Biochem J. (1980) 191, 429-436Printed in Great Britain

Identification of the subunits of bovine heart mitochondrial NADHdehydrogenase that are exposed to the phospholipid bilayer by photo-

labelling with 5-iodonaphth- l-yl azide

Fergus G. P. EARLEY* and C. Ian RAGANDepartment ofBiochemistry, University ofSouthampton, Southampton S09 3TU, U.K.

(Received 20 March 1980/Accepted 14 July 1980)

Mitochondrial NADH dehydrogenase may be isolated from bovine heart as alipoprotein complex (Complex I or NADH-ubiquinone oxidoreductase). Polypeptidesubunits that are exposed to the hydrophobic region of the phospholipid bilayer wereidentified by photolabelling with the hydrophobic probe, 5-['25lliodonaphth-1-yl azide.Chaotropic resolution of the labelled enzyme showed that the hydrophilic flavoproteinand iron-protein fragments of the enzyme were not in contact with the phospholipidbilayer. When Complex I that had been partially depleted of phospholipids wasphotolabelled, incorporation of radioactivity into certain polypeptides was increased,indicating either conformational changes in the protein or preferential association ofthese polypeptides with residual cardiolipin. A model of NADH dehydrogenasestructure is proposed on the basis of these results and those obtained with hydrophilicprobes by Smith & Ragan [(1980) Biochem. J. 185, 315-326].

Mitochondrial NADH dehydrogenase is an en-zyme of enormous complexity. It may be isolatedfrom bovine heart mitochondria as a detergent-soluble lipoprotein that catalyses the reduction ofubiquinone analogues by NADH (Complex I in thenomenclature of Hatefi et al., 1962a). This pre-paration consists of at least 26 different poly-peptides (Heron et al., 1979b). When treated withchaotropic agents such as NaClO4 the enzyme maybe resolved into three fractions: an insoluble residue,a water-soluble iron-sulphur flavoprotein (referredto below as the 'flavoprotein fragment') and awater-soluble iron-sulphur protein (referred tobelow as the 'iron-protein fragment') (Hatefi &Stempel, 1967). The flavoprotein fragmentcomprises three polypeptides in 1: 1: 1 molar pro-portions and contains all the FMN of the originalenzyme (Galante & Hatefi, 1979). The iron-proteinfragment consists of eight polypeptides (Heron et al.,1979b) and contains no FMN. The structuralorganization of the enzyme both in isolation and inthe mitochondrial inner membrane has been studiedby using two hydrophilic membrane-impermeableprotein labels, diazobenzene[ "Slsulphonate and thelactoperoxidase-catalysed incorporation of 'III(Ragan, 1976; Smith & Ragan, 1980). On the basis

* To whom correspondence should be addressed.

Vol. 191

of accessibility to these two labels, the subunits ofNADH dehydrogenase were divided into fivegroups: (a) those that are buried in the interior of theenzyme; (b) those that are accessible to labelling inthe isolated enzyme but not when the enzyme is inthe mitochondrial inner membrane; (c) those that are

exposed on the matrix side of the membrane; (d)those that are exposed on the cytoplasmic side of themembrane; (e) those that are transmembranous. Thesubunits of the flavoprotein fragment and all butthree subunits of the iron-protein fragment belong tothe first group, since they are inaccessible to bothlabels. Two subunits of the iron-protein fragmentare transmembranous. Many subunits fall into thesecond group, i.e. they can only be labelled in theisolated enzyme. This suggests (Smith & Ragan,1980) that, in the membrane, these polypeptides are

adjacent to the hydrophobic region of the phospho-lipid bilayer. In the isolated enzyme, which hasperhaps only a single shell of phospholipid around it,these subunits are more exposed to the aqueousphase and become accessible to membrane-impermeable reagents. This group of subunits are allfound in the insoluble residue from chaotropicresolution and appear to be very hydrophobic,consistent with their proposed location adjacent tothe fatty acid region of the bilayer.

