BiophysicalJournal Volume78 January2000 373–384 373 Atomic Force Microscopy and Electron Microscopy Analysis of Retrovirus Gag Proteins Assembled in Vitro on Lipid Bilayers Guy Zuber and Eric Barklis VollumInstituteandDepartmentofMicrobiology,OregonHealthSciencesUniversity,Portland,Oregon97201-3098USA ABSTRACT We have used an in vitro system that mimics the assembly of immature Moloney murine leukemia virus (M-MuLV) particles to examine how viral structural (Gag) proteins oligomerize at membrane interfaces. Ordered arrays of histidine-taggedMoloneycapsidprotein(his-MoCA)wereobtainedonmembranebilayerscomposedofphosphatidylcholine (PC) and the nickel-chelating lipid 1,2-di-O-hexadecyl-sn-glycero-3-(19-20-R-hydroxy-39N-(5-amino-1-carboxypentyl)imino- diaceticacid)propylether(DHGN).Themembrane-boundarrayswereanalyzedbyelectronmicroscopy(EM)andatomicforce microscopy (AFM). Two-dimensional projection images obtained by EM showed that bilayer-bound his-MoCA proteins formedcagessurroundingdifferenttypesofprotein-freecageholeswithsimilarcageholesspacedat81.5-Ådistancesand distancesbetweendissimilarcageholesof45.5Å.AFMimages,showingtopologicalfeaturesviewednearthemembrane- proximaldomainofthehis-MoCAprotein,revealedacagenetworkofonlysymmetricalhexamersspacedat79-Ådistances. These results are consistent with a model in which dimers constitute structural building blocks and where membrane- proximalanddistalhis-MoCAregionsinteractwithdifferentpartnersinmembrane-boundarrays. INTRODUCTION C-type retroviruses, such as the Moloney murine leukemia p12hasbeenshowntobecompatiblewithparticleassembly virus (M-MuLV), represent a family of RNA viruses that (HansenandBarklis,1995).Moreover,theonlypartofMA replicatethroughaDNAintermediate(Coffinetal.,1997). necessary for HIV or M-MuLV assembly appears to be its Theyarecomposedofcellularandviralcomponentsandare membrane-binding myristate anchor (Wang et al., 1993; enveloped by host-derived lipid membranes that contain Faeckeetal.,1993;Barklisetal.,1997).Furthermore,while proteins encoded by retroviral env genes. The virus core is the RNA-binding retrovirus NC domains contribute to the composed of 1000–5000 copies of the viral Gag proteins; efficiencyofparticleformation(CampbellandVogt,1995; 10–100copiesofpolgene-encodedproteins,whichinclude Wills et al., 1994), under some circumstances they can be the viral protease, reverse transcriptase, RNAse H, and replaced in vivo (Zhang et al., 1995), and they are not integrase; two copies of the viral RNA genome; and cellu- essentialtotheformationofparticle-likestructuresinvitro larly derived tRNAs that serve as primers during reverse (Grossetal.,1998;VonSchwedleretal.,1998).Thisleaves transcription (Coffin et al., 1997). The assembly of C-type thecapsid,oraportionoftheCAdomain,ascentraltothe retrovirus particles appears to occur at the plasma mem- particle production process. branes of infected cells and is directed by the Gag protein, Although some Gag protein functions seem to be well which has been shown to be necessary and sufficient for defined,thethree-dimensionalstructuresofretrovirusesare particleassembly.MostmammalianretrovirusGagproteins not. A major obstacle has been the heterogeneity of natu- are synthesized as precursor polyproteins (PrGag) and nor- rally occurring virus particles. Nevertheless, researchers mallyarecleavedintothematureprocessedGagproteinsby have determined the structures of some Gag protein do- the viral protease (PR) during or after budding. Processing mainsandtheirderivatives(Gittietal.,1996;Gambleetal., of PrGag results in a major morphological change in virus 1996;Hilletal.,1996;DeGuzmanetal.,1998;Dememeet particles, in which electron-dense material adjacent to the al., 1994; Fass et al., 1997), and certain Gag domains can peripheryoftheimmaturevirionreorganizesintothecentral assemble into rod- or sphere-shaped structures in vitro anddensesuperstructureofthematurevirus.ForM-MuLV, (Campbell and Vogt, 1995; Von Schwedler et al., 1998; processing of PrGag yields the four mature Gag proteins, Grossetal.,1998).However,tounderstandthemechanism matrix (MA), p12, capsid (CA), and nucleocapsid (NC). of C-type retrovirus assembly, it is important to analyze WhileeachmatureGagproteinservesanessentialfunction how Gag proteins organize on membranes, where virus for replication, several Gag domains are dispensable with particle assembly occurs (Coffin et al., 1997). One such regard to virus particle assembly. For instance, deletion of analysis was an electron microscopy (EM) study of HIV PrGag proteins assembled at the plasma membranes of bac- ulovirus vector-infected cells (Nermut et al., 1994). At low Receivedforpublication10November1998andinfinalform20Septem- resolution (40–50 Å), the PrGag proteins appeared to form ber1999. cage-like structures beneath the membranes. In an effort to AddressreprintrequeststoDr.EricBarklis,VollumInstituteandDepart- improveuponGag-membranestructurestudies,werecently mentofMicrobiology,OregonHealthSciencesUniversity,3181S.W.Sam JacksonParkRoad,Portland,OR97201-3098.Tel.:503-494-8098;Fax: devised an in vitro method for the analysis of Gag-mem- 503-494-6862;E-mail:[email protected]. brane interactions (Barklis et al., 1997, 1998). The ap- © 2000bytheBiophysicalSociety proach,whichisbasedonpreviouslipidmonolayerstudies 0006-3495/00/01/373/12 $2.00 (Darst et al., 1991; Uzgiris and Kornberg, 1983), employs 374 ZuberandBarklis histidine-tagged (his-tagged) Gag protein derivatives