Crosslinking and Mass Spectrometry : An Integrated Technology to Understand the Structure and Function of Molecular Machines

Crosslinking and Mass

Spectrometry: An Integrated Technology to Understand the Structure and Function of Molecular Machines

Alexander Leitner,

^1,

* Marco Faini,

¹

Florian Stengel,

^1,2

and Ruedi Aebersold

^1,3,

*

Inrecentyears,chemicalcrosslinkingofproteincomplexesandtheidentifica- tionofcrosslinkedresiduesbymassspectrometry(XL-MS;sometimesabbre- viated as CX-MS) has become an important technique bridging mass spectrometry(MS)and structuralbiology.By now,XL-MS iswell established andsupportedbypubliclyavailableresourcesasaconvenientandversatilepart of the structural biologist's toolbox. The combination of XL-MS with cryo- electronmicroscopy(cryo-EM)and/orintegrativemodelingisparticularlyprom- isingtostudythetopologyandstructureoflargeproteinassemblies.Amongthe targets studied so far are proteasomes, ribosomes, polymerases, chromatin remodelers, and photosystem complexes. Here we provide an overview of recentadvancesinXL-MS,thecurrentstateofthefield,andacursoryoutlook onfuturechallenges.

ChemicalCrosslinkingasa ToolforStructuralBiology

Structuralbiologymakesuseofmany differenttechniquesto elucidatethe3Dstructures of proteins and protein complexes. While high-resolution structures have traditionally been obtained by X-ray crystallography, cryo-EM is increasingly able to also generate (near) atomic-resolutionmodels.Inrecentyears,techniquesandapplicationsofMShavealsorapidly progressed.Earlierstudieswerelargelyfocusedonthelarge-scaleidentificationandquantifi- cationofproteins,whereasrecentmethodsalsosupportqueriesintothecomposition,stoichi- ometry, andspatialarrangementofsubunits ina complex. Thesedevelopments have now furtherprogressedtowardgeneratinginformationthatcontributes,aspartofhybridstructural strategies,to thestructureelucidationoflargemolecular assembliesincludingproteincom- plexesthat perform essential processes inthe cell. XL-MS is aparticularly powerful mass spectrometric technique in this respect, because it provides several layers of information.

Identifyingprotein–proteincontactsthroughXL-MSconfirmsphysicalproximitybetweensub- unitsbecausetheproteinsmustbecloseenoughinspacetobecrosslinked.Localizingtheside chainsthatareconnectedrestrictsthisproximitytocertainregions(e.g.,domainsorevensingle helicesorloops).Finally,thestructureoftheconnectedsidechainsandthecrosslinkermoiety impartadistancerestraintthatcanbeusedformolecularmodelingpurposesbecauseanupper

Trends

Chemicalcrosslinkingfollowedbythe massspectrometricanalysisofcross linkedpeptides(XL MS)identifiescon tactsites between residueswithina singleorbetweenmultipleproteins.

TheapplicationofXL MStomanybio logically relevantmolecularmachines hasbeenshown,witharapidlygrowing numberofsuccessfulstudiesreported inthepast2 3years.

Crosslinkingdataareusefulinintegra tivemodelingworkflowsbyproviding distance restraintsonthe surfaceof foldedproteins and complexes. XL MShasbeenshowntobeparticularly powerfulincombinationwith3Dcryo electronmicroscopy.

1DepartmentofBiology,Instituteof MolecularSystemsBiology,ETH Zürich,8093Zürich,Switzerland

2DepartmentofBiology,Universityof Konstanz,78457Konstanz,Germany

3FacultyofScience,Universityof Zürich,Zürich,Switzerland

*Correspondence:

leitner@imsb.biol.ethz.ch(A.Leitner) andaebersold@imsb.biol.ethz.ch (R.Aebersold).

20

Erschienen in: Trends in Biochemical Sciences ; 41 (2016), 1. - S. 20-32 https://dx.doi.org/10.1016/j.tibs.2015.10.008

boundofthephysicaldistancecanbecalculatedforaparticularcrosslinker.AnadvantageofXL- MSover otherstructural techniques isthatit candealwith limitedsampleheterogeneity or dynamiccomplexesasitprovidesanaveragedensemblemeasure.

ThegeneralapproachofXL-MSistochemicallycrosslinkproteinsintheirnativeornative-like statethengeneratecrosslinkedpeptidesbyenzymaticdigestionofthecrosslinkedsamplesand identifythesequenceofthecrosslinkedpeptidesviatandemMS.Mostcommonly,apurified proteincomplexisincubatedwithacrosslinkingreagentthatformscovalentbondsbetween reactivesurface-exposedaminoacidsidechainsandthesamplesaredigestedwithtrypsin.The resultingpeptidescanbeenrichedforcrosslinkedpeptidesandareanalyzedbyliquidchro- matographytandem MS (LC-MS/MS). Computationalanalysis of the MS/MS data enables sequence assignment of the crosslinked peptides as well as the localization of the exact crosslinkingsites.Anoverviewofthegeneralworkflowandrecentinnovationsarepresented inFigure1(KeyFigure).

Neither chemical crosslinking nor the use of MS for the identificationof single crosslinked proteinsisbyitselfanovelconcept.However,duetomultiplemajortechnicalobstaclesithad beenimpossibleuntilrecentlytodirectlyandreliablyidentifycrosslinkedpeptidesfromprotein complexesbyMS(see[1–4]forrecentreviews).Afterearlyworkonindividualproteins[5–7]and proteincomplexes[8,9],notablybytheSinzandRappsilbergroups,Aebersoldandcoworkers introducedthefirstrobustgeneralworkflowbyoptimizingwet-labprotocolsandthedevelop- mentofthepubliclyavailablexQuest/xProphetopen-sourcesoftwaresuitefortheanalysisand validationofcrosslinksfromlargeproteincomplexesbyMS[10–12].