In the present paper, evidence has been sought in

0306-3275/80/1 10429-08$01.50/1 (© 1980 The Biochemical Society

429

F. G. P. Earley and C. I. Ragan

support of these proposals. We have used as a probe5-[ '25lliodonaphth-1-yl azide, which is reputed tolabel protein subunits embedded in the hydrophobicregion of the phospholipid bilayer (Bercovici &Gitler, 1978). This compound partitions preferen-tially into the hydrophobic region of membranes,and on illumination it gives a highly reactive nitrene,which inserts into adjacent protein or lipidmolecules. Bayley & Knowles (1978) have objectedto the use of nitrenes for this kind of study on thegrounds that the nitrene is more polar than the azideand could, within the lifetime of the molecule,migrate to the polar surface of the membrane beforereaction. Thus it would not act as a probe forhydrophobic regions of the membrane. It is perhapsunfortunate that the model compound used by theseauthors, phenyl azide, is at least two orders ofmagnitude more soluble in aqueous solutions than is5-iodonaphth-1-yl azide. Moreover, the results ob-tained by Bercovici & Gitler (1978) with sarco-plasmic reticulum and erythrocyte membranes, thework of Cerletti & Schatz (1979) with cytochromeoxidase and the results described in the presentpaper are consistent with the idea that 5-iodo-naphth- 1-yl azide is a hydrophobic label.

Materials and methods

Biological preparationsComplex I (EC 1.6.99.3) was purified from bovine

heart mitochondria by the method of Hatefi &Rieske (1967). Chaotropic resolution of Complex Iwas performed with NaClO4 as described by Smith& Ragan (1980). The flavoprotein and iron-proteinfragments were purified by (NH4)2SO4 fractionationas described by Hatefi & Stempel (1969).

Complex I was depleted of phospholipids bycholate and (NH4)2SO4 treatment by the method ofHeron et al. (1977). This treatment decreased thephospholipid content from 0.20-0.23 to 0.04-0.05,umol of phospholipid P/mg of protein. Theresidual lipid was nearly all cardiolipin (Heron et al.,1977). Phospholipid-enriched Complex I was pre-pared by addition of purified soya-beanphosphatidylcholine (Ragan & Racker, 1973). Thephosphatidylcholine was dispersed by ultrasonicirradiation in 20 mM-potassium phosphate buffer,pH 8, containing 2% (w/v) sodium cholate to give afinal concentration of 100mM-phospholipid P. Com-plex I dissolved in 0.67M-sucrose/50mM-Tris/HClbuffer, pH 8, at a concentration of 20mg ofprotein/ml was mixed with the phosphatidylcholinedispersion to raise the phospholipid content from0.23 to 0.69,pmol of phospholipid P/mg of protein.

Endogenous phosphatidylcholine andphosphatidylethanolamine in Complex I were re-placed by 1,2-ditetradecanoyl-sn-glycero-3-phospho-choline as described by Heron et al. (1979a).

Cholate was removed from the lipid-replaced en-zyme by dialysis overnight against several hundredvolumes of 0.67M-sucrose/50mM-Tris/HCI buffer,pH 8.0, at 4°C.

Synthesis of5-[ '25I iodonaphth-J-yl azide5-f 125lllodonaphth-1-yl azide was synthesized by

method 2 of Bercovici & Gitler (1978) as modifiedby Cerletti & Schatz (1979). Carrier-free Na125Ifrom The Radiochemical Centre (Amersham,Bucks., U.K.) was diluted with Nal to a specificradioactivity of 8 Ci/mmol before use.

Photolabelling ofComplex IComplex I (4mg of protein/ml of 0.67 M-sucrose/

50 mM-Tris/HCl buffer, pH 8.0) was mixed with5- '25lliodonaphth-1-yl azide in 1.5 ml silica spectro-photometer cells of 1 cm light-path. The radioactivelabel was added as an ethanolic solution such thatthe final ethanol concentration did not exceed 1%(v/v). The amounts of radioactivity added are givenin the Figure legends. After incubation for 10min inthe dark at room temperature, the sample wasexposed to a 125W high-pressure mercury-vapourlamp (Thorn Electrical, Southampton, U.K.) at adistance of 30cm for 80s. The sample was stirredevery few seconds during exposure.

Determination ofprotein-bound radioactivityAfter the photolabelling, portions (20,ul) contain-

ing 80,ug of protein were mixed with 0.4 ml ofice-cold 10% (w/v) trichloroacetic acid. Precipitatedprotein was collected on glass-fibre filters (WhatmanGF/C) by suction and washed successively with 2 mlvolumes of 5% (w/v) trichloroacetic acid, acetone,ethanol/ether (1: 1, v/v) and ether. After the filtershad dried, their radioactivities were counted in aBeckman Biogamma counter.