and a andmicaAFMsubstrateswerefromDigitalInstruments(SantaBarbara, model membrane consisting of egg phosphatidylcholine CA)andTedPella,respectively,andcarbonEMgrids(300mesh)were (PC) and the novel nickel-chelating lipid 1,2-di-O-hexade- fromTedPella.WaterwasfilteredwithaMilliQpurificationsystem. cyl-sn-glycero-3-(19-20-R-hydroxy-39-N-(5-amino-1-carboxy- pentaiminodiaceticacid)(DHGN)(seeFig.1;Barklisetal., 1997). Using this system to produce samples for EM anal- Construction of the vector for bacterial ysis, we have found that membrane-bound HIV-1 capsid expression of M-MuLV capsid protein proteinsformedhexamer-trimercages(Barklisetal.,1998) AM-MuLVcapsid-codingregioncassettewasinsertedintotheBamHIsite consistentwithpreviouslowerresolutionstudies(Nermutet ofpET15B(Novagen).Thecassettewasconstructedbypolymerasechain al., 1994), while M-MuLV capsid proteins formed distinct reaction,generatingaBamHIsiteattheamino-terminusoftheCA-coding hexamer-hexamer cages (Barklis et al., 1997). To extend region,andaddingaBamHIlinkeratthecarboxy-terminalMscIsiteofthis theseresults,wehaveadaptedourprocedurestopermitthe region.TherespectiveN-andC-terminalendsofthecassetteareGGAT/ imagingoflipidbilayer-boundM-MuLVhis-taggedcapsid CCC,wheretheboldCistheM-MuLVviralnucleotide1266andTTG/ GCGGATCC, where the bold G is viral nucleotide 2055. The plasmid proteins(hisMoCA)bybothEMandatomicforcemicros- pET15B-MoCA then was introduced into Escherichia coli strain copy(AFM)inbuffersolution(Binnigetal.,1986;Hansma BL21(DE3)/pLysS(Novagen),andthebacteriawasstoredat280°Cin and Hoh, 1994; Muller et al., 1995; Shao and Yang, 1995; 50%glycerol. Brownetal.,1998;Czajkowskyetal.,1998;Fotiadisetal., 1998; Sato et al., 1998). In agreement with monolayer findings, our EM projections show that bilayer-bound his- Protein expression and purification MoCA forms a protein cage surrounding different types of cageholes,inwhichsimilarcageholesarespacedat81.5-Å CellsofE.colistrainBL21(DE3)/pLysScontainingpET15B-MoCAwere intervals,whiledissimilarholesoccurevery45–46Å.AFM grownat37°CinLBplus15mg/Lchloramphenicoland50mg/Lampi- cillin to an OD600 of 0.7. Protein expression then was induced by the images, showing topological features, revealed a cage net- additionofisopropyl-b-D-thiogalactopyranoside(IPTG)0.5mM,andafter work of only symmetrical hexamers spaced at 79-Å dis- 3 h of shaking at room temperature the bacteria were harvested by cen- tances. Our results are consistent with a model in which trifugationandstoredat280°C.Forpurification,frozenbacterialpellets dimers constitute structural building blocks and where wereresuspendedinlysisbuffer(50mMsodiumphosphate,pH7.8,300 membrane-proximal and distal his-MoCA regions interact mMNaCl,1mMphenylmethylsulfonylfluoride,2mMb-mercaptoetha- with different partners in membrane-bound arrays. nol) and disrupted in a French press. Cellular debris was removed by centrifugation (12,000 3 g; 15 min; 4°C), and his-MoCA proteins were purifiedbytwocyclesofnondenaturingaffinitynickel-chelatechromatog- raphy,usingastepwisegradientofimidazole(0,10,and250mMin50 MATERIALS AND METHODS mMsodiumphosphate,10%glycerol,0.5MNaCl,pH7.8forwashes,pH Materials 6.0 for elutions). Fraction purities were assessed by a combination of Coomassie staining and immunoblotting of electrophoretically separated Egg phosphatidylcholine (PC) was purchased from Avanti Polar Lipids. proteins.Onceidentified,purefractionsweredesaltedonSephadexG-25 DHGN was prepared by D. Thompson and charged with nickel as de- spin columns equilibrated with 5 mM 3-(N-morpholino)propanesulfonic scribed(Barklisetal.,1997).Highlyorientedpyrolyticgraphite(HOPG) acid(MOPS)(pH7.8),10%glycerol,andproteinswerestoredat280°C. FIGURE 1 Assembly of his-tagged proteinsonlipidmembranes.Shownis a schematic view of the system used forforminghistidine-taggedproteinar- raysonamonolayerofnickelchelating lipid and phosphatidyl choline at an interface.ThestructureofDHGN,our nickel-chelatinglipid,isshownonthe right. For analysis of monolayers by EM, arrays are lifted onto EM grids, stainedorfrozeninvitrouswater,and viewed. For examination of bilayer- boundproteins,liposomeswithbound proteinsaretransferredtoEMgridsor AFMsubstratesforviewing. BiophysicalJournal78(1)373–384 RetrovirusGagProteinAssembly 375 Preparation of vesicles containing Ni21-DHGN ;70%(50%inthe37–25Åresolutionshell).Thebestspacegroupsfor amplitudes plus phases (aph) files were determinated using the ALL- Ni21-DHGN(10nmol)andeggphosphatidylcholine(100nmol)inchlo- SPACE program, using hexagonally indexed diffraction patterns. (Note roformsolutionsweremixedanddriedunderafluxofnitrogen.Thelipids thatp6residualswereaberrantlyhighwhenorthogonallyindexedreflec- then were fully dried under vacuum and resuspended to 50 mM Ni21- tionswereusedastheALLSPACEinput.)Mergingofimagesanddeter- DHGN and mM PC in 23 PNG buffer (200 ml) by sonication in an minationofphaseresidualswereperformedusingunbentaphfilesandthe ultrasonic bath for 10 min. Vesicles were stored at 0–4°C for up to 2 program ORIGTILTB (Henderson et al., 1990). For all merges, the weeks.Alternatively,andwithsimilarsuccess,vesicleswerepreparedin r100316bfilmwasusedasthereferenceimage.Reconstructionofareal 10 mM MOPS buffer (pH 7.8), 50 mM NaCl, by detergent dialysis imagewasdoneusingtheprogramsCREATE TNFandFFTRANS.Files techniques(Tauskelaetal.,1992). inMRC-formatwerethenconvertedtoTIFFformat. ForAFM,300nm3300nmimageswereacquiredataresolutionof 512 pixel/line and were