Proc. Nati Acad. Sci. USAVol. 79, pp. 7175-7179, December 1982Biochemistry

Crystallization of mitochondrial cytochrome oxidase(affinity chromatography/electron diffraction/cardiolipin)

TAKAYUKI OZAWAt, MASASHI TANAKAt, AND TAKASHI WAKABAYASHIttDepartment of Biomedical Chemistry, Faculty of Medicine, University of Nagoya; and tDepartment of Pathology, Faculty of Medicine, Nagoya City University,Nagoya, 466 Japan

Communicated by David E. Green, August 30, 1982

ABSTRACT Cytochrome c oxidase (ferrocytochrome c:oxygenoxidoreductase, EC 1.9.3.1) was purified from beef heart mito-chondria. By washing the oxidase with detergent on a hydrophobicinteraction column, phospholipids were depleted to the level of 1mol of cardiolipin per mol ofheme a. Hydrophobic impurities andpartially denatured oxidase were separated from the intact oxidaseon an affinity column with cytochrome c as the specific ligand. Thefinal preparation of the oxidase contained seven.distinct polypep-tides. The molecular weight of the oxidase was estimated to be130,000 from its specific heme a and copper content and from thesubunit composition. Crystals ofthe oxidase were obtained by slowremoval of the detergent from the buffer in which the oxidase wasdissolved. The needle-shaped crystals were 100 Ium in averagelength and 5 ,um in width, -and they strongly polarized visible light.Electron diffraction patterns were obtained with an unstainedglutaraldehyde-fixed single crystal by electron microscopy using1,000-kV electrons. From electron micrographs and the diffrac-tion patterns of the crystal, it was concluded that the crystal ismonoclinic in the space group P21, with unit cell dimensions a =92 A, b = 84 A, and c = 103 A, and ai = (3 = 900, y = 1260.

Since the discovery (1) of cytochrome c oxidase (ferrocyto-chrome c:oxygen oxidoreductase, EC 1.9.3.1), this enzyme hasbeen extensively studied. Nevertheless, neither its structurenor its molecular weight has been established. The number ofsubunits and the specific heme and copper content of the oxi-dase have been reported in many papers but with wide vari-ations. In this study, we introduce a method of purification ofthe oxidase that permits crystallization of this intrinsic mem-brane complex. We have reported the crystallization of cyto-chrome c-cytochrome oxidase complex (2, 3); investigation ofthe three-dimensional structure of the complex by x-ray dif-fraction analysis is now possible. Comparative study ofthe crys-tal of cytochrome c-cytochrome oxidase complex and that ofcytochrome oxidase itself should add a new dimension to thex-ray diffraction analysis of the crystals and provide valuableinformation on the mechanism of electron transport.

In this paper, data will be presented that the oxidase can behighly purified by a combination of hydrophobic interactionchromatography and affinity chromatography. The purificationallows estimation of the molecular weight of the oxidase fromthree different aspects; its specific heme a content and coppercontent and its subunit composition. Also, the oxidase itselfcanby crystallized after the purification and depletion of its phos-pholipids. Evidence for a highly ordered three-dimensional ar-ray of the molecules in the crystal was provided by electronmicrographs and by electron diffraction studies.

15

1010

CD00cq

0 5-

0

c

-6a

Ce3

10 20 130 140 150Fraction

FIG. 1. Phenyl-Sepharose CL-4B hydrophobic interaction chro-matography of cytochrome oxidase. Cytochrome oxidase (106 mg ofprotein) was-applied to a phenylbSepharose CL-4B column (1 X 20 cm)equilibrated with 0.1 M bicarbonate buffer, pH 8.3, containing 1% de-oxycholate and 0.1M NaCl and washed with 500ml of the same buffer.At the point indicated by the arrow, 1% deoxycholate was replaced by1% Triton X-100. A280(-) andA598 (o-o) were monitored as mea-surements of protein and heme a, respectively. Four fractions shownby the horizontal bar were combined and purified further by affinitychromatography. Flow rate, 30 ml/hr; fraction size; 3.5 ml.

