Benesch et al. - 2007 - Protein complexes in the gas phase technology for structural genomics and proteomics. - Chemical reviews

8/7/2019 Benesch et al. - 2007 - Protein complexes in the gas phase technology for structural genomics and proteomics. - C

1/24

Protein Complexes in the Gas Phase: Technology for Structural Genomicsand Proteomics

Justin L. P. Benesch, Brandon T. Ruotolo, Douglas A. Simmons,, and Carol V. Robinson*

Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, U.K.

Received February 16, 2007

Contents

1. Introduction 35442. Generating Ions of Protein Complexes 3546

2.1. Electrospray Ionization 35462.1.1. Mechanism of Protein Ion Formation 35472.1.2. nESI Ionization 35472.1.3. Limitations of Electrospray Ionization 3549

3. Transmitting and Analyzing Ions of ProteinComplexes

3549

3.1. Transmitting Ions 35493.1.1. Acceleration of Gas-Phase Ions in the

Source Region3550

3.1.2. Transmission Efficiency of Massive Ionsin RF-Ion Guides

3550

3.1.3. Transmission Efficiency of Massive Ionsin Orthogonal Acceleration Time-of-FlightAnalyzers

3551

3.2. Analyzing Ions 35513.2.1. Quadrupole Analyzers 35523.2.2. Time-of-Flight Analyzers 35523.2.3. Tandem/Hybrid Instruments 3553

3.3. Instrument Modifications for NoncovalentComplexes

3553

4. Activation and Dissociation of Protein Complexes 35534.1. Collision-Induced Dissociation 35534.1.1. Activation of Macromolecular Assemblies 35544.1.2. Dissociation of Activated Macromolecular

Assemblies3556

4.1.3. The CID of a Large Oligomeric Protein:Case Study

3556

4.2. Other Activation Techniques 35584.3. Use of Gas-Phase Dissociation 3558

5. Companion Technologies for MS of ProteinComplexes

3558

5.1. Sample Isolation Technologies 35595.2. Sample Labeling Technologies 35605.3. Preionization Technologies 35615.4. Postionization Technologies 3562

6. The Future 35647. Acknowledgments 35648. References 3564

1. Introduction

How does the development of mass spectrometry (MS)for the study of protein complexes, intact in the gas phase,contribute to the twin fields of genomics and proteomics?Foremost, it is important to recognize the fact that the vastmajority of proteins do not exist as single entities in the cell,but rather interact noncovalently with additional copies ofthe same protein and/or other proteins.1 Furthermore, ad-ditional interactions can occur with nucleic acids, ligands,cofactors, or metal ions, such that the functional form of

many proteins is rarely the simple monomeric state. As such,while traditionally genomics and proteomics has focused ondetermining which proteins are encoded by the geneticmaterial, in order to understand their function, their interac-tion to form protein complexes must also be investigated.

Studying this highest state of protein organization thereforerepresents a crucial part of modern genomics and proteomicsefforts (Figure 1). Protein complexes themselves are oftencomposed of subcomplexes, that is multiprotein moduleswhich have some independent stability. Other articles in thisissue describe MS approaches for studying the individualprotein components which make up these noncovalentlybound species. These are typically performed by the cleavage

of the protein backbone into smaller constituent peptidesegments. In this review, however, we concentrate on theapplication of MS to the upper levels of protein organization,those in which noncovalent interactions are preserved (Figure1).

It is now well over a decade since intact protein complexeswere first successfully analyzed by means of MS.2-5 Theseearly reports stimulated much excitement and were swiftlyfollowed by a plethora of studies wherein protein and othernoncovalent complexes, largely from commercial or recom-binant sources, were maintained intact in the gas phase.6 Itsoon became apparent that the precision with which themasses of these species can be measured is unrivalled byany other approach. Furthermore, when coupled with theselective disruption of the intact protein complexes, eitherin solution or in the gas phase, to determine the masses ofthe constituents, stoichiometry can be established, even inthe absence of polypeptide sequence information. Suchapproaches have allowed the determination of the relativepopulations of the oligomers within polydisperse assemblies,7

have revealed substrate binding to molecular machines suchas the proteasome8 and GroEL,9 and defined the stoichiom-etry of subcomplexes within intact ribosomes.10 Coupled tothe realization that many aspects of solution-phase structureare maintained in the gas phase,11,12 these studies have

* Corresponding author. Telephone: +44 1223 763846. E-mail: [email protected]. These authors contributed equally to this work. Present Address: MDS SCIEX, 71 Four Valley Drive, Concord, OntarioL4K 4V8, Canada.

3544 Chem. Rev. 2007, 107, 35443567

10.1021/cr068289b CCC: $65.00 2007 American Chemical SocietyPublished on Web 07/25/2007


2/24

established MS of intact complexes as a complementaryapproach to many structural techniques.

Further MS developments, such as the inclusion of ionmobility spectrometry (IMS) in the analysis of proteincomplexes,13 or controlled gas-14 or solution-phase15 dis-sociation, have allowed the subunit architecture, overallshape, and subunit contacts of protein complexes to berealized. Moreover, the speed of analysis has enabled thestudy of the dynamics of these protein complexes.16 As such,a number of processes including assembly,17,18 disas-sembly,19-21 and exchange of subunits between proteincomplexes22-24 have been monitored by means of MS.Consequently, MS can now be used to examine all the levelsof protein organization (Figure 1), providing information asto the sequences of the proteins involved, the stoichiometryof their interactions, their connections, and the dynamics of

the assembled complex. Applying MS to the study of proteincomplexes, as well as more conventional applications toindividual proteins, is therefore vitally important to theburgeoning field of proteomics.

Very recently, combining the MS of protein complexeswith affinity purification techniques15 has highlighted thetantalizing possibility of identifying the components, con-nectivity, and shape of such species, expressed at endogenouslevels. In this review we chart the progress that has enabledthe MS of intact protein complexes to approach this vision.From early studies of recombinant protein complexes,through to the latest reports that couple MS of intact cellularcomplexes at endogenous expression levels, we describe theunderpinning technological advances. Rather than consider

Justin L. P. Benesch is a Postdoctoral Research Associate in theDepartment of Chemistry at the University of Cambridge. He obtained aMasters in Chemistry at Hertford College, University of Oxford, and thenundertook a Ph.D. at Gonville & Caius College, University of Cambridge.His doctorate was supervised by Professor Carol Robinson, and concerneddeveloping mass spectrometry techniques for the study of polydispersechaperone proteins. After receiving his Ph.D. in 2005, he spent 1 year asan MRC Discipline Hopping Fellow, which involved learning electronmicroscopy with Professor Helen Saibil at Birkbeck College, London. Hisprincipal research interests concern developing direct and hybrid massspectrometry approaches toward determining macromolecular proteinstructure.

Brandon T. Ruotolo received his B.S. degree in Chemistry from St. LouisUniversity in 1999. He then joined David H. Russells research group atTexas A&M University, where he was primarily involved with the design,development, and implementation of hybrid ion mobilitymass spectrometryinstrumentation for studying complex mixtures. He received his Ph.D. in2004 and is currently a Postdoctoral Research Associate at the Universityof Cambridge. His research is focused on both developing ion mobilitymass spectrometry as a tool for structural biology and understanding theinfluence of solvent on the stability of protein quaternary structure.

Douglas A. Simmons received his B.Sc. in Chemistry and EnvironmentalScience from Trent University in 1999. He then joined the group ofProfessor Lars Konermann at the University of Western Ontario, wherehis doctoral studies involved the use of mass spectrometry and H/Dexchange to study protein folding and conformational dynamics. Afterreceiving his Ph.D. in 2004, he took up a position as NSERC PostdoctoralResearch Fellow in the laboratory of Prof. Carol Robinson at the Universityof Cambridge. There his work focused on the development of novel massspectrometry instrumentation for the structural analysis of macromolecularcomplexes. In 2006 he moved to MDS Sciex in Toronto, where his currentresearch involves the application of mass spectrometry technologies tothe analysis of intact proteins and protein complexes.

Carol V. Robinson is Royal Society Professor of Biological Chemistry atthe University of Cambridge. She earned her Ph.D. from the Universityof Cambridge and subsequently was a Royal Society University ResearchFellow at the Oxford Centre for Molecular Sciences at the University ofOxford. She is a recipient of several awards, including the Biemann Medalof the American Society for Mass Spectrometry in 2003 and the RosalindFranklin Award of the Royal Society in 2004. In 2004, Professor Robinsonbecame a Fellow of the Royal Society. She has lectured extensively bothin the U.K. and internationally. Her current research interests lie indetermining protein interaction networks and in studying the properties ofprotein complexes in the gas phase more generally.

Mass Spectrometry for Structural Proteomics Chemical Reviews, 2007, Vol. 107, No. 8 3545


3/24

biological applications that have been described in detailelsewhere,6,25-32 this review focuses on the enabling toolsand technologies. We describe the processes of generating,transmitting, and analyzing ions formed by protein complexesand explain the underlying theory behind the experimentalconditions that are becoming standard for these investiga-tions. From the structure and dynamics of complexes insolution through to their overall topology in the gas phase,

we detail additional approaches that can be used in conjunc-tion with MS of intact complexes to broaden the scope ofthe information available. Overall, it is our aim to stimulateresearch in this area and to establish MS of intact complexesas an integral part of future structural genomic and proteomicinvestigations.