exported in TIFF formats. Images then were Specimen preparation convertedtoSPIDERformat(Franketal.,1988)forrealspaceaveraging steps.Foraveraging,rawimageswereGausianlow-passfiltered,and12 Thehis-MoCAproteins(5ml,1–5mg)weremixedgentlywith5-mlvesicle 140.6Å3140.6Åwindowswerepickedandsummedtoyieldacross- suspensions, and drops were transferred onto 5-mm wells of depression correlationreference.Thereferencethenwasusedtolocatecross-correla- wellslides(no.101005;CarlsonScientific).Theslidesthenwereplaced tion peaks using the SPIDER (Frank et al., 1988) operation CC, and into150-mmpetriplateshumidifiedwithafilterpaperwettedwith2.5ml 187.5 3 187.5 Å2 raw image areas, representing the top 100 cross- water. Dishes were sealed tightly with parafilm strips before overnight correlationpeaks,weresummedtogiveanaverageimage.Thequalityof incubationsat30°C.ForEManalysis,arraysweretransferredontoultra- thisimagewasevaluatedbyhalvingthe100imagedataset,averagingthe thin, formvar-removed carbon grids (no. 1882-F; Ted Pella) by placing half-sets,andcomparingtheFourierringcorrelation(FRC)valuesbetween grids on top of the drops for 1 min. Samples then were processed by the averages at different resolutions, using the SPIDER operation RFM placinggridsontopof100-mlwaterdropsfor30s,wickingfromtheside, (Franketal.,1988). stainingfor45son50ml1.3%uranylacetate(freshlydilutedandfiltered), followedbyblottingandairdrying.ForAFManalysis,dropsweredepos- itedonfreshlycleaved3mm33mmhighlyorientedpyrolyticgraphiteor Quantification of his-MoCA bound to vesicles micasheets.After1min,thesampleswererinsedwithwater(threetimes 50ml)andthenmaintainedunderaviewingbuffer(20mMTrisHCl,pH ForisolationofMoCA-boundmembranes,200-mlmixturescontaining25 8.0,100mMKCl). mMhis-MoCAand250mMPC:Ni21-DHGN10:1vesiclesinPNGbuffer, pH8.3,wereincubatedovernightat30°Candultracentrifugedfor20min at165,0003g(rotorBeckmanTLS55;50,000rpm)topelletmembranes. Transmission electron microscopy Supernatantswerecarefullydiscardedtoremoveunboundhis-MoCA,and membraneswereresuspendedin60ml10mMMOPS(pH7.8),50mM ElectronmicroscopywasperformedonaJEOLJEM1200EXoperatedat NaCl(bufferA).Ascontrols,twootherincubationswereperformed,with 100kV(PortlandVAHospital).Low-dosephotographywascarriedoutat theomissionofeithertheproteinorthelipidmembranes.Forquantifica- ambienttemperature,usingKodakSO163film.Searchingwasperformed tionofhis-MoCAboundtovesiclesusingthemouseanti-CAmonoclonal at a magnification of 50003, and focusing and photography were at antibodyHy187(Hansenetal.,1993),two10-mlaliquotsofthepreviously 40,000–60,0003. preparedhis-MoCA-boundmembraneswerewithdrawnandoverlayedon theupperfacesoftwofreshlycleavedHOPGpucks(2cm2)for30min, afterwhichthepuckswererinsedfourtimeswithbufferA(150ml).One Atomic force microscopy puckwasextracteddirectlywith40ml20mMTrisHCl(pH7.4),150mM NaCl, 1 mM EDTA, 0.1% sodium dodecyl sulfate (SDS), 1% Triton Afterspecimenadsorption,samplesweremountedonanE-piezoscannerof X-100,0.5%sodiumdeoxycholate(IPBbuffer),andtheotherpuckwas an atomic force microscope equipped with a fluid cell (Nanoscope III; incubatedwithanti-CA(500ml;0.3g/Linphosphate-bufferedsaline)for Digital Instruments, Santa Barbara, CA). Calibration of the scanner was 30minatroomtemperature,rinsed,andextractedasdecribedabove.As carried out with mica as the substrate reference. Cantilevers with oxide- sharpenedSiN tips(purchasedfromDigitalInstruments)were200mm controls,10-mlaliquotsofhis-MoCAboundmembranesandmembranes 4 4 alonewereincubatedinsolutionwiththeanti-CAantibody(500ml;0.3 longandhadnominalspringconstantsof0.06N/m.Initialtipengagements g/L)for30min,afterwhichtheantibody/his-MoCA/membranecomplexes were performed by setting the scan size to 0 nm to minimize sample werepelletedbyasecondultracentrifugationandweresuspendedinIPB deformation.Beforethesampleswerescanned,theoperatingpointofthe buffer. For protein analysis, samples were subjected to SDS-polyacryl- servosystemwassettoforcesbelow1nN. amidegelelectrophoresis(SDS-PAGE),andgelswereelectroblottedonto anitrocellulosefilter.Gagproteinswereimmunodetectedwiththemouse Image processing anti-CAantibodyHy187,whichwasrevealedbyusinganalkalinephos- phatase-conjugatedanti-mouseantibodyat1:12,550dilution,followedby ForEMimages,micrographs100315,100316,121215,and121216were acolorreaction(Hansenetal.,1993).Thesecondaryantibodypluscolor digitizedat6.53Å/pixelandconvertedtoMRCformatimagesr100315b, reactionstepsalsorevealednitrocellulose-boundanti-CAheavyandlight r100315c, r100316a, r100316c, r100316d, r100316e, r121215, and chainsfromtheincubationsdescribedabove. r121216(UnwinandHenderson,1975;Baldwinetal.,1988,Hendersonet al.,1990).WiththeuseoftheICEimageanalysispackage(Schmidetal., 1993),realspaceimageswereFouriertransformed,anddiffractionpatterns RESULTS wereindexedbyhand.LatticeswererefinedandunbentusingtheMRC- Preparation and analysis of lipid vesicles derivedprogramsMMBOXandUNBEND(Baldwinetal.,1988,Hender- son et al., 1990; Schmid et al., 1993). After these steps, the calculated Retrovirus particles have proved somewhat intractable to amplitudeandphase(aph)fileswereeditedmanuallytoremovealllow signal-to-noise reflections of IQ . 