, and (iii) the interaction is the oxidizing agent, the rate of enzymatic cholesterol of cholesterol and cholesterol oxidase under supercritical oxidation is greatly reduced in bone-dry carbon dioxide com- conditions. pared to that in water-saturated carbon dioxide. MATERIALS AND METHODS Supercritical carbon dioxide and supercritical carbon diox- ide cosolvent mixtures have been shown (1, 2) to provide a EPR spectra were recorded on an IBM ER 200 D-SRC EPR medium wherein enzymes maintain catalytic activity (also spectrometer with a custom-made quartz high-pressure cell. unpublished results of T.W.R., D. A. Miller, H.W.B., and Three spectra were recorded at each pressure with a sweep- J.M.P.). This supercritical-fluid medium is advantageous for time of 150 sec, and the spectra were averaged. Pressures enzyme-catalyzed reactions for a number of reasons: (i) high were recorded with an Omega pressure transducer (model diffusivities and low viscosities (relative to liquid solvents); PX 420-2KGI) calibrated with a dead-weight pressure gauge. (ii) simplified downstream separations of products, unre- Cholesterol oxidase from G. chrysocreas (Chemical acted substrates, and catalysts; and (iii) large changes in Dynamics, South Plainfield, NJ) was spin-labeled by incu- solvent power and dielectric constant caused by small bating the enzyme in a 100-fold molar excess of 2,2,5,5- changes in pressure and temperature. tetramethyl-l-pyrrolin-3-oxyl-carboxylic acid N-hydroxy The modification of steroids constitutes a class ofenzyme- succinimide (Kodak Chemicals; used as received) for 24 hr catalyzed reactions in which supercritical-fluid processing at room temperature. Unreacted spin label was removed by may be of particular interest. Since aqueous solubilities of dialysis for 48 hr at 0°C against a 50 mM phosphate buffer many steroids are low, reaction rates are also low in aqueous (pH 7.0). media. Substantial increases in steroid solubility may be The interaction of cholesterol and cholesterol oxidase obtained by using a supercritical fluid as a solvent. For from G. chrysocreas was examined by EPR spectroscopy by example, cholesterol is about 50 times more soluble in using 3-doxyl-5-a-cholestane, a cholesterol analogue spin- supercritical carbon dioxide at 123 bars (1 bar = 100 kPa) labeled with a nitroxide group (Aldrich; used as received). and 308 K (3) than in water at 298 K (4). Addition of small Cholesterol oxidase [Chemical Dynamics; dialyzed for 48 hr amounts of cosolvents to supercritical carbon dioxide (e.g., against 50 mM phosphate buffer (pH 7.0)] was first immobi- 3.5 mol % of methanol) may increase solubility by an lized on porous aminosilanized glass beads (Sigma). One additional order of magnitude (3). gram of beads was activated by treating the surface amine We have examined the enzyme-catalyzed kinetics of cho- groups with a 2.5% glutaraldehyde solution in 50 mM phos- lesterol oxidation by molecular oxygen in carbon dioxide. phate buffer (pH 7.0). The mixture was allowed to react for Although cholesterol oxidases from Streptomyces sp., Pseu- 1 hr at room temperature. Unreacted glutaraldehyde was domonas sp., Norcardia sp, and Gloeocysticum chrysocreas removed by washing on a Buchner funnel with 50 mM are active in supercritical carbon dioxide, this study is phosphate buffer (pH 7.0). Five milliliters of cholesterol oxidase solution [10 mg/ml of 50 mM phosphate buffer (pH The publication costs of this article were defrayed in part by page charge 7.0)] was added to the glass beads. After 3 hr of reaction at payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. *To whom reprint requests should be addressed.