2. Generating Ions of Protein Complexes

The first step of any mass spectrometry experiment is thegeneration of ions from a sample of interest. There are manychallenges associated with ion generation, a fact which ismirrored by the number of methods available for generatinggas-phase ions for analysis by mass spectrometry. For a

number of these ionization methods, it is necessary that theanalyte already be present in the gas phase (e.g., electronimpact ionization) or embedded in a nonphysiological matrix(e.g., fast atom bombardment, matrix-assisted laser desorp-tion/ionization (MALDI)) prior to ionization. The vastmajority of biological processes occur in solution, so itfollows that for species of biochemical interest it is oftennecessary to generate gas-phase ions directly from moleculesin solution. Furthermore, it is beneficial to do so in a mannerthat avoids fragmentation or other undesirable alteration ofthe analyte.

The above requirements effectively limit the scope ofmuch work in biological mass spectrometry to either elec-trospray ionization (ESI)33-35 or MALDI.36,37 Of the two

technologies, applications of ESI to the study of the structureand stability of intact protein complexes far outweighapplications of MALDI. This is primarily because thesample-preparation requirements for MALDI typically in-volve evaporation of solvent from an aqueous analytesolution that contains a 10-10,000-fold excess of a UV-absorbent organic acid. Such highly acidic conditions willundoubtedly perturb protein-protein interactions present insolution and will most likely denature the protein. In addition,

MALDI mass spectra often yield intense signals for proteinaggregates that are thought to be artifacts of the laserdesorption/ionization process. Several instrumental param-eters such as laser fluence and irradiation time,38,39 alongwith the localization of complexes within the MALDIdeposition,40 have also been shown to have a profoundinfluence on the observation of noncovalently bound proteincomplexes. Although some examples of using specializedmatrices that allow for MALDI sample preparations nearneutral pH (e.g., 6-aza-2-thiothyamine) have been reported,41-44

it is currently unlikely that UV-MALDI will be used as ageneral method for the desorption and ionization of intactprotein-protein complexes. In addition, although significanteffort has been put forth to elucidate the mechanism(s) of

ion formation in MALDI, it remains a hotly debated subject.MALDI is commonly used to identify precursor proteinsfrom the peptides produced by enzymatic digestion ofcomplex mixtures derived from a biological source, asdescribed in other articles in this issue. A detailed knowledgeof the desorption/ionization mechanism is not necessary inthe interpretation of these data (MS is used as a detector forexperiments carried out in solution only); therefore, we willnot cover it in depth here. Instead, we direct the reader toseveral outstanding reviews on the subject.45-49

2.1. Electrospray Ionization

The most commonly used ionization method for MSstudies of protein complexes is ESI.33-35 A typical electro-

spray setup involves passing a sample through a metalliccapillary held at high electrical potential, surrounded by aconcentric tube through which a parallel flow of inert gas ispassed to aid the nebulization of the emerging analytesolution. At the tip of the steel capillary, the applied potentialcauses charges to gather preferentially at the tip, forming aTaylor cone.50-52 At the tip of the cone, the stream ofsolvent is drawn out into highly charged droplets, generallyon the order of several micrometers in diameter.50 Emergingdroplets are subsequently drawn down a potential andpressure gradient toward the ion sampling interface of themass spectrometer. Aided by both parallel and then coun-terdirectional flows of nebulizing gas, solvent evaporationfrom the nascent droplets results in a reduction in dropletdiameter. This reduction in droplet size continues until theCoulombic repulsion between the increasingly crowdedcharges becomes strong enough to overcome the surfacetension holding the droplet together. At this point, termedthe Rayleigh limit, droplet fission occurs. The limitingcharge on a droplet, qR, is governed by the droplet diameter,D, and the solvent surface tension, , and is given by

where zR is the number of charges, e is the elementary charge,and 0 is the electrical permittivity of a vacuum.53

Equation 1 provides an upper limit of the charge densityrequired to bring about droplet fission. In practice, a droplet

Figure 1. Pyramid of protein organization states. There are variouslevels of protein organization which can be probed in proteomicanalyses by means of MS. A protein complex may be comprisedof several subcomplexes, which themselves are composed ofindividual protein chains. A further level, that of peptide segmentswhich are generated by proteolytic digestion, is the focus oftraditional proteomics. Combining investigations into all four levelsof organization however can allow for a far more extensive anddetailed characterization of proteins and their functional states. Inthis article we discuss methods and strategies for studying the uppertwo states of this pyramid, those wherein noncovalent interactionsare maintained.

qR ) zRe ) 80D3 (1)

3546 Chemical Reviews, 2007, Vol. 107, No. 8 Benesch et al.


4/24

undergoing evaporation and Coulombic explosion will doso at charge densities of approximately 70-100% qR.52 Closeobservation of this process has revealed that, rather thanundergoing symmetric division into two droplets of similardimensions, a fine jet of offspring droplets is generated fromthe parent. Each sequential fission event results in anoffspring droplet of approximately 2% of the mass and 15%of the charge of the parent droplet.52,54,55 Subsequent deple-tion of solvent by further evaporation results in several

generations of unequal droplet fission, until ultimately gas-phase ions are produced. During this process, ions are drawninto the ion-sampling interface of the mass spectrometer.Additional desolvation, often required to remove residualsolvent and solute particles from the protein ions, is broughtabout by collisions with gas molecules in the low-vacuumregion between the orifice and skimmer electrodes. Theenergy of these collisions is dictated by the difference inthe potentials applied to these two electrodes.

2.1.1. Mechanism of Protein Ion Formation

ESI usually results in the generation of intact, multiplycharged ions, generally in the form [M + nH]n+. While it ispossible to generate both positive (i.e., protonated) andnegative (i.e., deprotonated) ions during electrospray, themajority of protein mass spectrometry is performed in thepositive mode, and only this will be addressed explicitly inour following discussion of ESI. While the macroscopicaspects of electrospray are generally well-understood, themechanics of the final generation of desolvated (or nearlydesolvated) ions from a charged droplet remain incompletelyresolved. Two principal models have been proposed toaccount for this phenomenon. The charged residue model(CRM), conceived by Dole et al.,33 postulates that evapora-tion and Coulombic fission occur until a droplet containinga single residual analyte ion remains. Complete evaporationof the solvent comprising this droplet eventually yields a

naked analyte ion, the charged residue. A second mech-anism for gas-phase ion production, based on the work ofIribarne and Thomson,56 is termed the ion evaporation model(IEM). In this model, it is argued that, prior to completedesolvation of the droplet, the repulsion between the chargedanalyte ion and the other charges in the droplet becomesstrong enough to overcome solvation forces, and the ion isejected from the droplet surface into the gas phase.

The consensus in the literature is that neither the CRMnor the IEM can account for all experimental observations.57

Rather, it appears that one or the other, or both in combina-tion, can be invoked to describe ion formation, dependingon the type of analyte. For the case of proteins of mass>6000 Da, and thus those species most relevant to this

review, there is considerable evidence that Doles CRM isthe dominant mechanism of ion formation during ESI.52,55,58,59

By examining the data that had been generated for nativeproteins during earlier work, de la Mora found that themaximum charge acquired by globular proteins during ESImatches closely to the charge expected on spherical solventdroplets with similar radii.52 This is the behavior one expectsfrom a protein ion formed through the total evaporation ofan encompassing droplet which would presumably be onlyslightly larger than the protein itself. Furthermore, thisrelationship has been shown to hold even when the numberof strongly basic amino acids available for protonationexceeds the number of charges expected in the droplet, givenby eq 1.60,61 Additionally, ions formed by the CRM would

be expected to acquire a number of adducts due to thepresence of solutes (e.g., buffers) in the final precursordroplet. This phenomenon is not only a common empiricalobservation of any practitioner of ESI-MS of native proteins,but it has also been shown to be quantitatively predictablebased on CRM calculations.60

The immediate implication of the dominance of the CRMmechanism in protein ion formation is that the chargeacquired by a protein during ESI is dictated by the size of

the precursor droplet surrounding the protein immediatelybefore desolvation and, therefore, by the size of the proteinitself. This is consistent with the wealth of observations inthe literature concerning the relationship between proteinconformation and ESI charge state. In the early 1990s, Chaitand co-workers noticed that large changes in solvent condi-tions were accompanied by significant changes in proteinESI mass spectra.62 It is now well-documented that anunfolded protein in solution results in ions having highercharge states compared with the case of the same experimentperformed using solution conditions that promote foldedconformations. Moreover, the width of the charge statedistributions is also larger for unfolded proteins and isthought to be related to the structural heterogeneity of the

corresponding protein states in solution.61,63-67 Based on theseempirical relationships, but with the caveat that otherprocesses may be contributing to charge state variation,68-70

ESI-MS has become a standard qualitative method forprobing protein conformational changes in solution.62-65,71-76

More recently, in light of the strong evidence for the CRMmechanism, protein charge states have been used as quantita-tive indicators of protein structure. In particular, it has beenshown that the charge states acquired by native proteins canbe used as a means to estimate the surface area of proteins.61

From the results of two recent studies, however, the precisenature of the relationship appeared to be in some dispute.31

Kaltashov and Mohimen compiled a data set of the averageESI charge states, zavg, of native (or near-native) proteinsand compared those to protein surface areas, S, as determinedfrom crystallographic data. The resulting curve showed apower relationship well-described by zavg S0.69( 0.02.61 Thispower function is in contrast to the linear relationshipbetween zavg and S observed by Hautreux et al. in a previousstudy.77 This disagreement may be due to the differentmethods by which surface areas were calculated in the twostudies: whereas Kaltashov and Mohimen derived surfaceareas from crystallographic data, Hautreux et al. extrapolatedsurface areas from protein mass, assuming constant densityand a spherical shape.31 While the precise details of therelationship between charge state and surface area have yetto be completely resolved, it is apparent that this aspect ofESI-MS may represent a powerful means to probe the 3Dstructure of proteins in solution.