5. The acceptable reflections yielded structural analysis, as they are enveloped with a cellularly resolutions extending to ;26 Å, and the completeness of the data was derived lipid bilayer and are heterogeneous in size and BiophysicalJournal78(1)373–384 376 ZuberandBarklis shape (Fuller et al., 1997; Yeager et al., 1998). Because of contact mode (Fig. 2). After 4 h of adsorption to mica, this,wedesignedamethodfortheanalysisofhowthemajor vesicles appeared to be intact but heterogeneous in size, structural (Gag) proteins of the Moloney murine leukemia with diameters from 100 nm to 500 nm (Fig. 2 a). Heights virus(M-MuLV)organizeonmembranesinimmaturevirus rangedfrom7.3to7.6nm,withanoccasionalmeasurement particles (Barklis et al., 1997). Briefly, N-terminally his- of approximately twice that height (Fig. 2 b, left side). In tagged Gag protein derivatives assembled on monolayers thisregard,becauseofstrongelasticitiesobservedforthese containing the nickel-chelating lipid DHGN can be lifted samples,itisnoteworthythatimagingwasonlypossibleby onto electron microscope grids and analyzed by transmis- applicationof5–10-nNforces.Giventhesehighscanforces sion EM (see Fig. 1). This model system is faithful in that and assuming a bilayer thickness of 4–5 nm (Simon and in vitro assembly occurs on a membrane, with the his-tag/ McIntosh, 1984; Sackmann, 1983), our average apparent DHGNinteractionsubstitutingforthemembrane-anchoring height of 7.5 nm is consistent with those of flattened, functionofamyristategroup,whichmodifiestheM-MuLV unilamellar vesicles, while higher features may result from Gag protein amino-terminus in vivo (Rein et al., 1986). the stacking of two vesicles. While 4-h preparations ap- To complement EM analyses, we decided to adapt our peared as flattened but elastic vesicles, after aging of sub- systemforanalysisofmembrane-boundGagproteinsunder strate-bound vesicle preparations overnight, AFM imaging buffer solution by AFM. Initially, we attempted to transfer could be achieved with scan forces below 1 nN and lipids monolayers onto octadecyl-silanized mica and glass sub- spread over areas of 200-1000 mm in diameter (Fig. 2 c). strates. Silanization was by octadecyltrichlorosilane treat- Theheightsofsuchareaswererelativelyconstantat3.16 ment of cleaned glass or freshly cleaved mica surfaces 0.2nm(Fig.2d),withintheexpectedrangeoflipidbilayers (Eggeretal.,1990).However,silanizationofmicaappeared in a fluid state and imaged by AFM (Mou et al., 1995). incomplete, and hydrophobic glass substrates were subop- These observations suggest that initially substrate-bound timalforAFMasaconsequenceofpoormonolayertransfer vesiclesconvertedtosinglebilayersafterovernightincuba- and rapid loss of image quality during AFM scanning with tionsandareconsistentwithpreviousobservations(Mouet silicon nitride (Si N ) tips (data not shown). al., 1994). 4 4 Asanalternativetomonolayerlifts,andtomimiccellular plasmamembranes,wenextoptedtoprepareNi21-DHGN- containing lipid bilayers as Gag protein assembly targets. EM imaging of bilayer-bound M-MuLV capsid Unilamellar vesicles were prepared by sonication of dried protein two-dimensional arrays lipids to final lipid concentrations of 500 mM PC and 50 mMNi21-DHGN(seeMaterialsandMethods).Vesiclesso As noted above, while lipid bilayers could be imaged con- prepared were processed by 4 h of adsorption on freshly veniently by AFM (Fig. 2), monolayer imaging gave poor cleavedmica,followedbytwogentlerinseswithwaterand results. Consequently, any AFM images of membrane- thenviewingbuffer.Samplesthenwereimagedunderview- bound proteins we hoped to obtain would have to be com- ing buffer with an oxide-sharpened Si N tip at 2.0 Hz in pared with EM images of lipid bilayer-bound proteins, 4 4 FIGURE 2 AFMimagingofPC1Ni21-DHGNvesi- clesandbilayers.ShownareAFMimagesofvesicles containing Ni21-DHGN and phosphatidylcholine after 4h(a)andovernight(c)adsorptiontofreshlycleaved mica.banddshowcross-sectionalheightsofimagesa andc,respectively,atthelevelsmarkedbythearrows. Initial mica-bound vesicles (a) with apparent heights ranging from 7.3 to 7.6 nm (b) were converted over- nighttosinglebilayers(c)withthicknessesof3.160.2 nm (average of 14 mesurements from different sec- tions).Scanningwasperformedincontactmodeusinga fluid cell, employing 200-mm-long oxide-sharpened SiN tips,withnominalspringconstantsof0.06N/m. 4 4 Imaging was done under a viewing buffer (20 mM TrisHClpH8.0,100mMKCl),withappliedforcesof ;5–10nN(a,b)and1nN(c,d)andscanfrequencies of2.0Hz.Framesizeswere1mmforaand3mmfor c,andresolutionswere512pixels/line.Thefullgray- levelrangesofaandbwere30and25nm,respectively. BiophysicalJournal78(1)373–384 RetrovirusGagProteinAssembly 377 rather than the monolayer-associated protein images ob- Å2/molecule for a his-MoCA monomer (Barklis et al., tainedpreviously(Barklisetal.,1997).Becauseofthis,we 1997). undertook the EM analysis of his-tagged M-MuLV capsid Using the above conditions in overnight, 30°C incuba- (his-MoCA) proteins bound to lipid bilayers. To do so, tions, crystalline his-MoCA arrays made on vesicles were optimization experiments were undertaken. In agreement apparentbyEMofnegativelystainedsamples.Asshownin withpreviousdata(Barklisetal.,1997),optimalconditions Fig. 3 a, vesicle incubations in the presence of his-MoCA involved incubation in 50 mM sodium phosphate (pH 8.3), resulted in the appearance of large membranes with exten- 5 mM sodium acetate, 10 mM imidazole, 250 mM NaCl, sivehis-MoCAcrystallinearrays,whichweremoreobvious 20% glycerol (PNG buffer), using a PC:Ni-DHGN ratio of at higher magnification (Fig. 3 b). A number of calculated 10:1. We also observed that array formation was optimal diffractionpatternsfromsucharraysweretwinned,possibly at a lipid-to-protein concentration ratio of 10:1 (250 mM resulting from arrays formed on opposite sides of a vesicle lipid:25 mM his-MoCA), slightly lower than the ratio pre- orfromseparatebutadjacentcrystals.However,frequently dicted assuming a lipid surface area of 70 Å2/molecule untwinned patterns were observed, as in Fig. 3 c. The (Schmitt et al., 1994; Mingotaud et al., 1993) versus 910 diffraction patterns (Fig. 3 c) could be indexed in either a FIGURE 3 EMofliposome-boundhis-MoCAproteins.His-MoCAproteinsandsuspensionsofPC/Ni21-DHGNvesiclesinPNGbuffer,pH8.3,were incubatedovernightat30°C,transferedtoultrathincarbongrids,washed,anduranylacetatestained.