Architecture of Mammalian Respiratory Complex I

8/10/2019 Architecture of Mammalian Respiratory Complex I

1/17

ARTICLE doi:10.1038/nature13686

Architecture of mammalian respiratory

complex IKutti R. Vinothkumar1*, Jiapeng Zhu2*& Judy Hirst2

Complex I (NADH:ubiquinoneoxidoreductase) is essential foroxidativephosphorylation in mammalianmitochondria. Itcouples electron transfer from NADH to ubiquinone with proton translocation across the energy-transducing innermembrane, providing electrons for respiration and driving ATP synthesis. Mammalian complex I contains 44 differentnuclear- and mitochondrial-encoded subunits, with a combined mass of 1 MDa. The 14 conserved core subunits havebeenstructurally definedin theminimal, bacterial complex, but thestructures and arrangementof the30 supernumerarysubunits are unknown. Here we describe a 5 Aresolution structure of complex I fromBos taurusheart mitochondria, aclose relative of thehuman enzyme, determined by single-particle electron cryo-microscopy. We present thestructures

of themammalian coresubunits thatcontaineight ironsulphur clusters and 60 transmembrane helices, identify18 super-numerary transmembrane helices, and assign and model 14 supernumerary subunits. Thus, we considerably advanceknowledgeof thestructureof mammaliancomplex I andthe architecture of itssupernumerary ensemble around thecoredomains. Our structure provides insightsinto theroles of thesupernumerary subunitsin regulation,assembly and homeo-stasis, and a basis for understanding the effects of mutations that cause a diverse range of human diseases.

Mammaliancomplex I (ref. 1) is oneof thelargest andmost complicatedenzymes in the cell. Complex I from B. taurus (bovine) heart mitochon-driahas beencharacterizedextensively as a model forthehumanenzyme;both enzymes contain 44 different subunits (encoded by both the nu-clearand mitochondrial genomes)2,3 andnine redox cofactors(a flavinmononucleotide and eight ironsulphur clusters). Fourteen subunitsare the core subunits that are conserved in allcomplex I enzymes; they

contain allthe mechanistically critical cofactors and structural elementsand are sufficient for catalysis. Crystal structures of intact complex Ifromthe thermophilicbacteriumThermus thermophilus4,andofdomainsoftheprokaryoticenzymesfrom T. thermophilus and Escherichia coli57

have provided a wealth of information on the structures of these sub-unitsbuttheyrepresentonlyhalfthemassofthemammalianenzyme.Thecohort of 30 supernumerary subunitsparticular to the mammalianenzyme2,3 has been accumulated through evolution. The supernumer-ary subunits may have alternative functionsor be importantfor assem-bly, regulation, stability or protection against oxidative stresstheirstructures and arrangement around the core subunits are not known.

Owingto its size, L-shaped asymmetry, membrane-bound location,andmulti-component structure, mammaliancomplexI hasproved dif-ficult to crystallize, and its high-resolution structure has not yet been

determined. Crystallographicinformationon any eukaryotic complexIis currentlylimited to a medium-resolution mapof theenzymefromtheyeastYarrowia lipolytica, which has been described, but not modelled8.Conversely, the size andshapeof complex I make it an attractivetargetfor electron microscopy (EM), and the enzymes from several specieshavebeen visualizedto display theiroverall L-shapedstructures911,al-though at too low a resolutionto reveal detailed structural information.A high-resolution structure of the mammalian enzyme is essential forunderstanding how the 30 supernumerary mammalian subunits arearranged around the core domain, how they determine the properties,assembly andactivityof theenzyme, andhow mutations in both thecoreand supernumerary subunits cause human diseases12.

Imaging and reconstructionComplexI waspurifiedfrom B. taurusheart mitochondria in detergent13,and imaged in vitreous ice on holey-carbon grids with a Falcon directelectron detector (see Methods). The enzyme adopts different orien-tations on the grid, and reference-free two-dimensional class averagesclearlyshow thecharacteristic L-shape of theminimalprokaryotic formaugmented by extra domains from the supernumerary subunits (Ex-

tended Data Fig. 1).Refinementwas performedin RELION14

andmovieframes wereused to correctforbeam-inducedmovement15. Per-frame re-construction andB-factor weightingwerefollowedby three-dimensionalclassification, resulting in the final map (Fig. 1) obtained from 25,492particles withan overall resolution of,5 A(seeMethods and ExtendedData Fig. 2). Viewed at a low-density threshold the map is dominatedby a disordered detergentphospholipidbelt thatencircles thehydropho-bic domain and defines the position of the membrane. At intermediate-density threshold, the hydrophilic matrix domain and the extendedmembrane domain,containinga largenumberof transmembranea-helices(TMHs), areobserved. The highest-density peaks in themap revealtheeight ironsulphur (FeS) clusters that, as in theT. thermophilus4,7 andY. lipolytica8 enzymes, form a chain through the hydrophilic domain.

Structures of the core subunitsThe 14conserved core subunits ofcomplex I (refs 1, 4) catalyse the en-ergytransducing reactions:NADHoxidation, ubiquinonereductionandproton translocation (Extended DataTable1 summarizestheir nomen-clature).The seven nuclear-encoded hydrophilic coresubunits harboura flavin mononucleotide to oxidize NADH, FeS clusters for inter-substrate electron transfer,and the ubiquinone-binding site.The sevenmitochondrial-encodedmembrane coresubunits containfourantiporter-like domainsfor protontranslocation. Thestructuresof themammaliancoresubunits (Fig. 2) werefitted to thedensity map(see Methods)usingthestructure ofT. thermophilus complex I (ref. 4), secondarystructureanalyses and sequence alignments, and using structural features and

*These authors contributed equally to this work.

1MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge CB2 0QH, UK. 2MRC Mitochondrial Biology Unit, Wellcome Trust/MRC Building, Hills Road, Cambridge CB2 0XY, UK.

8 0 | N A T U R E | V O L 5 1 5 | 6 N O V E M B E R 2 0 1 4

Macmillan Publishers Limited. All rights reserved2014
http://www.nature.com/doifinder/10.1038/nature13686http://www.nature.com/doifinder/10.1038/nature13686


2/17

densities from aromatic side chains (Extended Data Fig. 3). Except fortheFeS-clusterligands they have been modelled as polyalanine chains,with the residue numbering optimizedto enableindividual residues tobe located (Extended Data Table 2). It is not possible to attribute den-sity to any bound ubiquinone species in the present map.

A comparison of thebacterial4 and mammalian core enzymesreveals

that the mammalian membrane domain is more strongly curved outof the membrane plane (Extended Data Fig. 4). However, within eachindividualsubunit the60 TMHs of themammalian coresubunits matchtheir T. thermophilus counterpartsclosely (Extended DataFig. 5)onlythepositionof TMH4 in subunit ND6is different, andthe extra carboxy-terminal TMH particular toT. thermophilusND1 is absent fromB.taurus (Fig. 2 andExtended Data Fig. 5).No notable density is observedin place of thethree amino-terminal TMHs(presentin T. thermophilusandY. lipolytica) that have been lost through evolution of mammalianND2 (ref. 16), so they have not been substituted structurally by othersubunits. Importantly, catalytically relevant features identified in theantiporter-like subunits of the bacterial complex5,6 are conserved. Theyinclude the loops in the six broken TMHs in ND2, ND4 and ND5 (seeExtendedData Fig.3 forexamples)thatmay constitute part of theproton-translocationmechanism,and thelong transverse helix in ND5, a pro-posed coupling element.

Of the seven hydrophilic core subunits (Fig. 2), the structures of theB.taurus51 kDasubunit(humanhomologue NDUFV1),49kDa(NDUFS2),24 kDa (NDUFV2), PSST (NDUFS7) andTYKY (NDUFS8) subunits,andthe smalldomain of the 75 kDa subunit (NDUFS1)are closely con-servedfromtheirT. thermophilushomologues4,7 (Extended Data Fig. 5),with marked variation only in the length and extent of some of their NandC termini.Consequently,the arrangements of theFeS cluster chainsarealso very similar (ExtendedData Table 3), except that, owing to ro-tation of the 51 and 24 kDasubunits, the superimposedchains divergewith increasing distance from the membrane (Extended Data Fig. 4).The sequence and structural conservation of the large domain of the75 kDa subunit, which contains an extra, catalytically redundant clus-ter inT. thermophilus7 and the 30 kDa subunit (NDUFS3), are lower

(Extended Data Table 2). As neither of them have any known catalytic

role, we conclude that the catalytically critical subunits and cofactorsare closely conserved in the mitochondrial and bacterial enzymes, sup-porting their common mechanism of catalysis.

The supernumerary ensembleOnce thecore subunits had been modelled, the map revealed that addi-tionaldensities form an open cage around the core (Fig. 3). These den-sities areattributedto thesupernumerary subunits,and theyarearrangedpredominantly around the membrane domain and lower hydrophilic

domain, wherethey may help to protect FeS-containing PSST andTYKYfromoxidativedamage.Conversely,the areaaround theNADH-bindingsite, where complex I produces superoxide17, is bare (Fig. 3), so super-numerary subunits do not shield it from O2to minimize superoxideproduction. The NADH dehydrogenase domain is added at the end ofthe complex I assembly pathway18, and the local paucity of supernu-merarysubunitsmay facilitate both its integration andits replacement(while retaining the rest of the protein) to mitigate the effects of oxi-dative damage19. Twolarge supernumerary domains, one capping ND5and part of ND4, the other capping ND2, are observed on the matrixsurface of the membrane domain. Facing the intermembrane space, asnotedin Y. lipolytica8, thesupernumerary subunits form a layer of pro-tein that may have a role similar to that of the stabilizingb-hairpinhelix structures observed in the prokaryotic enzyme5. Eighteen super-

numerary TMHs are distributed around the core membrane domain

Mitochondrial

inner membrane

Mitochondrial

matrix

Mitochondrial

intermembrane space

Figure 1| Overall map for complex I from B. taurusheart mitochondriadetermined by single-particle cryo-EM. Three distinct features of thecomplex are revealed by overlaying maps at different density thresholds. Themap at the highest threshold (red) reveals the FeS clusters. The map at

medium threshold (grey) reveals the overall architecture of the protein and the78 TMHsin the membrane domain. The detergentphospholipid beltobservedas a dominant feature at low density threshold (translucent blue) representsthe density that remains around the membrane domain after cutting out thefinal model of the protein, and denotes the position of the complex in themembrane. It is ,30 Athick, and 34 Athinner at the proximal end of thecomplex (left) than at the distal end (right). ND5

ND2ND4

ND4L ND6

ND1

ND3

PSST

24 kDa

75 kDa

51 kDa

49 kDa

30 kDa

TYKY

FeS clusters

N1aN3

N2

a

b

Membrane

1aN3

N2

Figure 2| Structures of the core subunits of mammalian complex I.a, Structural models of the 14 mammalian core subunits (cartoonrepresentation) and their density (transparent surface); the subunits arecoloured individually and labelled with text in the same colours. The chainof FeS clusters is shown modelled to the highest density peaks (blue mesh) inthe inset.b, The seven membrane-bound mammalian core subunits, viewedfrom the matrix. Arrows indicate the positions of the four TMHs in T.thermophilus that are not present in B. taurus: three N-terminal TMHs in

ND2 and one C-terminal TMH in ND1. The position of TMH4 in ND6 isdifferent inB. taurusandT. thermophilus(marked with stars). For a detailedcomparison of theB. taurusandT. thermophilusstructures see ExtendedData Figs 4 and 5.