METHODS AND RESULTSPurification and Phospholipid Depletion of the Oxidase.

Cytochrome c oxidase was isolated from beef heart mitochon-dria according to the method of Fowler et al. (4) and was re-fractionated with ammonium sulfate in the presence of3% cho-late by the method of Tzagoloff and MacLennan (5). Therefractionated oxidase (approximately 100 mg of protein) wasapplied onto a column (1 x 20 cm) of phenyl-Sepharose CL-4Bequilibrated with 0.1 M sodium bicarbonate buffer, pH 8.3,containing 1% deoxycholate and 0.1 M NaCl (6). The columnwas washed with 500 ml ofthe same buffer to remove impuritiesand phospholipids, and the oxidase was eluted with the samebuffer, containing 1% Triton X-100 in place of deoxycholate, asshown in Fig. 1. The oxidase was eluted as a sharp peak justbefore the appearance of Triton X-100 micelles. By this hydro-phobic chromatography, the heme a content of the oxidase wasincreased from 12.1 to 14.1 nmol/mg ofprotein, and the coppercontent was increased from 12.9 to 16.3 nmol/mg of protein,and the phospholipids content was decreased from 11.8 to 2.2mol of phosphorus per mol of heme a (Table 1). The specificactivity ofthe oxidase at this purification step, 31 ,umol of0 permin per mg of protein at 30'C, was in rough accord with that,26-25 ,umol/min per mg at 380C, reported by Fry and Green(10) for phospholipid-depleted oxidase.

7175

The publication costs ofthis article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertise-ment" in accordance with 18 U. S. C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

7176 Biochemistry: Ozawa et al.

Table 1. Purification of beef heart cytochrome oxidasePhosphorus,t Activity,§

Heme a*, Coppert mol/mol ,umol O/minStep nmol/mg nmol/mg heme a per mg

Fowler et ali (4) 8.5 9.5 45.0 39Tzagoloff and MacLennan (5) 12.1 12.9 11.8 40Phenyl-Sepharose CL-4B 14.1 16.3 2.21 31Cytochrome c-Sepharose 4B 15.2 16.1 2.2w 34

* Heme a was determined from the reduced spectrum, using E65 - E630 = 16.5 mM-1 * cm-' (3). Proteinwas determined by the biuret method (7) with bovine serum albumin as a standard. In order to correctfor the contribution of alkaline hemochromogen of heme a to absorbance at 540 nm, 0.6% sodium po-tassium tartrate tetrahydrate solution in 3% sodium hydroxide was used in place of biuret reagent.

t Copper was measured with an atomic absorption spectrophotometer, Hitachi 170-10.t Phosphorus was determined by the method of Chen et al. (8).§ The activity was measured by oxygen uptake at 300C in 3 ml of 50 mM sodium phosphate buffer, pH7.4, containing 0.02% 1-oleoyl L-a-lysophosphatidylcholine (Sigma, synthetic), 30 mM ascorbate, and40 jLM horse cytochrome c (Sigma, type HI), using a Clark electrode (9). Before the assay, approximately0.1 mg of enzyme protein was mixed with 1 mg of soybean phospholipids in 2 ml of 50 mM sodium phos-phate buffer, pH 7.4, containing 0.1% cholate. After incubation at 4°C for 30 min, 100 ,ul of the mixturewas added to the reaction mixture.

¶ The phospholipid was identified as cardiolipin by thin-layer chromatography as shown in Fig. 3.