Polyacrylamide-gel electrophoresisAfter the photolabelling, portions (0.5 ml) contain-

ing 2mg of protein were mixed with 4.5 ml ofice-cold acetone, and the precipitated protein wascollected by centrifugation in a bench centrifuge.The pellets were dried and redissolved in 20mM-sodium phosphate buffer, pH 8.0, containing 2%(w/v) sodium dodecyl sulphate and 1% (v/v)2-mercaptoethanol by heating at 1000C for 2min.Acetone-precipitation largely removed a band ofhighly radioactive material that ran with the trackingdye on electrophoresis. This material was probablyphotolabelled phospholipids. The dissociated proteinsamples (100,ug per gel) were analysed by themethod of Weber & Osborn (1969) on cylindricalpolyacrylamide gels (12cm long x 0.6cm internaldiam.) containing 12.5% (w/v) acrylamide and0.34% bisacrylamide. Samples were also analysedby discontinuous slab-gel electrophoresis after

1980

430

Photolabelling ofNADH dehydrogenase

dialysis overnight against 100vol. of 25mM-Tris/0.192 M-glycine containing 1% (w/v) sodiumdodecyl sulphate and 1% (v/v) 2-mercaptoethanol.Running gels (12cm long x 1mm thick) of the abovecomposition were overlaid by a stacking gel (1cmlong x 1 mm thick) containing 3% (w/v) acrylamideand 0.24% bisacrylamide. Samples containing 50,ugof protein were applied to each track. The buffersystem used was that of Laemmli (1970).

Gels were stained with Brilliant Blue R (Sigma,Poole, Dorset, U.K.) and destained as described byWeber & Osborn (1969).

Determination ofradioactivity in gelsCylindrical gels were sliced into 1 mm-thick

sections with a Mickle gel slicer (Mickle EngineeringCo., Gomshall, Surrey, U.K.). The radioactivities ofthe slices were then counted directly in a BeckmanBiogamma counter.

Slab gels were dried down on to filter paper forradioautography. Kodak X-OMAT H film wasexposed to the dried gel at -20°C for 7 days anddeveloped in accordance with the manufacturers'instructions.

AssaysProtein was measured by the method of Lowry et

al. (1951), with bovine serum albumin (fraction Vfrom Sigma) as a standard. NADH-ubiquinone-1oxidoreductase was assayed as described by Ragan(1976).

Results

Time course of incorporation of S-iodonaphth-J-ylazide

Under the conditions of illumination described inthe Materials and methods section, photolabelling ofComplex I was nearly maximal after 2min (Fig. 1),at which time approx. 9% of the added radioactivityhad been incorporated into protein. Also shown inFig. 1 is the inactivation of the enzyme byirradiation. The extent of inactivation was the samein the presence and in the absence of 5-iodo-naphth- 1-yl azide. To minimize labelling ofdenatured enzyme, which might have had an alteredconformation, we adopted a standard illuminationperiod of 80 s.The experiment of Fig. 1 was performed after a

preincubation period of 10min in the dark to allowthe reagent to equilibrate with the enzyme. A similarprocedure was adopted by Bercovici & Gitler (1978)and Cerletti & Schatz (1979). In practice, we foundthat equilibration was very rapid, and incorporationby a 10s illumination period was the same forpreincubation periods ranging from 15s to 5min(results not shown).

Vol. 191

00 25

-

q 20C)

;*^- *Ev 1 5

C)0.-0

5x

V)

100

E;^

80 _

C)

60 *- ;3S. ._:0

a.- C40 0 c

20 ra)C)a)

V I ' I I J O0 20 40 60 80 100 120

Illumination time (s)Fig. 1. Time course ofphotolabelling ofComplex IComplex I was photolabelled with 5-[l251iodo-naphth-1-yl azide (2.6 x 107c.p.m./mg of protein) asdescribed in the Materials and methods section.Portions were removed at intervals for measure-ment of protein-bound radioactivity and NADH-ubiquinone- 1 oxidoreductase. No radioactivity wasfound in a parallel sample incubated in the dark. 0,Radioactivity; A, enzyme activity.