Downloaded by guest on September 27, 2021 2979 2980 Biophysics: Randolph et al. Proc. Natl. Acad. Sci. USA 85 (1988) room temperature, uncoupled enzyme was washed away on When a solution of 3.9 mM 3-doxyl-5-a-cholestane in a Buchner funnel with the phosphate buffer. The final wash carbon dioxide at 112 bars and 308 K was examined by EPR contained 0.01% sodium azide as a preservative. spectroscopy, the expected sharp three-peak signal was not Spin-labeled cholesterol was added to the high-pressure observed. Instead, the spectrum (Fig. 2) is composed of EPR cell in a methylene chloride solution. The methylene three broad peaks on a sharply sloping baseline. Reducing chloride was evaporated under vacuum leaving the spin- the 3-doxyl-5-a-cholestane concentration to 65 ILM did not labeled cholesterol derivative deposited on the walls of the give a typical "dilute solution" EPR signal; the spectrum in EPR cell. Spin-labeled enzyme was added as a solid immo- Fig. 2 resembles a "solid-like" form typical of much higher bilized on glass beads. spin-label concentrations. Although the overall concentra- tion of slin-labeled cholesterol had been greatly reduced, the RESULTS local concentration remained sufficiently high for spin-spin interactions to cause merging of the three peaks. Therefore, Conformation of Cholesterol Oxidase from G. chrysocreas in it appears that the nitroxide-labeled cholesterol molecules Supercritical Carbon Dioxide. Cholesterol oxidase from G. exist as aggregates in supercritical carbon dioxide under chrysocreas was derivatized by using 2,2,5,5-tetramethyl- these conditions. pyrrolin-1-oxyl-3-carboxylic acid N-hydroxy succinimide es- Krafft Pressure Behavior in Supercritical Carbon Dioxide. ter, a spin label that is reactive towards lysine residues (6). Further evidence of cholesterol aggregation was provided by Nitroxide groups were attached to the enzyme with an the EPR spectrum of 3-doxyl-5-a-cholestane in carbon diox- average stoichiometry of eight per enzyme. Fig. 1 shows ide at lower pressures, where carbon dioxide's solvent power EPR spectra of spin-labeled cholesterol oxidase at atmo- spheric conditions and under supercritical conditions in is low. In liquid-micellar systems, there often exists a carbon dioxide with addition of 3% methanol or 2-methyl-2- "Krafft" temperature, where the solubility ofthe surfactant is propanol. Surprisingly little difference in the spectra is seen equal to the critical micelle concentration. Below the Krafft as conditions change from atmospheric to supercritical or temperature, surfactant molecules are not sufficiently con- upon addition of2-methyl-2-propanol. Some distortion ofthe centrated to form micellar aggregates. Similarly, in a super- spectrum occurs upon addition of methanol; the peak split- critical fluid, a Krafft pressure may be defined as the pressure ting decreases somewhat and the shoulder on the first peak (at a given temperature) at which the solubility of a surfactant is more pronounced. While these spectra cannot rule out a is equal to the critical micelle concentration. small conformational change in the enzyme caused by co- For a solution containing 0.27 mM 3-doxyl-5-a-cholestane solvent addition, such a change seems unlikely, considering and 1.4 mM cholesterol, the EPR spectrum changed dramat- that the magnitude of the observed changes in the EPR ically as the pressure was reduced from 112 bars. The three spectra was small and that the enzyme had been labeled at nitroxide peaks are clearly distinguishable in Fig. 3 above 89 eight different sites. Changes in the EPR spectrum with bars, although they are quite broad and exhibit a steeply addition of methanol do not necessarily reflect enzyme- sloping baseline. In this region, the local concentration of conformation changes; methanol addition changes the super- unpaired electrons is significantly higher than the overall critical solvent's properties (e.g., dielectric constant and bulk-phase concentration. viscosity), which in turn can affect the EPR spectrum The local nitroxide concentration increases further as the without causing conformational changes. pressure is lowered from 89 to 85 bars. In this region in Fig. Self-Association of Cholesterol in Supercritical Carbon Di- 3, the peaks are so broad that three peaks can scarcely be oxide. A rapidly tumbling nitroxide group in dilute solution defined. This unexpected increase in local concentration (of the order of 1 mM or less) gives an EPR signal composed may be due to the formation of an aggregate that is different ofthree sharp peaks ofapproximately equal height separated (from those observed elsewhere) in size or packing struc- by a flat baseline. Under more concentrated conditions, the ture. peaks widen until, at concentrations of approximately 0.1 M The region between 83.2 and 81 bars may be defined as the or more, the three peaks merge into a single broad peak (7, Krafft-pressure region. Here an abrupt change occurs in Fig. 8). 