The oxidase purified by the hydrophobic interaction chro-matography was further purified by affinity chromatography,using cytochrome c as the specific ligand. Cytochrome c-Seph-arose 4B was prepared according to the method of Ozawa et aL(11) with some modification as reported (12). A typical chro-matographic pattern is shown in Fig. 2. In thi's step, impurities(fractions 6-50 in Fig. 2) and partially denatured enzyme, whichhad the a band at 598 nm (fractions 70-80) were separated fromthe intact oxidase. The final preparation exhibited a heme acontent of 15.2 nmol/mg of protein, copper content of 16.1nmol/mg of protein, phosphorus content of 2.2 mol/mol ofheme a, and a specific activity of 34 ,umol of0 per min per mgof protein (Table 1). The phospholipid retained in the prepa-ration was identified as cardiolipin by thin-layer chromatogra-phy as shown in Fig. 3. From the specific heme a and phos-

E O."80.

°, O."O.

cq 0.

CZ

3 25 mM 100 mM2- NaCl NaCl [.1 I 1I 1

-15

0

10N

20 40 80 100 120Fraction

FIG. 2. Affinity chromatography of cytochrome oxidase on cyto-chrome c-Sepharose. Cytochrome c-Sepharose 4B was prepared ac-cording to the method of Ozawa et al. (11). Ligand concentration was0.32 ,umol of cytochrome c per ml of the gel. The column was equili-brated with 10 mM Tris-HCl, pH 7.5, containing 0.1% Triton X-100.The oxidase (70 mg of protein) purified by the phenyl-Sepharose col-umn (Fig. 1) was diluted 1:10 with the buffer and was applied to thecolumn (1.5 x 11 cm). The column was washed with the buffer and thenwith the buffer containing 25 mM NaCl. Impurities (fractions 6-50)and partially denatured oxidase with the a band at 598 nm (fractions70-80) were washed out. The intact oxidase was eluted from the col-umn with 100 mM NaCl in the buffer. Flow rate, 34 ml/hr; fractionsize, 3.2 ml. The left ordinate is for fractions 1-50; the right ordinateis for fractions 65-129. Three fractions shown by the horizontal barwere combined.

pholipid content, it is concluded that 1 mol of cardiolipin permol of heme a binds tightly to the enzyme.

Subunit Composition and Molecular Weight of the Oxidase.To analyze the subunit composition, the oxidase samples pu-rified by the hydrophobic interaction chromatography (Fig. 1)and by the affinity chromatography (Fig. 2) were tested in twodifferent NaDodSOdurea gel electrophoresis systems (13, 14).The densitometric tracings of gel profiles (Fig. 4) indicate thatthe final preparation contains seven distinct subunits separableby using the gel system described by Merle and Kadenbach(13). Their system showed better resolution of the oxidase sub-units than did the other NaDodSO4 gel systems. The oxidasepreparation ofTzagoloffand MacLennan (5) showed seven sub-units in the ordinary NaDodSO4 gel systems and 9 or 10 sub-units in the Merle-Kadenbach system. Comparison of Fig. 4A and B indicates that two polypeptides retained in the oxidasepreparation after the hydrophobic interaction chromatography

A

'lC

Pe

_ Sp4 0

Pi

FIG. *3. Phospholipids in the crude oxidase and in the phospholipid-depleted cytochrome oxidase. (A) Chromatogram of the phospholipidsextracted from cytochrome oxidase prepared according to the methodof Fowler et al. (4). (B) Two-dimensional thin-layer chromatogram ofthe phospholipids extracted from the oxidase sample that was depletedof phospholipids on the hydrophobic interaction column. The phos-pholipidswere extracted for4 hr at roomtemperature with chloroform/methanol/2 M KOH, 200:100:3 (vol/vol) and were spotted on a pre-coated thin-layer chromatography plate of silica gel 60 (Merck). Thesolvent system for the first dimension (right to left) was chloroform/methanol/water/28% NH40H, 130:70:8:0.5 (vol/vol), and that forthe second dimension (bottom to top) was chloroform/methanol/ace-tone/acetic acid/water, 100:20:40:20:10 (vol/vol). Spots were visu-alized with iodine vapor. C, cardiolipin; N, neutral mitochondrial phos-pholipids; Pe, phosphatidylethanolamine; Pc, phosphatidylcholine; Pi,phosphatidylinositol; Sp, sphingomyelin; and 0, origin. The arrow in-dicates the solvent front.