ARTICLE RESEARCH

6 N O V E M B E R 2 0 1 4 | V O L 5 1 5 | N A T U R E | 8 1



3/17

(Figs 3 and 4), consistent with the predictions of sequence analyses for14 to 18 TMHs from these subunits (Extended Data Table 4). In total,therefore, we observe 78 TMHs in themammalianenzyme.TwoTMHsare on the outside of the ND5 transverse helix, appearing to strap it tothe core domain, and four more are positioned close to the end of it,appearing as a restraint for its lateral movement (Fig. 3). These obser-

vations raise the question of whether large-scale piston-like motionsof this helix during catalysis, as postulated from theT. thermophilusstructure6, are feasible.

Assignment of 14 supernumerary subunitsTo identify and assign individual supernumerary subunits to the mapfor mammalian complex I (Extended Data Table 4 summarizes their

nomenclature)we used biochemical, sequence and structural informa-tion. Homology models for six of the hydrophilic supernumerary sub-units were createdusingknownstructures (ExtendedData Tables4 and5).Human B8 (NDUFA2) adopts a thioredoxin fold20 and its structure(Fig.5) was located at the tipof the large domain ofthe 75kDa subunit(Fig. 4), so (contrary to current models18) B8 is likely to be assembledinto complex I after(or with) the 75kDa subunit. B8 is extensively de-graded in brainmitochondria frompatientswith Parkinsons disease21,and,alongwithother NADH dehydrogenase domain subunits, it is rap-idly exchanged under steady-state conditions19. Therefore, it may helpto protect the coreenzyme against oxidativedamage. Similarly, regionsofdensity consistentwithtwo subunitsimportantforcomplexI assembly18,the18 kDa(NDUFS4) and13 kDa(NDUFS6) subunits(Extended DataTables 4 and5), were located (Fig. 4).However, they aresmallproteinswith no predicted dominant secondary structure and it cannot be ex-cluded that other supernumerary subunits have similar structures. Inthe current map, the 18 kDa subunit has been modelled into a densityin a cleft between the 75kDa subunitand the 49kDa,30 kDa and TYKYsubunits; the density attributed to the 13 kDa subunit suggests that itinteractswiththe 75 kDa, 49kDa andTYKYsubunits (Fig.4). Theselo-cations may explain why clinically identified mutations in the18 kDa and13 kDasubunits leadto accumulation of late-stage interrupted-assemblyintermediates lacking the NADH-dehydrogenase module22,23.

The 42 kDa subunit (NDUFA10), a member of the nucleoside ki-nase family24, was easily located as the density on top of ND2, on thematrix side of themembrane (Figs 4, 5 and Extended Data Tables4, 5).Its location is neatly confirmed by its absence from the density map ofY. lipolyticacomplex I (ref. 8), which lacks this mammalian-specificsubunit. Phosphorylation of a serinein the42 kDa subunit by a PINK1-

dependent mechanism hasbeen proposed to be required forcomplex I

Membrane

ND1ND5 and ND4

75 kDa

NADH-

binding site

ND2, 4L, 6 and 3

30 and 49 kDa,PSST and

TYKY

51 and

24 kDa

ND5

em r ne

ND15 an 4

75 kDa

ND2, 4L, 6 and 3

4 kDa,n30T an

and1

kDa

Figure 3| Architecture ofmammalian complex I showing thedensities of the supernumerarysubunits enclosing the coredomain. The models for the coresubunits are in light colours (aslabelled) in surface representation,and density attributed to thesupernumerary subunits, forming a

cage around the core subunits, is indark red. The supernumerarysubunits are concentrated on eachside of the membrane domain, andaround the lower section of thehydrophilic domain. The NADH-binding site in the 51 kDa subunit isindicated, with the predictedpositions for the flavin isoalloxazine(orange spheres) and threeconservedphenylalanines at the entry to thesite (yellow); the vicinity of this site isdevoid of supernumerary subunitdensity.

SDAP-42 kDa

SDAP-

B8

39 kDa

B16.6

B14.7

(18 kDa)

(13 kDa)

PGIV

(MWFE

+ B9)

a

b

(B13)B14

B16.6

.

B16.6

Figure 4| Structural assignments of supernumerary subunits inmammalian complex I. a, A semi-transparent surface for the density map formammalian complex I is shown in pale grey, with the surface from the coresubunitsin wheat. Structural models for the supernumerarysubunits are shownin colour and labelled accordingly (dashed lines indicate subunits on the backof the structure). Subunits labelled with brackets are those with less certainassignments, and structural elements, which cannot be assigned confidentlyin the current map, are in blue. b, Arrangement of TMHs, viewed from thematrix. The core subunits are in light colours (wheat for ND1, ND4 and ND5,green for ND2, ND3, ND4L and ND6). The supernumerary subunits are

coloured as ina.

RESEARCH ARTICLE

8 2 | N A T U R E | V O L 5 1 5 | 6 N O V E M B E R 2 0 1 4



4/17

activity25, implying bothits regulatoryroleand a molecular linkbetweenPINK1 dysfunction andcomplex I activity in Parkinsons disease. Fur-therelucidation of thisregulatorypathway must nowbe reconciled withthematrix locationof the42 kDasubunit.The 39 kDa subunit (NDUFA9),a member of thenucleotide-binding short-chaindehydrogenase/reductase

family26 (Extended Data Tables 4 and 5), was readily located adjacentto PSST (Figs 4 and 5), and observed to contain a density consistentwitha bound dinucleotide (Extended Data Fig.3)27. Furthermore, it par-tially encloses the long ND3 loop (resolved only in T. thermophilus4)that is critical for coupling electron and proton transfer, and which,in a conformational transition now known to also involve the 39 kDasubunit28, switches theenzyme intoa deactivestateduring ischaemia.Notably, these proposed regulatory elements areall located close to the

junction between the hydrophilic and membrane domains wherethe en-ergy from the redox reaction is used to initiate proton translocation 1.