3 from a spectrum corresponding to high aggregation to a spectrum typical of rapidly tumbling "monomeric" spin label. Further evidence that at higher pressure cholesterol is indeed found in aggregate form was provided by the appear- ance of this sharp signal corresponding to nonaggregation at lower pressures. Below 81 bars (not shown), the signal remained sharp but decreased with decreasing 3-doxyl-5-a- a cholestane solubility until, below about 75 bars, the 3-doxyl- 5-a-cholestane concentration was too low for detection. b Transient EPR Spectra of 3-Doxyl-5-a-cholestane After Pressure Jumps in Supercritical Carbon Dioxide. The EPR spectra described in the previous section represent steady- C

d

3.275 3.305 3.335 3.365 3.395 3.425 GAUSS (Thousands) FIG. 1. EPR spectra of spin-labeled cholesterol oxidase from G. chrysocreas under atmospheric and supercritical conditions. Spec- 3.270 3.310 3.350 3.390 3.430 tra: a, atmospheric pressure (308 K); b, in carbon dioxide at 104 bars Gauss (Thousands) (308 K); c, in carbon dioxide and 3% (vol/vol) 2-methyl-2-propanol at 104 bars (308 K); d, in carbon dioxide and 3% (vol/vol) methanol FIG. 2. EPR spectrum of 65 ,uM 3-doxyl-5-a-cholestane in at 104 bars (308 K). supercritical carbon dioxide at 308 K. Downloaded by guest on September 27, 2021 Biophysics: Randolph et al. Proc. Natl. Acad. Sci. USA 85 (1988) 2981

, 3.305 3.335 3.360 3.95 3.425 3.2753305'3.335 3.365 3.39E GAUSS (Thousands) GAUSS (Thousands) FIG. 3. EPR spectra of0.27 mM 3-doxyl-5-a-cholestane with 1.4 mM cholesterol in carbon dioxide at 308 K. Pressure in bars is indicated. state measurements. Fig. 4 shows the EPR spectrum of 3-doxyl-5-a-cholestane as a function of time after the pres- sure was rapidly decreased (during a period of about 2 sec) from 84.1 to 82.7 bars and from 82.7 to 82.4 bars, respec- tively. The spectra show large changes for approximately 2 min until steady state was reached. Such changes may represent transient aggregate packing structures or aggregate sizes. The large magnitude of such spectral changes is surprising, considering the small pressure changes that pro- duced them. These results indicate the extreme sensitivity of density-dependent solvent characteristics to pressure changes in the region near the solvent's critical point. Cosolvent Effects on Cholesterol Aggregation in Supercrit- ical Carbon Dioxide. Addition of cosolvents to supercritical carbon dioxide solutions of 3-doxyl-5-a-cholestane caused large changes in the extent of aggregation. Fig. 5 shows the EPR spectrum of 3-doxyl-5-a-cholestane in supercritical carbon dioxide at 104 bars (308 K) with 3% (vol/vol of 3300 3350 3400 various cosolvents added. Addition of the branched buta- GAUSS nols-2-methyl-1-propanol and 2-methyl-propanol-pro mo- FIG. 4. Transient EPR spectra of 3-doxyl-5-a-cholestane after a ted cholesterol aggregation; only one broad peak is evident rapid pressure drop. (Upper) Pressure drop from 84.1 to 82.7 bars. in the EPR spectrum in Fig. 5. Ethanol promoted aggrega- Spectra were recorded after 12 sec (spectrum A), 24 sec (spectrum tion less effectively, as indicated by the appearance of three B), 36 sec (spectrum C), 48 sec (spectrum D) and 60 sec (spectrum poorly resolved peaks in the EPR spectrum. E). (Lower) Pressure drop from 82.7 to 82.4 bars. Spectra were recorded after 12 sec (spectrum A), 24 sec (spectrum B), 36 sec Addition of cosolvents 1-butanol acetone, and methanol (spectrum C), 48 sec (spectrum D), 60 sec (spectrum E), 72 sec caused progressively less aggregation. Although still not as (spectrum F), and 9 min (spectrum G). sharp as would be typical of nonassociating nitroxide groups at this concentration, the spectral line shapes were