Proc. Natl. Acad. Sci. USA 79 (1982)

v

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Proc. Natl. Acad. Sci. USA 79 (1982) 7177

VI

Migration

E B

IV

VI~

VII

Migration

FIG. 4. Densitometric tracings of gel profile of cytochrome oxi-dase, Coomassie blue stain. Cytochrome oxidase (30 pg of protein) pu-rified either by the hydrophobic interaction chromatography or by theaffinity chromatography was analyzed by polyacrylamide gel electro-phoresis in the presence of 3.6M urea, 0.1% NaDodSO4, and 13% (vol/vol) glycerol (13). (A) The oxidase sample from the chromatogramshown in Fig. 1. (B) The oxidase sample from the chromatogram shownin Fig. 2.

(Fig. 1) are eliminated by the affinity chromatography (Fig. 2).The molecular weights for subunits I, II, III, IV, V, VI, and VIIwere 46,000, 24,300, 22,800, 15,100, 9,800, 8,300, and 4,800,respectively. Essentially the same results (15) were obtained byusing the gel system described by Downer et al. (14). Althoughsubunit III stains less intensely with Coomassie blue than theothers do, its presence in the oxidase is clear. By summing upthe molecular weights ofthe seven subunits, a molecular weightof 131,000 was obtained, on the assumption that there is oneof each of the subunits in the oxidase.The molecular weight of the oxidase calculated from its sub-

unit composition (131,000) (Fig. 4) is comparable to the valuecalculated from the specific heme a content (132,000) (Table 1)and that calculated from the specific copper content (124,000)(Table 1). From these results, it seems reasonable to concludethat the oxidase consists of seven subunits and has a molecularweight of about 130,000. Verheul et at (16) have proposed a 12-subunit picture for the oxidase, with a minimal molecularweight of 172,000. This discrepancy might indicate that thepresent preparation is highly purified, and contaminating poly-peptides, which are not essential with respect to binding ofprosthetic groups and catalytic function, are eliminated from thepreparation.

Crystallization of Cytochrome Oxidase. Crystallization ofcytochrome oxidase itself was carried out with the oxidase pu-rified by affinity chromatography (Fig. 2) and then slowly di-alyzed to remove detergent. During the dialysis of the oxidaseagainst 10 mM Tris-HCl buffer, pH 8.5, green needle-shapedcrystals began to form on the seventh day, and they increasedin size and in number on further dialysis. As shown in Fig. 5A,crystals of the oxidase (100 um in average length and 5 p.m inwidth) were obtained, with some smaller crystals and a smallamount of amorphous precipitate. The crystals showed strongpolarization of light when observed between crossed polarizers(Fig. 5B).