Two regions of density corresponding to the structure of the SDAPsubunit (NDUFAB1), which is identical to the acyl carrier protein inthe mitochondrial matrix29,30, were identified in the mammalian en-zyme (Fig. 4).This resultis supportedby thepresence of SDAP in both

subcomplex Ia (which contains the hydrophilic domain) and subcom-plex Ib (the distalportion of the membrane domain) ofB. taurus com-plex I (ref. 3), and with the presence of two SDAP homologues in Y.lipolyticacomplex I (ref. 31). One SDAP is located at the distal end ofthe enzyme, above ND5, the other in a peripheral region of the hydro-philicdomain, in a subdomainthat interacts with the49 kDaand 30 kDacore subunits through a three-helix bundle (Figs4 and5). From a recentstudyinY.lipolytica32thesehelicesareassigned tosubunit B14(NDUFA6),a protein with an LYR motif that, when deleted inY. lipolytica, resultsin loss of catalytic activity. Notably, subunit B22 (NDUFB9) also con-tains an LYR motif and it is in subcomplex Ib, so it is possible that thedistalSDAPmolecule interactswith it in a similar fashion. Finally, sub-units B13(NDUFA5)and B14 have similar predicted secondary struc-tures so we ascribe the second three-helix bundle observed on the side

ofthe 30 kDasubunit, adjacent to the42 kDa subunit, to B13(Figs 4, 5).

Continuous density links the four supernumerary TMHs at the endof thetransverse helix. Onlyonesubunit,B14.7(NDUFA11),is predictedto contain more than two TMHs (Extended Data Table 4); secondarystructure analyses predict that the first two are unusually long (,30residues), so they probably correspond to the two highly tilted TMHs.Therefore, these four TMHs are assigned to subunit B14.7 (Figs 4, 5and Extended Data Table 5), a proteinthat is important forthe assem-blyand/orstability of themembrane domain33.Asecondclusterofthree

TMHs (opposite B14.7)mayinclude twoTMHsfrom B14.5b(NDUFC2),but the connectivity between them is ambiguous and a clear assign-ment cannot be made. The11 remaining TMHs arespread aroundthemembrane domain (Fig. 4). Three TMH-containing subunits, B16.6(NDUFA13), MWFE (NDUFA1) and B9 (NDUFA3), remain assoc-iated with the hydrophilic domain in subcomplex Iaafter fractiona-tion ofB. tauruscomplex I with zwitterionic detergents3, and they aremissing fromY. lipolyticasubcomplex Id, which lacks core subunitsND1,2,3and4L34.Therefore,theyareassignedtothethreeTMHden-sities next to ND1 (Extended Data Table 4 and 5). Sequence analysespredict a single 67-residue helix in subunit B16.6, with the first 20 res-idues forming a TMH. Correspondingly, one of the three densities is

very long and modelled as a single 63-residue helix that interacts withtheN terminusof TYKY on thematrix side, spans themembrane, then

bends into the intermembrane space and is anchored under the heel(Figs 4 and 5). Therefore, this density is assigned to B16.6, a proteinidentical to thecell death regulatory geneproduct GRIM19 (ref. 35). Itis currentlynot possibleto confidently deduce assignments fortheTMH-containing subunits of subcomplex Ib.

ThePGIV (NDUFA8), 15 kDa (NDUFS5) andB18 (NDUFB7) sub-units contain twin CX9C motifs that form two intramolecular disul-phides within CHCH domains36, and they have been assigned to theintermembrane surface of complex I (ref. 37). A double L-shaped den-sity, resembling two CHCH domains at right angles, is clearly visibleon the heel, clamping B16.6 (Figs 4 and 5) onto the core. PGIV, a sub-unit present in subcomplex Ia (Extended Data Table 4), contains twoCHCH domains, so it is assignedto theL-shaped density feature (Figs4and 5), consistent with the position of an antibody label to its homol-ogous subunit in Y. lipolytica34. Ourstructure thus reveals the architec-ture ofthe 400 kDaassemblyintermediate ofhuman complexI (refs18,33)that contains the core hydrophilic subunits 49 kDa, 30 kDa, PSST andTYKY, core membrane subunit ND1, and the supernumerary subu-nits PGIV, B9, B16.6 and B13.

Conclusions and perspectivesWe have described a 5 Aresolution cryo-EM density map for mam-malian complex I, and used it to produce structural models for the 14core subunits that are conserved in all complex I enzymes,plus14 ofthesupernumerary subunits of themammalian enzyme. Thecoresubunitscomprisethe catalytically active centre of theenzyme, and(as expected)theyclosely resembletheircounterparts describedby theatomic-resolutionstructure of bacterial complex I (ref. 4). The 14 supernumerary sub-units assigned include two copies of subunit SDAP, bringing the totalnumberof subunits in themammalian complex upto 45.We have usedour structural models for the supernumerary subunits to support anddiscuss their roles in assembly, homeostasis and regulation. Higher-resolution maps arerequiredfor assignment of the remaining17 super-numerary subunits.