(a)Electronmicrographshowinghis-MoCAbound membrane.Thesizebarrepresents200nm.(b)High-magnificationimageofhis-MoCAproteinsarrayedonalipidbilayer.Thescalebarrepresents50 nm.(c)Shownisthecalculateddiffractionpatternforthestainedhis-MoCA2Dimage.Thepatterncanbeindexedinahexagonalororthorhombicfashion. Theinnermostsixreflectionsandhigherresolutionreflectionsarebarelyvisibleinthiscontrastrange,whilethesixbrightestreflectionsat0.0254Å21 correspondtothe1,1;-1,2;-2,1;-1,-1;1,–2;and2,–1reflectionsforagamma560°unitcellor2,0;1,3;-1,3,-2,0;-1,-3;and1,–3reflectionsfora gamma590°unitcell. BiophysicalJournal78(1)373–384 378 ZuberandBarklis hexagonal (a* 5 b* 5 0.0142 Å21, g* 5 60°) or orthog- spacing of PrGag proteins in immature M-MuLV particles onal (a* 5 0.0123 Å21, b* 5 0.0074 Å21, g* 5 90°) (Yeager et al., 1998). fashion,withreflections(notvisibleinFig.3c)outto25.4 Å (orthogonal reflection 1, 5), similar to what we have obtainedwithnegativelystainedmonolayerarrays.Averag- AFM imaging of his-MoCA bound to lipid bilayers ing of eight untwinned diffraction patterns indexed in an orthogonal fashion yielded a mean unit cell of a 5 81.5 6 BasedonourimagingoflipidmembranesbyAFM(Fig.2), 0.3Å,b5135.361.8Å,andg589.161.4°,whilethe weoptedtoimagehis-MoCAproteinsonlipidbilayers.To correspondinghexagonalunitcellswerea581.260.5Å, doso,his-MoCAproteinswereincubatedovernightat30°C b 5 79.6 6 1.7 Å, g5 117.9 6 1.5° (see Table 1). As withPC/Ni21-DHGNvesicles,depositedonmicaorhighly shown in Table 1, his-MoCA proteins formed arrays that oriented pyrolytic graphite (HOPG), and imaged under 20 wereconsistentwithhexagonal(p6)spacegroupsymmetry, mM TrisHCl, 200 mM KCl (pH 8.0). Scanning was per- giving an average phase residual of 18.1 6 4.5° to 15-Å formed in contact mode with a Nanoscope III fluid cell, resolution.Phaseresidualvaluesfortrigonal(p3)symmetry operated without the O-ring, employing 200-mm-long ox- calculationswereslightlybetterthanthoseforthep6space ide-sharpened Si N tips, with nominal spring constants of 4 4 group (13.5 6 5.2 at 15-Å resolution; Table 1), suggesting 0.06N/m.Aftertipengagement,tolocatemembraneareas, that it may be a more appropriate space group designation slow scans (2 Hz) were performed on 1–25 mm2 areas. As for his-MoCA crystals; implications with regard to protein illustrated (Fig. 5 a), bilayer areas as large as several hun- packing will be discussed below (see Discussion). dred nanometers to microns were observed. In contrast to Comparison of calculated diffraction patterns of bilayer- bilayers composed only of lipid (Fig. 2 c), areas of his- bound his-MoCA, assuming no symmetry constraints (p1), MoCA-boundbilayers(Fig.5a)revealedfeaturesastallas showedgoodagreementbetweenimages,withphaseresid- 8.5–10.6nm,consistentwithmembraneheightsof3.0–4.0 uals less than 30° at 26 Å or lower (Table 1). Back- nm plus M-MuLV capsid protein heights of ;6 nm. Slight transformation of unbent, filtered diffraction patterns re- differencesbetweenheightmeasurementsmadeondifferent sultedin2Dprojectionreconstructions,asshowninFig.4, dates may be related to the utilization of different tips, A–C.Asillustrated,theM-MuLVcapsidproteinassembled which yielded different compression forces on the mem- into cage-like structures on PC/Ni21-DHGN bilayers. On brane. At higher topologies, his-MoCA-bound membranes inspection, there appeared to be distinct types of protein- (Fig. 5 c) showed patches of arrays, comprising apparent free cage holes (dark), which were surrounded by six elec- 0.33 6 0.06 nm depressions spaced at ;8.0-nm intervals, tron-dense (white) units, apparently representing protein and calculated diffraction patterns (Fig. 5 d) demonstrated monomers. Reconstructions showed two or three types of anapparentsixfoldsymmetry(tworeflectionsareobscured cageholes(numberedinA–C).Inallpanels,theno.1cage by the y axis), corresponding to a unit cell of a 5 b 5 7.9 holes appeared distinct, as observed previously (Barklis et nm, g5 60°. al., 1997). However, it was unclear whether the no. 2 and Thehis-MoCAarraysasimagedbyAFM(Fig.5)didnot no. 3 holes were distinct, as predicted by p3 symmetry, or yield crystals that were compatible with conventional 2D similar, consistent with p6 symmetry (see Discussion). In diffraction analysis. However, it was possible to perform any case, each cage hole was spaced 45.5 Å from its real space averaging operations using the SPIDER suite of nearest-neighborcageholes,andspacingsbetweenputative image reconstruction programs (Frank et al., 1988). To do distinctcageholetypes(Fig.4,A–C,1to1,2to2,and3to so, 12 140.6 3 140.6 Å2 windows were added from low- 3) were 81.5 Å. These distances are consistent with those pass filtered images to yield a cross-correlation reference observed for M-MuLV capsid proteins assembled on lipid image, which was used to identify the top 100 cross-corre- monolayers (Barklis et al., 1997), as well as the apparent lationpeaksfromhis-MoCAAFMscans.Afterthisstep,the TABLE 1 ComparisonofEMimages Orthorhombiccell Spacegroupfitphaseresidual(°) Interimagephase Filename a(Å) b(Å) g(°) p1 p2 p3 p6 residual(°) r100316cU 81.7 135.4 90.0 25.5 31.4 11.2 18.3 — r100316aU 82.1 133.1 89.7 26.0 22.0 9.4 14.2 15.5 r100316dU 81.1 133.1 89.7 23.8 28.3 9.9 17.3 17.0 r100316eU 81.5 136.2 87.0 31.8 22.7 12.1 16.7 28.1 r121215U 81.1 133.7 90.4 26.9 32.9 21.1 24.3 26.1 r121216U 81.6 136.1 90.8 28.9 23.9 6.9 10.8 24.7 r100315bU 81.3 137.6 88.4 31.4 36.8 18.4 23.6 10.1 r100315cU 81.5 137.3 87.6 31.9 32.7 19.0 19.8 14.3 BiophysicalJournal78(1)373–384 RetrovirusGagProteinAssembly 379 FIGURE 4 EMprojectionstructureofvesicle-boundhis-MoCA.Shownare2Dprojectionreconstructionsofhis-MoCAproteinsboundtoPC/Ni21- DHGNvesicles.Thescannedimagesr121216c(A),r121216(B),andr121215(C)wereFouriertransformed,unbent,filtered,andback-transformedwith nosymmetryconstraintstoyieldthe2Dprojections.Notethatinallpanelsofthefigure,proteinregionsareinwhite,andtheprotein-freeareasappear dark,whileputativedifferentcageholetypesarenumbered.ThesizebarinCcorrespondstoadistanceof80Åinallthreepanels. corresponding 187.5 3 187.5 Å2 image areas from unfil- HOPG,whichservedasasubstrateforourAFMstudieson tered scans were added to give an averaged AFM image of his-MoCA arrays (Fig. 5). Our procedure involved binding membrane-bound his-MoCA proteins. As illustrated (Fig. proteins or vesicles to HOPG, postbinding steps with anti- 6), the topology of the bilayer-bound proteins shows a MoCA, and detection of his-MoCA and anti-MoCA pro- lattice of higher features surrounding shallow depressions teins by electrophoresis and immunoblotting after release spaced at 7.9-nm distances, reminiscent of hole-to-hole from the HOPG substrate. As indicated in Fig. 7, the free spacing between similar cage holes from EM micrographs his-MoCAproteinhassomecapacityforbindingtoHOPG (Fig. 4). The quality of our AFM reconstruction was as- by itself (lane 10), while his-MoCA vesicles appeared to sessed by halving our data set, generating two independent bindquiteefficientlytoHOPG(lane9).Significantly,anti- reconstructions, and comparing the averaged images with MoCA,whichboundwelltohis-MoCAvesiclesinsolution each other by determination of Fourier ring correlation (lane 4), bound poorly to substrate-bound his-MoCA vesi- (FRC) values. By this method, we found FRC values cles (lane 9). Indeed, while the amount of substrate-bound droppedfrom0.93(highlycorrelated)to0.53at0.032Å21, anti-MoCA in lane 9 exceeded the level of direct anti- yielding a practical resolution limit of ;31 Å. MoCA binding to HOPG (lane 11), it was approximately equaltotheamountthatadheredtoHOPG-boundPC/Ni21- DHGN vesicles (lane 12). Furthermore, on a proportional Quantification of MoCA bound to membrane. basis, the amount of anti-moCA bound to HOPG via his- Although the above AFM results clearly showed the exis- MoCA vesicles (lane 9) appeared to be much reduced tenceofacage-likelattice,itwasunclearwhetherthelattice relativetothelevelofanti-MoCAboundviafreehis-MoCA corresponded to the protein or lipid sides of his-MoCA to HOPG (lane 13). These results indicate that membrane- bound membranes. To disciminate between these possibil- bound his-MoCA proteins on AFM supports were not ities,wetestedwhetherhis-MoCAproteins(Fig.7,lane1) readily accessible to antibody binding and thus imply that on substrate-bound membranes were accessible to binding these retroviral Gag proteins were sandwiched between bytheanti-MoCAmonoclonalantibodyHy187(Fig.7,lane PC/Ni21 DHGN membranes and HOPG substrates during 2;Hansenetal.,1993).Fortheseexperiments,weprepared AFM imaging. PC/Ni21-DHGN vesicles and PC/Ni21-DHGN vesicles with bound his-MoCA proteins (his-MoCA vesicles). Not DISCUSSION surprisingly, while 35-kDa his-MoCA proteins in his- MoCAvesicleswerepelletedbycentrifugation(Fig.7,lane Moloney murine leukemia virus, like its C-type retrovirus 3), free his-MoCA proteins were pelleted much less effi- conterparts,assemblesattheplasmamembranesofinfected ciently (Fig. 7, lane 6). Also, as expected, the 50-kDa and cells,andtheexpressionoftheM-MuLVGagpolyproteinis 25-kDa heavy and light chains of the anti-MoCA antibody sufficient for the assembly of immature virus particles. bound well to his-MoCA containing vesicles (lane 4) but Normally, during or after budding, cleavage of the Gag not nearly as well to the naked PC/Ni21-DHGN vesicles proteins by the viral protease results in a morphological (lane 5). Analysis of the orientation of his-MoCA proteins changefromanimmaturetoamaturevirusform(Coffinet on substrate-bound bilayers involved binding reactions on al., 1997). Recent cryo-EM studies of immature retrovirus BiophysicalJournal78(1)373–384 380 ZuberandBarklis FIGURE 5 AFMimagingofbilayer-boundhis-MoCAproteinarrays.Samplesofhis-MoCAboundtoPCplusNi21-DHGNbilayerswereadsorbedonto AFMsubstratesasdescribedinMaterialsandMethods.Scanningwasperformedatambienttemperature(21–23°C)incontactmode,usingaNanoscope IIIfluidcellandemploying200-mm-longoxide-sharpenedSi N tips,withnominalspringconstantsof0.06N/m.Imagingwasdoneunderabuffer(20 4 4 mMTrisHCl,pH8.0;100mMKCl)withappliedforcesof;0.5nNandscanfrequenciesof2.0Hz.(a)Alow-resolutionAFMscanshowingvariously sizedhis-MoCA/PCplusNi21-DHGNmembranepatches.Thesizebarindicates200nm.