Labelling ofNADH dehydrogenase polypeptidesFig. 2(b) shows the distribution of radioactivity

between Complex-I polypeptides (Fig. 2a) afterphotolabelling by 5-iodonaphth-1-yl azide. Most ofthe labelling occurred in the smaller polypeptides ofmol.wt. less than 30000. Clearly the labelling is quiteselective. This selectivity was largely abolished whenthe enzyme was dissociated with sodium dodecylsulphate before the photolabelling (Fig. 2c). Thus,when each subunit is surrounded by a hydrophobiclayer of sodium dodecyl sulphate molecules, it canbe readily labelled with 5-iodonaphth- 1-yl azide.Similar findings were reported by Cerletti & Schatz(1979) for cytochrome oxidase.As isolated, Complex I forms optically clear

solutions at high concentration because of cholateand (NH4)2SO4 present from the last step ofpurification. On dilution into detergent-free media,the enzyme becomes insoluble and forms mem-branous aggregates. The photolabelling experimentof Fig. 2 was performed at a protein concentrationof 4 mg/ml, which is high enough to keep aconsiderable proportion of the enzyme in mono-meric form. To test whether the labelling pattern wasdependent on the state of aggregation of the enzymeor the presence of detergent, cholate and (NH4)2S04were removed by dialysis before the photolabelling.Small effects were observed in the high-molecular-weight region of the gel (Fig. 3), where the degree oflabelling of the subunits of mol.wts. 75 000 to 39 000was increased. Since these subunits were labelled toonly a small extent anyway, it may be that theirdegree of labelling- was more sensitive to variationsin the partition of 5-iodonaphth-1-yl azide betweenthe lipoprotein and cholate micelles. In the experi-

431

F. G. P. Earley and C. I. Ragan

V QIV vvv p

4 X: 11

I

_ 1 8I4a aj

0 20 0-0-0-'0

-4

5 ad

a a20 0 60 of pt

naht-aylaad

described in the Materials and methods section andanalysed by polyacrylamide-gel electrophoresis withthe buffer system of Weber & Osborn ( 1969).Radioactivity was measured in I mm-thick slices ofthe gel. (a) shows the densitometer scan of thestained gel; (b) gives the radioactivity of the gelslices and (c) the radioactivity of a sample ofComplex I that was labelled in the presence of 2%(w/v) sodium dodecyl sulphate. The polypeptides areindicated by their molecular weights (in thousands).

ments of Figs. 2 and 3, a peak of radioactivity was

found at the dye front in a region of the gel thatcontained no stained protein bands. Since photolysisproducts of the reaction are removed by the gelstaining and destaining procedure, it is likely thatthis peak is due to labelled phospholipids, which

12

(a) 18

10

I.

9.'~~ ~ ~ ~ ~~2.

4 8~~~~~~~~;

0.

i 7549 26 rc 23.

[<l >R 2 ala

.a . . .

o bi .toIj9. 1

0 20 40 60 80 1Slice no. or migration distance from origin (mm)

100

Fig. 3. Effects ofdialysis on the labelling profileComplex I (a)or Complex I dialysed for 16h against500vol. of 0.67 M-sucrose/50mM-Tris/HCl buffer,pH8.0, at 40C (b) was labelled with 5-f125IIiodo-naphth-1-yl azide (2.4 x 107c.p.m./mg of protein)and analysed by the Weber & Osborn (1969)electrophoretic procedure. The positions of the poly-peptides are indicated by the molecular weights (inthousands).

would not be completely extracted by 90% (v/v)acetone (Bercovici & Gitler, 1978).

Distribution of label between the products ofchaotropic resolution

In the experiment of Fig. 4, photolabelled Com-plex I was treated with NaClO4, and the iron-protein and flavoprotein fragments were isolated by(NH4)2SO4 fractionation. None of the constituentpolypeptides of either of these fragments waslabelled, and all radioactivity remained in theinsoluble hydrophobic residue.

Effects of hydrophobic ligands of Complex I on thelabelling pattern

Inhibitors of NADH-ubiquinone oxidoreductaseactivity that probably bind to hydrophobic regions

1980

432

Photolabelling ofNADH dehydrogenase

SD:S (a)

Origin

42 =q

2*22

D:ye front--- f"''-

}*....::.

.:::::: :.:

..............