Recent developments in direct electron detectors, microscopy andimage processing algorithms have enabled high-resolution structuresof biological macromolecules to be determined by single-particle cryo-EM at resolutionsthat havepreviously onlybeen routinely possiblewithX-ray crystallography3840. Thus, we believe that it will be possible toextend ourcurrentstudyto produce a high-resolutionstructure forcom-plex I inthe nearfuture, to allow usto identifyand model all the super-numerary subunits,andto characterizethestructural changesthatoccurduring catalysis, a crucial step in defining the mechanism of electron-

coupled proton translocation.

b

B8

a 39 kDa

42 kDa

(B13)

B16.6B14.7

ND5

PGIV

SDAP39 kDa

42 kDa

B14

Figure 5| Structural models for supernumerary subunits in mammaliancomplex I. a, Models for three supernumerary subunits in cartoonrepresentation,colouredfromblueto red(N to C termini). b, Structural modelsand relationships of supernumerary subunits to the core structure. B14.7 is

located at the end of the transverse helix, next to ND5TMH16. The densityassigned to PGIV forms an L-shaped clip over B16.6, which bends aroundthe heel at ND1. Finally, the supernumerary subunits around the lower sectionof the hydrophilic domain are viewed in cartoon representation from thematrix. The core membrane subunits (white) and four core hydrophilicsubunits, the 49kDa (blue), 30 kDa (green), PSST (yellow) and TYKY (cyan)subunits, are shown in surface representation.

ARTICLE RESEARCH

6 N O V E M B E R 2 0 1 4 | V O L 5 1 5 | N A T U R E | 8 3



5/17


6/17

METHODSProtein preparation. Complex I waspurifiedfrom B. taurus heart mitochondrialmembranes by solubilizationand anion exchangechromatographyin n-dodecyl-b-D-maltoside (DDM), andsize-exclusionchromatography inDDM or 7-cyclohexyl-1-heptyl-b-D-maltoside (Cymal 7) as described previously13.Cryo-EM specimen preparation and imaging. Aliquots of complex I (3ml, 34.5mgml21) wereapplied to glow-dischargedholey-carbonQuantifoil R 0.6/1grids,blotted for 1518 s, then plunge-frozen in liquid ethane using an environmentalplunge-freeze apparatus41. The grids were transferred into cartridges, loaded into

anFEI Titan Krioselectron microscope,andimageswererecorded at 25mm under-focus on a Falcon II CMOS (complementary metal oxide semiconductor) directelectron detector at 300keV at381,495magnification(nominally347,000), withthespecimen temperature at2186uC usingthe EPUsoftware (ExtendedDataFig.1).The detector pixel size of 14mm corresponds to a sampling density of,1.72Apixel21. Eachimage wasexposedfor 4 s (totaldose,6 4 eA22)and72frameswerecapturedas previously described15. Fortilt-pairanalysis, thesamearea wasimagedtwice, at 0ufor 0.8 s and then at 10ufor 2.0 s, and analysed with FREALIGN42.Image processing and three-dimensional reconstruction. An initial data set forcomplex I in DDM was obtained by manually picking particles with XIMDISP 43.Seven-thousand six-hundredand thirty particles,from 366micrographs,were usedto generateinitialmapsin EMAN2(ref.44).Particleswere boxed in2803280 pixelsand contrast transfer function (CTF) parameters estimated internally. Reference-freeclassification wasperformed using thedefaultEMAN2parameters, andclasseswith distinct orientationsselected(seeExtendedData Fig. 1bfor an example) tobuildinitial models. The initial model that best matched the class averages was selectedand two cyclesof refinementperformedin EMAN2. Subsequently, comparisonofthemodelwiththestructureofcomplexIfromT. thermophilus (ProteinDataBank(PDB) accession 4HEA4) suggested that it had the wrong hand; this observationwas verified using tilt-pair analysis45 andcorrected (ExtendedData Fig. 2).All fur-ther refinements were performed in RELION14, starting with maps that were low-pass filtered to 60 A.

Typical micrographs prepared from complex I in Cymal 7 exhibited, on aver-age, twice as many particles (,40 per micrograph) than those from complex I inDDM, so a larger data set was collected using Cymal 7. The reason for the differ-ence in particle distribution is not clearit may simply be a product of the gridpreparationand freezing protocols. The class averages were used as a reference topickparticles automatically using RELION but manyfalse positiveswere included,so all the images were inspected manually and particles too close to each other,aggregates and ice contaminants were deleted. The final data set contained 45,618particlesfrom 1,154micrographs.The CTF wasdetermined withCTFFIND3 (ref.46)

using the images summed from all 72 frames. Subsequently, refinement was per-formed using frames 132of eachimage (thelast 40 frames werediscarded), to pro-duce a map with resolution of 5.86 Aand orientational accuracy of 1.2u. To checkforoverfitting,phases wererandomized beyond 10 Aon individual images (frames132), followed by refinement as forthe normalimages47. The results clearly showthepresence of informationbeyond10 A(ExtendedDataFig. 2).Note thatRELIONdivides the data set into two halves at the initial step and calculates the resolutionusing a gold-standard Fourier shellcorrelation(FSC), so the phaserandomizationprocedure serves here only as an additional control.