Because the final preparation ofthe oxidase used for the crys-tallization retained 2 mol of tightly bound cardiolipin per molof enzyme (Fig. 3, Table 1), we intended to assess the degreeof ordered array of molecules in the crystal by electron diffrac-tion study according to Unwin and Henderson (17). Because theunstained crystals had thickness (0.5-5 pm) far beyond the suit-able thickness (400-600 A) for 100-kV electrons (17), electrondiffiaction was carried out at 1,000 kV with a Hitachi HU-1000D ultra-high-voltage electron microscope, using a field-limit-ing aperture of 50-pym diameter. When 1,000-kV electrons areused, the deterioration of molecules caused by radiation can becut down to one third of that by 100-kV electrons. The imageof a negatively stained crystal on a carbon-coated grid is shownin Fig. 5C. The electron diffraction pattern from this (hk0) planeof the crystal is shown in Fig. 5D. The reciprocal lattice shownby the diffraction spots corresponds to the image. The anglebetween the a* and b* axes, which are perpendicular to the band a axes, respectively, is 54°. Thus, it was concluded that theangle between the a and b axes-namely y-of the crystal is126°. In Fig. 5D, diffraction spacings along the a* axis could bemeasured. The longitudinal spacing of 7.4 A is located on the(h,0,0) line, the spacings of 10.6 and 24.7 A on the (h,20,0) line,those of 12.3 and 18.5 A on the (h,40,0) line, and those of 9.25and 37A on the (h,60,0) line. Thus, the minimum repeating unithaving a 74-A dimension along the a* axis is expected in thereciprocal lattice-namely a = a*/sin y = 92 A. Twentieth-order diffractions are located on b* axis with the spacing of 3.4A-namely b* = 68 A and b = 84 A. Thus, it can be presumedthat the lattice with a = 92 A, b = 84 A, and y = 1260 existsin the crystal. Fig. 5E demonstrates the arrays of individualmolecules, shown by arrows, and the dimensions and the anglesof the crystal lattice. The electron diffraction pattern from the(Okl) plane of the crystal is shown in Fig. 5F. The outermostreflections corresponded to 1.2-A resolution as calculated bythe Bragg equation. In Fig. 5F, tenth-order diffractions withthe spacing of 10.3 A are located on c* axis. From the diffractionpattern, the minimum repeating unit in the (Okl) plane of thecrystal was expected to have the dimensions b = 84 A, c = 103A, and a = 900. The corresponding image of the lattice satisfiesthe expected values as shown in Fig. 5G. Microcrystals with athickness below 1,000 A were prepared by brief washing of thecrystals (Fig. 5A) with Triton X-100. An electron diffractionpattern from microcrystals is shown in Fig. 5H.From these results, it was concluded that the crystal has pg

symmetry with the crystallographic space group P21, the di-mensions of the monoclinic unit cell being a = 92 A, b = 84A, and c = 103 A, with a = f3 = 900 and y = 1260.

DISCUSSIONSince we reported the crystallization of intrinsic membraneproteins, the cytochrome c oxidase-cytochrome c complex (2,3), and the cytochrome bc,-cytochrome c complex (12), otherintrinsic membrane proteins such as bacteriorhodopsin (18) and

Biochemistry: Ozawa et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

7178 Biochemistry: Ozawa et al.

A.A

..V~ ~ ~ 4

Proc. Natd. Acad. Sci. USA 79 (1982)

C

a

1 000A

FIG. 5. Crystals of cytochrome oxidase and their electron diffraction patterns. (A) Crystals of the oxidase under ordinary light. (x 100.) (B) Sameas A, except under polarized light. (C) Crystals of the oxidase were fixed with 1% glutaraldehyde overnight then negatively stained with 2% am-monium molybdate, pH 7.4. Photographs were taken at different angles with a Hitachi H-800 microscope. One crystal is shown. (D) A stained singlecrystal on a carbon-coated grid was observed with 1,000-kV electrons in a Hitachi HU-1000D ultra-high-voltage electron microscope with a field-limiting aperture of 50-,um diameter. An electron diffraction pattern from the (hO) plane of the crystal was photographed under an illuminationlevel of -0.25 electrons/A2 per sec. 1 cm = 0.43 A-1. (E) An image in the (hk) plane of the crystal was taken as described for C. Arrows indicateindividual oxidase molecules. (Inset) Negatively stained ferritin molecules (dark core is 55 A in diameter) under the same conditions. (F) Electrondiffraction pattern from the (Okl) plane of the crystal. The outermost reflections correspond to 1.2-A resolution. (G) Image of the (Okl) plane of thecrystal. Arrows indicate individual oxidase molecules. (H) Electron diffraction pattern of microcrystals. The microcrystals were prepared by briefwashing of the crystals with Triton X-100.

the pore-forming protein from Escherichia coli (19) were crys-tallized, probably in the form of the protein-detergent complex(18). X-ray diffraction patterns of these membrane proteinswere recorded to a resolution of 8 A in the case of bacteriorho-

dopsin and of 3.8 A in the case of the pore-forming protein. Bycombination of low-dose electron microscopy and electron dif-fraction (17), Michel et al. (20) have found a limit of 6.5-A res-olution in the case ofthe purple membrane. The resolution lim-