*: :.:::

.::...i: .::.: :.:'°:.: .:;

_E

'

,.:. :<::;:.:............. ...:::

.: ..

..: ::::.::" .o ..

:2:Q::. :.:

s;SSi

(c).:,;o

4-49-42

U-

2

Fig. 5. Effects of lipid/protein ratio on the pattern oflabelling

Complex I was depleted of lipids or supplementedwith lipid as described in the Materials and methodssection. All samples were dialysed before photo-labelling with 5-1 125l1iodonaphth- 1-yl azide(2.5 x 107c.p.m./mg of protein). Labelled proteinwas analysed by discontinuous polyacrylamide-gelelectrophoresis and radioautography. The Figureshows radioautography of (a) Complex I, (b)lipid-depleted Complex I and (c) lipid-supplementedComplex I. A sample of Complex I that had been-treated with sodium dodecyl sulphate before thelabelling is included as a marker (SDS). The majorlabelled polypeptides in native Complex I areindicated by their molecular weights (in thousands).Those polypeptides whose labelling was increased inlipid-depleted Complex I are indicated on the right oftrack (c). Molecular weights were taken fromCrowder & Ragan (1977).

0 20 40 60 80 100

Slice no. or migration distance from origin (mm)

Fig. 4. Chaotropic resolution ofphotolabelled Complex I

Complex I (20mg of protein in 2ml of 0.67M-sucrose/50 mM-Tris/HCI buffer, pH 8.0) was il-luminated for 80s in the presence of 5-1 1251-

iodonaphth-1-yl azide (1.27x 108c.p.m.) as other-wise described in the Materials and methods section.A portion (100,ul) was removed for analysis, and tothe rest 4 M-NaCIO4 was added to give a finalconcentration of 0.5 M. Resolution was carried out at350C for 10min as described by Smith & Ragan(1980). The iron-protein and flavoprotein frag-ments were purified from the supernatant obtainedby centrifugation of the resolved Complex I asdescribed by Hatefi & Stempel (1969). Samples were

analysed by polyacrylamide-gel electrophoresis withthe Weber & Osborn (1969) buffer system. (a)Complex I (100,ug of protein); (b) iron-proteinfragment (derived from lO0,ug of Complex-I protein);(c) flavoprotein fragment (derived from 200,g of

of the enzyme are rotenone (Lindahl & Oberg, 1961)and diphenyleneiodonium (Ragan & Bloxham, 1977).The distribution of radioactivity after the photo-labelling was unaffected by the presence of either ofthese compounds at concentrations of 1 nmol/mg ofprotein and 100nmol/mg of protein respectively(results not shown).

Effects of varying the lipid/protein ratio on thedistribution oflabel

Complex I contains phosphatidylcholine,phosphatidylethanolamine and cardiolipin. The first

Complex-I protein). , Densitometer scans ofthe stained gels; 0-0, radioactivity. The radio-activity scale has been increased for the fragments.The polypeptides are indicated by their molecularweights (in thousands).

Vol. 191

V 1.0

CUW._.20.5

x

00

1.0

0.5

433

F. G. P. Earley and C. I. Ragan

3.0 r

(a)

2.5 _

2.0

E

C3 1.5

C>

o 1.0._

Cd

x

0.5

0

0.5

42

-!l if

II

I

I49

,1I j.EIv

.,

I, I

It20.5, 1,I2 .1

of

;i23.50,f. I II .

26 i' ;.W j

t;-

J~~~~~

i11II

iI

.h.i

.-

(b)49

It,pa4

'~~~~~~~~~~t~~~~~~~~~~~~~~~~~~~0 20 40 60 80 100

Slice no. or migration distance from origin (mm)Fig. 6. Chaotropic resolution of lipid-depleted Complex I

Dialysed lipid-depleted Complex I (5.5 mg of proteinin 0.55 ml of 0.67 M-sucrose/SOmM-Tris/HCl buffer,pH 8.0) was photolabelled with 5-[I 251liodonaphth-1-yl azide (1.27 x 108c.p.m.). The enzyme wasresolved with 0.5 M-NaClO4 and centrifuged toseparate the soluble and insoluble fractions. (a)Insoluble fraction (derived from 100,ug of Complex-I protein); (b) soluble fraction (derived from 100,ugof Complex-I protein). Analysis was by polyacryl-amide-gel elctrophoresis with the Weber & Osborn(1969) system. The positions of the polypeptidesare indicated by the molecular weights (inthousands).