Overthepastfewyearsthefieldhasseensignificantprogressandseveralmethodstoenrich crosslinks [13–15], various crosslinking chemistries [9,10,15–22], and the introduction of multipledetectionandidentificationstrategies[9–11,23–28].Statisticalmodelsthatdifferentiate truefromfalseidentificationshavealsobeendeveloped[10].Recentadditionstothefieldalso includevarioustoolsto visualizecrosslinksasnetworksofconnectedresiduesorasspatial restraintsonthesurfaceofproteinstructures,agreatimprovementoverthemanualannotation andmappingoftheearlydays[29–32].Althoughspacelimitationsdonotallowustocoverallof theexcitingadvancesinthefieldinthisreview,wehighlightheresomeparticularlynoteworthy methodological developments. We then discuss how these advances have provided new insights into the structure and function of complexes such as proteasomes, ribosomes, polymerases,chromatinremodelers,photosystemcomplexes,andtheirassociatedfactors.

RecentAdvancesinXL-MSMethodology

ThemainbottlenecksinXL-MShavelongbeenthelackofdedicatedmethodstospecifically enrichcrosslinkedpeptidesfromthecomplexpeptidemixturegeneratedbythedigestionof largeproteincomplexesand,evenmoreso,thereliableidentificationofcrosslinkedpeptides [1].Overthepastfewyearstherehasbeengreatprogressinmakingworkflowsmorerobust andnowmultiple relativelymatureapproaches exist,at leastfor theanalysis ofabundant recombinantlyexpressedorreconstitutedproteincomplexes(seeApplicationsofXL-MSfor ElucidationoftheStructureandFunctionofLargeProteinComplexes).Numerousapplications ofXL-MShaveshownthatthemethodsarerobustbutthenumberofobtainedcrosslinksis sometimesmarginal.Thishascreatedgrowinginterestinmethodsthatincreasethearsenalof availablecrosslinking chemistries and thus the number of identified crosslinks.The most commonlyusedcrosslinkingreagentsarehomobifunctionalactiveesterssuchasdisuccini- midylsuberate(DSS)orbis(sulfosuccinimidyl)suberate(BS³)thatinducenucleophilicattacks onprimaryaminesandthusrelyonthecouplingoflysineresidues.Oneexcitingwaytoexpand beyondthis established strategyis the recently developedcrosslinking chemistry specific forcarboxyl groupsusing homobifunctionaldihydrazides ascrosslinking reagents[18,33].

Key Figure

An Overview of the Crosslin king Mass Spectrometry (XL-MS) Workflow

[

Sample type

] -d ^~

Complexes with up to 100 subunits Pull downs, In vivo applications

'

Complementary chemistry targeting

l

Cross linking acidic residues

reaction Affinity tagged and cleavable reagents

Increased use of size exclusion, ion

[

Fractionation,

l ^{--- --- =r:: .:::::c::}

exchange, and affinity chromatography

enrichment

- :c

to enrich crosslinked peptides

[

MS acquisition

) lhl lhl

Faster and more sensitive spectrometers

...

¹⁰ New software for differential quantification

[

Data analysis

) lL: ^·- ·- ·-

calculation of false discovery rate

Q

Dedicated software for sequence graphs

l

Visualization,

] ^::Ll£[

Structural mapping and filtering software modeling

- i n lliochemic!ll Sci!ncet

Agure 1. A schematic representation of the crosslri<ing wor1<11ow ~eft and middle) isaocompanled ~ recent nnovations In the field (right). XL MS is generally perfcrmed on purified ~es. llltematively, cells can be ncwated with crosslii<ers qn

W\.o crossli1ki1g). The ~e is then digested by a protease Into peptldes and crossJnked peptides (with black lnes) are eniched or ractlonated by siZe excluslon dYomat~. ion exchange dYomat~. or affinity chromatography.

Errt:tled or fractionated samples are separated by liqlld dYomatography and analyzed by tandem MS. The crossJnked peptides are Identified (I D) and an estimate lor the false discovery rate is obtained (FOR). A list of validated Identifications Is complied. The vaidated crossJii<s can be displayed as sequenoe graphs, visualiZed on atonic structures, or used br assembly modeing.

This crosslinking chemistry can provide distance restraints that are highly complementary to those obtained from lysine crosslinking [19,34,35], thereby considerably expanding the num- ber of crosslinks and thus the amount of structural information that can be obtained from a protein complex. However, carboxyls are inherently less reactive than amine groups and require a more elaborate procedure for crosslinking. Zero-length crosslinking, using carbo- diimides to bridge amino and carboxyl groups directly without insertion of a spacer, has recently also been applied to larger protein complexes [36,37]. although data analysis in such experi- ments remains more challenging.

The abovementioned reactions are specific for one or several amino acids, but more promiscu- ous reagents also exist, including those with photoreactive groups [38,39]. Expanding the list of reactive residues increases the chances that proteins with fewer acidic or basic residues, such as membrane proteins, can also be crosslinked. A completely unspecific chemistry would make

thetechniquelesssuitedfortheanalysisoflargercomplexes.Theuseofheterobifunctional reagents containing one specific and one photoreactive group is an attractive solution to circumventsuchproblems[40]andpromises highcrosslinkcoverageforindividualproteins [41].A somewhatdifferent applicationof photochemical crosslinking was used ina recent breakthrough study that introduced the first semiautomated workflow for assigning RNA– proteinbindingsites[42].Inthiscasetheintrinsicphotochemicalreactivityofthenucleobases uraciland5-thiouracilcanbeexploited.Eveniftheanalysisofsuchdata remainsextremely demanding,thisstudydoes opennewhorizons forextendingtheconceptbeyondprotein– proteininteractions.