(b)Across-sectionalheightplottakenfromthelowerlefttoupper leftsectionofa.Theabscissamarksindicate200nm,andthe610nmheightmarksareshownontheordinateaxis.Thefullgray-levelrangeofthescan is20nm,andtheapparentmembraneheightinthisscanis8.5nm,althoughscanheightswereobservedtovaryfrom8.5to10.6nm.(c)Ahigh-resolution, smallscansizeimageofhis-MoCA/PCplusNi21-DHGN,takenatafullgray-levelheightrangeof2nmisshown.Thecage-likearrayofproteinwith 0.3360.06nmdepressionsspacedat;8.0-nmintervalsisvisibleatthelowerleft.Thesizebarindicates50nm.(d)Thepowerspectrumfromthelower left-handthirdofc,calculatedusingtheNanoscopeIIIpowerspectrumoption,appearshexagonal,althoughtworeflectionsareobscuredbytheyaxis.The reflectionscorrespondtoarealspacedistanceof7.9nm. particles have shown that Gag polyproteins associate to capsid domain because it mediates critical Gag-Gag con- form paracrystalline sheets, but the subsequently formed tacts, and it is not sensitive to proteolysis. sphericalvirusshellsapparentlylackicosahedralsymmetry As observed in monolayer experiments (Barklis et al., (Fuller et al., 1997; Yeager et al., 1998). Unfortunately, 1997), his-tagged Gag proteins formed regular arrays on natural retrovirus particle pleomorphism and difficulties in PC/Ni21-DHGN vesicles (Fig. 3). Diffraction analysis in- preparation of homogeneous virus preparations have ham- dicated that vesicle-bound his-MoCA proteins formed con- pered the analysis of immature and mature retrovirus par- sistent, 2D crystals (Table 1). Two-dimensional projection ticles(Nermutetal.,1994;Fulleretal.,1997;Yeageretal., reconstruction shows that the proteins formed cages com- 1998). To circumvent the above difficulties, we designed a posedofcageholesspacedat45–46-Åintervals(Fig.4),in modelsystemforthestudyoftheassemblyofGagproteins agreement with studies on immature M-MuLV particles onthefaceofalipidmembrane(Barklisetal.,1997,1998). (Yeager et al., 1998). The appearance of reconstructions The approach employs a his-tagged retrovirus capsid do- suggeststhattheno.1holesinFig.4,A–C,arehexagonally main and membranes consisting of PC and a nickel-chelat- symmetrical, consistent with a p6 space group assignment. ing lipid, DHGN. Here, we have focused on the M-MuLV However, if slight differences in no. 2 and no. 3 holes are BiophysicalJournal78(1)373–384 RetrovirusGagProteinAssembly 381 FIGURE 6 AveragedAFMimageofhis-MoCAproteinsboundtolipid bilayers. Twelve 140.6 3 140.6 Å2 low-pass-filtered AFM windows of FIGURE 7 Orientation of membrane-bound his-MoCA proteins on membrane-bound his-MoCA arrays were added using the SPIDER real AFM substrates. The 35-kDa his-MoCA and Hy187 anti-MoCA heavy spaceADoperationtoyieldacross-correlationreferenceimage,whichwas chain(50kDa)andlightchain(25kDa)proteinsboundinincubationsto employed in the CC operation to identify the top 100 cross-correlation membranesandsubstratesweredetectedaftertheextractionofproteins, peaksfroma3003300nm2his-MoCAAFMscan.Cross-correlationpeak gelelectrophoresis,electroblottingontonitrocellulose,andimmunodetec- locations were used to cut the corresponding 100 187.5 3 187.5 Å2 tion as described in Materials and Methods. Incubations included only unfilteredimagewindowsusingtheSPIDERWIoperation,andthewin- proteins(lanes1,2,6),proteinsplusPC/Ni21-DHGNvesicles(lanes3–5), dowswereadded(SPIDERoperationASR)toyieldtheaveraged187.53 orproteinandmembranecombinationsonHOPGAFMsubstrates.Lane 187.5Å2AFMimage,wherehigherfeaturesarewhiteandlowerfeatures designationsindicatesamplesasfollows:1,4mg(120pmol)his-MoCA; aredark.Notethatspacingbetweendepressionsis;8.0nmandthatthe 2, 12 mg (80 pmol) anti-MoCA; 3, his-MoCA vesicles, purified by cen- size bar corresponds to 4 nm. The quality of the reconstruction was trifugation after a 0.2-ml, 250 mM PC/Ni21-DHGN, 25 mM his-MoCA assessedbyhalvingthe100-imagedatasetandcomparingthetwohalf- incubation;4,his-MoCAvesicles,postincubatedwith1nmolanti-MoCA averagesbyFourierringcorrelation(FRC),usingtheSPIDERoperation and repurified; 5, PC/Ni21-DHGN vesicles (containing no his-MoCA), RF M. By the use of this method, the 50 image averages showed FRC postincubatedwith1nmolanti-MoCAandrepurified;6,29mg(0.83nmol) valuesof0.93(highlycorrelated)at0.011Å21,droppingto0.53at0.032 his-MoCA,processedtosimulatevesiclepurificationconditions;7,protein Å21,givinganeffectiveresolutionlimitof;31Å. size standard; 8, his-MoCA vesicles bound to HOPG; 9, HOPG-bound his-MoCAvesicles(if100%recoveryinpellet:0.83nmolhis-MoCA;8.3 nmol PC/Ni21-DHGN) postincubated with 1 nmol anti-MoCA and verifiedinfuture,higherresolutionstudies,thelowersym- washed;10,29mg(0.83nmol)his-MoCAbounddirectlytoHOPGand metry p3 assignment may prove more appropriate. washed; 11, 150 mg (1 nmol) anti-MoCA bound directly to HOPG and Assuming a p6 packing arrangement, Fig. 8 A shows a washed; 12, HOPG-bound PC/Ni21-DHGN vesicles (containing no his- model for M-MuLV capsid protein assembly at a mem- MoCA;8.3nmolPC/Ni21-DHGNif100%pelletrecovery)postincubated with 1 nmol anti-MoCA and washed; 13, 29 mg (0.83 nmol) his-MoCA brane.Thismodelshowshexagonalandtrigonalcageholes, bound directly to HOPG, washed, then postincubated with 1 nmol anti- each surrounded by six CA monomers. Features of the MoCAandwashed.Themolecularmassesoftheproteinsinthestandard model are that it accounts for p6 symmetry, and head-to- laneare97,68,43,29,and18kDa. headCAhomodimers(Gambleetal.,1997;shownashexa- gonpairsjoinedbyno.4–no.4interfaces)andCAsurfaces (idealized as numbers on hexagons) occupy constant posi- In comparison with EM results, AFM analyses gave