two phospholipids can be almost completelyremoved by (NH4)2SO4 precipitation of the enzymein the presence of high concentrations of cholate(Heron et al., 1977). The cardiolipin content(approx. 20% of the total phospholipid P) isunaffected.When lipid-depleted Complex I was photolabelled,

several differences in the labelling pattern wereapparent (Fig. 5). Most polypeptides were labelled toexactly the same extent as they were in undepletedComplex I (Fig. 5a). However, three subunits ofmol.wts. 49000, 42000 and 20000 were labelled toa much greater extent in lipid-depleted Complex I(Fig. Sb). The converse experiment, in which thelipid content of Complex I was increased, revealedno changes in the labelling pattern (Fig. 5c).

The increased labelling of the 49 000-mol.wt.polypeptide prompted us to repeat the chaotropic

resolution experiment. As shown in Fig. 6, some ofthe radioactive label in this position is solubilized byNaClO4, and therefore is associated with the49 000-mol.wt. polypeptide of the iron-protein frac-tion. The remaining label in this position, which isnot solubilized, is probably associated with one ofthe high-molecular-weight subunits of ubiquinol-cytochrome c oxidoreductase (Complex III; Hatefiet al., 1962b). This enzyme is always present insmall amounts in Complex-I preparations, and itslargest subunit, of mol.wt. 50000 (Gellefors &Nelson, 1975), is a major identifiable contaminentin Complex I (Heron et al., 1979b).The effects on the labelling pattern of lipid-

depletion are rather hard to interpret, since it is notcertain how 5-iodonaphth-1-yl azide acts. Theobservation that the degree of labelling of manyof the subunits is unaltered could be explained ifthe reagent binds directly to hydrophobic regionsof the protein whether they are in contact with thelipid bilayer or not. Bercovici & Gitler (1978)showed that bovine serum albumin can be photo-labelled in this way. However, they regard this as anunlikely mechanism for labelling of membraneproteins, because the rapid and near completepartition of the compound into the lipid phase wouldleave very little over to bind from the aqueous phaseto exposed hydrophobic regions of the protein. Wecould not test this directly by removing all phos-pholipids from Complex I, because this cannot bedone in a non-destructive manner. Assuming thatlabelling does occur exclusively from the lipid phase,then decreasing the lipid/protein ratio would merelyserve to increase the concentration of 5-iodonaphth-1-yl azide in the lipid that remained. If this lipid wererandomly distributed over the hydrophobic surfaceof the enzyme, no change in the labelling patternwould be expected. The increased labelling of certainpolypeptides in the depleted enzyme can be ex-plained in two ways. Firstly, conformational changesin the protein could bring these polypeptides intocloser or more extensive contact with the residuallipid. Alternatively, the distribution of the remainingcardiolipin might not be completely random. Thoseproteins associated with this lipid would thereforebecome more heavily labelled because of the higherconcentration of 5-iodonaphth-1-yl azide in thevicinity. At present we cannot distinguish betweenthese possibilities.An attempt was made to provide evidence that

5-iodonaphth-1-yl azide does label proteins from thelipid phase by using Complex I whose endogenousphosphatidylcholine and phosphatidylethanolaminehad been. replaced with 1,2-ditetradecanoyl-sn-glycero-3-phosphocholine (Heron et al., 1979a).This lipid in the pure state undergoes a phasetransition between the liquid-crystalline and gelstates at 240C (Hinz & Sturtevant, 1972). It was

1980

434

A

Photolabelling of NADH dehydrogenase

hoped that at low temperatures lipid-replaced Com-plex I might show decreased photolabelling, becausepartition into the lipid phase would be slower andless complete. In addition, if protein labellingoccurred after migration of the nitrene to thelipid/water interface (Bayley & Knowles, 1978), thismight be prevented in gel-phase lipid. Unfortunately,we were not able to show any temperature-dependent effects of this type. The pattern oflabelling of lipid-replaced Complex I (under anyconditions) was very similar to that of nativeComplex I, except that there was increased labellingof certain high-molecular-weight polypeptides,reminiscent of the results obtained with lipid-depleted Complex I. Since lipid-replacement alwaysresults in some denaturation of the enzyme, thisfinding supports the possibility discussed above thatthe increased labelling of certain polypeptides inlipid-depleted Complex I is a consequence ofconformational changes in the protein.