Sofar,XL-MShasbeenprimarilyusedtostudythestructureofproteincomplexesasstatic entities.BecauseMSprovidesnotonlyqualitativebutalsoquantitativeinformation,thedevel- opmentof ‘quantitative’XL-MS (qXL-MS) workflowsis anobviousnext step. Additionof a quantitativedimensionislikelytoenabledeeperinsightintothecompositionalandconforma- tionaldifferencesoftheproteincomplexespresentindifferentbiologicalstates.Thesechanges wouldbereflectedinchangesintheabundanceofspecificcrosslinks.ThepotentialofqXL-MS remainsunderexplored,partlyduetothelackofadaptedexperimentalworkflowsandalackof software tools that can handle qXL-MS data [43]. Initial studies have used manual data extractionandquantificationandwerethereforenoteasilyscalabletomoreambitiousprojects [43–45].However,Schmidtetal.haveprovidedanexcitingearlyexampleofthepotentialofthis methodbycomparingunphosphorylatedandphosphorylatedstatesofaspinachchloroplast ATPase[44].Recently,theAebersoldgrouphasintroducedanewsoftwaresolution forthe automatedanalysisofqXL-MSdatasets[46].Theyalsodemonstrateditsuseforthestructural analysisofproteincomplexesthatexistindistinguishablestableconformationalstatessuchas thechaperoninTRiC/CCT.

Despitetheserecentadvances,crosslinkingrestraintsalonearenotsufficienttoelucidatethe precisestructureofaproteincomplex.Thus,itwasrealizedthatcombiningXL-MSwithother structuralapproachesinahybridsettingoffersgreatadvantages(see TheRoleofXL-MSin IntegrativeStructuralBiology)[47].Forexample,itwasdemonstratedthatamoreintegrated viewofthesubunitarrangementofanintactproteincomplexcanbeobtainedbyexamininga sample with XL-MS, native MS, and ion-mobility spectrometry coupled to MS so that the complementarystructuralMSmethodsprovideadditionalandorthogonalrestraintscompared withusingXL-MSalone[48].Advancedcomputationalpipelinescansupportintegrationofthe differenttypes ofdata fromcrosslinkingand fromothersources –suchas complementary experimentalorcomputationalmethods–tofacilitatetheunbiaseduseofcrosslinkingandother restraintstogeneratereliablemodelsforproteincomplexes[49–52].Thisledtothegenerationof agenerichybridapproachfortheaccuratemodelingoflargeproteinassemblies,whereXL-MS togetherwithpartialcrystalstructuresofsubunitsweresuccessfullyintegratedwithmodeling.

This integrated method generally permits alarge number of low-resolutionrestraints to be effectivelyintegratedandoptimized.Asakeyexampleofitsusefulness,itwasusedtoelucidate thestructureofanintacttranslationinitiationfactorthathadevadedcrystallizationeffortsfortwo decades,boundtotheribosome[49].

XL-MSisnotlimitedbythesizeofaproteincomplexandprovidesawealthofinformationon theconnectivity,interaction,andrelativeorientationofsubunitswithinacomplex.Acrosslink contactalsocontains(relativelylow resolution)spatial informationinitself.Recenttechno- logical advances [53,54] have now rendered this method applicable to physiologically importantbiologicalassembliesthataredifficulttoexamineusingmoretraditionalstructural techniques.Thistechnologyisthereforeoptimallysuitedforstudyingthearchitectureoflarge proteincomplexesandtheirinteractionpartnersintheirnativeenvironmentandonamedium tolarge scale.

The Role of XL-MS in Integrative Structural Biology Integrative Structural Biology

Common structural techniques such as X-ray crystallography, NMR spectroscopy, and 30 EM have strengths and specific application fields but no technique by itself may perform as well as a combination of methods when complex molecular machines are being analyzed. Therefore, in recent years structural biology has seen the rise of a

new

methodological paradigm In the form of integrative structural biology and modeling. It was pioneered by creating a model for one of the largest and complex macromolecular machines: the nuclear pore complex [55]. Integrative structural biology determines structure in three stages (Figure 2). First, a comprehensive list of 'parts' of the complex of interest is compiled. These parts can include, for example, atomic models of single subunits in addition to the sequence of the component proteins. Flexible regions or regions for which no structural information is available can be represented by beads (as placeholders) occupying the space of single amino acids, secondary structure elements, or entire protein domains [56]. Additionally, knONn interactions between parts, such as copur- ification data derived from affinity-purification MS (AP-MS) and data from yeast two-hybrid experiments can be integrated. Available data from XL-MS or other structural MS techniques, and from NMR spectroscopy or electron paramagnetic resonance, are also translated into spatial restraints. In the second stage, these restraints are combined into a single scoring function. Multiple sampling of such functions generates a set of assembly models that have

Parts and interactions

Atomic models

Bead models

EM densities or SAXS rurves

AP·MS

XL·MS

NMR,EPR

Sampling

Olsta.nc:e restraints exduded volume overlap with densities uncertainty parameters

lo'

Conformational sampling

Assessment

b

,,' ' ^{, '} ,'

.

^/Score

_,'

:

C....l •

Localization density

ll"endlln lllochemlcal!lc!encel Figure 2. The Role of Crossllnklng Mass Spectrometry (XL MS) In Integrative Structural Biology. Integrative structural J:Mologycan be Sttxllvided In me consecutive steps moon from left to right. Flrst, structural oompooents such as atcrolc models and electron mlcrosoopy (EM) denslty maps are gathered. Parts that are not descriled at atomic resolution are represented by beads oocopyng the spaoe of arnnoaddsor prot

en

dcmans (lett). lnterocticn data between parts are cdlected tern all'lity puiftcation MS lAP MS), XL MS, NMR, cr electron paramagnetic resonance (EPA) experiments. Interactions are then translated ntodlstance restraints and a1 restraints are oombl1ed i1toa SCOring function.