tions within the network. An alternative model, which al- similar yet slightly different results (Figs. 5 and 6). For lowsforthreedifferenttypesofcageholes,isshowninFig. AFM studies, we found that his-MoCA-bound PC/Ni21- 8 B. The model is compatible with p3 symmetry, but does DHGNvesiclesadheredtosubtratestoyieldimageheights not directly account for the existence of head-to-head of 8.5–10.6 nm (Fig. 5), consistent with a lipid bilayer dimers(Gambleetal.,1997);requiresthatCAsubdomains heightof3.0–4.0nm,plusahis-MoCAlayerheightof;6 mustfulfilltwodifferentroles(aswherehexagonno.1and nm. As demonstrated in Fig. 7, the his-MoCA layer ap- no.2facesbothformdimerinterfacesandcageholeedges); peared inaccessible to antibody probing, supporting the and presents an ambiguous assembly pathway. notion that his-MoCA proteins were sandwiched between BiophysicalJournal78(1)373–384 382 ZuberandBarklis FIGURE 8 Modelsofmurineretroviruscapsidassemblyonalipidbilayer.(A)AmodelofM-MuLVCAproteinsorganizedonamembraneandshowing p6symmetry.Proteinmonomersaredepictedashexagonswithnumberedfaces,andprotein-freecageholesareeitherhexagonally(blackhexagons)or trigonally(blacktriangles)symmetrical.Themodelpredictshead-to-headdimerbuildingblocks(arrowsattheupperright)anduniqueenvironmentsfor each monomer face. It is compatible with a p6 space group assignment. (B) The trigonal (p3) model of his-MoCA assembly on membranes features monomers(face-numberedhexagons)surroundingthreedifferenttypesofcageholes(blacktriangles,whitetriangles,andtrianglesinhexagons,which signifycageholesthatappearmosthexagonal).Assemblyispostulatedtooccurfromasymmetricalhead-to-headdimerbuildingblocks(arrowsattheupper right),wheredimerformationisinitiatedatverticesbetweentheno.1andno.2faces.Anassemblyconstraintisthatadjacentno.3andno.6sideson dimerunitsareunavailableforinterproteincontacts,whileseparatedno.3andno.6sidesmediateformationofhexamerbuildingblocks.(C)Inthismodel tocorrelateEMandAFMresults,capsidmonomersaredepictedascommas(picture1)orasdeformedrodsinthreedimensions(pictures2,3).Notethat themembrane-proximalCAdomainsareorientedupinpictures2and3andaredepictedastheglobularportionsofcommasinpicture1.(1)Schematic model of his-MoCA assemblies viewed from the bilayer side. Two different cage holes, corresponding to the triangle and hexagon holes from A, are indicated,althoughthemodelisconsistentwiththeschemeinB,assumingthatcageholessimilartotheonemarkedbytheXareadistincttype.(2)An arrangementofhis-MoCAmonomerssurroundingacageholethatcanbedetectedbyAFM.Inthisdepiction,thecageholerepresentsthehexagonholes fromA,withmembrane-proximalproteinregionsorientedupward.Asillustratedabovetheproteinunits,AFMscanningofarraysisexpectedtoidentify shallowdepressions.(3)Thepredictedarrangementofhis-MoCAproteinssurroundingasecondtypeofcagehole,withproteinunitsorientedasinpicture 2.ThesecageholelocationsarenotexpectedtobeobservedbyAFM.Notethatforatrigonalmodel,athirdtypeofcagehole(markedbytheXinA) wouldappearsimilartothecageholesshownhere,exceptthatdarklyandlightlyshadedmonomerunitswouldbeswitched. bilayers and substrates during AFM imaging. Our observa- are numerous ways in which AFM and EM results can be tions suggest that his-MoCA-decorated vesicles adhered to reconciled.However,otherobservationsreducethenumber supports by protein-substrate binding, followed by vesicle of possible models. Notably, CA proteins are composed of breakage,andinside-upbilayerunrolling,ahypothesisthat two domains (Fuller et al., 1997; Yeager et al., 1998; Gitti isconsistentwithpreviousresults(Mouetal.,1994)andour et al., 1996), and cage holes appear to be formed by hex- own observations with PC/Ni12-DHGN vesicles (Fig. 2). amerrings(Barklisetal.,1997,Fig.4).Giventheseresults Although his-MoCA proteins appeared to be covered by and making the assumption that AFM cage hole positions bilayersduringAFManalyses,byusingmoderate(0.5nN) correspond to one type of EM cage hole, models must forcewithlowspringconstant(0.06N/m)cantileversitwas explain how one type of cage hole appears to be more possible to image membrane-bound arrays under buffer electrondensethantheothers,whiletwo-thirdsofthecage (Fig. 5). Under these conditions, proteins would not be holes are supported well enough to resist deformation by expected to be imaged directly. Rather, bilayer regions AFMtips.Onemodelthatsatisfiestheaboverestrictionsis supported by proteins would appear as higher features, depicted in Fig. 8 C. As shown, each CA monomer is while membrane regions that covered protein-free regions modeled as a hockey stick or golf club, with the club ends of his-MoCA cages could be deformed by tip forces, ap- adjacent to lipid monolayers. Club heads point away from pearingasdepressions.IncontrasttoEMresults(Fig.4),we one type of cage hole (Fig. 8 C, 1, 2) and toward the other found that the membrane surfaces of his-MoCA arrays holes(anexampleisshowninFig.8C3).Whilethemodel showed evidence of only one type of cage hole. Specifi- obviously remains hypothetical, several of its implications cally,shallowdepressionsof0.3360.06nmwereobserved are pertinent for consideration. In particular, the two do- tobespacedat7.9-nmhole-to-holedistances(Figs.5and6). mainsofCAmonomersinteractmostcloselywithdifferent Because AFM images show surface topologies and EM partnersinmembrane-boundarrays.Moreover,whilemem- imagesrepresent2Dprojectionsofelectrondensities,there brane-bound his-MoCA EM projections show cages with BiophysicalJournal78(1)373–384