Discussion

The inability of 5-iodonaphth-1-yl azide to labelthe subunits of the iron-protein and flavoproteinfragments of Complex I is consistent with the ideathat these components are mostly buried within theenzyme. The two largest iron-protein subunits, ofmol.wts. 75 000 and 49 000, are exposed to theaqueous phase on either side of the mitochondrialinner membrane (Smith & Ragan, 1980). Theirinaccessibility to photolabelling by 5-iodonaphth-1-yl azide suggests that they do not come intocontact with the lipid bilayer, but are surrounded byother polypeptides except where they make contactwith the aqueous phases. The labelling of these twopolypeptides clearly distinguishes the mode of actionof 5-iodonaphth-1-yl azide from that of hydrophilicprobes such as diazobenzenesulphonate.

The polypeptides that are most heavily labelled inComplex I are those of mol.wts. 42000, 39000,22000 (more than one of very similar mol.wts.),18000, 16500, 15500 and 8000. Some of these,and possibly all, are subunits that are inaccessible tohydrophilic labels such as diazobenzene[ 35S]-sulphonate when Complex I is in the membrane, butaccessible when the enzyme is in isolation (Smith &Ragan, 1980). Exact identification is impossible insome instances because of the multiplicity ofsubunits of similar molecular weights. The results inthe present paper therefore support the conclusionthat these subunits are in contact with the hydro-phobic region of the bilayer and surround the morehydrophilic flavoprotein and iron-protein frag-ments.

Since NADH dehydrogenase contains far morepolypeptides than redox centres, most of thesubunits must be catalytically inactive. Of the six

Vol. 191

redox centres in the enzyme (Ohnishi, 1979), threeare probably located in the flavoprotein fragment(Galante & Hatefi, 1979), and at least one more inthe iron-protein fragment. Thus most of thecatalytic machinery is associated with the fragmentsthat are solubilized by chaotropic agents and verylittle with the large number of more hydrophobicpolypeptides that are not solubilized by chaotropicagents. The flavoprotein fragment retains the abilityto oxidize NADH in the presence of artificialelectron acceptors (Hatefi & Stempel, 1969). Kineticanalysis of this process both in the flavoproteinfragment and in Complex I showed that theinteraction of the protein with NADH was verysimilar in both preparations (Dooijewaard & Slater,1976a,b). Any similarity is remarkable, since theflavoprotein fragment is a truly water-soluble protein,whereas the native enzyme is an insolublemembrane-associated protein. The likely explanationis that the environment of the flavoprotein fragmentis very similar in both the soluble and bound states,except that the electron acceptor in the lattercondition is a protein-bound redox centre. Thisexplanation now provides a role for the catalyticallyinactive subunits that surround the flavoproteinfragment in NADH dehydrogenase. Their functionis to provide a hydrophilic environment for theflavoprotein fragment within the membrane. Thelarge number of such subunits can be predicted fromsimple geometrical- considerations. The flavoproteinfragment has a mol.wt. of 70000 (Galante & Hatefi,1979). To cover a globular protein of this sizecompletely with globular proteins of mol.wt. 25000requires about 20 such proteins. There are in NADHdehydrogenase at least 23 subunits other than thoseof the flavoprotein fragment, and these have anaverage mol.wt. of 25000. The expected number ofsubunits would be increased if the iron-proteinfragment is included with the flavoprotein fragment

Matrix 2H+

Cytoplasm

Fig. 7. Structure of NADH dehydrogenase in the mito-chondrial inner membrane

435

436 F. G. P. Earley and C. I. Ragan

in the central core of the enzyme and decreased ifthose subunits of these fragments that are exposed tothe aqueous phases (e.g. the two largest subunits ofthe iron-protein fragment) are excluded.