~ng translations and rotaticns to the ndMdual parts Is equivalent to extensive4y sarnping the sccrng Ulctlon. Models With higher scores satisfy more of the restrants (cent~. Fnaly, models are ranked by their score and the best models are dustered byther structUral simlarity [root mean square deviation (RMSD)). A representative duster (red broken ine square) Is then averaged Into a localiZation density map that lh.Jstrates the most probable location of the ndvidual parts (ri!tlt).

SAXS, smal angle X ray scatterng. Pdapted from [49].

differentdegreesofconsistencywiththeinputdata.Inthethirdstage,modelsarescoredbased onhowwelltheysatisfyspatialrestraintsandonecanthengenerateaconsensusmodel[56].

Thisworkflowiscurrentlyimplementedingenericsoftwaresuchas theIntegrativeModeling Platform(IMP)[57],HADDOCK[58,59],andROSETTA[60].Thus,byintegratingdifferentlevels ofheterogeneousstructuralinformation,integrativestructuralbiologyisabletoprovidestructural modelsofassemblieswithlargenumbersofsubunitsandsomedegreeofheterogeneityand flexibility.

RestraintsfromXL-MS

XL-MSplaysapivotalroleinintegrativestructuralbiology.Therestraintsderivedbycrosslinking areparticularlyimportantto determinethe topologyandrelativeorientationoftheindividual subunits.XL-MSprovidesdefinitivebinaryinteractiondata(e.g.,subunitAiscloseinspaceto subunitB)andspatialrestraintsbetweenproteinswitharesolutionofseveralaminoacidsatthe primarysequencelevel(limitedbythelocationofcrosslinkableresidues).Theserestraintsarein therange7–30Å,withamediandistanceofapproximately15Åforthemostcommonlyused lysine-reactive reagents[14,61,62] and slightly shorter forcarboxyl-reactive hydrazides and zero-length crosslinks [18]. This allows the determination of the proximity and, if multiple crosslinksincomplementaryregionsarefound,therelativeorientationofthesubunits.Addi- tionally,thetechniquecandetectinteractionsofthesamesitewithmultipletargetsites(one-to- manyconnections)andthusidentifymutuallyexclusiveassemblies.Consequently,crosslinks canprovideastrongdiscriminantbetween falseandtrue solutionsgeneratedbyintegrative modeling.

ApplicationExamples

XL-MSdataprovideorthogonalinformationtolow-resolution(10–20Å)3DEMdensitymaps.For example,theChaitgroupproposedanintegrativemodelforthenuclearporesubcomplexNup84 combiningdatafromX-raycrystallography,EM,andXL-MS[50].Withthegoalofobtainingalarge setofcrosslinks,theycombinedthewidelyusedreagentDSS,whichtargetslysineresidues,witha carbodiimidetogeneratezero-lengthcrosslinks.Bycombiningvalidatedcrosslinksfromthetwo chemistriestheyobtainedasetof59and47intermolecularcrosslinks,respectively.Theresulting modelisconsistentwiththerecentlypublishedcrystalstructure[63].Acombinationoftwoflavors of3DEM–cryo-electrontomographyandsingle-particleEM–andXL-MSwasemployedbythe BeckgrouptodefinerestraintswithinthepurifiedNup87complexandbetweendifferentcopiesof thesamecomplexwithintheassemblednuclearporecomplex[64].

Anotherexampleofintegrativestructuremodelingappliedtoalargemacromolecularmachineis themolecularstructureofeukaryoticinitiationfactor(eIF)3[49].HeretheauthorscombinedX- raycrystallography,single-particleEM,andXL-MS.Alargesetofexperimentsallowedcom- prehensivemappingofthelysine–lysineinterproteininteractionsbetweentheeIF3complexand the40Sribosomalsubunit(155crosslinks)andwithinthe40Sribosomalproteins(461cross- links).Giventhesizeofthedataset,therecurrenceofaparticularcrosslinkcouldbeusedasa furtherconfidencemetrictoassessthevalidityofthecreatedstructuralmodel[49].

Limitations

ThemajorlimitationofXL-MSfromthestructuralpointofviewisthatitcannotdirectlydetermine therelativestoichiometryofsubunitsinacomplex,althoughthisinformationcanbederivedfrom complementaryquantitativeMSmethods.XL-MSalsocannoteasilydistinguishbetweenintra- subunitcrosslinksandcrosslinksbetweenmembersofahomomericinteraction.Homodimeric complexescanbereconstitutedfrommixturesof¹⁴N-and¹⁵N-labeledsubunits[65,66]butthis approachisnot easilyextensible tolarger complexes.Forthesame reason,interactionsof differentcopiesofthesamesubunitwithdifferentpartnerscannotyetbedeconvoluted;thatis, attributedtosinglesubunits.

ApplicationsofXL-MSforElucidationof theStructureandFunctionof Large ProteinComplexes

WhereasthefirstapplicationsofXL-MSintheearly2000sfocusedonindividualproteinsorsmall complexes,theRappsilbergroupdemonstrateditsuseforlargerproteinassembliesinseminal studiesonthetetramerickinetochoresubcomplexNDC80[8]andRNApolymeraseII(PolII)in complexwithtranscriptioninitiationfactorIIF(TFIIF)[9],a15-subunit,670-kDacomplex.Inthe latterstudy,morethan90high-confidencelysine–lysinecontactsbetweenPolIIandTFIIFwere usedtopositiontheinitiationfactorontheknownstructureofthepolymerase(Figure3A).Inthe followingyears,morelaboratoriesspecializinginMSandproteomicsadoptedthetechnique, althoughittooktimeforthetechniquetobecomeacceptedinthestructuralbiologycommunity.