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

Proc. Natl. Acad. Sci. USA 79 (1982) 7179

its mentioned above may be due to the detergent molecule orphospholipids present in the crystal or in the membrane. Thesemolecules may allow movements of the protein molecules thatdecrease the resolution. However, a resolution higher than 3A is required to identify the positions of amino acid residues ina protein structure. For the understanding of biological mem-brane phenomena such as proposed proton translocation in cy-tochrome c oxidase, three-dimensional crystals with highly or-dered array of the molecules are essential for the resolution ofthe protein structure in atomic and chemical detail. For thispurpose, we purified cytochrome c oxidase extensively by thecombination of hydrophobic interaction chromatography (Fig.1) and affinity chromatography (Fig. 2). These procedures allowthe depletion of phospholipids from the oxidase (Table 1). Only1 mol of cardiolipin (Fig. 3) per mol of heme a remained in thefinal sample (Table 1). After the oxidase preparation was ex-posed to affinity chromatography on cytochrome c-Sepharose,the intact oxidase was identified in the eluate from the phenyl-Sepharose column (Fig. 2). Comparison ofthe polypeptide com-position (Fig. 4) and the specific heme and copper content ofthe oxidase before and after the affinity chromatography (Table1) indicates that some contaminating polypeptides and partiallydenatured oxidase cannot bind to the specific affinity ligand,cytochrome c. The molecular weight of the oxidase was esti-mated to be 130,000 from the subunit composition (Fig. 4) andthe specific heme a and copper contents (Table 1). After thepurification the oxidase itself can be crystallized by slow re-moval ofthe detergent (Fig. 5A and B). An electron micrographof negatively stained crystals is shown in Fig. 5C. Electronmicroscope and diffraction results were obtained at differentspecimen tiltings. Electron diffraction analysis ofour crystal wascarried out by following the general procedure of Unwin andHenderson (17), except using 1,000-kV electrons (Fig. 5 D, F,and H). From the electron micrographs (Fig. 5 E and G) anddiffraction patterns, it was concluded that the crystal is mono-clinic in the space group of P21 with cell dimensions a = 92A, b = 84 A, and c = 103 A, with a = (3 900, y = 1260. Thesevalues yield a unit cell volume [abc(sin y)] of 6.44 X 105 A3.Assuming two oxidase molecules (Mr = 130,000) in a unit cell,the unit cell volume also yields a reasonable value (2.48 A3jdalton) for the volume-to-mass ratio (21).

Henderson et aL (22) have noted that "two-dimensional crys-tals" of the oxidase reported by Wakabayashi et aL (23) and Van-derkooi et aL (24) show characteristic pgg symmetry and a"green membrane fraction" of the oxidase reported by Seki etaL (25) shows pg symmetry. Fuller et aL (26) also obtained "two-dimensional crystals" showing either pgg symmetry or pg sym-metry. In the case of three-dimensional crystal reported here,the oxidase molecules show pg symmetry. Thus, oxidase mol-ecules seem to have tendency to array themselves in either pggor pg symmetry, and the depletion of phospholipids from theoxidase may allow the shift in the molecules from two-dimen-sional to three-dimensional arrays. The Bragg diffractions ex-tended to spacing of 1.2 A (Fig. 5F), indicating that the oxidasemolecules exist in the crystals in a highly ordered three-di-mensional array, even in the presence of cardiolipin. In fact,negatively stained images ofthe crystals (Fig. 5E and G) supportthis notion.

These observations indicate that 2 mol ofcardiolipin per molofthe oxidase maintain a high order ofstructural integrity, prob-ably with a specific association ofsome tenacity with the proteinsuggested by Fry et al (27), or with a specific role as a hoopproposed by Ozawa (28).The authors express their thanks to Prof. N. Tanaka (Protein Research

Institute, Osaka University) for his advice about our analysis, and Mr.C. Morita and Mr. S. Arai in the Optical Laboratory (University ofNagoya) for their technical assistance with the electron diffraction study.This work was supported in part by Grants-in-Aid for Special ProjectResearch (311909) from the Ministry of Education, Science and Cul-ture, Japan.1. Warburg, 0. (1925) Ber. Dtsch. Chem. Ges. 58, 1001-1011.2. Tanaka, M., Suzuki, H. & Ozawa, T. (1980) Biochim. Biophys.