Modelling of the beam-induced movement of the complex I particles (using arunning average of 11 movieframes in RELION) provideda modest improvementin resolution to 5.16 A. The parameters from this analysis were then used to carryouta per-frame reconstruction in RELION(particle-polish),and a B-factor weight-ing was applied to each frame, resulting in a 4.8A map. The B-factor-weightedparticles were subjected to 25 iterations of three-dimensional classification into 4classes; this separated a major class containing 55% of the particles from smaller

classes of 24, 14 and 6%. Difference maps revealed minor localized variations insome of the peripheral regions of themolecule, but no large-scale conformational

variation was observed.The major class with 25,492 particles was refined and, afterpost-processing withRELION,a shape-mask, correctionfor the modulation trans-fer function (MTF) of the detector and a B-factor of2152 (ref. 48) were applied,and filtered to 4.95 A resolution (Extended Data Fig. 2C, note that the magnifica-tion andCTF valueshave notbeen refined). Despite containing a lower numberofparticles, the maps from this major class and the whole data set were comparable.Analysis of the local resolution by ResMap49 showed that the core sections of themolecule, particularly the TMHs, have higher resolutions than the peripheral sec-tions, and that (as expected)the detergentphospholipidbelt is at lowerresolution(Extended Data Fig. 2e).Model building. Model building was performed using Coot50. All the models de-scribedhave beenbuiltas polyalaninechains,except forthe residues thatcoordinatethe FeS clusters. Note that the present model has not been refined, so it inevitablycontainssome errors and inaccuracies.Examplesof the modelfittedto the electron

density are shown in Extended Data Fig. 3, and figures were created using the

PyMOL Molecular Graphics System or UCSF Chimera (http://www.cgl.ucsf.edu/chimera/).

The seven core hydrophilic subunits ofB. tauruscomplex I were modelled ini-tially using the coordinates ofT. thermophiluscomplexI (PDB accession 4HEA4)as a template, trimmed where the densities were ambiguous, and adjusted manu-ally. FeSclusters were locatedusing the highest peak densitiesin the unsharpenedmap, and the subunits were built around them. The 24 and 51 kDa subunits wereeasilybuilt as they arewellconserved andthe connectivityin thedensities isclearlyresolved. The 49 kDa subunit has dominant secondary structures and, except for

the N-terminal peptide, could be completelytraced.Similarly, the PSSTand TYKYsubunits, with three FeS clusters, were readily built. The central part of the 30kDasubunit, containing a mixture ofa-helices andb-strands, could be traced, but thepath ofthe long unstructuredC terminusis unclear.The 75kDa subunitis theleastconserved hydrophilic core subunit. The small domain containing the three FeSclusters iswell resolvedand couldbe tracedeasilyusingthe T. thermophilus model,but considerable portions of the large, peripheral domain have poor density, lowsecondarystructure content andlow sequencesimilarity to T. thermophilus,andsocouldnot be tracedconfidently.The seven core subunits inthe membrane domaincould all be readily traced, except for a few loop regions, and assigned using theirsimilarity to theT. thermophilussubunits. The long transverse helix at the C ter-minus of subunit ND5 is well ordered and extends over ND4 and ND2. In betterresolved regions of themap,protrudingdensities that arelikely to be side chainsofaromatic residuesare observed(see ExtendedData Fig.3), andthese features,alongwith secondary structureinformation and sequence alignments, were usedto pro-duce an optimized assignment for the B. taurusresidue numbers in the modelledsubunits (Extended Data Table 2), for use as a guide to the positions of individualresidues.

Once the electron density for the core subunits had been assigned, models forthe TMHs of the supernumerary subunits were built. A total of 18 TMHs weremodelled, and when the density was clear they were extended. Connectivity wasobservedbetweenfour TMHsadjacent to subunit ND4so theywere combinedintoa single chain. To aid in supernumerarysubunitassignments,the secondarystruc-ture of each subunit was predicted using PSIPRED51 and TMHs were predictedusing TMHMM2 (ref. 52), HMMTOP2 (ref.53) andthe TOPCONS suite54 (sevenmethods in total) (Extended Data Table 4). Known structures of soluble proteinswith high homology to the complex I supernumerary subunits were identified byHHpred55 andused to build homologymodelsin Modeller56 and SwissModel57 (Ex-tended Data Table 5). Regions of the density map with features corresponding tothe predicted structures were locatedmanually. Long loop regions were trimmed,then the models were placed in the density, jiggle fit in Coot was used to find the

best fit, and the models were adjusted manually. Finally, several additionaltubulardensities in the map were built as a-helices. Most of them are located close toTMHs from thesupernumerarysubunits,but theconnectivity to them isnot clear;a higherresolution mapwill be necessaryto assign these helices to their respectivesubunits.

41. Bellare, J. R., Davis, H. T., Scriven, L. E. & Talmon, Y. Controlled environmentvitrification system: an improved sample preparation technique.J. ElectronMicrosc. Tech.10,87111 (1988).

42. Grigorieff, N. FREALIGN: high-resolution refinement of single particle structures.J. Struct. Biol.157,117125 (2007).

43. Smith, J. M. XIMDISPa visualization tool to aid structure determination fromelectron microscope images.J. Struct. Biol.125,223228 (1999).

44. Tang, G.et al. EMAN2: an extensible image processing suite for electronmicroscopy. J. Struct. Biol.157,3846 (2007).

45. Henderson,R. etal. Tilt-pairanalysisof images froma range ofdifferentspecimensin single-particle electron cryomicroscopy.J. Mol. Biol.413,10281046 (2011).

46. Mindell,J. A.& Grigorieff,N. Accuratedeterminationof local defocusand specimentilt in electron microscopy.J. Struct. Biol.142, 334347 (2003).

47. Chen, S.et al.High-resolution noise substitution to measure overfitting a ndvalidate resolution in 3D structure determination by single particle electroncryomicroscopy. Ultramicroscopy135,2435 (2013).

48. Rosenthal, P. B. & Henderson, R. Optimal determination of particle orientation,absolute hand, and contrast loss in single-particle electron cryomicroscopy.J. Mol. Biol.333,721745 (2003).

49. Kucukelbir, A., Sigworth, F. J. & Tagare, H. D. Quantifying the local resolution ofcryo-EM density maps.Nature Methods11,6365 (2014).

50. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development ofCoot. Acta Crystallogr. D66,486501 (2010).