Themoregeneralacceptanceresultedfromimprovementsinwet-labprotocols,theavailability ofmoresensitiveMSinstrumentation,andmorerobustsoftwarethatwasabletoroutinelyand robustlydealwithlargerdatasets,asoutlinedabove.

Inrecentyears,themethodologyhasbeenappliedtoseveralmolecularmachinesinvolvedinthe mostessentialcellularprocesses.Wediscussselectedexamplesinthefollowingparagraphs.

Box1discussestheemergingexpansionoftheXL-MSconcepttotheproteomescale.

Proteasomes

Theproteasomeisthemostimportantcellularmachineryforproteindegradation.Itcomprises two main parts, the19S cap particle and the 20S core particle,which constitutethe 26S holocomplex.Althoughthestructureofthe20Scoreparticlewasdeterminedbycrystallography twodecadesago[67],thestructureofthecomplete26Sassemblyhasremainedelusiveasthe wholecomplexremainedrefractorytocrystallography.Instead,newinsightsintothearchitec- tureoftheproteasomeweregainedbyXL-MS.ThefirstapplicationofXL-MStotheproteasome datesbackto2006,whenRobinsonandcoworkersusednativeMSandcrosslinkingtoobtain partialinteractionsbetweensubunitsinthe19SsubcomplexoftheSaccharomycescerevisiae proteasome[68].However,theactualcrosslinkingsiteswerenotdeterminedinthiscase.Spatial restraints obtained fromXL-MS were used to eventuallycharacterize the complete subunit architectureofthe26Sproteasome[69,70].Thearchitectureoftheintactproteasomederived fromahybridmethodwith contributionsfromXL-MS,EMandothermethodsproved tobe nearlyidentical tothe architectureobtained froman independentcryo-EM project thatwas publishedatthesametime[71].Inaddition,theproteasomehasservedasamodelcomplexfor the development of new experimental and computational workflows connected to XL-MS [11,14,18,72].

RibosomesandAssociatedProteins

Becauseof their essential and diversefunction, ribosomes are amongthe most frequently studiedtargetsinstructuralbiologyoflargeproteincomplexes,includingstructuralproteomics andXL-MS.Lauberand Reillydemonstrated theutility ofanew lysine-reactivecrosslinking reagent,diethylsuberthioimidate,byapplyingittotheEscherichiacoliribosome[73].Later,the samegroupusedthismethodtolocalizethebindingsiteofribosomalproteinS1onthecomplex [74].Dueto itsdynamics,X-raycrystallographystudies cannotobserveS1, asituationthat highlightstherelevanceofXL-MSasanalternativeapproachforprobingsubunitrelationshipsin flexibleassemblies.Theribosomeisdependentoninteractionswithotherproteinsandprotein complexesforitsbiogenesisandfunction.Forexample,associationwitheIF3isrequiredfor translationinitiationineukaryotes.SpatialrestraintsfromXL-MSwereusedincombinationwith EM, X-ray crystallography of selected subunits, and integrative modeling to elucidate the architectureofthe40Sribosome–eIF1–eIF3complexofS.cerevisiae[49](Figure3B).Inthis study,965intra-andinterproteincrosslinkswereobtainedonthecompleteassembly,makingit themostcomprehensiveXL-MSdatasetfromindividualcomplexestodate.Additionalcross- linkingstudiesoneIFswereperformedbytheRobinsongroup[75,76].

(A)

(B)

Localization of subunits in hybrid model ' '

Manual docking of interacting proteins onto core structure

Fitting of subunits in EM maps

ern

ern

Statistical modeling of subunit arrangement

n.a In lllodlemlcal Sclencel Agure 3. Examples d Various Uses of Crosslinking Restraints for Elucidation of the Structure of Protein Complexes. (A) Mooual posltloring of the transcr~tion lritiation factor IIF (TFIIF) di'neri:zalion domail onto the structue of RNA polymerase II. Reproduced, with perrrisslon, from [9}. (B) Localization of the tllordered C terminal domain of 91J<aryotic lritiation factor 3a (eiF3a) on the 40S ribosome by crossMii<lng glided h~d modeing. Reproduoed from (49}. (q Pootioni'lg of subuits of the mttochCI'ldrial ribosome on ayo electrCI'l rricrosoopy (cryo EM) density maps. Broken ines comect crossli'lked rasldues. Reproduced from (34}. (D) Statistical derivation of the stburit armngement of the TCP 1 ring ocrnplex (mlq/chaperCrli'l CCI'ltalni'lg TCP 1 (CCl) chaperoni'l. CCT suburits am nurrbel'ed tom 1 to 8. Sdld ines i'ldlcate crossinked st.burits; bfd<en Iiles Indicate v..tllch pal's of homctypic Inter ring CCI'ltacts occu-. Rep-oduoed, wtth permission, tOll [97}.

The mitochondrial ribosome (mitoribosome) differs substantially from the cytosolic form of the complex in eukaryotes in terms of both structure and cellular function. Specifically, its RNA content is only about one-third of the total mass, compared with two-thirds for the cytosolic ribosome, and it is responsible for the synthesis of only a limited number of hydrophobic

membraneproteins.Unfortunately,itslowabundancehasmadethemitoribosomepractically inaccessible to structureelucidation by crystallography.Thus, high-resolution cryo-EM was usedinseveralrecentprojectstodecipherthearchitectureofthemitoribosomeofyeastand mammals[34,35,77–80].Forthreeofthesestudies[34,35,77],XL-MScontributedrestraints thatwerecriticalinlocalizingindividualsubunitswithinthismassivecomplex(Figure3C).