Acta 612, 295-298.3. Ozawa, T., Suzuki, H. & Tanaka, M. (1980) Proc. Nati Acad. Sci.

USA 77, 928-930.4. Fowler, L. R., Richardson, S. H. & Hatefi, Y. (1962) Biochim.

Biophys. Acta 64, 170-173.5. Tzagoloff, A. & MacLennan, D. H. (1965) Biochim. Biophys.

Acta 99, 476-485.6. Ozawa, T., Tada, M. & Suzuki, H. (1979) in Cytochrome Oxi-

dase, eds. King, T. E., Orii, Y., Chance, B. & Okunuki, K. (El-sevier/North-Holland, Amsterdam), pp. 39-52.

7. Gornall, A. G., Bardawill, C. J. & David, M. M. (1949) J. BiolChem. 177, 751-766.

8. Chen, P. S., Jr., Toribara, T. Y. & Warner, H. (1956) AnalChem. 28, 1756-1758.

9. Vik, S. B. & Capaldi, R. A. (1980) Biochem. Biophys. Res. Com-mun. 94, 348-354.

10. Fry, M. & Green, D. E. (1980) Biochem. Biophys. Res. Commun.93, 1238-1246.

11. Ozawa, T., Okumura, M. & Yagi, K. (1975) Biochem. Biophys.Res. Commun. 65, 1102-1107.

12. Ozawa, T., Tanaka, M. & Shimomura, Y. (1980) Proc. Natl Acad.Sci. USA 77, 5084-5086.

13. Merle, P. & Kadenbach, B. (1980) Eur. J. Biochem. 105, 499-507.

14. Downer, N. W., Robinson, N. C. & Capaldi, R. A. (1976) Bio-chemistry 15, 2930-2936.

15. Tanaka, M. & Ozawa, T. (1982) Biochem. Int., in press.16. Verheul, F. E. A. M., Draijer, J. W., Dentener, I. K. & Muij-

sers, A. 0. (1981) Eur. J. Biochem. 119, 401-408.17. Unwin, P. N. T. & Henderson, R. (1975) J. Mol Biol 94, 425-

440.18. Michel, H. & Oesterhelt, D. (1980) Proc. Nati. Acad. Sci. USA 77,

1283-1285.19. Garavito, R. M. & Rosenbusch (1980) J. Cell Biol. 86, 327-329.20. Michel, H., Oesterhelt, D. & Henderson, R. (1980) Proc. Nati

Acad. Sci. USA 77, 338-342.21. Mathews, B. W. (1968) J. Mol. Biol 33, 491-497.22. Henderson, R., Capaldi, R. A. & Leigh, J. S. (1977)J. Mol Biol

112, 631-648.23. Wakabayashi, T., Senior, A. E., Hatase, O., Hayashi, H. &

Green, D. E. (1972) Bioenergetics 3, 339-344.24. Vanderkooi, G., Senior, A. E., Capaldi, R. A. & Hayashi, H.

(1972) Biochim. Biophys. Acta 274, 38-48.25. Seki, S., Hayashi, H. & Oda, T. (1970) Arch. Biochem. Biophys.

138, 110-121.26. Fuller, S. D., Capaldi, R. A. & Henderson, R. (1979)J. Mol Biol.

134, 305-327.27. Fry, M., Blondin, G. A. & Green, D. E. (1980) J. Biol Chem.

255, 9967-9999.28. Ozawa, T. (1982) in Transport and Bioenergetics in Biomem-

branes, ed. Sato, R. (Plenum, New York), pp. 3-38.

Biochemistry: Ozawa et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0