2.1.2. nESI Ionization

A conventional electrospray apparatus uses a spray capil-lary on the order of 0.5 mm in diameter and requires flowrates of several microliters per minute to maintain the stableTaylor cone necessary for droplet formation. Consequently,it is usually necessary to have a minimum of50 L ofsample for most ESI-MS analyses. In 1994, however, Wilmand Mann introduced an important variant of conventionalESI, termed nanoelectrospray (nESI), a name chosen toreflect the low flow rates involved.78 While this techniqueuses the same fundamental sequence of charged-droplet



5/24

generation followed by multiple asymmetric Coulombicfission events, and finally ion formation, it is distinguishedfrom regular ESI in a number of important ways. nESI istypically performed using glass or quartz capillaries whichhave been pulled to a fine tip (1 m inner diameter) andgiven a metallic (usually gold) coating to hold the electricpotential in the place of the metallic capillary used forconventional ESI. Approximately 1-3 L of sample isinjected into the glass capillary and electrosprayed at flow

rates in the range of1 nL/min to several tens of nL/min.79,80

Flow is driven primarily by the approximately 0.5-1.5 kVpotential applied to the capillary, although it is oftennecessary to provide an auxiliary backing gas pressure tothe sample in order to initiate and/or maintain a steady streamof the solution through the tip.

nESI addresses a number of key issues which limit thesuitability of ESI for the analysis of biological samples andmore specifically noncovalent protein assemblies. First andperhaps most obvious is the reduction in the amount of sam-ple consumed during analysis. Protein complexes are oftengenerated via painstaking molecular biology or isolationtechniques and are consequently often limited in samplequantity. Using nESI, it is possible to produce mass spectra

using only a few picomoles of native protein complex.Furthermore, attomolar levels of the individual componentsof a complex are detectable when the analysis is performedunder ideal electrospray conditions with organic cosolvents(i.e., when preserving noncovalent interactions is unimpor-tant).81 This particular characteristic of nESI alone has mademass spectrometric analysis possible for a number ofbiochemical species which are difficult to produce in thequantities required for conventional ESI.

Second, it has been shown in a number of studies thatnESI is more tolerant of salt contamination of the proteinsolution than conventional ESI.79,82 This is particularlyimportant for the study of noncovalent protein complexes,

since these compounds often require the presence of certainbuffer salts to remain stable in solution. An increasedtolerance of the presence of nonvolatile salts can be explainedthrough examination of the subtle differences in the ESIprocess between nanospray and conventional ESI. With bothtechniques, the first generation of charged droplets undergoevaporation of solvent to reach the Rayleigh limit necessaryto bring about droplet fission (eq 1). This results in anincrease in the concentrations of the analyte as well as anynonvolatile salts which do not undergo field evaporation. Itfollows that the higher the number of fission events whichare necessary to form a charged residue, the greater the saltconcentration in the final droplet containing a single proteincomplex molecule. The extent of salt enrichment during the

multiple steps of evaporation/asymmetric fission is thereforeaffected greatly by the size of the primary droplets formedfrom the Taylor cone at the capillary tip. As mentionedabove, the initial droplets formed during conventional ESIare of the order of several micrometers in diameter. Incontrast, the diameters of the primary droplets formed duringnESI have been estimated to be on the order of 150-200nm.79,82,83 Since offspring droplets are generally 1 order ofmagnitude smaller than primary droplets during ESI, thedifference in droplet size achieved by swapping from ESIto nESI removes approximately one round of Rayleighfission, and therefore one salt concentration step, prior toreleasing the desolvated gas-phase ions. This has beenverified experimentally by comparing the size of salt

[(NaCl)nNa+] clusters which are formed using both tech-niques; the greater salt enrichment experienced by dropletsduring ESI was found to result in larger, more highly chargedsalt clusters than those formed during nESI.82

Third, the nESI desolvation process has been shown tobe a gentler, more reliable, method of introducing extremely

labile protein assemblies into the gas phase intact. It isgenerally accepted that the small initial droplet sizesproduced by the nESI source reduce the number and theenergy of the collisions required to desolvate the macro-molecules of interest. Figure 2 shows two mass spectra ofthe 800 kDa E Coli. chaperone GroEL acquired using thesame MS instrument. The upper spectrum was acquired usingnESI for ion generation. A number of well-resolved chargestates are observed, yielding sufficient information tomeasure the intact mass of the GroEL assembly. The lowerspectrum was acquired using conventional ESI, and evenafter reoptimizing the instrument settings to desolvate thelarger droplets, a very different spectrum is obtained. Severalpeaks are observed which vary in shape and width. Using

Figure 2. Conventional and nanoelectrospray MS of a protein-protein complex. MS of the GroEL complex ionized by means ofESI (lower) and nESI (upper). The nESI spectrum displays a seriesof peaks around 11500 m/z which correspond to the 800 kDatetradecamer. Conventional ESI of the same solution results inpoorly resolved humps centered on 12500, 16000, and 18500m/z. These are assigned to the tetradecamer, a dimer of tetradecam-ers, and a trimer of tetradecamers, respectively. There is also asignal at low m/z which corresponds to the GroEL monomer.Though a signal is observed corresponding to intact tetradecamer,this signal differs from that observed by nESI in three main ways:the peaks are less well resolved, the charge state distribution isbroader and bimodal, and the formation of nonspecific oligomersis increased. These results indicate some of the benefits of usingnESI. A smaller initial droplet size leads to less nonspecificaggregation (both protein-protein and protein-salt), and the gentlerinterface conditions possible, while still allowing adequate desol-vation, lead to less dissociation and disruption of oligomericstructure. Solution conditions were 200 mM ammonium acetate,pH 6.9, and a protein concentration of 2 M tetradecamer. Spectrawere obtained on a modified Q-ToF 2 (Waters/Micromass).105



6/24


7/24

occurred when it was found that increasing the pressures inthe ion guides in the first vacuum stages of the massspectrometer resulted in increased signal intensities for high

m/z ions.87-89

Prior to these reports, Douglas and French hadshown a clear correlation between ion mass and the pressurerequired for optimal ion transmission.90

Early systematic investigations of the impact of sourcepressure on the observation of noncovalent protein assemblieswere performed in 2001.88,89 In one study, under standardpressure conditions (2 mbar), only trace signals were assignedto dimeric protein. However, upon increasing the pressureto 3.5 mbar, a significant improvement in the intensity ofthe peaks assigned to the protein dimer was observed. Furtherincreases in pressure, up to 7.0 mbar, resulted in theemergence of a charge state distribution corresponding toan octameric protein assembly which had not been observedat lower pressures. This phenomenon was generally rational-

ized as being due to a reduction in dissociation by thesomewhat vague concept of gentle MS conditions.2,4 Theseconditions enabled efficient desolvation of macromolecularions while maintaining their noncovalent interactions. Thegeneral opinion at this time was that the higher pressureresulted in frequent low-energy collisions and that thisresulted in less dissociation than fewer higher energycollisions.

In recent years, however, significant progress has been

made in understanding this phenomenon. These studies havefocused not on how higher pressures and hence morecollisions might increase the transmission of noncovalentcomplexes by reducing the amount of dissociation, but ratheron how they might act to improve the focusing of these ionssuch that more pass through the mass spectrometer.91,92 Wedescribe below the various steps along this collisionalfocusing pathway.