PolymerasesandAssociatedComplexes

RNAtranscriptionisanintricateprocessthatinvolvestheinterplayofseveraldifferentprotein complexes,amongthemRNApolymerasesthemselvesandtheMediatorcomplexthatactsasa transcriptionalactivator,aswellasvarioustranscriptionfactorsthatassemblewithRNAPolIIin thepreinitiationcomplex(PIC).Asmentionedabove,thePolII–TFIIFcomplexwasthefirstlarge assemblystudiedindetailbyXL-MS[9].Inrecentyears,XL-MShasprovidedimportantinsights intothearchitectureofmanyofthesecomplexes.Thisincludesinformationaboutthesubunit organizationofthepolymerasesthemselvesaswellastheorganizationofpolymeraseswith additionalproteinsintohigher-orderfunctionalunits.StudieshavetargetedRNAPolI[81,82],II [83],andIII[84]aswellasthePolII–PICcomplex[85,86]andthePolII-cappingenzymecomplex [87].AsubcomplexoftheMediatorcomplex(theso-calledmiddlemodule)[88]andtheMediator headmoduleincomplexwiththeC-terminaldomainofPolII[89]werealsoinvestigatedbyXL- MS.Morerecently,XL-MSwasusedtoprovidespatialrestraintsontheMediatorcoreandthe PolII–Mediatorcoreinitiationcomplexinthemostcomprehensivestudyonpolymerasesand associatedcomplexestodate[90].

ChromatinRemodelers

Chromatin remodeling complexes are responsible for the reorganization of nucleosomes throughvariousmechanisms. Recentcrosslinkingstudieshavetargetedseveralfamiliesof remodelerenzymes.Forexample,theAebersoldgroupcontributedtotwohybridstructural biologyprojectsfocusing on theremodelers INO80 [91]and Swi2/Snf2-related1 (SWR1) [92].Inbothcases,XL-MSdataaidedinpositioningindividualsubunits intoEMmaps.For INO80,crosslinksbetweentheremodelerandthenucleosomewereobtained,providinga firstglimpseintotheregulatoryactivityofthecomplex.Recently,Vermeulenandcoworkers usedAP-MStoexaminetheinteractionnetworkofthenucleosomeremodelinganddeace- tylasecomplex(NuRD)remodelerandusedXL-MStostudythearchitectureofthehuman NuRD[93].

Box1.CrosslinkingofProteinNetworksandWholeProteomes

Recently,thefirststudiestoapplyXL MStoalargersetofcomplexesthatwereaffinitypurifieddirectlyfromcellshave emerged[48,110].Herzogetal.studiedtheinteractionnetworkofhumanproteinphosphatase2A(PP2A)[110]by purifyingproteinsassociatingwithatotalof14differentbaitproteins(PP2Asubunitsandknowninteractors)and supportedtheresultingnetworkwithXL MS derivedspatialrestraintsobtainedfromcrosslinkingdirectlyontheaffinity beads.TheAebersoldandRobinsongroupstudiedtheassemblyoftheproteasomelidinyeastusing,amongother methods,XL MSonpull downsfromtaggedproteasomesubunits[48].Theextensionofcrosslinkingtosuchcomplexes expressedatendogenouslevelswillmaketheconceptmoreaccessibletostudyproteininteractionsinunprecedented detail.Currentlimitationsofthisapproacharethelimitedamountofmaterialthatcanberecoveredbyaffinitypurification andthelowthroughputforlargernumbersofbaits.

TheextensionofXL MStotheproteomescalehasmadetheprobingofproteininteractionsdirectlyinlivingcellsorincell lysatespossibletosomeextent.Usingtheproteininteractionreporterconceptbasedongas phasecleavable,affinity taggedcrosslinkingreagents,Bruceandcoworkershaveshownapplicationstoadiverserangeoforganismssuchas Shewanellaoneidensis[111],Escherichiacoli[112 114],Pseudomonasaeruginosa[115],andevenhumancelllines [116];othergroupshavereportedsimilarconcepts[117 119].Recently,thegroupofBrucealsodemonstratedthata quantitativedimensioncanbeaddedtoproteome widecrosslinkingdatawiththehelpofmetabolicstable isotope labeling[120].Thesepotentiallyexcitingapproachesstillhavedifficultyinidentifyinglargernumbersofinteractions betweenproteinsorproteincomplexesthatarenothighlyabundantliketheribosome,certainchaperones,orhistones, butimprovementsinexperimentalprotocolsandinstrumentationwillincreasethecoverageinthefuture,asexemplified byrecentworkfromHeckandcoworkers[121].

ComplexesInvolvedinPhotosynthesis

Photosynthesisinorganismssuchascyanobacteriaandredalgaeinvolveslightabsorptionand energytransferviaproteincomplexes.SeveralrecentstudieshavetakenadvantageofXL-MSto obtaininformationaboutthespatialproximityofindividualproteinsinthesemolecularmachines.

Forexample,Liuetal.studiedaphycobilisome–photosystemI–photosystemII‘megacomplex’ inthecyanobacterium Synechocystis[94].ThearchitectureofphycobilisomesfromThermo- synechococcusvulcanuswasalsostudiedbycrosslinking[95].Finally,thecomplexbetweenthe Fenna–Matthews–OlsonantennaproteinandthereactioncentercorecomplexinChlorobac- ulumtepidumwasalsoprobedbyXL-MS[96].