consid- Interaction of Spin-Labeled Cholesterol with Cholesterol erably narrower than those observed when either of the Oxidase from G. chrysocreas. The interaction of cholesterol branched butanols was added. with cholesterol oxidase from G. chrysocreas was examined Addition of cosolvents also changed the Krafft-pressure behavior for nitroxide-labeled cholesterol. Krafft-pressure behavior was observed near the critical pressure of carbon dioxide for the cosolvents investigated, but the Krafft- pressure transitions were less sharp in cosolvent-containing mixtures. Transitions were sharper when aggregate- promoting cosolvents (e.g., 2-methyl-2-propanol) were added than when cosolvents like methanol were added. Fig. 6 shows the Krafft-pressure transition for carbon dioxide solutions containing 3% (vol/vol) 2-methyl-2-propanol and methanol, respectively. The broadened Krafft-pressure re- gion may be the result of a broadened distribution of aggregation number. The degree of aggregation of cholesterol in carbon dioxide cosolvent mixtures correlates well with the observed rate of FIG. 5. EPR spectra of 0.31 mM 3-doxyl-5-a-cholestane in enzymatic cholesterol oxidation (5). The cosolvents that supercritical carbon dioxide at 103 bars and 308 K with 3% (vol/vol) promote cholesterol aggregation also provide the largest of various cosolvents added. Spectra: a, methanol, b, acetone; c, increases in the rate of enzymatic oxidation in supercritical 1-butanol; d, ethanol; e, (2-methyl-2-propanol); f, 2-methyl-i- carbon dioxide. propanol. Downloaded by guest on September 27, 2021 2982 Biophysics: Randolph et al. Proc. Natl. Acad. Sci. USA 85 (1988) The EPR cell was then pressurized at 308 K with bone-dry carbon dioxide. The sample was allowed to equilibrate for 1 1134 hr at a pressure of 104 bars. EPR spectra were then recorded at various pressures. The spectrum recorded at 104 bars is shown in Fig. 8. The spectrum appears to be a composite comprising a broad, exchange-broadened spectrum with the maximum peak split- ting AN of about 35 G and a sharper spectrum with an AN of about 17.5 G. The sharper spectrum appears to be asymmet- ric, with the third peak smaller than the first two, indicating restricted motion of the spin label. In Fig. 8, as the pressure decreases to 90.6 bars, both 81.2 spectra narrow slightly. The broad signal decreases in inten- 15 3 335 3.365 3.395 3 425 sity relative to the sharper signal, and the peaks of the GAUSS (Thousands) GAUSS (Thousands) sharper spectrum become more even. These trends continue with decreasing pressure in Fig. 8 until at 76.9 bars the broad FIG. 6. Krafft-pressure behavior of 3-doxyl-5-a-cholestane in spectrum is almost undetectable. The sharp spectrum dis- supercritical carbon dioxide with 3% cosolvent. (Left) With 2- plays a splitting AN of about 15 G, and the third peak has methyl-2-propanol. Pressure in bars as indicated. (Right) With grown to about 80%o of the size of the first peak. This methanol. Pressure in atmospheres (1 atm = 101.3 kPa) is indicated. lower-pressure spectrum is due to monomeric, freely tum- by EPR spectroscopy by using 3-doxyl-5-a-cholestane and bling 3-doxyl-5-a-cholestane. enzyme covalently attached to 50-gum glass beads. EPR There was a sharp contrast between the EPR spectra of spectra were recorded at 308 K and various pressures. The spin-labeled cholesterol with immobilized cholesterol oxi- spectra were very broad, with only one peak present (Fig. 7). dase in dry supercritical carbon dioxide and the correspond- As pressure was reduced through the critical region, the ing spectra in water-saturated carbon dioxide. Although signal did not sharpen to three peaks, in contrast to previous 3-doxyl-5-a-cholestane adsorbed to the enzyme at 104 bars experiments with no enzyme present (see Fig. 3). Since there in both cases, the adsorption apparently was much stronger was no evidence of monomeric spin-label signal, it appears when water was present. As the pressure was lowered below 104 bars in the case ofdry supercritical carbon dioxide, there that more than one spin-labeled cholesterol molecule ad- was significant (almost complete) desorption of 3-doxyl-5-a- sorbs onto the enzyme surface at a time, and the polar cholestane from the enzyme; on the other hand, the spin- nitroxide groups remain in sufficiently close proximity to labeled cholesterol