3.1.1. Acceleration of Gas-Phase Ions in the SourceRegion

As described above, ESI, the preferred ionization methodfor the study of noncovalent complexes by MS, is performedat atmospheric pressure. Carried along in the flow of bathgas, ions enter the first vacuum stage through a smallaperture. The consequent free jet expansion of the gas streamduring this traversal from atmospheric pressure to roughvacuum, as well as space-charge effects, results in most ofthe ions deviating from the ideal trajectory. Furthermore, theexpansion of the gas stream results in ion velocities of severalhundred meters per second. Even very massive ions, suchas those of protein complexes, exhibit little variation fromthe velocity of the gas stream. The corresponding mass-dependent kinetic energies of such ions range from 1 eVfor ions of 1 kDa up to 1 keV for ions of 1 MDa.91,92 Assuch, the ions formed from protein complexes at this stageare not only defocused but are traveling along these inap-propriate trajectories with high kinetic energies. Typical ionguide parameters which are used to focus ions back ontothe intended route are generally ineffective, as proteincomplexes carry fewer charges than denatured proteins ofsimilar mass. As a result, their manipulation is highlyproblematic in the optics of a conventional mass spectrom-eter, which has electrostatic lens voltages operating generallyat less than 100 V. Therefore, ion trajectories are poorlyfocused with respect to the center line of the instrument,resulting in poor transmission efficiency at the apertures thatseparate the initial stages of differential pumping. Variousstrategies for focusing massive ions have been proposed, suchas aerodynamic focusing,93 but currently the most widelyused method is collisional focusing in a radio frequency (RF)ion guide.

3.1.2. Transmission Efficiency of Massive Ions in RF-IonGuides

For some time it has been recognized that collisions withneutral gas molecules can be used to focus the trajectoriesof ions stored in three-dimensional Paul traps. Collisionsbetween comparatively large ions and small neutral gasmolecules serve to dissipate ion kinetic energy into thesurrounding gas. This is functionally important because itallows more efficient trapping of ions in the potential wellat the center of the ion trap. This effect has often been termedcollisional cooling, which is in some ways misleading, as,during a collision with a neutral, though the ions kinetic

Figure 4. Adduction of solvent molecules and buffer ions toproteins. (A) The theoretical deconvoluted spectrum of a pureprotein complex would appear at a mass value governed solely byprimary sequence, with a peak width defined by the isotopicdistribution and instrumental resolution (sequence, pink). Themasses of protein complexes invariably are higher than those dueto the sequence alone, which is attributed to the retention of solventmolecules and buffer ions. The higher the accelerating voltages(low, blue; medium, indigo; high, violet) and consequentlythe more energetic the collisions, the smaller this shift becomes,and there is a concomitant narrowing of the distribution. (B)Examination of the 68+ charge state of GroEL (see Figure 2) atdifferent activation voltages demonstrates that the amount ofadditional mass and the width of the distribution can be ap-proximated by a linear relationship.86 The amount of acceleratingvoltage (low to high) required to achieve the indicated peak widthis indicated by the color of the points in panel B.



8/24

energy is decreased, its internal energy, or temperature, isincreased. Therefore, the ion is not cooled, but rather isheated. As such, we will use the expression collisionalfocusing throughout this article.

Long after this collision focusing became standard operat-ing procedure for quadrupole ion-trap mass spectrometers,Douglas and French extended this methodology to linearmultipole devices, specifically the RF quadrupole ion guide.90

In this study it was found that the transmission efficiency of

ions produced by ESI is pressure-dependent and increasedpressure in the RF ion guide can result in increasedtransmission efficiency. While this effect was at firstcounterintuitive, when stopping curves (plots of barriervoltage vs ion transmission efficiency) were generated forions at different pressures, it was discovered that the appliedpotential necessary to stop ions at low pressure was muchhigher than that required to stop ions under high-pressureconditions. This result indicated that at higher pressure theions possessed diminished kinetic energies and, therefore,velocities.

As the axial component of the velocity (along the ionbeam) is reduced by the increased collisions with bath gasat higher pressures, the same must also be true for the radialcomponent. It was hypothesized that the increased iontransmission was therefore due to collisional damping ofradial trajectories, allowing ions to be more efficientlycaptured in the dynamic potential well along the center axisof the RF quadrupole.90 The focusing effect in turn allowedions to be more efficiently transmitted through the exitaperture and into the subsequent ion optics. The trajectoriesof ions undergoing collisional focusing in an RF multipolewere first calculated in silico by Krutchinsky et al.;91

however, such an approach has only recently been appliedto examine the transmission of large protein complexes.92

In this study, it was shown that while normal operatingpressures (10 bar) in the RF ion guide resulted insufficient collisional focusing such that small protein ionswere successfully transmitted through the aperture im-mediately after the RF quadrupole, the ions of the 692 kDa20S proteasome were poorly focused. Increasing the pressureto 43 bar, however, resulted in sufficient collisionsbetween the proteasome and the background gas to enableefficient focusing of the ion beam.92 An alternative approachtoward increasing the number of collisions experienced whileoperating at the same pressure is to increase the effectivelength of the ion guide. This can be achieved, without havingto incorporate an ion guide which is physically longer, bytrapping the ions under conditions where they undergonumerous collisions with neutrals.92

Figure 5 describes this collisional focusing principle, as

it applies to a quadrupole-time-of-flight (Q-ToF) instrument(Figure 5A). The simulated trajectories (in a plane perpen-dicular to the quadrupole axis) of the 147 kDa tetramericprotein ADH at three different pressures are shown (Figure5B). As the ions enter the RF ion guide, their oscillationsresult in large deviations from the center of the ion axis (theirideal position for transmission through the instrument). Asthey experience collisions with bath gas, the velocity of theions decreases and, hence, their oscillations deviate less fromthe central axis. At the higher simulated pressures, there aresufficient collisions such that the focusing results in a verynarrow ion beam. At the lower pressure (9.3 bar, standardoperating pressure), the collisional focusing is insufficientand results in a broad ion beam leading to losses of ions

that fail to pass through the various small apertures down-stream.

3.1.3. Transmission Efficiency of Massive Ions inOrthogonal Acceleration Time-of-Flight Analyzers

In addition to the improvement in transmission efficiencygained by collisional focusing of the ions, there are furtherconsequences to the deceleration of protein complex ions inthe initial stages of the mass spectrometer. Most of theexperiments conducted on these molecules are performedwith instruments incorporating an orthogonal time-of-flight(o-ToF) mass analyzer. The trajectories of ions in an o-ToFanalyzer are determined not only by the acceleration impartedin the orthogonal extraction source but also by the axialvelocity of the ions as they enter this region (Figure 5). Ionswhich enter the orthogonal extraction source with excessvelocity may overshoot the detector and are consequentlynot recorded.92

To demonstrate this effect, we recorded the arrival positionof tetrameric ADH ions on a four-part segmented anodemicrochannel plage (MCP) detector as a function of gaspressure in the primary RF ion guide of a Q-ToF instrument

(Figure 5C). At the lowest pressure (10 bar), the overallsignal intensity is low and the majority of the ion current isdetected by anode segments furthest from the extractionsource (anodes 3 and 4). As the pressure is increased to 20bar, total ion current increases and presents a more uniformdistribution of impacts across all detection surfaces. At thehighest pressures, ion current is primarily detected on anodesegments nearest the extraction source (plates 1 and 2), andthe total ion current had again decreased. The data in Figure5C indicate that large macromolecular ions may not besufficiently decelerated at low pressures. Consequentially,many ions retain excess velocity as they enter the ToForthogonal extraction source, and overshoot the position ofthe detector during their passage through the ToF analyzer.

The opposite problem, where ions have insufficient axialvelocity to reach the ToF detector, becomes significant atvery high source pressures. Though other possibilities, suchas scattering by background gas at higher pressure, may alsocontribute,92 these observations provide a plausible explana-tion as to why there appears to be an optimal pressure forion detection, rather than an approach toward the 100%transmission value with increasing pressure.88,89,94,95

3.2. Analyzing Ions

Although a wide range of mass analyzers has beendeveloped and applied to the analysis of biomolecules, onlya small subset of these has been successfully applied to the

study of large protein complexes. Typically, large proteincomplex ions appear at high m/z, which limits the scope ofapplicable mass analyzers considerably. Specifically, somelarge multiply charged protein complexes have been detectedat>20000 m/z,10,96-98 whereas the ions typically formed fromindividual peptides and denatured proteins have considerablylower m/z values (


9/24

instruments. As such, we cover the elements of theseinstruments in detail below.

3.2.1. Quadrupole Analyzers

While quadrupoles can be used in RF-only mode as ionguides, Vide supra, the application of a simultaneous DCvoltage enables them to be used as mass analyzers. Theirlow cost, relatively high operating pressures, and compat-ibility with the continuous ion beam generated by ESI madethem the mass analyzer of choice in early MS experiments

of protein complexes. The maximum m/z that can be resolved([m/z]max) is given by

where Vm is the RF amplitude, is the RF frequency, and r0is the inner radius between the rods in meters. Lowering theoperating frequency of the quadrupole therefore increasesthe [m/z]max in mass-resolving mode.103,104 Theoretically, Vmcould be increased and/or r0 decreased, as an alternativestrategy, but the increased chance of high-voltage breakdownbetween the rods makes this impractical. The main disad-vantage in operating quadruopole mass analyzers at suchreduced frequencies, however, is a decrease in resolution.