OtherComplexes

Inadditiontothefamiliesofmolecularmachinesdiscussedabove,severalotherlargeprotein complexes thathave constitutedlongstanding problems instructural and cell biologyhave alreadybeenstudiedbyXL-MS.Somenotableexamplesinclude:thegroupIIchaperoninTCP-1 ringcomplex(TRiC)/chaperonincontainingTCP-1(CCT)[97–99],includingtheuseofXL-MSto identifyitssubstrate-bindingsites[99](Figure2D);theSAGAtranscriptioncoactivatorcomplex [100,101];polycombrepressivecomplex2[102];spinachF-typeATPase[44];thePyrococcus furiosusCmrcomplex,partoftheCRISPRsysteminprokaryotes[20];andthemetabolon,an assemblyofmitochondrialenzymesupercomplexes[103].

Very large macromolecular assemblies that may comprise hundreds of individual proteins currentlyremaininaccessibletoacomprehensiveXL-MSanalysisintheirentirety.Thisispartly becauseofthedifficultiesinisolatingsufficientamountsofthemashomogeneousentities,but alsobecauseexperimentalchallengeswiththecurrentlyavailableinstrumentationdonotallow comprehensivecoverage. However, suchverylarge machinescanbetargetedbystudying partialassemblieswiththeintenttothencomputationallyassemblethesubcomplexesintothe structureofthewholesystem.Forexample,severalsubcomplexesofthenuclearporecomplex havebeenstudiedusingcrosslinkingstrategies[50,64,104,105].Similarly,complexesassoci- atedwiththekinetochorehavebeenstudiedbyXL-MS[8,106–109].

Insummary,theapplicationofXL-MStothisdiversegroupofmolecularmachineshasprovided crucial information about their hierarchical organization and has supported complementary experimental and computational structural techniques. XL-MS, therefore, has been firmly establishedasanimportantcomponentofthehybridstructuralbiologytoolbox.

ConcludingRemarks

XL-MShasmadeessentialcontributionstostructuralbiologythatarereflectedinthesignificant increaseinthenumberofpublishedapplicationsofthistechnology.Itisencouragingtosee increasinginterestinandabroadrangeofsuccessfulapplicationsofamethodthatonlyafew yearsagowasrestrictedtoproof-of-principleexperimentsonmodelproteins.Itcanbeexpected thatthemorewidespreadacceptanceofXL-MSasaviablepartofstructuralbiologyprojectswill resultinexciting newdirectionsforthistechnique.The immediatechallengesforXL-MS are relatedtotwodifferentpartsoftheworkflow(see OutstandingQuestions).Thefieldremains heterogeneous,withvariousexperimentalprotocolsandsoftwareapplications,andthisvariety mayseemintimidatingtonewcomers. Itcanbeexpectedthatinthecomingyearspreferred workflowswillemerge,andthiscouldbeacceleratedbycommunityeffortsrelatedtostandardi- zationandbenchmarking. Samplesofever-increasingcomplexity willbestudied byXL-MS;

however,whethertheeventualgoalofcomprehensiveinteractionprofilingofwholeproteomesis achievableremainstobeseen.Nevertheless,XL-MS studiesofcomplexes thatarepartially purifiedfromtheirnativeenvironmentwillcertainlyprovidenewinsightsaboutproteininteraction networksandtheirchangesonperturbation;forexample,asaresultofmutationsconnectedto diseases.Ifsuchassembliescanbeprobedroutinelyatconsiderabledepthandfromlimited

OutstandingQuestions

How can we further increase the amount of information from XL MS experimentsforpurifiedcomplexes?

Howcanwebetter integrateXL MS datainmodelingpipelines?

Howcan weexpandtheconceptof XL MS to characterize cellular com plexesintheirnear nativestateatthe samedepthofcoverageachievedfor purifiedcomplexes?

How can we make this technology moreaccessibletonon experts?

sample amounts, this will solidify the relevance of crosslinking-based methods not only in structuralbiologybutalsoinsystemsbiology,thusadvancingtheconvergenceofstructural andcellbiology.

Acknowledgments

XL MSresearchinthelaboratoryoftheauthorswas/issupportedbytheEuropeanUnion7thFrameworkProgram (PROSPECTS,HEALTH F4 2008 201648),theEuropeanResearchCouncil(ERCAdvancedGrants233226and670821), andtheInnovativeMedicinesInitiativeJointUndertaking(ULTRA DD,grantnumber115366).M.F.wassupportedbya LongTermFellowshipfromtheEuropeanMolecularBiologyOrganization(EMBO).

References

1. Leitner,A.etal.(2010)Probingnativeproteinstructuresby chemicalcross-linking,massspectrometry,andbioinformatics.

Mol.Cell.Proteomics9,1634–1649

2. Rappsilber,J.(2011)Thebeginningofabeautifulfriendship:

cross-linking/mass spectrometry and modelling of proteins andmulti-proteincomplexes.J.Struct.Biol.173,530–540 3. Walzthoeni,T.etal.(2013)Massspectrometrysupporteddeter-

minationofproteincomplexstructure.Curr.Opin.Struct.Biol.

23,252–260

4. Sinz,A.(2014)Theadvancementofchemicalcross-linkingand massspectrometryforstructuralproteomics:fromsinglepro- teinstoproteininteractionnetworks.ExpertRev.Proteomics11, 733–743

5. Young,M.M.etal.(2000)Highthroughputproteinfoldidentifi- cationbyusingexperimentalconstraintsderivedfromintramo- lecularcross-linksandmassspectrometry.Proc.Natl.Acad.Sci.