Fig. 7 summarizes our conclusions and alsoshows how NADH dehydrogenase might act as aproton-translocator. The flavoprotein fragment isshown completely buried within a shell of otherproteins except that access for NADH has to beprovided by a channel in the shell. This channel doesnot allow access for hydrophilic or hydrophobicprobes (Smith & Ragan, 1980, and the presentpaper). Electron transfer to the transmembranousiron-protein fragment is accompanied by loss of twoprotons per pair of electrons to the cytoplasmicphase. The iron-protein fragment is also excludedfrom direct interaction with the lipid phase of themembrane by the shell, but is exposed to theaqueous phases on both sides of the membrane(Smith & Ragan, 1980). This fragment catalysestransfer of electrons across the membrane andeventually to a further iron-sulphur centre in theshell. Reduction of ubiquinone in the membrane thentakes place with the uptake of protons from thematrix. Smith & Ragan (1980) showed that therewere several polypeptides other than the iron-protein subunits that were exposed on the matrixside of the membrane and could fulfil this last step.

The ordering of the electron-transfer sequence asa 'loop' (Mitchell, 1968) is essentially the same asthat proposed by Lawford & Garland (1972) forNADH dehydrogenase. Although this mechanism isadequate to explain a proton/electron stoicheio-metry of 1: 1, it would require revision such as theaddition of a separate proton pump to account forother proposed stoicheiometries such as 1.5 or 2protons per electron.

This work was supported by the Science ResearchCouncil. We are grateful to Dr. G. Schatz, Biozentrum,University of Basel, Switzerland, for helpful advice on thesynthesis of 5-iodonaphth- l-yl azide.

References

Bayley, H. & Knowles, J. R. (1978) Biochemistry 17,2414-2419

Bercovici, T. & Gitler, C. (1978) Biochemistry 17,1484-1489

Cerletti, N. & Schatz, G. (1979) J. Biol. Chem. 254,7746-7751

Crowder, S. E. & Ragan, C. I. (1977) Biochem. J. 134,295-301

Dooijewaard, G. & Slater, E. C. (1976a) Biochim.Biophys. Acta 440, 1-5

Dooijewaard, G. & Slater, E. C. (1976b) Biochim.Biophys. Acta 440, 16-35

Galante, Y. M. & Hatefi, Y. (1979) Arch. Biochem.Biophys. 192, 559-568

Gellefors, P. & Nelson, B. D. (1975) Eur. J. Biochem. 52,433-440

Hatefi, Y. & Rieske, J. S. (1967) Methods Enzymol. 10,235-239

Hatefi, Y. & Stempel, K. E. (1967) Biochem. Biophys.Res. Commun. 26, 301-308

Hatefi, Y. & Stempel, K. E. (1969) J. Biol. Chem. 244,2350-2357

Hatefi, Y., Haavik, A. G. & Griffiths, D. E. (1962a) J.Biol. Chem. 237, 1676-1680

Hatefi, Y., Haavik, A. G. & Griffiths, D. E. (1962b) J.Biol. Chem. 237, 1681-1685

Heron, C., Corina, D. L. & Ragan, C. I. (1977) FEBSLett. 79, 399-403

Heron, C., Gore, M. G. & Ragan, C. I. (1979a)Biochem.J. 178,415-426

Heron, C., Smith, S. & Ragan, C. I. (1979b) Biochem. J.181,435-443

Hinz, H. J. & Sturtevant, J. M. (1972) J. Biol. Chem. 247,6071-6074

Laemmli, U. K. (1970) Nature (London) 227, 680-685Lawford, H. G. & Garland, P. B. (1972) Biochem. J. 130,

1029-1044Lindahl, P. E. & Oberg, K. E. (1961) Exp. Cell Res. 23,

228-237Lowry, 0. H., Rosebrough, N. J., Farr, A. L. & Randall,

R. J. (195 1)J. Biol. Chem. 193, 265-275Mitchell, P. (1968) Chemiosmotic Coupling and Energy

Transduction, Glynn Research, BodminOhnishi, T. (1979) in Membrane Proteins in Energy

Transduction (Capaldi, R. A., ed.), pp. 1-87, Dekker,New York

Ragan, C. I. (1976) Biochem. J. 154, 295-305Ragan, C. I. & Bloxham, D. P. (1977) Biochem. J. 163,

605-615Ragan, C. I. & Racker, E. (1973) J. Biol. Chem. 248,

2563-2564Smith, S. & Ragan, C. I. (1980) Biochem. J. 185,

315-326Weber, K. & Osborn, M. (1969) J. Biol. Chem. 244,

4406-4412

1980