The implications of this drawback can, however, be largelynegated by introducing a subsequent analysis step, as is thecase in the tandem instruments described below.105

3.2.2. Time-of-Flight Analyzers

Time-of-flight (ToF) analysis was first coupled to anelectrospray source in 1991,106 and relatively shortly there-after, the first ESI ToF spectra of noncovalent proteincomplexes were recorded in 1994.107 ToF analyzers providedseveral immediate benefits to the study of macromolecularassemblies, with the most significant being their theoreticallyunlimited m/z range. Due to the pulsed nature of the ToFexperiment, ESI-ToF experiments are usually conducted inan orthogonal mode, where the primary ion beam is deflectedfrom its original direction of motion into the ToF analyzer.This arrangement typically exhibits a poor duty cycle,especially when large masses (having long flight times) areobserved. This limitation has been addressed by includingan ion trapping step prior to orthogonal extraction of theprimary ion beam.108,109 This step modulates the originallycontinuous ion beam into a source of ions pulsed at afrequency timed to match the extraction timing of the ToFanalyzer. This trapping increases the overall sensitivity and

Figure 5. A Q-ToF type instrument customized for the transmission and analysis of protein complexes. (A) Ions are introduced into thespectrometer by nESI. The ion beam is focused by RF-only ion guides, before entering a quadrupole analyzer. This analyzer is modifiedsuch that it operates at a reduced RF frequency, allowing the selection and transmission of high m/z ions. Ions then traverse a collision cell,into which gas can be leaked in order to allow collision-induced dissociation. The dissociation of protein complex ions requires highercollision cell gas pressures and accelerating voltages relative to normal parameters. Ions exiting the collision cell enter a pusher region,whereupon they are pulsed orthogonally in packets into a ToF analyzer. Ions are detected by means of an MCP detector, which in someinstruments is split into four segments along the ion axis. The two primary areas for protein-complex ion loss, once they have entered thevacuum stages, are in the RF-only ion guides and the apertures between them, and by over- or undershooting the ToF detector. Improvedtransmission is achieved by adjusting the velocity of the ions, via altering the pressures in the instrument. Simulation of the ion trajectoriesof a 147 kDa protein complex (ADH tetramer) at different pressures (B) shows how, at the standard pressure of 9.3 bar, the ions arepoorly focused. Increasing the pressure to 40 bar results in a much narrower ion beam and consequently improved transmission. Thisamelioration is reflected if one monitors the ion intensity across the MCP detector (C). As the pressure is increased from 9.3 bar, the totalsignal intensity (width of the bars) increases. Once the pressure exceeds 40 bar, however, the signal intensity starts to decrease again.Examining how the ion signal divides between the four segments of the MCP plate shows how at low pressures the ions mainly impact onthe anodes furthest from the pusher (3 and 4), whereas at the highest pressures they primarily impact anodes 1 and 2. This suggests that,at pressures in the source region below and above the optimum, ions over- and undershoot the detector, respectively. Simulations in partA were performed as described previously,92 assuming an initial kinetic energy of 122 eV, a parallel gas flow of 50 m s-1, and a collisioncross section of 7000 2 for ADH (based on ref 247). Data in part B were recorded for the 25+ charge state of ADH, at 200 mM ammoniumacetate, pH 6.9, and 3 M tetramer, on an MDS Q-Star XL instrument with a 4-channel MCP detector.

[m/z]max ) 7 106

Vm/(2r0

2) (2)



10/24

limit-of-detection of the instrument, as a larger proportionof the total ion current is mass analyzed.

3.2.3. Tandem/Hybrid Instruments

One of the main strengths of MS is the ability not only tomeasure the mass of the intact species under investigationbut also to selectively dissociate the species in the gas phaseand to mass measure the product ions. Tandem MS, or MS/MS, has widespread applications in both bottom-up and

top-down proteomics experiments, and it also has tremen-dous potential in the study of protein complexes. Differenttypes of tandem mass spectrometers have been used in thestudy of macromolecular complexes. Although some earlyexperiments were conducted using triple quadrupole instru-ments,2,110,111 the most prevalent tandem mass spectrometerin this field today is the Q-ToF instrument. The arrangementof such an instrument is shown in Figure 5A.

In Q-ToF instruments, quadrupole and ToF analyzers arearranged in series, separated by a collision cell. Ions ofinterest can be selected according to their m/z ratios in thequadrupole analyzer, subsequently activated in the collisioncell, with their products being analyzed in the ToF analyzer.Activation is typically performed by ion-neutral collisions;

however, ion-surface collisions112 and laser irradiation113have also been successfully implemented on these instru-ments. In order to be able to select ions at high m/z ratios,quadrupoles operating at low frequency have been imple-mented in Q-ToF instruments by several groups.92,105,114 Thereduction in resolution that results from operating at thislowered frequency generally does not impair the functionof the mass spectrometer, as the product ion spectrum isacquired by the ToF analyzer.105 When a quadrupole is notin mass-resolving mode, i.e., it is not being used to select aparticular ion, ions up to approximately five times the[m/z]max (eq 2) can be transmitted.105 As such, lowering theoperating RF frequency of the quadrupole not only allowsthe selection of high m/z ions but also improves the

transmission of ions such that their detection at over 85000m/z has been reported.105

3.3. Instrument Modifications for NoncovalentComplexes

The modification of various components and conditionswithin a mass spectrometer can contribute strongly towardsuccessful mass analysis of noncovalent complexes. Higherpressures in the front end of the instruments are required tofocus and decelerate high m/z ions, low-frequency quadru-poles are often required for their selection and/or transmis-sion, ToF analyzers need to be operated at low samplingfrequencies, and benefits can be gained from implementing

detectors with high sensitivity at high m/z values. Higherion guide pressures have been achieved in several ways:reducing the pumping efficiency by throttling the vacuumlines,105 leaking additional gas into the source region,105 orincorporating a flow-restricting sleeve between the ion guideand the pump orifice.92 The flow-restricting sleeve has theadvantage of not increasing the load on the turbo pumps,though this benefit might be outweighed by the reduction inflexibility it affords in finding the pressure for optimaltransmission of a particular species, Vide supra.

The use of low-frequency/high-m/z quadrupoles for pur-poses of tandem mass spectrometry105 has led to the abilityto selectively dissociate very large complexes, and furthermodifications have since been implemented to aid in this

dissociation process. Recently constructed Q-ToF instru-mentation has allowed significantly higher acceleratingpotentials upstream of the collision cell than is possible inanalogous commercial instruments.115 This access to highervoltages provides the ability to perform higher energydissociation experiments and, consequently, obtain additionalproduct ions. Early results show that both the loss of furtherprotein subunits and covalent fragmentation of individualprotein subunits can be accessed at these high collision

energies (>200 V accelerating potential).115

In summary, since the first spectra of noncovalent proteincomplexes were obtained over 15 years ago, numerousmethodological and instrumental developments have takenplace. It is now possible to control and interrogate ions thatare significantly more massive and heterogeneous than everbefore. As the desire to study samples of greater complexitycontinues to drive developments in tandem mass spectrom-etry, further progress will be required to effect greater controlof protein complex dissociation while an increased mecha-nistic understanding is necessary to garner the maximumamount of information from macromolecular assemblies.

4. Activation and Dissociation of ProteinComplexes

While much effort over the past decade has been directedat maintaining noncovalent assemblies intact in Vacuo forsubsequent mass spectrometric detection, a lot of work isnow focused on their gas-phase activation and dissociation.31

Just as data derived from protein fragmentation has madeMS a driving force in traditional proteomics,116 the gas-phasedissociation of protein complexes is being used to delivercompositional information which may prove to be equallyimportant.

Numerous different ion activation techniques have beendeveloped over the years, only a relatively small subset ofwhich has been successfully used to dissociate noncovalent

protein assemblies. The activation techniques amenable tothis research are necessarily limited to those that can beimplemented in MS instrumentation which itself is capableof transmitting and analyzing noncovalent complexes. Assuch, the majority of work so far has been performed usingcollision-induced dissociation (CIDssometimes known asCAD, or collisionally activated dissociation). However,blackbody infrared radiative dissociation (BIRD)100 and, morerecently, electron-capture dissociation (ECD)117 and surface-induced dissociation (SID)112 have also been used. Whilethe dissociation of protein complexes in the gas phase isbecoming ever more widespread, methodological challengesremain. In particular, further elucidations as to the mechanismof dissociation, understanding the relationship between gas-

phase dissociation and solution-phase disruption, and estab-lishing the limits of structural information that can beobtained require further attention. Once these questions areanswered, it is our belief that the use of protein complexdissociation in the gas phase will become an even morepowerful tool for structural genomics and proteomics.

4.1. Collision-Induced Dissociation

The majority of studies in which noncovalent complexesare disassembled in the gas phase have employed collision-induced dissociation (CID). This is performed by collidingthe ions of interest with neutral gas atoms or molecules.Activation occurs as a portion of an ions kinetic energy is



11/24

converted into internal energy during each collision event,and this may lead to its subsequent dissociation if sufficientinternal energy is accumulated. The use of CID in MS iswell established, and several excellent reviews exist on thissubject.118-120 The majority of examples which use CID fordissociating noncovalent protein complexes employ Q-ToFmass spectrometers. It is this instrumental arrangement onwhich we base our discussion below.