51. Jones, D. T. Protein secondary structure prediction based on position-specificscoring matrices.J. Mol. Biol.292,195202 (1999).

52. Krogh, A., Larsson, B., von Heijne, G. & Sonnhammer, E. L. L. Predictingtransmembrane protein topology with a hidden Markov model: application tocomplete genomes.J. Mol. Biol. 305,567580 (2001).

53. Tusnady,G. E. & Simon, I. Principlesgoverningamino acidcompositionof integralmembrane proteins: applications to topology prediction.J. Mol. Biol.283,489506 (1998).

ARTICLE RESEARCH

http://www.cgl.ucsf.edu/chimerahttp://www.cgl.ucsf.edu/chimerahttp://www.cgl.ucsf.edu/chimerahttp://www.cgl.ucsf.edu/chimera


7/17

54. Bernsel, A., Viklund, H., Hennerdal, A. & Elofsson, A. TOPCONS: consensusprediction of membrane protein topology.Nucleic Acids Res. 37,W465W468(2009).

55. Soding, J., Biegert, A. & Lupas, A. N. The HHpred interactive server for proteinhomology detectionand structure prediction. NucleicAcids Res.33,W244W248(2005).

56. Eswar, N.et al. Comparative protein structure modeling using Modeller.Curr.Protoc. Bioinform.Chapter 5, Unit 5.6 (2006).

57. Arnold,K.,Bordoli,L., Kopp,J.& Schwede, T.The SWISS-MODELworkspace:a web-based environment for protein structure homology modelling.Bioinformatics22,195201 (2006).

58. Efremov, R. G. & Sazanov, L. A. Respiratory complex I: steam engine of the cell?Curr. Opin. Struct. Biol.21,532540 (2011).

59. Krissinel, E. & Henrick,K. Secondary-structure matching (SSM), a newtool forfastprotein structure alignment in three dimensions. Acta Crystallogr. D60,22562268 (2004).

60. Balsa, E.et al.NDUFA4 is a subunit of complex IV of the mammalian electrontransport chain.Cell Metab.16,378386 (2012).

61. Johansson, K.et al.Structural basis for substrate specificities of cellulardeoxyribonucleoside kinases.Nature Struct. Biol.8,616620 (2001).

62. King,J. D. etal. Predictingprotein function fromstructure - theroles of short-chaindehydrogenase/reductase enzymes inBordetellaO-antigen biosynthesis.J. Mol.Biol.374,749763 (2007).

63. Parris, K. D.et al.Crystal structures of substrate binding toBacillus subtilisholo-

(acyl carrier protein) synthase reveal a novel trimeric arrangement of moleculesresulting in three active sites. Structure8,883895 (2000).

RESEARCH ARTICLE



8/17

Extended Data Figure 1| Single-particle cryo-EM analysis ofB. tauruscomplex I. a, Typical micrograph of complex I particles imaged after freezingin vitreous ice on a holey-carbon grid. Some of the selected particles aremarked with red boxes. Scale bar, 50 nm.b, Two-dimensional referenceclassification showing particleslying in differentorientations in the ice.The sizeof each box is 280 pixels and the two-dimensional classification was madein RELION14.

ARTICLE RESEARCH



9/17

Extended Data Figure 2| Validation of the map and resolution. a, Tilt-pairanalysis45 of complex I in Cymal-7. One-hundred complex I particles fromeight image pairs, recorded with a relative tilt angle of 10u, were extractedand subjected to tilt-pair analysis with FREALIGN42. The outer radius of theplotis 40u andthe orangecircle centred at theexpected tilt angle hasa radiusof6u.b, Phase randomization to check for overfitting. Phases that are beyond10 Ain each of the micrographs used in the final data set (frames 132) wererandomized, and then refinement was performed as for a normal data set(FSC summed image corresponding to frames 132). As expected, the graphshows a drop in the Fourier shell correlation (FSC) curve at 10A, validating thepresence of information beyond 10 Ain the images. Note that the use ofgold-standard refinement procedures in RELION14 prevents any overfitting,

and this test was done only as an additional control. c, An overview of the final

map and the model built into it. d, FSC curves of the final map and of themodel versus the map. The curve in red is the gold-standard FSC of the finalmap (after classification) and the resolution at FSC5 0.143 is ,4.95 A. Thecurve incyanis the FSC between the final map andthemodel, and atFSC5 0.5the resolution is 6.7 A. Note that the present model is not complete sinceit is only a polyalanine model without any side chains, and loop regions in anumber of subunits have not been modelled.e, The final map of mammaliancomplex I was analysed with ResMap49. The left-hand panel (with lowerdensity threshold) shows that the detergentphospholipid belt is of lowerresolution, and most of the protein regions of the map show resolutiondistributed from 5 to 6A. In the right-hand panel the map is shown at a higherdensity threshold, so the detergentphospholipid belt is not visualized. Some of

the interior parts of the map have resolution of 4.85 A.

RESEARCH ARTICLE



10/17

Extended Data Figure 3| Example regions of the density map with themodel fitted to the map. a, ND2 is shown from the membrane plane,highlighting the densities for three aromatic side chains and one of thehelix-breaking loops. b, Subunit ND4viewed from thematrix. c, Thedensityfora [4Fe4S] cluster and surrounding protein is shown in the PSST subunit.d, A region of the 49 kDa subunit shows a well resolveda-helical stretch andaromatic side chains, andthe b-strands are beginningto be resolved. e, SubunitB8 is an example of a supernumerary subunit in a peripheral region of themolecule.f, Density consistent with a bound nucleotide is observed in the39 kDa subunit, in a similar position to in homologous structures and asexpected from analysis ofY. lipolyticacomplex I (ref. 27). However, thepresentresolution of the map precludes the inclusion of this nucleotide in thefinal model.