U.S.A.97,5802–5806

6. Rappsilber,J.etal.(2000)Agenericstrategytoanalyzethe spatialorganizationofmulti-proteincomplexesbycross-linking andmassspectrometry.Anal.Chem.72,267–275 7. Dihazi, G.H. and Sinz,A. (2003) Mapping low-resolution

three-dimensionalproteinstructuresusingchemicalcross- linkingandFouriertransformion-cyclotronresonancemass spectrometry.RapidCommun.MassSpectrom.17,2005–

2014

8. Maiolica,A.etal.(2007)Structuralanalysisofmultiproteincom- plexes by cross-linking, massspectrometry, and database searching.Mol.Cell.Proteomics6,2200–2211

9. Chen,Z.A.etal.(2010)ArchitectureoftheRNApolymeraseII– TFIIFcomplexrevealedbycross-linkingandmassspectrometry.

EMBOJ.29,717–726

10. Rinner,O.etal.(2008)Identificationofcross-linkedpeptides fromlargesequencedatabases.Nat.Methods5,315–318 11. Walzthoeni,T.etal.(2012)Falsediscoveryrateestimationfor

cross-linkedpeptidesidentifiedbymassspectrometry.Nat.

Methods9,901–903

12. Leitner,A.etal.(2014)Lysine-specificchemicalcross-linkingof proteincomplexesandidentificationofcross-linkingsitesusing LC-MS/MSandthexQuest/xProphetsoftwarepipeline.Nat.

Protoc.9,120–137

13. Trester-Zedlitz, M. et al. (2003) A modular cross-linking approachfor exploringproteininteractions.J.Am.Chem.

Soc.125,2416–2425

14. Leitner,A.etal.(2012)Expandingthechemicalcross-linking toolboxbytheuseofmultipleproteasesandenrichmentbysize exclusionchromatography.Mol.Cell.Proteomics11,http://dx.

doi.org/10.1074/mcp.M111.014126M111.014126 15. Fritzsche,R.etal.(2012)Optimizingtheenrichmentofcross-

linkedproductsformassspectrometricproteinanalysis.Rapid Commun.MassSpectrom.26,653–658

16. Müller,M.Q.etal.(2010)Fragmentationbehaviorofathiourea- based reagent for protein structureanalysis by collision- induceddissociativechemicalcross-linking.J.MassSpec- trom.45,880–891

17. Kölbel,K.etal.(2012)Analysisofpeptidesecondarystructures byphotoactivatableaminoacidanalogues.Angew.Chem.Int.

Ed.Engl.51,12602–12605

18. Leitner,A.etal.(2014)Chemicalcross-linking/massspectrom- etrytargetingacidicresiduesinproteinsandproteincomplexes.

Proc.Natl.Acad.Sci.U.S.A.111,9455–9460

19. Yu,C.etal.(2014)Developingnewisotope-codedmassspec- trometry-cleavablecross-linkersforelucidatingproteinstruc- tures.Anal.Chem.86,2099–2106

20. Benda,C.etal.(2014)StructuralmodelofaCRISPRRNA- silencingcomplexrevealstheRNA-targetcleavageactivityin Cmr4.Mol.Cell56,43–54

21. Ihling,C.etal.(2015)Dissociationbehaviorofabifunctional tempo-activeesterreagentforpeptidestructureanalysisbyfree radicalinitiatedpeptidesequencing(FRIPS)massspectrometry.

J.MassSpectrom.50,396–406

22. Burke,A.M.etal.(2015)Synthesisoftwonewenrichableand MS-cleavablecross-linkerstodefineprotein–proteininteractions bymassspectrometry.Org.Biomol.Chem.13,5030–5037 23. Chu,F.etal.(2010)Findingchimeras:abioinformaticsstrategy

foridentificationofcross-linkedpeptides.Mol.Cell.Proteomics 9,25–31

24. Xu,H.etal.(2010)Databasesearchalgorithmforidentificationof intactcross-linksinproteinsandpeptidesusingtandemmass spectrometry.J.ProteomeRes.9,3384–3393

25. Hoopmann,M.R.etal.(2010)Improvedstrategiesforrapid identificationofchemicallycross-linkedpeptidesusingprotein interactionreportertechnology.J.ProteomeRes.9,6323–6333 26. Kao,A.etal.(2011)Developmentofanovelcross-linkingstrat- egyforfastandaccurateidentificationofcross-linkedpeptidesof proteincomplexes.Mol.Cell.Proteomics10,http://dx.doi.org/

10.1074/mcp.M110.002212M110.002212

27. Götze,M.etal.(2012)StavroX–asoftwareforanalyzingcross- linkedproductsinproteininteractionstudies.J.Am.Soc.Mass Spectrom.23,76–87

28. Yang,B.etal.(2012)Identificationofcross-linkedpeptidesfrom complexsamples.Nat.Methods9,904–906

29. Zheng,C.etal.(2013)XLink-DB:databaseandsoftwaretoolsfor storingandvisualizingproteininteractiontopologydata.J.Pro- teomeRes.12,1989–1995

30. Combe,C.W.etal.(2015)xiNET:cross-linknetworkmapswith residueresolution.Mol.Cell.Proteomics14,1137–1147 31. Kosinski,J.etal.(2015)XlinkAnalyzer:softwareforanalysisand

visualizationofcross-linkingdatainthecontextofthree-dimen- sionalstructures.J.Struct.Biol.189,177–183

32. Grimm,M.etal.(2015)xVis:awebserverfortheschematic visualization and interpretation of crosslink-derived spatial restraints.NucleicAcidsRes.43,W362–W369

33. Novák,P.andKruppa,G.H.(2008)Intra-molecularcross-linking ofacidicresiduesforproteinstructurestudies.Eur.J.Mass Spectrom.14,355–365

34. Greber,B.J.etal.(2014)Thecompletestructureofthelarge subunitofthemammalianmitochondrialribosome.Nature515, 283–286

35. Greber,B.J.etal.(2015)Thecompletestructureofthe55S mammalianmitochondrialribosome.Science348,303–308 36. Sriswasdi,S.etal.(2014)Probinglargeconformationalrearrange-

mentsinwild-typeandmutantspectrinusingstructuralmass spectrometry.Proc.Natl.Acad.Sci.U.S.A.111,1801–1806