4.1.1. Activation of Macromolecular AssembliesDuring typical CID experiments, ions will undergo many

collisions, depending primarily on the size of the ion andthe gas pressure. The number of collisions experienced bythe ion during its transit through the collision cell, n, can becalculated from the following relationship, which is derivedfrom the mean free path equation and includes a unitconversion factor:

where l is the collision cell length in meters, is the collisioncross section in square angstroms, p is the pressure inmillibars, and T is the temperature in kelvin.

In cases where the collision cross section is not known,estimates can be used. For example, the collision crosssection can be estimated by approximating the shape of themolecule to a sphere, at which point the collision crosssection (2) is given by

where Ri and Rg are the radii of the ion and the gas moleculein angstroms, respectively. When structural informationsufficient to perform the above approximation is not avail-able, one can assume a spherical shape and a typical densityF (in Da/3) of proteins and their complexes. For the collisioncross section estimates reported in this review, we use a value

of 0.84 Da -3

for F, though different values for proteindensity can be found throughout the literature.52,77,121 Fromthese assumptions, a spherical volume may be calculatedfrom the mass Mi of the ion and subsequently the radius.Therefore, eq 4 becomes

In Figure 6 we give an overview of the number ofcollisions experienced and the length of time spent in thecollision cell for different proteinaceous species under typicalexperimental conditions for the CID of large noncovalentcomplexes on a modified Q-ToF instrument.92,105,114 The toppanel shows the number of collisions experienced by the

different species (based on eqs 3 and 5) and the massdependency of the trend. It can be seen that large proteincomplexes typically experience from thousands to tens ofthousands of collisions as they pass through the cell and thatthe number of these collision can be modulated by adjustingthe pressure (inset).

The bottom panel shows how long the ions spend in thecollision cell, i.e., the amount of time during which theyexperience the collisions. This residence time depends onthe initial kinetic energy of the ions and, hence, on the chargestate. Therefore, to model a mass dependency, we used thefollowing relationship between ion mass and average chargestate (Zav):52 Zav 0.0778m. By modeling all the collisionsexperienced by an ion, and the resulting reductions in

velocity, one can see that the amount of time spent in thecollision cell is dependent on mass and is typically on theorder of about 200-400 s (Figure 6). Only at the low endof the mass scale does this residence time deviate consider-ably, reaching, in the case of cytochrome c, over 1700 s(under the conditions used to generate Figure 6). Thisdramatic increase in residency time in the collision cell isdue to the considerably higher transition efficiency of kineticto internal energy for the relatively small cytochrome c. Thisallows the ion to reach a steady-state velocity (determined

largely by the kinetic energy of the neutrals and space-charge effects) early in the transit of the ion through thecollision cell. It is important to note that at no time are ionsassumed to be held stationary in the collision cell under theconditions described in Figure 6. The residence time willalso depend on both the accelerating voltage and the pressurein the cell. The inset of the lower panel is a surface plot ofthese dependencies. In all of the conditions displayed, theions experience collisions over a time period on the orderof hundreds of microseconds.

Each of the thousands of collisions a protein complexexperiences in the collision cell causes its internal energy toincrease. The maximal increase in internal energy ac-cumulated by an ion, Eint (assuming no energy dissipation

n ) 102430(lp/T) (3)

) (Ri + Rg)2 (4)

) (33Mi/4F + Rg)

2 (5)

Figure 6. Number of collisions experienced and time spent in thecollision cell. The upper panel shows the calculated number ofcollisions experienced, plotted relative to mass, by cytochrome cmonomer (violet), transthyretin tetramer (blue), MjHSP16.5 24mer(green), GroEL tetradecamer (orange), and the 70S ribosome fromThermus thermophilus (red) (see eq 3). The inset shows the lineardependence of the number of collisions on pressure for MjHSP16.5.The time spent in the collision cell undergoing these collisions isplotted in the lower panel (using eqs 3 and 5). The inset shows theaccelerating voltage and pressure dependency of this time (forMjHSP16.5, 47+ charge state). These calculations are based on a

collision cell length of 18.5 cm, a gas pressure of 30 bar argon,and an accelerating voltage of 200 V.



12/24

in between collisions), after it has experienced n collisionscan be given by the following relationship, where the initialtranslational energy is assumed to be determined only by

the acceleration into the collision cell:

where Mg and Mi are the masses of the gas and the ion,respectively, z is the charge state of the ion, and Va is thevoltage accelerating the ions into the collision cell (adaptedfrom refs 122 and 123).

To demonstrate the transfer of translational energy intointernal modes of ions as they traverse the collision cell, wehave performed a series of explicit individual-collisionsimulations based on eq 6. The top panel of Figure 7demonstrates the conversion of energy as a function ofdistance traveled along the collision cell for the same fivespecies shown in Figure 6, ranging from the 12 kDacytochrome c to the 2.3 MDa 70S ribosome from Thermusthermophilus. It can be seen how, for the relatively smallcytochrome c, essentially all the translational energy isconverted into internal modes by halfway along the collisioncell, after less than 1000 collisions. At this distance into thecollision cell, however, the much larger 70S ribosome does

not achieve such complete conversion, despite havingexperienced over 47000 collisions. This demonstrates that,under these conditions, very large species will retainsignificant velocities as they exit the collision cell, whichwill affect their transmission efficiency to the MCP detec-tor.92 To overcome this problem, from eq 6 we can see thereare several possibilities: either increasing the number ofcollisions n, which in turn can be affected by increasing thegas pressure or the length of the cell (eq 6), or increasingthe mass of the gas Mg. The middle and lower panels ofFigure 7 show how the effective conversion efficiencybetween kinetic and internal energy is altered by changes inpressure or neutral mass, respectively. Clearly, higherpressures and heavier target gases are beneficial for morerapid energy conversion. However, the insets of Figure 7show how altering these variables meets with diminishingreturns, and as such, the practical benefit is somewhat limited.

Ion activation can also be achieved in the skimmer-coneregion of the instrument through careful management of thepressure and accelerating potential experienced by the ionsin the first regions of differential pumping within theinstrument. The dissociation products formed here fromprotein complex ions are similar to those generated usingthe collision cell.14 Dissociation of protein complexes issometimes also observed if the pressure in the first regionof differential pumping is too low, which is attributed to theions undergoing a small number of high-energy collisionsrather than a large number of very low-energy collisions.88Upon first inspection, this result may seem counterintuitivewhen viewed in the context of Figure 6. This is becausevibrational/rotational relaxation which occurs in betweencollisions is not taken into account in these simulations.Therefore, while the calculations shown in Figure 6 offer auseful description of the number of collisions experiencedby an ion in the collision cell and the ion kinetic energylosses experienced, the same approach is limited in describingthe internal energy of the ions. In general, the activation ofions in this region of the instrument is affected most easilyunder conditions of high accelerating potentials and lowerpressures.

Figure 7. Energy conversion during collisional activation. Simula-tions showing the maximum percentage conversion of energy fromkinetic to internal modes as ions pass through the gas-filled collisioncell, and the variations due to mass (upper), gas pressure (middle),and collision gas (lower). The upper panel shows the energyconversion profiles for cytochrome c monomer (violet), transthyretintetramer (blue), MjHSP16.5 24mer (green), GroEL tetradecamer(orange), and the 70S ribosome from Thermus thermophilus (red),at 30 bar argon collision cell pressure. In order to determine atrue dependency on mass, eq 6 was used. The distance at whichhalf of the energy is converted (dashed line) for the different species

allows the extraction of the dependency of energy conversion onmass (inset). The middle panel shows the effect of varying pressureon this energy conversion process for a single species, MjHSP16.5.Different pressures (labeled from 50 bar, violet, to 10 bar, red)of argon are simulated. Similarly, the lower panel shows (usingthe same ion as above) the effect of the mass of the gas (the noblegases from radon, violet, to neon, red). Inset into these panels arethe dependencies for 50% conversion. These simulations show thatin order for sufficient conversion of kinetic energy into internalmodes to occur, in this length of collision cell, the use of higherpressures and/or heavier target gas is preferable.

Eint ) zVa(1 - [(Mi2+ Mg

2)/(Mi+ Mg)2]n) (6)



13/24

4.1.2. Dissociation of Activated MacromolecularAssemblies

In the previous section, we have dealt briefly with the wayin which macromolecular ions are activated by gas-phasecollisions. In this section, we turn our attention to describingthe unimolecular decay process for noncovalent proteincomplexes. Collisional activation has been used to dissociategas-phase protein-protein complexes since the very earlieststudies that demonstrated their detection.2,3 In these studies,

the activation of the noncovalent complexes was shown tobreak the noncovalent bonds, rather than causing covalent-bond cleavage. While initial studies were conducted ondimers, when complexes comprised of more than twosubunits were dissociated by CID, some unexpected char-acteristics of the dissociation process were observed. Thedissociation products of tetramers in the gas-phase werefound not to be dimers, as expected, but rather complemen-tary monomers and trimers.4,5,124 Furthermore, it was notedthat the charge partitioning between the dissociation productsalso appeared to be surprisingly asymmetric. For one of theproteins studied, avidin, the monomers and trimers both werecentered on the 8+ charge state despite there being a 3-folddifference in mass.5 This dissociation pathway of proteincomplexes into highly charged monomers and relativelylowly charged oligomers missing one subunit (strippedoligomers) has been demonstrated for a large variety ofprotein complexes ranging from dimers to species with morethan 100 subunits.14 As such, the general scheme of thereaction can be described by

where n is the number of subunits in the oligomer of speciesP, q is the number of charges on the oligomer, x is theaverage charge carried by the monomer P1, and the chargedensity by mass is much higher in P1 than P(n-1).