ARTICLE RESEARCH



11/17

Extended Data Figure 4| Global comparison of the core subunit structuresof bacterialand mammalian complex I. The core subunits fromB. taurus arein blue, and from T. thermophilus(PDB accession 4HEA4) in orange. Thestructures have been superimposed using ND1 (the heel subunit). Top: theND2, ND4 and ND5 domain is rotated in B. taurusrelative to inT.thermophilus, increasing thecurvature in the B. taurus membrane domain. Thecomplex is viewed along the 11urotation vector (orange) that maps the T.thermophilus ND2, ND4 and ND5 domain to theB. taurus domain, along witha small 5 Atranslation to superimpose the domain centres. Correspondingly,the ND3, ND4L and ND6 domains are superimposed by a 4urotation and a1 Atranslation. Rotation of ND2, 4 and 5 about the long axis of the domain, asnoted forY. lipolytica58, is not observed. Bottom: the NADH dehydrogenasedomain containing the 51 and24 kDa subunits is rotated by 23uand translated

by 14 A inB. taurus, relative to inT. thermophilus, causing the FeS chains todiverge as the distance from ND1 increases. A similar rotation was observed inY. lipolytica58. The complex is viewed from behind ND1. Correspondingly,the 49 kDa, PSST and TYKY subunits are superimposed by a 6urotation and a2 Atranslation. The structures were analysed using Superpose from theCCP4 suite59 and the 75 kDa and 30 kDa subunits were not included due totheir lower structural conservation.

RESEARCH ARTICLE



12/17

Extended Data Figure 5| Comparison of the individual structures of thecore subunits of bacterial and mammalian complex I. a, The structure ofeach subunit fromT. thermophilus(wheat) (PDB accession 4HEA4) hasbeen superimposed separately on its corresponding subunit fromB. taurus(coloured as labelled) with the transverse helix plus TMH16 of ND5 alsoaligned separately. Thecomplexes areviewed from behindND1 (top),from theside (middle) and from the matrix (bottom, ND subunits only). b, Observed

differences in the structures of the core subunits ofB. taurusandT. thermophiluscomplexes I. Grey, conserved structure from B. taurusandT. thermophilus(PDB accession 4HEA4); red, structural elements present onlyinT. thermophilus; blue, structural elements present only in B. taurus. TheC-terminal domain of the 75 kDa subunit is not resolved inB. taurus, but itsstructure is clearly different to inT. thermophilus.

ARTICLE RESEARCH



13/17

Extended Data Table 1| Reference table for the nomenclature of the core subunits of complex I

In the text the names of the subunits fromB. taurusare used, with the names from the human enzyme presented alongside as appropriate.

RESEARCH ARTICLE



14/17

Extended Data Table 2| Summary of the models of the core subunits of B. tauruscomplex I

*For proteins with a mitochondrial-targeting pre-sequence, residue 1 is the first residue of the mature protein 2,3.

{Thepercentage identityand theroot meansquared deviation(r.m.s.d., calculated using PDBeFOLD59) arebetweenthe sequences andstructuresof thesubunits of B. taurus and T. thermophilus (PDB accession

4HEA) complex I.

ARTICLE RESEARCH



15/17

Extended Data Table 3| Distances between the redox cofactors in structural models of complex I

*The [2Fe2S]clusterinthe 24kDasubunit (knownasN1a)is ontheother side oftheflavin fromthe maincofactor chain. The [4Fe4S]clusterin the51kDasubunit(knownasN3) isthefirstclusterinthe chain

and the [4Fe4S] cluster in subunit PSST (known as N2) is the last (seventh) cluster in the chain.

{The distancesare inA, between thegeometriccentres ofthe Feand S cluster coresor theflavinisoalloxazineringsystem (centre),or between thecentresof thetwoclosest atoms(edge) ascommonly used in

calculations of electron transfer rates. Distances are estimated to be accurate to within 1 A

.{The position of the flavin inB. taurusis poorly resolved and has been approximated using its position in PDB accession 4HEA.

RESEARCH ARTICLE



16/17

Extended Data Table 4| Knowledge about the supernumerary subunits of B. tauruscomplex I

*The former subunit MLRQ (NDUFA4) is no longer considered a subunit of complex I (ref. 60).

{SubcomplexIl, whichcontains the7 hydrophiliccoresubunits and89 supernumerarysubunits,representsa considerableportionof thehydrophilic domainof complexI. SubcomplexIa, which containsallthe

subunits of subcomplex Ilplus core subunit ND6 and 910 additional supernumerary subunits, represents the hydrophilic domain of complex I plus associated membrane subunits. Subcomplex Ib, which

contains ND4 and ND5 and 1213 supernumerary subunits, represents part of the membrane domain3.

{TMHswere predictedusingTMHMM2 (ref.52), HMMTOP2(ref. 53)and theTOPCONSsuite54 (sevenmethods intotal) andare presentedas consensus values withlessrepresented values inbracketsand single

outliers discarded.

ARTICLE RESEARCH



17/17

Extended Data Table 5| Summary of the models of the supernumerary subunits of B. tauruscomplex I

*For proteins with a mitochondrial-targeting pre-sequence, residue 1 is the first residue of the mature protein 2,3.

{Known structures with high homology to the complex I subunits20, 36, 6163 were identified by HHpred55.

{The percentage identity and the r.m.s.d. (calculated using PDBeFOLD59) are between the sequences and structures of the subunits of B. tauruscomplex I and the PDB models.

1Subunit with less certain assignment.

"Residue numbers are arbitrary and not assigned to the sequence.

RESEARCH ARTICLE

Architecture of Mammalian Respiratory Complex I

Documents

Transcript of Architecture of Mammalian Respiratory Complex I