remained on the enzyme surface, even cause spin-spin broadening of the EPR spectrum. The very near the critical point, when the carbon dioxide was spin-labeled cholesterol remains on the enzyme surface even saturated with water. in pressure regions where spin-labeled cholesterol does not In a water-restricted environment, the diminished capabil- form aggregates in supercritical carbon dioxide. No freely ity for the enzyme to bind to multiple molecules tumbling, monomeric spin-label signal could be detected, correlates well with the observed loss of oxidative activity of even at 81.7 bars (Fig. 7). the enzyme in bone-dry carbon dioxide. Enzyme binding to Earlier experiments in a continuous enzyme reactor a cholesterol-containing membrane may be a necessary first showed that reducing the water content of a supercritical step for enzymatic activity, and this binding appears to be carbon dioxide solution has a substantially negative effect on hampered when the enzyme is dry. the rate of enzymatic oxidation: a reversible decrease by a factor of 10 in enzymatic activity was found when the carbon dioxide/cholesterol solution was dried over a molecular CONCLUSIONS sieve (T.W.R., H.W.B., and J.M.P., unpublished results). A Damp cholesterol oxidase binds more than one cholesterol corresponding situation was investigated with EPR spectros- molecule. Dry cholesterol oxidase loses the ability to bind copy. A 0.29 mM solution of 3-doxyl-5-a-cholestane in multiple substrate molecules. This loss of binding ability is methylene chloride was evaporated onto the walls of a accompanied by a large reduction in cholesterol oxidase quartz high-pressure EPR cell. One hundred milligrams of activity. damp enzyme on glass beads was added, and the entire cell was dried for 5 hr at 3.3 x 10-5 bar and room temperature.

3.31 3.33 3.35 3.37 3.39 3.41 3.4 GAUSS (thousands) FIG. 7. EPR spectra of 3-doxyl-5-a-cholestane (0.29 mM) in FIG. 8. EPR spectra of 3-doxyl-5-a-cholestane (0.29 mM) in carbon dioxide at 308 K in the presence ofdamp cholesterol oxidase bone-dry carbon dioxide at 308 K in the presence of dry cholesterol from G. chrysocreas. Pressure in bars is indicated. oxidase from G. chrysocreas. Pressure in bars is indicated. Downloaded by guest on September 27, 2021 Biophysics: Randolph et al. Proc. Nadl. Acad. Sci. USA 85 (1988) 2983

Cholesterol aggregation provides an explanation for the financial support, we are grateful to the National Science Founda- varied effects ofcosolvents on cholesterol oxidase activity in tion (Grant CBT8513642) and to the Center for Biotechnology supercritical carbon dioxide. EPR spectroscopy shows that Research, San Francisco, CA. the cosolvents that cause large reaction-rate increases also 1. Randolph, T. W., Blanch, H. W., Prausnitz, J. M. & Wilke, promote cholesterol aggregation. An alternative explana- C. R. (1985) Biotech. Lett. 7, 325-328. tion, cosolvent-induced enzyme-conformation changes, may 2. Hammond, D. A., Karel, M., Klibanov, A. & Krukonis, V. J. be discounted. Only small changes in the EPR spectrum of (1985) Appl. Biochem. Biotechnol. 11, 393-400. spin-labeled cholesterol oxidase are seen in supercritical 3. Wong, J. M. & Johnston, K. P. (1986) Biotechnol. Prog. 2, carbon dioxide with cosolvent addition. 29-39. Cholesterol aggregation in supercritical carbon dioxide is 4. Haberland, M. E. & Reynolds, J. A. (1973) Proc. Natl. Acad. affected by cosolvent addition and the solution pressure. Sci. USA 70, 2313-2316. Changes in the EPR spectrum of spin-labeled cholesterol 5. Randolph, T. W., Clark, D. S., Blanch, H. W. & Prausnitz, J. M. (1988) Science 239, 387-390. may be a result of changes in cholesterol-aggregate packing 6. Twinning, S. S., Sealy, R. C. & Glick, D. M. (1981) Biochem- or size. At the "Krafft pressure," cholesterol exists in istry 20, 1267-1272. monomeric form in supercritical carbon dioxide; cosolvent 7. Likhtenshtein, G. I. (1976) Spin Labelling Methods in Molec- addition broadens the Krafft pressure region. ular Biology (Wiley, New York), pp. 40-45. 8. Jost, P. & Griffith, 0. H. (1976) in Spin Labeling: Theory and We are grateful to Mr. Paul S. Skerker for assistance with the Applications, ed., Berliner, L. J. (Academic, New York), pp. EPR measurements and for aid in interpreting EPR spectra. For 251-272. Downloaded by guest on September 27, 2021