Furthermore, it has also been shown that the asymmetricdissociation process can reoccur, such that several monomericunits can be removed.14 This has been shown to proceed ina sequential manner as described by

where n and q are the number of subunits and charges in

the original oligomer, and x, x, and x are the averagecharges carried by the monomers P1 at different dissociationsteps.14 The number of dissociation steps which are observedis governed by the initial kinetic energy of the ions. Underthe highest energy CID conditions currently accessible on amodified Q-ToF instrument, at 350 V acceleration, up to fourdissociation steps have been observed.115

The many observations made of the dissociation of proteincomplexes, including the typically high charge states for themonomeric product ions (indicative of unfolded gas-phaseconformations),62 led one early report to suggest that dis-sociation of the [oligomer] may occur by a Coulombically-driven process in which a monomer species becomesunravelled and ejected ... with a disproportionately large

share of the charge.5 Compelling evidence that an unfoldingevent occurs during the asymmetric dissociation pathwayobserved during CID of a multimeric assembly comes fromstudies performed on nonspecific cytochrome c and R-lac-talbumin dimers with and without the presence of confor-mational restraints. When the individual protein chains werecovalently cross-linked, thereby restricting their unfolding,dissociation was shifted from an asymmetric pathway to asymmetric one.125,126 Furthermore, it has been shown that

the estimated increase in surface area induced by monomerunfolding correlates well with the charge partitioning ob-served, indicating that the observed charge redistribution issymmetric with respect to surface area while asymmetric withrespect to mass.14 Recently, this CID process has beenvisualized by means of ion-mobility mass spectrometry,127

wherein the activated state of the transthyretin tetramer wasshown to populate multiple states much larger than theinactivated form. The sizes measured for these ions areconsistent with a structure comprised of a single unfoldedprotein subunit and three folded protein subunits.

Current thinking on the mechanism of dissociation of amultimeric protein complex stands as follows.14,100,125 Anoligomeric complex is produced by ESI with the charges

distributed, on average, evenly over its surface. The ion isaccelerated into a gas-filled collision cell, with an increasein kinetic energy dependent on its charge state and theaccelerating voltage. During each of the approximately1 103 to 1 104 collisions with neutral gas molecules,some of this translational energy is converted into internalenergy, where it is distributed among the vibrational androtational modes of the proteins that comprise the complex.The increase in internal energy enables local unfolding/disordering events to occur. Local unfolding increases thesurface area, and mobile charge carriers redistribute to thefreshly exposed area to minimize Coulombic repulsion.Regions close to the locally unfolded section are destabilizedby this process, and further unfolding follows, with conse-

quent further charge migration. This continues until a subunitis essentially fully unfolded, and its interactions with theoligomer are broken. An intermediate state is reached whenthe intraoligomer forces holding the monomer to the rest ofthe oligomer are equal to the Coulombic forces pushing themapart. With the activation barrier having been overcome, thetwo charged species then separate, with the total chargepartitioned according to their surface areas. The magnitudeof the activation energy is therefore dependent on the energyrequired to overcome both intramonomer and intraoligomerinteractions.

4.1.3. The CID of a Large Oligomeric Protein: CaseStudy

In this section we demonstrate the CID of a largeoligomeric protein complex from both an experimentalstandpoint and the theoretical background described above.Figure 8A shows the CID of the 200 kDa, 12 subunitoligomer formed by TaHSP16.9.128,129 The 32+ charge statewas selected by using the quadrupole analyzer of a Q-ToFinstrument and accelerated into a collision cell containing30 bar of argon. At the lowest acceleration voltages, onlythe signal corresponding to the intact oligomeric species isobserved. Therefore, the ions do not accumulate sufficientinternal energy through transfer from translational energyto dissociate on the time scale of their passage through themass spectrometer. When the accelerating voltage is in-

[Pn]q+f [P(n-1)]

(q-x)++ [P1]

x+ (Scheme 1)

[Pn]q+f [P(n-1)]

(q-x)++ [P1]

x+f

[P(n-2)](q-x-x)+

+ [P1]x++ [P1]

x+f

[P(n-3)](q-x-x-x)+

+ [P1]x++ [P1]

x++ [P1]

x+f

(Scheme 2)



14/24

creased to 80 V, with the ions therefore entering the collisioncell with 2560 eV, the 12mer signal is reduced and peakscorresponding to monomers and 11mers are observed. Thesefirst dissociation products become of increasingly high

intensity as the voltage is increased, until at 120 V twoadditional series are observed: one corresponding to 10mers,and a second monomer distribution at lower charge states(higher m/z) than the first. As the voltage is increased stillfurther, up to the maximum possible on this instrument (200V), no further dissociation steps occur, but the intensity ofthe 10mer peaks increases while those of the 11mersdecrease.

Figure 8B shows how the relative intensities of the parent12mer and the different stripped oligomers vary as a functionof acceleration voltage. This plot clearly demonstrates thatprotein complex dissociation is a sequential reaction, asdescribed in Scheme 2.14 By setting a definition of onsetfor dissociation as being the accelerating voltage at which

the dissociation product amounts to 10% of the total ioncurrent, we can use this graph to read off the thresholdvoltage at which dissociation occurs. In this case, the 11merforms at 80 V and the 10mer at 145 V. As the selected ion

was the 32+

charge state, these acceleration voltages cor-respond to initial translational energies of 2560 and 4640eV, respectively. As the process is sequential, with the 10merbeing formed from the 11mer, a crude estimate of the energyrequired for the 11mer to 10mer process is the difference ofthe two, namely 2080 eV.

Having established the dissociation threshold for the 12merto the 11mer, we examine the effect of using differentaccelerating voltages on energy deposition. Figure 8C showssimulations of the distance along the cell (left panel) andthe time taken (right) for the 12mer32+ ion to reach thedissociation threshold. We can see that, with the voltagesaccessible in this instrument, activation of these ions can beaccomplished on the order of 11-341 s. From the inset it

Figure 8. Dissociation pathway of a multiprotein complex. (A) CID of the 32+

charge state of the TaHSP16.9 dodecamer (violet) resultsin the formation of complementary monomers at low m/z and 11mers at high m/z (both blue). At higher accelerating voltages (>100 V),a second distribution of monomers as well as decamers is observed (both cyan). This is indicative of a sequential dissociation reaction(white arrows). Plotting the relative abundance of the different oligomeric species as a function of accelerating voltage (B) further emphasizesthis reaction sequence and, moreover, allows the determination of dissociation thresholds (set at a relative intensity value of 10%). Thesevalues are then used in simulations as in Figure 6, demonstrating that, at the lowest voltage that still incurs CID (80 V), the entire lengthof the collision cell is used and that activation takes approximately 342 s. At the highest accelerating voltage used in this experiment, 200V, the time taken to reach the dissociation threshold is 11 s. This demonstrates that the activation time for such a molecule is on themicrosecond time scale but can vary by more than an order of magnitude. Experiments were performed on a modified Q-ToF 2.105 Thesolution of TaHSP16.9 was infused by nESI at a concentration of 1.4 M in 200 mM ammonium acetate, pH 6.9.14



15/24

is clear that, even if higher energies were attainable, as onother instrument platforms described in the literature,115 theactivation process would not occur on a significantly fastertime scale.

4.2. Other Activation Techniques

Though the vast majority of studies in which proteincomplexes have been dissociated have employed CID, therehave been a few reports of the successful application of otheractivation techniques, notably SID,112 BIRD,100,112,130 andECD.117 SID can be viewed almost as an extension of CID,but with a solid surface rather than gas molecules being thecolliding neutral. The immediate benefits of SID in ef-fectively increasing the mass of the collision partner arehowever difficult to realize, and it is only very recently thatthe first examples of successful SID of multiprotein com-plexes have been demonstrated. Wysocki and co-workershave implemented an SID system into a Q-ToF, to providean alternative technique to CID.131 An initial study usingnonspecific cytochrome c dimers compared the symmetryof charge partitioning of the product ions created using SIDor CID as the activation technique.112 In the case of the 11+

ion, asymmetric partitioning was observed by CID whereassymmetric dissociation was observed by SID. The authorssuggest that this could be explained by the much faster timescale of activation of SID (on the order of picoseconds)relative to CID (tens to hundreds of microseconds). SID oflarger, specific, protein complexes has also revealed thatsymmetric pathways can be accessed: SID of tetramers ofTTR and concanavalin A revealed a significant amount ofdimeric dissociation products.132 The dodecameric sHSPsTaHSP16.9 (dissociation data shown in Figure 8) andPaHSP18.1 dissociated with the same asymmetry in massas observed using CID, though differences in the chargepartitioning were observed.131

ECD has seemed an attractive approach to the dissociation

of protein complexes, as several reports have suggested thatECD can initiate covalent fragmentation while maintainingnoncovalent interactions.133-135 This has been successfullyapplied to locate a ligand binding site on a protein implicatedin Parkinsons disease, R-synuclein.136 ECD of larger proteincomplexes has also recently been implemented on an FT-ICR system.117 A relatively intense signal for charge-reducedproducts and some limited fragmentation was observed, withanalytically useful ECD product ions only being producedfrom a few charge states. In those cases where fragment ionswere observed, they were identified as resulting from bothdimer and monomer ions ejected from the heptamericprecursor ion.

BIRD, an activation technique where ions are slowly

heated by absorption of blackbody photons, is an alternativeapproach to ion activation typically implemented on ion trap-type mass analyzers.137,138 Due to the intrinsically tempera-ture-resolved nature of the BIRD approach, temperature-dependent rate constants can be easily extracted from thedissociation data acquired. The first study to employ BIRDfor gas-phase dissociation of a multiprotein complex wasreported on the pentameric Shiga-like Toxin I.100 Asymmetricdissociation into monomers and tetramers was observed, evenwith this activation happening over seconds, and Arrheniusparameters for the dissociation were determined. Of particularnote are the remarkably large pre-exponential factors deter-mined for the dissociation, which indicate that the loss ofthe monomeric subunit results in a substantial favorable

entropy gain. This is consistent with a dissociation mecha-nism that proceeds via the unfolding of a monomeric unit,as originally suggested by Smith et al.5 Further insights intothe mechanism of dissociation of multimeric proteins130,139,140

and the stability of the shiga toxins101,141 have been obtainedby additional BIRD experiments.

4.3. Use of Gas-Phase Dissociation

Strikingly, despite the different mechanics and time scalesinvolved in the various methods described above, mostproduct ions generated by protein complex dissociationremain very similar. The dominant dissociation pathway ishighly asymmetric with respect to mass (Scheme 1) and verydifferent from that which is observed in solution.19 In thelast year, however, gas-phase dissociation pathways havebeen observed in which the loss of a highly chargedmonomer ion is not the chief decay pathway.114,132 In thesetwo cases, noncovalently bound tetrameric proteins werefound to undergo symmetric dissociation into dimers. Suchexceptions provide us with the opportunity to increase ourunderstanding, and possibly manipulate the mechanism ofprotein complex dissociation in the gas phase. Based on an

increased understanding, a future challenge is to use gas-phase activation and dissociation to establish the location ofsubunits within protein interaction networks.8,15

As more information on the mechanism of dissociationcomes to light, it is becoming clear that considerablestructural information can be obtained for protein complexesvia such dissociative approaches.31,32 Information can begenerated regarding the nature of intra- and inter-subunitinteractions, on both global and local levels, as well as detailsof the overall organization of subunits within an oligomer(Figure 9).14 Coupled with the thermodynamic informationattainable from BIRD,100 and the differing dissociationpathways now being accessed,112,114,132 the extra dimensionafforded by dissociative approaches is as an essential part

of future MS investigations in structural proteomics.

5. Companion Technologies for MS of ProteinComplexes

In addition to insight provided by the mass measurementof intact protein complexes and their fragments in the gasphase, technologies can be employed in conjunction withMS to provide further insight into their composition,structure, dynamics, and topology. In this section, we limitour scope to those technologies that have been used inconjunction with MS to investigate protein-protein interac-tions. We have organized these technologies according tothe order in which they occur relative to the typical MS

experiment (Figure 10). Some of the technologies, forexample hydrogen/deuterium exchange, are long established,while others have only recently been coupled to MS;however, in all cases, the technology described providesfurther information on some aspect of the protein complex.For instance, affinity-based purification not only providesan effective means of obtaining protein complexes for directanalysis by MS but also provides information on bindingpartners within uncharacterized protein complexes expressedat endogenous levels. While many of the technologiesdescribed in this section do not involve the detection of intactcomplexes themselves, though in many cases this extensionis theoretically possible, the information garnered from theirapplication can provide complementary insight into protein



16/24

complex structure and function. In some cases, thesetechnologies overlap considerably in terms of their capabili-ties; however, in general Figure 10 demonstrates the comple-mentarity of the enabling technologies covered here. Whilethe progress of MS as described in the previous sections ofthis review will play an important role in proteomics, it iswhen coupled with these companion technologies that thefull scope and potential of MS in this field can be realized.

5.1. Sample Isolation Technologies

The majority of MS studies of protein complexes havebeen performed using recombinantly expressed proteins,purified using standardized molecular biology procedures.

However, an exciting frontier in MS is the isolation andanalysis of previously unknown protein-protein complexesdirectly from cells at endogenous levels of expression.Tandem affinity purification (TAP)142,143 is a widely usedmethod for MS analysis of such protein assemblies.144 Inthis experiment, a protein is expressed containing a tagconsisting of two sections, at least one of which is chemicallycleavable. Both sections exhibit a high degree of affinitytoward resin-bound substrates.145 Various additions andrefinements have been made to the original TAP-tagprotocol.146-149 For example, several popular variants of thestrategy incorporate polyhistidine tags (for Ni2+ bindingaffinity) as either the primary or secondary section of thetag.146,147

Figure 9. Applications of collisional activation to the study ofprotein complexes. When a protein complex is collisionallyactivated, both the reaction products and pathway allow conclusionsto be drawn about its structure. Information as to oligomericcomposition (green) can be obtained from the identity of dissocia-tion products (purple), as well as from beneficial effects regardingthe parent ion (pink). A detailed examination of the reactionpathway can reveal certain parameters (blue) which can allow thedetermination of interaction strengths (red) and information as tooligomeric organization and size (orange).

Figure 10. Technologies associated with MS for the study of pro-tein complexes. Chart demonstrating the benefits of some of thenumerous technologies which have been coupled to MS for thestudy of protein complexes. Here they are classified in their orderrelative to the MS experiment, namely under sample isolation,sample labeling, and online pre- and postionization categories.



17/24

Since the initial landmark TAP-tag-MS experiments werereported,150,151 several applications of the technology havebeen reported. These experiments have provided valuabledata on the global organization of proteins for severalorganisms including examples from yeast147,152-154 andhuman155 cells. Typically, these experiments follow protocolssimilar to those used in bottom-up proteomics experiments.Proteins are isolated after TAP-tag purification, separatedby gel-electrophoresis, enzymatically digested, and then

identified by a combination of accurate peptide masses andsequence tags generated by tandem MS. This isolationstrategy has been used to generate the most comprehensivemap of the yeast interactome, totaling 547 protein-proteininteractions, averaging 4.9 proteins per complex.156

In addition to defining interacting proteins within a givenorganism, several studies have taken a more focused ap-proach: using affinity purification in conjunction with MSto study a single signaling pathway, a single regulationsystem, or even a single protein complex. As before, thecomposition of the protein complex or complexes involvedis frequently the principal target of the analysis. Using suchan approach, the components of the yeast nucleoporecomplex were identified using MALDI-ToF-ToF analysis

following TAP-tag isolation.157 Isolation of intact proteincomplexes using the TAP-tag strategy and MS analysis ofthe noncovalent complex is also having important implica-tions for structural genomics. By studying MS of intactprotein assemblies, along with subcomplexes generated undermildly denaturing conditions, a high-confidence model forthe subunit architecture of the yeast exosome was recentlyproposed, derived entirely from MS experiments.15

The ability to examine protein complexes expressed atendogenous levels has enormous implications for MS instructural proteomics. The high sensitivity and consequentlylow sample requirement relative to other structural biologytechniques means that MS is well placed to provide initiallow-resolution structural data on novel protein complexes.

Moreover, the fact that substoichiometric binding of proteinsubunits is readily apparent in spectra enables this techniqueto be applied to heterogeneous complexes that are often notamenable to other structural biology approaches.

5.2. Sample Labeling Technologies

Nuclear magnetic resonance spectroscopy (NMR) or X-raydiffraction analysis requires relatively high concentrationsof homogeneous protein complexes for successful structuredeterminations. In addition, NMR and X-ray diffraction aredifficult to automate and operate in a high-throughput mode.Consequently, in situations where high-resolution structuresare difficult to obtain due to the scarcity or low purity of

the available protein complex, MS can contribute signifi-cantly to structural data. A general MS approach towardobtaining low-resolution structural data involves the use ofchemical labeling, with the most common forms beingchemical cross-linking, hydrogen/deuterium exchange, andoxidati

Benesch et al. - 2007 - Protein complexes in the gas phase technology for structural genomics and proteomics. - Chemical reviews

Documents